International Differences in Well-Being
Series in Positive Psychology Christopher Peterson, Series Editor Well-Being for Public Policy Ed Diener, Richard E. Lucas, Ulrich Schimmack, and John Helliwell Oxford Handbook of Methods in Positive Psychology Anthony D. Ong A Primer in Positive Psychology Christopher Peterson A Life Worth Living: Contributions to Positive Psychology Mihaly Csikszentmihalyi Handbook of Positive Psychology, 2nd Edition Shane J. Lopez and C. R. Snyder
International Differences in Well-Being
Edited by Ed Diener John F. Helliwell Daniel Kahneman
1
2010
1 Oxford University Press, Inc., publishes works that further Oxford University’s objective of excellence in research, scholarship, and education. Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With offices in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam
Copyright 2010 by Oxford University Press, Inc. Published by Oxford University Press, Inc. 198 Madison Avenue, New York, New York 10016 www.oup.com Oxford is a registered trademark of Oxford University Press, Inc. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press. Library of Congress Cataloging-in-Publication Data Diener, Ed. International differences in well-being / Ed Diener, John F. Helliwell, Daniel Kahneman. p. cm. ISBN 978-0-19-973273-9 1. Well-being—Cross-cultural studies. 2. Quality of life—Cross-cultural studies. 3. Social indicators—Cross-cultural studies. I. Helliwell, John F. II. Kahneman, Daniel, 1934– III. Title. HN25.D54 2010 306.090 0511—dc22 2009026402
9 8 7
6 5 4
3 2 1
Printed in the United States of America on acid-free paper
Acknowledgments
We gratefully acknowledge the support of The William and Flora Hewlett Foundation (grant 2007–1180), Richard Suzman and the National Institute of Aging of the National Institute of Health (grant P30 AG024928), and the Roybal Centers for Translational Research on Aging for their generous support of the conference on which this book was built. We also thank Katherine Ryan for her diligent and thoughtful work putting together the volume, and Debbie Nexon for her invaluable assistance in organizing the Princeton Conference.
v
This page intentionally left blank
Contents
Introduction
ix
Section I. Measuring Well-Being in an International Context 1
2
3
Income’s Association with Judgments of Life Versus Feelings Ed Diener, Daniel Kahneman, William Tov, and Raksha Arora
3
The Structure of Well-Being in Two Cities: Life Satisfaction and Experienced Happiness in Columbus, Ohio; and Rennes, France Daniel Kahneman, David A. Schkade, Claude Fischler, Alan B. Krueger, and Amy Krilla
16
Culture and Well-Being: Conceptual and Methodological Issues Shigehiro Oishi
34
4
Life Satisfaction Arie Kapteyn, James P. Smith, and Arthur van Soest
70
5
Life (Evaluation), HIV/AIDS, and Death in Africa Angus Deaton, Jane Fortson, and Robert Tortora
105
Section II. International Comparisons of Income and Well-Being Through Time 6
Does Relative Income Matter? Are the Critics Right? R. Layard, G. Mayraz, and S. Nickell vii
139
viii
Contents
7
Happiness and Economic Growth: Does the Cross Section Predict Time Trends? Evidence from Developing Countries Richard A. Easterlin and Onnicha Sawangfa
8
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’ Rafael Di Tella and Robert MacCulloch
9
The Easterlin and Other Paradoxes: Why Both Sides of the Debate May Be Correct Carol Graham, Soumya Chattopadhyay, and Mario Picon
166
217
247
Section III. International Differences in the Social Context of Well-Being 10
International Evidence on the Social Context of Well-Being John F. Helliwell, Chris Barrington-Leigh, Anthony Harris, and Haifang Huang
291
11
How Universal Is Happiness? Ruut Veenhoven
12
Faith and Freedom: Traditional and Modern Ways to Happiness Ronald F. Inglehart
351
The Impact of Time Spent Working and Job Fit on Well-Being Around the World James K. Harter and Raksha Arora
398
13
14
Index
Work, Jobs, and Well-Being Across the Millennium Andrew E. Clark
328
436
469
Introduction
A decade ago Kahneman, Diener, & Schwartz (1999) assembled thirty contributions from around the world to make a case for the scientific study of well-being. They argued in their introduction that the new field would require changing the balance of research efforts in psychology to pay more attention to the sources and consequences of positive states of mind, and relatively less to the study of depressed states of mind. To emphasize this difference, they suggested describing the new field as ‘‘hedonic psychology.’’ A central feature of hedonic psychology, and of the broader science of well-being of which it would be a part, would be to take subjective assessments of emotions and of life as a whole more seriously than they had been during the long reigns of behaviorism in psychology and of revealed preference in economics. There has been in the intervening years a strong growth of interest in the multi-disciplinary science of well-being. This growth, of breadth and scale, makes it impractical for a single conference and volume to do full justice to the current state of the field. Our chosen focus for this volume, and the Princeton conference at which the initial draft chapters were presented and discussed, is the measurement and explanation of international differences in well-being. There were a number of central questions: 1. What are the best ways of measuring well-being to facilitate international comparisons? 2. How much do various types of well-being measures differ among individuals, across countries, and over time, and why? 3. Do various measures of well-being, and the circumstances that appear to contribute to their explanation, differ in the extent to which their variation is within rather than among nations? ix
x
Introduction
In the light of the answers to these general questions, two more specific questions were the focus of most of the research presented in this volume: 1. How much does income matter in explaining differences in well-being within countries, across countries, and over time? 2. Are measured international differences in well-being explained by common reactions to differences in local circumstances—for example, social conditions—or by culturally different assessments of the same circumstances of life? If both factors play a part, where does the balance lie?
In attempting to answer these questions, the chapter authors have been aided by the growing availability of internationally comparable measures of well-being and their likely correlates. Earlier work was made possible principally by the successive waves of the European and World Values Surveys (WVS). More recently the Gallup Organization sponsored a continuing evaluation of well-being through annual samples of 1,000 randomly selected respondents in each of more than 140 countries. Both the WVS and Gallup World Poll have used different measures of well-being, thus making it easier to answer some of the key questions outlined above. Without the wide availability of internationally comparable data, the questions posed above would have had to be answered by guesswork and conjecture. Although the answers we suggest here were made possible by access to these data, it is fair to say that all the chapters in this volume can be regarded as reports of work in progress. There are many unsettled issues that will require more data and more analysis, with the structure of surveys and studies adjusted to build on and test ideas suggested by the results now available. Some at the conference suggested that we should determine if there was enough consensus apparent in the data and discussions to support a ‘‘Princeton Manifesto’’ of suggestions for future conduct of international well-being surveys and analysis. Following are several of the candidates for inclusion in such a document. First, there was a strong consensus that it was important that several measures of subjective well-being need to be comparably collected to better understand the nature and consequences of international differences in subjective well-being. Three types of difference were considered of primary importance. One is between life evaluations and measures of mood or emotion. The second relates to the frequency of assessments, and the third embraces current versus retrospective evaluations. Many combinations are possible, and further research with comparable data should guide decisions about which particular measures and collection
Introduction
xi
methods can be combined to produce the most useful assessments at a reasonable cost. Accepting, collecting, and using multiple measures of well-being will be of greatest value if surveyors always try to adopt standard terminology, while continuing to test its validity. Research will be much aided by the use, where possible, of multiple measures within the same survey or set of studies, in order to better understand their similarities and differences and to examine how to best combine the insights from each. At this still-early stage of an empirical science of well-being, it would be unwise to try to reduce subjective well-being to a single measure. On those occasions where more general surveys might be able to include only a single measure, however, this is much better than nothing. Adding well-being questions or modules to existing mainline surveys would provide more frequent, geographically fine-grained, and contextually rich assessments of subjective well-being than would otherwise be available. Second, there were discussions and some general consensus about terminology. The core idea was to use well-being as the generic name for all well-being measures. The science of well-being has for many researchers, authors, and reporters been the science of ‘‘happiness.’’ However, this term is imprecise in referring to a variety of specific different states, as well as to the entire area of well-being. While there are good reasons for using the term happiness, from marketing through intuitive appeal to a desire to have a title that carries an important message, there remains a corresponding need for scientific precision to keep the research both understandable and in accord with the phenomena that we now understand are to some degree separable. Using our preferred terminology, measures of ‘‘happiness’’ would include only those derived from questions that explicitly ask about happiness. Happiness measures are always measures of well-being, but not vice versa. Other forms of well-being include evaluations of life (e.g., life satisfaction) and specific domains of life such as job satisfaction, positive feelings, and absence of negative feelings. We are concerned in this volume with two main sorts of international differences: the first relating to international differences in average well-being (and their explanations), and the second relating to differences among nations in the ways in which people frame their experiences, and in the ways in which life circumstances influence their reported well-being. To help answer the question about the extent to which measures of wellbeing and their determinants vary within and among nations, Figure 1 shows the international (between-nation) shares of total variance of the individual responses from the first three waves (2006–2008) of the Gallup World Poll. These ratios are all within the range from zero to one, with the
Introduction
Proportion of variance Due to Nations
xii
0.7 0.6 0.5 0.4 0.3 0.2 0.1
H
us
ho g(
lo
ou
se
ho
ld
co
nv en Sa eho ien ce tis ld s Im fa in po cti co rta on me nc wi ) th e Pe of life rc ep Ca relig tio nt io n n ril of lad N c At ot or de en ten rup r ou d t gh ed ion m chu o D ne rch on y Li (fo a fe t ch ed od) m o Fr on i ie ce nd fre ey s ed t H el o c om pe ou d nt a o Af stra n fe ct nge D ba r Ti ona lan m te ce e us d ti m e fre e ed om
0
Figure 1 Between-Nation Shares of Variance.
former applying to variables having identical distributions in all countries (freedom to spend one’s time as one wishes, at .04, is the closest among the variables we have chosen), and 1.0 for the variables measured only at the national level, and hence having the same value for each individual within the same country. Because the circumstances of life differ much more between countries than do patterns of emotion, we would expect to find the between-country share of variance to be higher for life evaluations than it is for measures of emotion. That difference is apparent in Figure 1. Comparing the three different measures of subjective well-being, life evaluations (life satisfaction and the Cantril Ladder) have between-country shares of variance exceeding .25, whereas affect balance has a between-country share just below .06. The emotions that are averaged in the affect balance—enjoyment, laughter, worry, sadness, depression, and anger during the preceding day—are largely universal in their incidence, with relatively little difference from country to country. By contrast, there is much more international variation in average life assessments. When respondents from more than 140 countries covered in the first three waves of the Gallup World Poll were asked to evaluate their lives as a whole (using the Cantril Ladder, on a scale of 0 to 10), national averages range from an average of 3.3 for the bottom group of Togo, Burundi, Sierra Leone, and Zimbabwe, to an average of almost 7.7 for the top four: Denmark, Finland, Norway, and the Netherlands.
Introduction
xiii
Is it really possible that people living in the countries with the lowest life evaluations could have had on average as much enjoyment and as little sadness as those living in the highest-ranking countries? The Gallup data say no. Across all countries, the correlation between national averages of affect balance and life evaluations is +.58. For the bottom four life-evaluation countries, the average affect balance is .32, compared to .65 for the four top countries. Thus emotions and life evaluations are telling a broadly consistent story about the level of well-being. Figure 1 shows that some life circumstances, such as the possession of household conveniences, and (the log of) household income, are even more international in their structure than are the two life-evaluation measures. Household consumption levels, as represented by the presence of a variety of household conveniences (see Chapter 1 for the details), differ more among nations than does household real income, and both have more between-country shares of variance than do the life-evaluation measures. One measure of basic needs—having sufficient money for food—has a larger degree of within-country variation. Measures of the social context, while also correlated with life evaluations, have more of their variation within than among countries. Importance of religion and religious observance both differ more among countries than does having friends or family to count on in times of need. The three measures of individual benevolence, all positively correlated with life evaluations (see Chapter 10), have declining international shares as they move from donating money through helping a stranger to donating time. Self-assessed freedoms are importantly correlated with life evaluations (see Chapters 1 and 10), with freedom to make life choices being more international than freedom to choose how one spends one’s time. Thus Figure 1 shows that various measures of well-being, and the circumstances that appear to contribute to their explanation, differ very greatly—by more than ten times in some cases—in the extent to which their variation is between nations. A natural question at this stage is whether the countries at the bottom of the life-evaluation ladder are there because of their obviously low average incomes, or because their material disadvantages are accompanied by social ones. The Gallup data show clearly that the latter is the case. Even though the social variables have relatively more within-country variance, their international differences are nonetheless large enough to contribute significantly to the explanation of average life evaluations across countries. For example, in the four lowest-ranking countries, only 55% of the respondents have
xiv
Introduction
relatives or friends that they can count on to help in times of trouble, compared to over 95% in the four top countries. Turning to the more specific empirical questions we have posed above, it is perhaps too soon to be offering conclusions in ‘‘manifesto’’ terms, or even in terms of consensus, because some of the chapter authors present different and diverging lines of evidence. It may nonetheless be useful to summarize some of the emerging outlines of research results reported in the chapters to follow. Below we present a brief outline of the chapters of the book. Although the book is divided into sections, readers will note that most chapters have implicit and explicit links to research reported in other sections. The first section primarily contains chapters assessing the nature and cross-cultural validity of different measures of well-being. The chapters in the second section (and some of those in other sections) attempt to assess the links between income and well-being, both across countries, and over time within countries. In Section III are grouped chapters in which non-economic determinants of well-being are assessed, both in the workplace and in life as a whole. In Section I of this book, the chapters address issues related to assessing well-being across cultures. The chapter authors ask about the different forms of well-being, and whether they show different patterns in various cultures. The chapters also raise the question of how to correct for biases in responding that might occur in different cultures, making comparisons of well-being results problematical. Thus, the first section of the book is ‘‘basic’’ in the sense of addressing definitional and measurement issues. In the first chapter in Section I, Diener, Kahneman, Tov, and Arora show that one should separately analyze judgments of life versus ongoing positive and negative feelings. The authors report that both income levels and changes in income are more strongly associated with people’s judgments of life than with their daily feelings of well-being. The Kahneman, Schkade, Fischler, Krueger, and Krilla chapter contrasts on-line feelings linked to time use and activities with global judgments of life. For example, they found that the rich and married are somewhat more satisfied with their lives, but not much happier from moment to moment. They also discovered that feelings of well-being in a city in France are similar to those in a city in the United States, and that time-use patterns, with several notable exceptions, were also remarkably similar across the two cultures. Kahneman et al. offer the theoretical suggestion that time-use patterns are very important to well-being, and that attention is also important. Thus, over time, people may attend less to situations that recur frequently in their lives, although habitual preoccupations as well as major life events can move people’s attention away from their immediate
Introduction
xv
situations, and therefore make the immediate situation less important to their feelings of well-being than is normally the case. Finally, the authors suggest that one reason we have overestimated the positive effects of states such as marriage or being rich is that we have, like our research subjects, focused primarily on the benefits of these states and have ignored the costs involved. Oishi looks more deeply at cultural differences in what is considered ‘‘well-being,’’ and the measurement and methodological issues arising from this variety in definitions. Oishi recommends that various forms of well-being be measured comparably in many nations so we may better understand differences in levels and correlates across societies. Next, Kapteyn, Smith, and Soest present a promising approach for calibrating well-being scores across cultures so that they are more likely to have the same meaning. Using this approach, they find that social relations are more important and income less important to global life satisfaction in both the United States and the Netherlands. They also find that income is more highly associated with life satisfaction in the United States than in the Netherlands. Finally, Deaton, Fortson, and Tortora use Gallup World Poll data to assess the well-being consequences of family deaths from various diseases, both within and among African countries. In most cases they fail to find the expected large negative effects on life evaluations of recent diseaserelated deaths in the family, although there are large effects on sadness and depression. Africans appear to care about poverty much more than they care about health, a result that poses serious problems for health policy in Africa. The authors consider the use of their results to derive monetary estimates of the value of life, and conclude, for a variety of reasons, that this would not be appropriate. In Section II of the book, the chapters are focused even more strongly on the correlations between income and well-being within and across societies, although chapters addressing this topic can be found in the first and last sections as well. The chapter authors address the degree to which income correlates with well-being in different cultures, and importantly, the degree to which income change correlates with changes in well-being. In 1974, Easterlin framed what has become known as the ‘‘Easterlin Paradox,’’ the seemingly contradictory findings that changes in income are not associated with changes in well-being, although cross-sectional analyses show that higher-income individuals do report greater well-being. This classic paper has framed the debate about income and ‘‘happiness’’ for the last several decades. The first chapter in this section, by Layard, Mayraz, and Nickell, uses time-series data from the richer industrial countries of Europe and North America to compare the importance of absolute and relative incomes
xvi
Introduction
as determinants of subjective well-being. They conclude that incomecomparison effects, rather than adaptation, provide the primary reason why, in the United States and many countries of Western Europe, life evaluations have not improved despite substantial increases in average per capita real incomes. In the Easterlin and Sawangfa chapter, time trends of income and wellbeing are compared for 13 developing countries. They find no significant relationship between changes in subjective well-being and those that would be predicted from cross-sectional relationships. Nor do they find evidence, across their sample countries, that the countries with higher growth rates have had greater increases in subjective well-being. Making use of the basic-needs hypothesis, Di Tella and MacCulloch conclude that the subjective well-being effects of income changes in wealthy nations are small, owing to adaptation, although such adaptation takes a number of years. Graham, Chattopadyay, and Picon review evidence showing that the effects of income growth depend on a number of factors, from the time frame selected, to the specification of the income variable, to the rates of economic growth. In addition, Graham et al. suggest the importance of aspiration levels, leading to the paradoxes of happy peasants and frustrated achievers. Thus, whether income growth is accompanied by increases in well-being seems to depend on a number of other variables, such as the type of well-being, the rate of changes in people’s aspirations, and previous levels of economic development. Chapters in other sections also focus on the empirical linkages between income and life satisfaction. In particular, Diener et al. in Section I and Inglehart in Section III find positive linkages over time between changes in life evaluations and changes in per capita incomes. In addition, the Gallup World Poll cross-section data analyzed by Helliwell et al., as well as in recent papers by Stevenson & Wolfers (2008) and Deaton (2008), show insignificant differences between within-country and between-countries estimates of the effects of income on life evaluations. Because the analyses are based on a different sampling of nations, different measures, and different time frames and types of analysis, we leave it to the reader to evaluate the current state of play in the continuing debate about the relative size and significance of the linkages between income changes and changes in well-being. In our view, the linkages are complicated and contextdependent, with the debate best regarded as unsettled. Unscrambling these complexities will be greatly aided by systematic periodic collection of a variety of measures of subjective well-being. In Section III of this volume are chapters that are primarily concerned with how social factors influence well-being across nations. The Helliwell,
Introduction
xvii
Barrington-Leigh, Harris, and Huang chapter demonstrates the importance of institutional and social variables for well-being, and calculates the large income-equivalent values (or ‘‘compensating differentials’’) of several aspects of the social context. They also find that Cantril’s Self-Anchoring Striving Scale (generally described here as a ladder) and a measure of life satisfaction provide consistent explanations of the determinants of wellbeing, and can be combined to provide an even more robust explanation of well-being. The authors also conclude that the meaning and causes of a good life are largely the same across cultures, with the large cross-country differences in life evaluations being driven by correspondingly large differences in life circumstances. In the following chapter, Veenhoven echoes this sentiment, concluding that well-being depends on similar conditions across the globe, reflecting the degree to which human needs are met. Inglehart focuses on personal and political freedom as important causes of well-being. He concludes that income increases can raise well-being, but that rising social tolerance and political freedom have been more important causes of increasing well-being. He also discusses the impact of belief systems and ideologies; for example, religiosity, on well-being. Harter and Arora describe analyses of the Gallup World Poll that show that good job fit—being in work that allows one to use one’s best skills—predicts both life evaluations and daily affect. Furthermore, working more hours is more detrimental to feelings of well-being for those with poor job fit. Clark pursues the nature of satisfying work, and trends over time. He concludes that there was a dip in job satisfaction during the 1990s, but that positive attitudes bounced back by 2005. However, Clark also finds that people now give more importance to the social aspects of jobs. Most people say they would prefer to be self-employed, although the rates of self-employment have dropped over time, suggesting to Clark that the barriers to selfemployment have grown. Based on the results in all chapters, and especially in the three chapters to which they have themselves directly contributed, the editors conclude that international differences in reported well-being are not just larger for life evaluations than for reports of affective experience, but are more readily explained by differing economic and social circumstances. Furthermore, we find that international differences in life evaluations are quite readily explicable in terms of shared evaluations of life circumstances across societies. Hence, the large international differences in average life-evaluation scores (whether measured by satisfaction with life or the Cantril Ladder of life) are to be explained primarily by the different economic and social circumstances in which people live, and much less by differences in their ways of viewing life.
xviii
Introduction
References Deaton, A. (2008). Income, health and well-being around the world: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22, 53–72. Easterlin, R.A. (1974). Does economic growth improve the human lot? In Paul A. David and Melvin W. Reder (Eds.), Nations and households in economic growth: Essays in honor of Moses Abramovitz. New York: Academic Press, Inc. Kahneman, D., Diener, E., & Schwarz, N. (Eds.). (1999). Well-being: The foundations of hedonic psychology. New York: Russell Sage Foundation. Stevenson, B., & Wolfers, J. (2008). Economic growth and subjective well-being: Reassessing the Easterlin Paradox. Brookings papers on economic activity, Spring 2008, 1–87.
Section I
Measuring Well-Being in an International Context
This page intentionally left blank
Chapter 1 Income’s Association with Judgments of Life Versus Feelings Ed Dienera, Daniel Kahnemanb, William Tovc, and Raksha Arorad a
University of Illinois Princeton University c Singapore Management University d The Gallup Organization b
Attention has recently been drawn to the fact that ‘‘happiness’’ is actually not a single entity, and can be divided into distinct elements. Kahneman (1999) suggested that global judgments such as an evaluation of ‘‘life satisfaction’’ computed and reported at a single moment in time are fundamentally different than the pleasantness of people’s emotional lives. In support of this distinction, Lucas, Diener, & Suh (1996) found that various forms of subjective well-being are empirically separate. Thus, it is no longer sufficient to simply discuss and study well-being; the various forms of well-being must be assessed and analyzed. The major distinction that Kahneman described was between global evaluative judgments and people’s feelings of pleasure and displeasure summated over time. We suggest that the various self-report measures of subjective well-being are saturated to varying degrees with judgment and affect. Although perhaps no well-being measure is totally free of either of these components, it is plausible that a global measure of ‘‘life satisfaction’’ taken at one point in time might be more heavily weighted with judgment, whereas reports of ‘‘happiness’’ might be more saturated with affect. In the present study discussed in this chapter, we were fortunate to have two measures that appeared a priori to be toward the two ends of the judgment–affect dimension, Cantril’s Self-Anchoring Striving Scale (1965), which for brevity we refer to as the ‘‘Ladder’’ throughout this chapter, as well as a report of yesterday’s affect. We analyzed two additional measures, life satisfaction and happiness, that we predicted would fall between the Ladder and affect scales on the judgment–affect dimension, but each closer to the opposite poles. 3
4
Section I: Measuring Well-Being in an International Context
One goal of our study was to determine whether judgment and affect measures perform differently and to identify the mix of the two processes reflected in certain scales. We examined the intercorrelations among the measures at both the individual and national levels, as well as their correlations with external variables such as income. We also examined how the measures changed over time in response to movements in income. In this way we aimed to explore the correlates of affect versus judgment measures of well-being. In a classic 1974 article, Richard Easterlin asked whether economic growth improves ‘‘the human lot,’’ and he focused on ‘‘happiness’’ to answer the question. The ‘‘Easterlin paradox’’ is the paradoxical fact that differences between people in income are usually correlated with reports of well-being, but as national incomes grew, there was seldom substantial growth in wellbeing. Much debate has ensued about whether nations have in fact risen in average well-being over time in response to increasing income. Hagerty & Veenhoven (2003), Stevenson & Wolfers (2008), and Inglehart, Foa, Peterson, & Welzel (2008) all claimed that, on the whole, the evidence suggests that there was increasing subjective well-being in many nations, and that this was associated with rising incomes. An examination of the data reported in these articles, however, indicates variability in the findings. For example, in a response to Hagerty and Veenhoven, Easterlin (2005) pointed out that many nations in fact grew in income over time and did not increase in well-being. Easterlin presented reports of happiness over the decades in the United States, which were essentially flat, and contrasted that pattern with the substantial economic growth the country experienced during the same period of time. He concluded that across nations there are ‘‘disparate trends in happiness, suggesting that factors other than growth in income are responsible for the differential trends in happiness’’ (p. 429). In response, Veenhoven & Hagerty (2006) suggest that, on average, happiness increases occurred in nations where income rose the most. Stevenson and Wolfers argue that increasing income has led to increases in happiness, but they also point to the substantial statistical uncertainty in a few of their conclusions. Inglehart (2008) suggested that life satisfaction might be more influenced by economic conditions than is happiness, and this suggestion forms the starting point for our income analyses. We examine the possibility that various forms of well-being vary in their responsiveness to income change. Specifically, it could be that judgments and affective well-being vary in how much they are associated with economic growth. We analyze the correlation of four well-being variables that we hypothesize lie along the judgment versus affect dimension, with several economic variables— income, income change, and the ownership of modern conveniences such as televisions. Thus, we explore whether some of the past differences in
Income’s Association with Judgments of Life Versus Feelings
5
conclusions about the money–happiness relation are due to the differential association of various types of well-being with income. To examine income change and well-being change, we examined income changes over the interval from an early survey to a later survey for each type of well-being. The minimum for inclusion in our analysis was a period greater than or equal to five years, and the average for each of the measures was an interval of several decades. We selected the surveys for each type of well-being measure that were the most distant from each other in time, and used the incomes for the corresponding years for each nation. We were fortunate in this study to have a representative sample of virtually the entire adult population of the world. Unlike many previous studies, the Gallup World Poll (GWP) includes many less economically developed nations, and a representative sample of rural residents outside the major metropolitan areas of these nations. We were also fortunate that the survey included both a global judgment of life measure and an assessment of emotions experienced ‘‘yesterday.’’ One issue with past research is that the wording of questions was different in various surveys conducted over the decades. Thus, we focused on surveys that used very similar or identical wording and the same response formats. In our analyses we relied most heavily on the analyses of national data, not individual data, for several reasons. First, we have longitudinal data over time only for nations, and this is a keystone of our analyses. Second, the Easterlin claim about income change is that, in the aggregate, at the societal level income increases do not raise well-being, because as the income of everyone rises, the standard for adequate income also rises. Thus, our analyses focus on the national level, but we also examined individual data in the Gallup World Poll to determine whether the same dimensionality can be found in the well-being measures, and whether the predictors are the same as at the national level. In sum, there were two goals in our study. First, we analyzed measures of well-being to determine whether they are separable, but associated in a way reflecting an underlying dimension related to global evaluative judgments versus the ongoing experience of affect. Second, we examined income and other predictors to determine whether they are most related cross-sectionally to the judgment versus affective ends of the well-being dimension, and whether income changes relate more to judgment or to affect.
Methods
The Gallup Organization initiated its World Poll in 2005, and the first wave, conducted from late 2005 to 2006, includes representative surveys of
6
Section I: Measuring Well-Being in an International Context
132 societies accounting for about 96 percent of the world’s population. The poll features a consistent set of standard questions in all surveys and uses nationwide samples (with the exception of Angola, Myanmar, and Cuba, where only urban populations were surveyed; and Afghanistan, where there were representative provinces). Two sampling procedures were used in the World Poll: random-digit-dial (RDD) telephone surveys and face-to-face interviews. The RDD design was used in countries where the vast majority of the population has access to land line telephones. In all other countries, face-to-face surveys were conducted with clusters of households (obtained from census tract listings) serving as the basis for random sampling. The typical World Poll survey in a country consisted of approximately 1,000 respondents. The total sample size for the wave 1 World Poll was 140,658. Wave 2 of the survey (2007) presented an additional well-being question on life satisfaction that was not included in the first wave. Thus, we analyzed the association of the well-being measures at the individual level using wave 2, which included 113,872 respondents from 105 nations. Approximately 55,497 respondents were presented with the life satisfaction question. We used two measures of well-being from wave 1 of the Gallup World Poll. The Ladder measure asks respondents to evaluate their life on a scale from 0 (worst possible) to 10 (best possible). The item queries respondents as to which step on the Ladder they feel that they personally stand at the present time. Another measure assesses the recent experience of emotions; namely, positive feelings (enjoyment and smiling/laughter) and negative feelings (sadness, anger, worry, and depression). To reduce the extent of bias in recalling past experiences, respondents reported (yes or no) whether they experienced lots of these feelings during the previous day. We averaged enjoyment and smiling and subtracted the average of the four negative emotions to create an Affect Balance score. We averaged the individual scores to create nation-level scores after weighting the individual scores by sample demographic weights to bring the national scores as close as possible to representativeness. The additional measure of well-being in wave 2 on Life Satisfaction asked respondents how satisfied they were with their lives on a scale ranging from 0 (Dissatisfied) to 10 (Satisfied). The Gallup World Poll also queried respondents about their ownership of modern household conveniences, and we averaged four of these to create a composite Conveniences score—electricity, telephone, television, and computer. In addition, we analyzed a question that asked how free subjects were in deciding how to spend their time. The question, answered on a binary Yes–No response scale, asked: ‘‘Were you able to choose how you spent your time all day yesterday?’’
Income’s Association with Judgments of Life Versus Feelings
7
In addition to the Gallup World Poll, we also obtained well-being measures for nations from Veenhoven’s (2008) ‘‘World Database on Happiness.’’ We searched for those nations where the same four-point happiness question was asked at two points in time separated by more than five years. When the scale had been administered more than two times, we used the first and last administrations. The question asked: ‘‘Taking all things together, would you say you are: 4 – Very happy, 3 – Quite happy, 2 – Not very happy, or 1 – Not at all happy?’’ We used variants of this scale where the identical scale and responses had been used at both points in time. For example, in some administrations of the scale, the 3 response is labeled ‘‘Pretty happy.’’ We used scores from these variations if the identical scale was used in that nation at two points in time at least five years apart. We also obtained Time 1 scores for nations where the Ladder had been administered previously, and used the oldest date where the 0 to 10 response format was employed. There were often small changes in the Ladder scale from the earlier to the later administration. For instance, the early administration often showed steps ascending a mountain, whereas the later administration simply showed a ladder with steps. For life satisfaction Time 1, we used the oldest administration in each nation where the same 0 to 10 format was used as in the Gallup World Poll. However, in order to increase our number of nations for life satisfaction, we also used instances where the response format was 1 to 10, as long as earlier and later dates were available using the same response scale, which were separated by at least five years. One nation, the Dominican Republic, was dropped from the analyses of the Ladder scale because its score was an extreme outlier, far lower than any score ever reported. The score was so low that we suspect that it is an error, or a temporary response to some acute disaster. Although there are several sources that provide GDP per capita data, they differ in the years for which data are available. Given that our data set spanned over three decades, we sought a single source of GDP per capita estimates rather than drawing from different sources that may differ in terms of their estimation methods. Therefore, we used income data from Maddison (2007), who provided a broad coverage of estimates for GDP per capita adjusted for purchasing power parity in 1990 international dollars.
Results
Our general plan was to first use the large GWP to examine how well-being measures are related to one another cross-sectionally at a point in time. We next analyzed how they relate to predictors such as income. These analyses were computed at both the national and individual levels. In the second set
8
Section I: Measuring Well-Being in an International Context
of analyses, we examined how national changes in income over time are associated with country-level changes in the well-being measures. We also analyzed the magnitude of the changes in well-being. Cross-Sectional Analyses
Our analyses begin with cross-sectional correlations among the well-being measures themselves, to explore their relations to each other. We correlated the nation-level well-being averages for four measures of well-being at the most recent time of the surveys. We further explored at the nation-level the composition of Life Satisfaction and Happiness by predicting them simultaneously with both the Ladder and Affect Balance scores of nations. Life Satisfaction was predicted most strongly by the Ladder score (Beta = .67, p < .01), although Affect Balance also significantly added to the prediction (Beta = .27, p < .01). When the predictors are entered in a sequential manner, the Ladder accounted for 33 percent of the variance in Life Satisfaction beyond Affect Balance, whereas Affect Balance accounts for only 5 percent when the Ladder is entered first. Together these two measures account for 71 percent of the variance in the Life Satisfaction of nations. Happiness was also predicted by the Ladder (Beta .36, p < .01), with the Affect Balance also predicting positively (Beta = .48, p <.01). When entered sequentially, the Ladder accounted for 8 percent of the variance in Happiness beyond Affect Balance, whereas Affect Balance accounted for 15 percent beyond the Ladder. Together Affect Balance and the Ladder predict 57 percent of the variance in the Happiness of nations. The intercorrelations of the well-being variables at the individual level are shown in the lower portion of Table 1.1. Note that the happiness scale was not included in the Gallup World Poll. As can be seen, the correlations are lower than those for nations, perhaps because there is more measurement error and situational variability in individual responses. The intercorrelations of the measures indicate a consistently strong correlation between the Ladder and Life Satisfaction, and a weaker correlation of those variables with Affect Balance. A set of regression analyses in which Life Satisfaction was predicted by the other two variables at the individual level indicated that it was most closely associated with the Ladder, but that Affect Balance added significantly to its prediction as well (all p’s < .001). When the Ladder was entered first as a predictor, it accounted for 32 percent of the variance, and Affect Balance added 3 percent more to the prediction when it was added. In contrast, when Affect Balance was entered first as a predictor, it explained 10 percent of the variance in Life
Income’s Association with Judgments of Life Versus Feelings
TABLE 1.1
9
Intercorrelations of Well-Being Measures
Well-Being Variables
Ladder
Life Satisfaction
Happiness
Across Nations Life Satisfaction Happiness Affect Balance
.82 N = 63 .64 N = 61 .55 N = 126
.68 N = 41 .62 N = 61
.57 N = 55,057 .25 N = 105,126
.31 N = 51,485
.70 N = 60
Across Individuals Life Satisfaction Affect Balance
Note. Correlations across individuals consist only of data from Wave 2 of the Gallup World Poll.
Satisfaction, but the Ladder added 25 percent additional variance. Thus, at the individual level as at the national level, Life Satisfaction was both a judgment and an affect variable, but much more strongly saturated with judgment. Happiness also has a substantial judgment component, but a considerable amount of affect as well. This conclusion is strengthened when the correlates of well-being with other types of measures are examined. We next analyzed the correlations of the four well-being measures with three predictors, and these associations are shown in Table 1.2. All income values were transformed to Log10 values. As can be seen, the Ladder correlated significantly more highly with income and conveniences and significantly lower with choosing how to spend one’s time than did the other SWB variables. In some cases the correlations of Life Satisfaction with the predictors differed from those for Happiness and Affect Balance, and in some cases not. Affect Balance and Happiness never differed significantly from each other. The pattern of correlations clearly indicates that income and conveniences are more strongly associated with well-being judgments, and that feelings tend to be more strongly associated with the perceived freedom to choose how to spend one’s time. As can be seen, the relationship of income to the well-being variables was similar at the two points in time. Although we are uncertain what ‘‘feeling free’’ in terms of spending one’s time indicates, the important point here is that it correlated in a pattern opposite to that of the material variables. Table 1.3 presents the correlations at the individual level of the wellbeing variables with the predictors shown in Table 1.2. Because of the very
10
Section I: Measuring Well-Being in an International Context
TABLE 1.2 Nation-Level Correlates of Well-Being Well-Being Variables
Income Per Capita
Choose How to Spend Time
Possession of Modern Conveniences
.74a N = 19 .83a N = 126
.30a N = 130
.78a N = 91
.69b N = 29 .56b N = 64
.49b N = 63
.42b N = 32
.31c N = 53 .42c N = 60
.43ab N = 61
.03bc N = 33
Affect Balance
.31c
.56b
.14c
(Time 2 only)
N = 122
N = 126
N = 90
Ladder Score Time 1 Time 2 Life Satisfaction Time 1 Time 2 Happiness Time 1 Time 2
Note: Correlations for the same time period in the same column who do not share a subscript letter in common differ by p < .05 or less.
TABLE 1.3 Individual Level Correlates of Well-Being Well-Being Variables Ladder Score Life Satisfaction Affect Balance
Income
Choose How to Spend Time
Possession of Modern Conveniences
.40 N = 77,213 .33 N = 34,771 .14 N = 72,572
.09 N = 109,393 .15 N = 53,353 .32 N = 104,942
.41 N = 93,070 .30 N = 51,892 .13 N = 87,643
All correlations are significant at p < .01, and all correlations in the same column are significantly different from one another at p < .01
large sample sizes, all correlations shown in the table differ significantly from one another by p < .001. The correlations with the material variables and well-being, as well as feelings of autonomy and well-being, all mirror the pattern of associations found at the nation level. The correlations, however, are again lower than those found with the nation-level variables. Taken together, these results indicate that Life Satisfaction is closer to the Ladder, and Happiness is closer to Affect Balance.
Income’s Association with Judgments of Life Versus Feelings
11
Longitudinal Analyses
We examined the correlations of each of the well-being measures with income and income change over periods of time greater than or equal to five years. In Table 1.4, we present the means for income and the well-being variables at the two points in time, as well as the average year of the surveys. As can be seen, on average the surveys were many years apart, with intervals of 37, 21, and 18 years for three wellbeing measures. Furthermore, there were large increases in income over those periods of time. Thus, if rising income has a long-term effect on well-being, it should be apparent during the prosperous periods of time we analyzed. We correlated the long-term change in log per capita income with the change in well-being, and found associations of: Ladder, r = .54, N = 19, p < .05; Life Satisfaction, r = .25, N = 50, p = .08; Happiness, r = –.13, N = 58, p = .34. How large and consistent were the changes in well-being? Because income substantially increased, there ought to be a recognizable overall increase in well-being, not simply a correlation with changes in income, if income influences well-being. All three measures of well-being increased significantly from T1 to T2, all p’s < .01. When the scale score changes are expressed in terms of the between-nation standard deviations in scores, well-being changed the following amount: Ladder = .69 SD units; Life
TABLE 1.4
Means and Standard Deviations of Key Variables for Both Waves
Wave & Variables
Ladder
Life Satisfaction
Happiness
Wave 1 Well-Being
5.60 (1.06)
6.32 (1.28)
2.96 (.29)
Wave 2 Well-Being
6.30 (.96)
6.63 (1.08)
3.08 (.26)
Wave 1 Year
1969 (10.1)
1985 (10.2)
1987 (10.5)
Wave 2 Year
2006 –
2006 (1.8)
2005 (2.8)
Wave 1 GDP/capita
4,524 (2.69)
7,279 (2.02)
7,244 (2.14)
Wave 2 GDP/capita
10,506 (2.52)
11,277 (2.16)
10,513 (2.29)
Number of Nations at Both Waves
19
50
53
Note: For GDP/capita we present the geometric mean (geometric SD) because logged values were used in all analyses.
12
Section I: Measuring Well-Being in an International Context 8 Time 1
7.5
Time 2 Mean Ladder Score
7 6.5 6 5.5 5 4.5 4 3.5 3 2.5
3
3.5 4 Log GDP per Capita
4.5
5
Figure 1.1 National Mean Ladder Scores by Log GDP per Capita at Times 1 (dotted line) and 2 (solid line)
Satisfaction = .25 SD units; and Happiness = .41 SD units. As percentages of the total possible range of the scales, the differences between Times 1 and 2 were: Ladder = 7 percent; Life Satisfaction = 3 percent; and Happiness = 4 percent. Thus, the Ladder changes over time were larger in standard deviation units, but also in terms of moving across more of the range of the scale. In keeping with the conclusions of Inglehart (2009), however, we did find that, in general, well-being rose on average during the several-decade period we studied. Our final analysis examined the regression of GDP on the Ladder at both Time 1 and 2. A comparison of the intercepts of the two regressions provides a test of whether there is adaptation to income. If there is adaptation, the same level of income should provide less well-being at time 2 than at time 1. As can be seen in Figure 1.1, the regressions are effectively identical. Although GDP more than doubled between the two measurements, the relationship between the Ladder and national income did not change. The analyses reported by Deaton (2008) and Stevenson & Wolfers (2008) suggested strongly that in the recent GWP people all over the world compared themselves to a common standard of material wellbeing. The findings of Figure 1.1 suggest further that this standard has not changed appreciably over the last three decades. This does not mean that the standard cannot move, but it does suggest that, at least in this case, there was a relatively consistent material standard for the ideal life in material terms.
Income’s Association with Judgments of Life Versus Feelings
13
Discussion
Our findings indicate that measures of well-being vary along a dimension that is anchored by judgments about one’s life at one end and by affect at the other. Selected measures can be placed on this continuum based on the relative amount they are influenced by the two types of subjective well-being. The Ladder appears to reflect a judgment about one’s life, whereas reports of emotions during the previous day are located toward the other end of the dimension. Life Satisfaction is between the two anchors, but close to the Ladder, primarily reflecting a judgment; and this conclusion is consistent with the findings of Helliwell et al. (2009). Reports of ‘‘happiness’’ also fall toward the middle of the dimension, but closer to the affective end than does Life Satisfaction. We suspect that the feelings end of the dimension is anchored by experience-sampling measures of momentary feelings. Not only do the measures differ in their relations with each other, but they also differ in their strength of association with variables such as income and the ownership of modern conveniences. The Ladder was mostly strongly correlated with these material variables, and the Affect Balance measure was least strongly associated with them. Life Satisfaction was significantly more strongly related to the material variables than was Affect Balance. The strength of associations for Happiness and the material variables was significantly weaker than the Ladder correlations. In contrast to the material variables, feelings of autonomy in everyday life were more strongly associated with affect and less strongly associated with the Ladder. Thus, at both the individual and national levels the pattern of correlations with the predictors confirms the dimensional ordering derived from the intercorrelations of the well-being variables with each other. The correlations suggest that material prosperity is strongly associated with judgments of life but much less correlated with affective well-being. An examination of changes in well-being and income over time again supports the separability of the measures along the judgment–affect dimension. Changes in the Ladder scores over time showed a clear association with changes in income, whereas the strength of this association for happiness and life satisfaction was weaker. Is Easterlin or are his critics correct? Over the long-term, life judgments of life were strongly related to income and rose with income. On the other hand, affect benefited less from long-term rising income. One can point to the increases in well-being that have occurred in most nations, or one can point to a number of nations that have declined in well-being even as their incomes have risen. Clearly, more factors influence well-being than simply changes in income. For example, Inglehart
14
Section I: Measuring Well-Being in an International Context
(2009) in this volume points to the fact that political freedom and income can move in different directions, and therefore produce countering forces on well-being. Our findings are consistent with those of Inglehart in suggesting that the well-being of nations can indeed change over time, and that certain forms of well-being are more likely to change in association with changes in income. In addition, other factors in societies besides income must be considered, such as social trust and urbanization, and psychological factors such as rising aspirations might also play a role. Easterlin was correct in his claim that rising incomes do not invariably increase subjective well-being. However, his critics are correct in their claim that rising incomes have on average been associated with some increases in at least some forms of well-being. The challenge now is to understand when income helps well-being and when it does not, as well as how changes in other societal characteristics influence the various types of well-being. One conclusion that is certain from our findings is that it is no longer productive to talk about income and general ‘‘happiness’’; at the very least, ‘‘well-being’’ must be parsed into the judgmental versus affective components. Whether rising income improves the human lot appears to depend at least in part on the types of well-being being assessed. In a related paper (Kahneman, Diener, Arora, Muller, Harter, & Tov, 2009), we separately analyze positive and negative affect, and show that even for feelings the associations with other variables can differ systematically. The current findings raise many issues for future study. An important question is what are the factors most responsible for changes in affect in nations. The fact that the same regression line described the relationship between the Ladder and GDP in measurements taken, on average, 37 years apart also deserves detailed study. Although we have focused on the different types of well-being measures, more sophisticated measures of wealth, income, and consumption are also needed. It will also be important to determine whether there are factors associated with rising income such as democratization that produce some of the well-being effects, and conversely, whether rapidly rising national incomes are associated with negative changes in some aspects of the quality of life, including residential dislocation. References Cantril, H. (1965). The pattern of human concerns. New Brunswick, NJ: Rutgers University Press. Deaton, A. (2008). Income, health, and well-being around the world: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22, 53–72.
Income’s Association with Judgments of Life Versus Feelings
15
Easterlin, R. A. (1974). Does economic growth improve the human lot? Some empirical evidence. In P. A. David & M. W. Reder (Eds.), Nations and households in economic growth. New York: Academic Press. Easterlin, R. A. (2005). Feeding the illusion of growth and happiness: A reply to Hagerty and Veenhoven. Social Indicators Research, 74, 429–443. Hagerty, M., & Veenhoven, R. (2003). Wealth and happiness revisited—growing national income does go with greater happiness. Social Indicators Research, 64, 1–27. Helliwell, J. F., Barrington-Leigh, C., Harris, A., & Huang, H. (2010). International evidence on the social context of well-being. In E. Diener, J. F. Helliwell (Eds.), & D. Kahneman, International differences in well-being. Oxford, UK: Oxford University Press. Inglehart, R., Foa, R., Peterson, C., & Welzel, C. (2008). Development, freedom, and rising happiness: A global perspective (1981–2007). Perspectives on Psychological Science, 3, 264–285. Inglehart, R. (2010). Faith and freedom: Traditional and modern ways to happiness. In E. Diener, J. F. Helliwell (Eds.), and D. Kahneman, International differences in well-being. Oxford, UK: Oxford University Press. Kahneman, D. (1999). Objective happiness. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology. New York: Russell Sage Foundation. Kahneman, D., Diener, E., Arora, R., Muller, G., Harter, J., & Tov, W. (2009). Prosperity and well-being on Planet Earth. Unpublished paper, Princeton University. Proceedings of the National Academy of Sciences (PNAS), U.S. Lucas, R. E., Diener, E., & Suh, E. (1996). Discriminant validity of well-being measures. Journal of Personality and Social Psychology, 71, 616–628. Maddison, A. (September 3, 2007). Statistics on world population, GDP, and per capita GDP, A.D. 1–2006. In Angus Maddison (faculty homepage). Retrieved November 29, 2008, from http://www.ggdc.net/maddison/. Stevenson, B., & Wolfers. J. (April 15, 2008). Economic growth and subjective well-being: Reassessing the Easterlin Paradox. Brookings Papers on Economic Activity, Spring, 1–87. Veenhoven, R. (2008). World database of happiness, Trends in nations. Rotterdam: Erasmus University, September 10, 2008. Available at http:// worlddatabaseofhappiness.eur.nl/. Veenhoven, R., & Hagerty, M. (2006). Rising happiness in nations 1946–2004. Social Indicators Research, 79, 421–436.
Chapter 2 The Structure of Well-Being in Two Cities: Life Satisfaction and Experienced Happiness in Columbus, Ohio; and Rennes, France Daniel Kahnemana, David A. Schkadeb, Claude Fischlerc, Alan B. Kruegerd, and Amy Krillae a
Princeton University University of California, San Diego c Centre National de la Recherche Scientifique, Paris d Princeton University e Princeton University b
The authors thank the National Institute on Aging, the William and Flora Hewlett Foundation, and Princeton University’s Woodrow Wilson School for financial support; and Dan Gilbert, Sonja Lyubomirsky, Norbert Schwarz, and Arthur Stone for helpful comments. We thank Raksha Arora and the Gallup Organization for making available results from the Gallup World Poll.
The questions that motivated this study are familiar: Are the rich happier than the poor? Are the married happier than the unmarried? We will show that the answer to these questions depends on how happiness is defined— by the prevalence of positive affect in everyday experience or by the satisfaction that people report when asked to evaluate their life. The conditions and achievements that are associated with being satisfied are not necessarily associated with being happier moment to moment. The distinction we draw between experienced happiness and life satisfaction reflects the consensus that well-being has both judgmental and affective aspects. In Diener’s (1994) authoritative statement, ‘‘Subjective well-being (SWB) comprises people’s longer-term levels of pleasant affect, lack of unpleasant affect, and life satisfaction’’ (p. 103). This definition implicitly requires separate measurement of affective experience and life satisfaction and separate analysis of their determinants. Surprisingly, this requirement has 16
The Structure of Well-Being in Two Cities
17
been neglected in well-being research. Keeping multiple aspects of wellbeing in mind is difficult, intellectually costly, and a major hindrance to communication. It is far easier to speak of ‘‘happiness’’ as if it were a single concept, and this is what many researchers have chosen to do. This preference for conceptual and terminological simplicity is manifest in many chapters of the present volume, including several that explicitly distinguish life evaluation from experienced happiness. We describe a study conducted in two cities in different countries, which was intended to highlight the distinct determinants of the two aspects of well-being. Our study is made possible by a recent development in the measurement of affective experience. The gold standard for such measurements is the Experience Sampling Method (ESM), in which the participant is prompted at irregular intervals to record her current circumstances and feelings (Csikszentmihalyi & Larsen, 1987; Stone, Shiffman, & DeVries, 1999). Experience sampling minimizes the role of memory and interpretation, but it is expensive and difficult to implement in large samples. Consequently, we use the Day Reconstruction Method (DRM), in which participants are instructed to think about the preceding day, break it up into episodes, and describe each episode by selecting from several menus (Kahneman, Krueger, Schkade, Schwarz, & Stone, 2004), including a list of feelings. The DRM involves memory, but the technique is designed to increase the accuracy of emotional recall by inducing retrieval of the specifics of the successive episodes that make up the day (Robinson & Clore, 2002; Belli, 1998). Evidence that the two methods can be expected to yield similar results was presented elsewhere (Kahneman et al., 2004). A major advantage of the DRM is that it provides data on time-use—a valuable source of information in its own right, traditionally the province of sociologists (Michelson, 2006)—which has rarely been applied to the study of wellbeing (exceptions are Lawton et al., 1999; Robinson & Godbey, 1999; Robinson, 2008). We report a DRM study of one day in the lives of women, conducted concurrently in Columbus, Ohio, and in Rennes, France. A comparison of the level of well-being in the two cities has been reported elsewhere (Krueger et al., 2009a,b). This paper addresses the structure of well-being, which we find to be remarkably similar in the two countries, and the content of well-being, where we find some differences.
Theory
Our view of the structure of well-being is shown in Figure 2.1, which identifies two qualitatively distinct constituents. (1) Experienced happiness
18
Section I: Measuring Well-Being in an International Context
is the temporal average of a dimension of subjective experience reported in real time over an extended period. (2) Life satisfaction is a global, evaluative judgment that the participant makes about her life. A potent determinant of well-being is not shown in Figure 2.1. Much of the variance of both experienced happiness and life satisfaction is explained by variation in a participant’s personal disposition that has a significant genetic component (Lykken, 1999). We focus here on two other determinants: the general circumstances of people’s lives (marital status, age, income), and the specifics of how they spend their time. Determining one’s life satisfaction requires a difficult judgment. Like other hard judgments, the evaluation of one’s life is accomplished by consulting heuristics—the answers to related questions that come more readily to mind (Kahneman, 2003). Experimental demonstrations of priming and context effects provide the evidence for the role of such heuristics in reports of life satisfaction (Schwarz & Strack, 1999). For example, priming people to think of a particular life domain, such as their love life as students (Strack, Martin, & Schwarz, 1988) or marital satisfaction as adults (Schwarz, Strack, & Mai, 1991) causes global judgments of life
AROUSAL (Activation) TIME USE (from DRM)
AFFECTIVE EXPERIENCE HAPPINESS (Difmax)
FORTUNATE CIRCUMSTANCES (GFI)
Figure 2.1
EVALUATION LIFE SATISFACTION (SWLS)
The Structure of Well-Being in Two Cities
19
satisfaction to be dominated by the primed aspect. Without being aware that they do so, people answer the specific question that has been primed and project the answer onto the more inclusive dimension of life satisfaction. Figure 2.1 identifies two heuristic questions: ‘‘How fortunate am I?’’ and ‘‘How good is my mood these days?’’ The first involves a comparison of the individual’s circumstances to conventional or personal standards, while the second calls her attention to recent affective experience. These heuristics are not the only ones that people use, but it is safe to assume that many respondents implicitly answer one or both of these questions while they ponder their overall satisfaction with life. Individuals differ in their basic level of mood; some people are generally much happier than others. However, the affective experience of each person varies substantially during the day, depending on the activities in which she engages and on the social context. Most people are much happier sharing lunch with a close friend than driving alone in heavy traffic. We define an individual’s happiness on a given day by its average value over the day. Happiness, so defined, is influenced by the individual’s allocation of time: a longer lunch and a shorter commute make for a better day. A person’s use of time, in turn, reflects her circumstances. Age and employment, income and education, marital status and child-rearing obligations all impose a structure on an individual’s use of time, and indirectly on her experiences. As we shall see, the activities that are associated with the highest life satisfaction are not necessarily associated with a good mood. Figure 2.1 does not show a direct link from life circumstances to experienced happiness—because we found none. Contrary to a widespread belief, we argue that favorable life conditions do not automatically induce a generally sunny mood (Kahneman, Krueger, Schkade, Schwarz, & Stone, 2006; Schkade & Kahneman, 1998). Instead, the circumstances and activities that go with high life satisfaction tend to be moderately stressful.
Method Participants.
Professional survey firms recruited 810 women in Columbus, Ohio, and 819 in Rennes, France, using random-digit dialing. The acceptable age range was 18–60, and all participants spoke the dominant language at home. Full-time students under age 30 were excluded from the present analysis; the remaining Ns were 770 (Columbus) and 700 (Rennes). Respondents were paid about $75 each. The demographics of the samples generally matched those for the two cities.1
20
Section I: Measuring Well-Being in an International Context
Procedure.
The DRM protocol described by Kahneman et al. (2004) was followed. Groups of participants were invited for a weekday evening to a central location, where they completed a series of questionnaires. The first questionnaire included general satisfaction and demographic questions. The second asked respondents to construct a diary of the previous day2 as a series of episodes, noting the content and the beginning and ending time of each. The average number of episodes described was 13.3 in Columbus and 14.4 in Rennes. In the third questionnaire, respondents completed a form for each of the episodes they had previously listed. The form included a list of 22 activities and 8 interaction partners, with an instruction to mark all that apply. In a significant refinement of the original DRM protocol, respondents who had checked multiple activities were required to indicate the one that ‘‘seemed the most important to you at the time’’ (we call it focal). Unless specifically noted, all analyses below refer to focal activities.3 The form also requested ratings of 10 feelings on a scale from 0 (not at all) to 6 (very strongly).4 Results and Discussion
We first introduce our measures of well-being and discuss their relationship. Subsequent sections describe the correlates of life circumstances and of time-use. The findings in Columbus and Rennes were qualitatively similar, and except where noted, we report pooled results. The final section discusses some differences between the cities. Experienced Happiness (DIFMAX).
We defined Experienced Happiness for an episode as: DIFMAX ðepisodeÞ ¼ ‘‘Happy’’ Max ð‘‘Tense; ’’ ‘‘Depressed; ’’ ‘‘Angry’’Þ
ð1Þ
The range of DIFMAX is –6 to +6. We chose DIFMAX over the more conventional measure of ‘‘Net Affect’’ (the difference between averages of positive and negative affect ratings) because it reflects the intuition that an episode can be very aversive if even one of the negative feelings is intense. The conclusions we report are generally robust to the choice of definition, because DIFMAX and Net Affect are highly correlated: .95 over the 20,277 episodes in the data set. The happiness experienced by an individual for the day is an average of the values of DIFMAX(episode), where each episode is weighted by its duration. The average DIFMAX is 2.34 in Columbus and
The Structure of Well-Being in Two Cities
21
2.51 in Rennes (SD = 2.08 and 1.85, respectively). As already noted, the differences in the level of experienced happiness have been analyzed elsewhere (Krueger et al., 2009a, b). In the present analyses, the values of individuals’ DIFMAX are standardized separately within each country. Activation (ACT).
The circumplex model of affect shown in Figure 2.2 is an influential psychological theory, which identifies two major dimensions of affective experience (Russell, 1980, 2003). The principal dimension is Evaluation; the second is Activation/Arousal. The circumplex model implies an important distinction between two types of positive experience: joy and elation are both positive and high in activation; contentment is positive but low in activation. There are also two types of negative experience: anxiety is associated with higher arousal than depression. As suggested by this analysis, we defined a measure of Affective Activation for each episode as: ACT ðepisodeÞ ¼ ð‘‘Tense’’ ‘‘Depressed’’Þ þ ð‘‘Happy’’ ‘‘Calm’’Þ ð2Þ The measure of experienced activation/arousal for an individual is the duration-weighted average of the episode scores (standardized separately ACTIVATION Tense Jittery
Exited Ebullient
Excited Happy
Upset Disconnected
PLEASURE
DISPLEASURE Sad Gloomy
Serene Connected
Tired Lethargic
Placid Calm
DEACTIVATION
Figure 2.2
22
Section I: Measuring Well-Being in an International Context
within each country for analysis). As expected from the circumplex structure, the linear correlation between DIFMAX and ACT is close to zero (r = –.04). However, there is a highly significant quadratic relationship (r = –.30, p < .001). The extremes of the activation scales are associated respectively with high levels of tension or depression. On the other hand, the happiest women report very low levels of both variants of negative affect, and their activation scores are generally moderate. Life Satisfaction (SAT).
Life satisfaction was measured by the Diener-Emmons scale (Diener et al., 1985), which consists of five items (e.g., ‘‘My life is close to ideal’’), rated from 1 (strongly disagree) to 7 (strongly agree).5 Very similar results were obtained with the frequently used single question, ‘‘Taking all things together, how satisfied are you with your life as a whole these days?’’ We used the longer and more reliable scale for analysis. The correlation between Life Satisfaction (SAT) and Experienced Happiness (DIFMAX) is positive, but modest: r = .36. The correlation drops to .21 when two questions that we used as markers for depression are controlled (‘‘During the past month, how would you rate your overall sleep quality?’’; ‘‘During the past month, how much of a problem has it been for you to keep up enough enthusiasm to get things done?’’ Life Satisfaction is also significantly correlated with both components of the Activation measure (r = .14 with ‘‘Tense – Depressed,’’ p < .001, and r = .17 with ‘‘Happy – Calm,’’ p < .01), and with the combined measure (r = .19). However, the correlation between Life Satisfaction and Activation is not significantly reduced (to r = .17) when controlling for both Experienced Happiness and the depression markers. In accord with the main theme of this paper, these results establish Life Satisfaction and Experienced Happiness as distinct aspects of well-being. In addition, they identify a moderately high level of affective activation as a concomitant of both happiness and life satisfaction. In these two prosperous Western countries, a satisfying life is not necessarily more enjoyable, but it is likely to be high in activation—tense as well as happy, less depressed but also less calm.
Correlates of Life Circumstances
The first column of Table 2.1 displays the relationship between various life circumstances and SAT. The results confirm previous findings (Argyle, 1999). Consistent with the concept of a Good-Fortune heuristic, the correlates of life satisfaction are plausible reasons for a person to consider herself relatively fortunate or privileged. For women in the United States
The Structure of Well-Being in Two Cities
TABLE 2.1
23
Relationship of Life Circumstances to Well-Being Measures Life Satisfaction (SAT)
household income has mate education employed lives with child < 6 yrs no medical treatment age BMI *p < .05 ** p < .01
Experienced Happiness (DIFMAX)
Activation (ACT)
Correlation
Beta
Correlation
Correlation
.33** .32** .23** .18** .17** .10** .03 .10**
.19** .18* .13** .11** .12** .06* .02 .06*
.09** .08** .03 .06* .03 .03 .02 .02
.10** .16** .09* .06* .17** .05 .13** .05
and France, household income and having a mate (married or cohabiting) are the strongest predictors of life satisfaction. Education, employment, and children also contribute substantially. Good health and a trim figure also help. The list of correlates of life satisfaction reads like a conventional definition of a desirable life for a woman in a developed country. Table 2.1 also shows standardized multiple regression coefficients, which we used to define a Good Fortune Index (GFI). Although few of the individual coefficients are impressively large, the overall amount of the variance of life satisfaction explained by life circumstances (20%) is substantial in the present study. This is somewhat higher than the 8%– 15% reported in previous research (Argyle, 1999; Diener et al., 1999), perhaps because our samples are relatively homogeneous in gender, age, and location. The last two columns of Table 2.1 present correlations of individual characteristics with the two dimensions of affective experience, Happiness and Activation. The lay theory of well-being, which does not discriminate sharply between life satisfaction and good mood, incorporates the common myth that life conditions such as being rich or divorced have a pervasive effect on daily mood (Kahneman et al., 2006; Schkade & Kahneman, 1998). Table 2.1 shows that the lay theory is wrong: the correlations between favorable life circumstances and DIFMAX are positive, but consistently much smaller than the corresponding correlations with SAT (see also Kahneman et al., 2004). Life circumstances account for very little of the variance in DIFMAX (1.6%). Furthermore, additional analyses showed that the small correlations between individual characteristics and Experienced Happiness vanish altogether when the depression markers are
24
Section I: Measuring Well-Being in an International Context
controlled. The question that introduced this article has a simple answer: being rich or married makes women more satisfied with their lives, but contributes little to the happiness they experience. There is an obvious objection to our conclusion that life circumstances are better predictors of life evaluation than of experienced happiness. Because life satisfaction is the answer to a global question but experienced happiness is measured on a single day, we could expect large differences in the reliability of these measures. The observed difference in predictability could be an artifact of differential reliability. This objection is plausible, but not supported by the data. Krueger & Schkade (2008) examined the test-retest reliability of affect and satisfaction measures over a two-week period for the same respondents. They found that the reliabilities of DIFMAX (r = .60) and the common four-category life satisfaction question (r = .59) were virtually identical. The Diener-Emmons satisfaction scale we use here is composed of multiple questions and could be slightly more reliable, but a large difference is unlikely. Favorable life circumstances are more strongly correlated with Activation than with Experienced Happiness. The correlation between GFI and ACT is highly significant (r = .21), and is not reduced when the depression markers are controlled. This association between tension and favorable circumstances is not an isolated finding. A large-scale study using experience sampling in the workplace (n = 374) showed no relationship of income with ratings of Happiness (r = .01), but significant correlations with Angry/Hostile (r = .14), Anxious/Tense (r = .14), and Excited (r = .18).6 We can now refine our earlier statement—women whose circumstances are favorable (married, better off financially, employed) are relatively more likely to feel tense than depressed and more likely to feel happy than calm. The relationship between Activation and major elements of GFI supports this conclusion. Respondents in more than 140 countries were asked to report on their feelings, using the following question: ‘‘Did you experience the following feelings during A LOT OF THE DAY yesterday? How about . . .’’ The answers were yes/no. The list of feelings included Stress and Depression, which enabled us to define an index of Activation by the difference (Stress - Depression). To analyze the separate contribution of Activation, we used the three measures (Stress, Depression, and Activation) to predict income and education separately within each country. Our question was whether the difference would contribute to the prediction of these outcomes. The results were unequivocal. Activation was positively related to income (within country) in 83% of countries and was positively related to education in 92% of countries. The same result was also found when we looked separately at France and the U.S. We also find that the relationship between Activation and favorable life
The Structure of Well-Being in Two Cities
25
circumstances holds across countries. When the same analysis was applied (using country means of Stress, Depression, and Activation to predict GDP and average education), the coefficients for Activation were positive and significant. The results complement and reinforce the findings of Ng and Diner (2009) who reported, using the same data, that average stress level in a country is positively correlated with GDP. These authors also found, as we do, that at the individual level, reports of stress are negatively correlated with income (and we find the same with respect to education). Stress (like ‘‘Tense’’ in the DRM) belongs to the cluster of negative emotions, and by itself does not contribute to well-being. However, the conjunction of stress and no depression is associated both with high education and with high income—important precursors of a satisfying life. Ng et al. (2009) have independently reached the same conclusions from an analysis of the Gallup World Poll, using only reports of stress. The difference between stress and depression, which we use here, is a cleaner measure of activation because it is less highly correlated with negative evaluation.
Analyses of Time Use
Correlations between life circumstances and affective experience arise in different ways. The model of Figure 2.1 focuses on a particular path, in which life circumstances influence affective experience through their effects on the use of time. To introduce the analyses of this section, we consider a particular condition of life: having a mate. Having a mate involves several changes in how time is spent—some for the better and some for the worse (Table 2.2). Compared to others, women who have a mate spend less time alone, but also less time with friends. They spend more time making love, doing compulsory activities, preparing food, and caring for children. On weekends, they spend less time eating alone, but also have less time for passive TABLE 2.2
Comparison of Time Use for Women With and Without a Mate
Activity alone with friends making love doing compulsory activities preparing food caring for children eating alone (weekends) passive leisure (weekends)
Has Mate
No Mate
Units
p(diff)
3.2 21 9 4.5 32 63 5 1.8
5.4 75 6 3.7 20 31 23 2.2
hrs min min hrs min min min hrs
.001 .001 .01 .001 .001 .001 .001 .001
26
Section I: Measuring Well-Being in an International Context
leisure. Evidently, the lack of correlation between marriage and DIFMAX that we reported above occurs, not because marriage makes no difference, but because the affective consequences of various time use differences balance out. Kahneman et al. (2006) show a similar result for time-use shifts associated with higher incomes. Table 2.3 presents experienced affect and time results for several categories of activity. The ranking of activities by DIFMAX confirms previous DRM findings (Kahneman et al., 2004)—work, commuting, compulsory activities at home, and being alone are associated with lower experienced happiness. A different pattern is found for ACT, where work, compulsory activities, and social interaction are all associated with high Activation. The last column of Table 2.3 shows the average difference in time allocation between the top and bottom thirds of the distribution of the Good Fortune Index. As mentioned earlier, satisfying life circumstances are associated with a mix of hedonic costs and benefits. Work and Obligatory activities are distinctly less pleasant than the Passive Discretionary activities that they mostly replace. On the other hand, spending more time with others provides a hedonic gain. In contrast, the time use of high-GFI women (more obligations, more social interaction) is consistently associated with relatively high Activation. Thus, the puzzle of the affective correlates of satisfying circumstances is largely solved by the analysis of time use. A more satisfying life comes with shifts in time use that increase a person’s activation level, but on balance make them no happier.
Well-Being in Two Cities
The analyses we reported here did not attempt to solve the problem of translating expressions of feelings across the language barrier, and focused on the correlational structure of results. For analytical convenience we assumed that SAT and DIFMAX were similarly distributed in the two cities and standardized the main variables separately in the two samples. We had expected to find substantial differences between the determinants of life satisfaction and experienced happiness in the two cities, and were instead surprised by their remarkable similarity. For example, the correlation between the two cities across the means for the 22 activities is .90 for DIFMAX and .92 for the proportion of focal time spent in the activity. The top half of Table 2.4 shows the similarity between the two cities in the pattern of relationships depicted in Figure 2.1.
TABLE 2.3
Time Use and Experienced Affect for a Weekday Mean Daily Time Spent (min) Mean Experienced Happiness (zDIFMAX)
Mean Activation (zACT)
Overall
GFI Top 3rd Bottom 3rd
active leisure eating talking passive leisure compulsory work/commute other
.29 .25 .17 .18 .21 .50 .09
.19 .24 .01 .67 .14 .38 .08
90.6 67.2 79.7 99.0 244.5 271.5 66.7
10.6 .4 .1 50.9 23.7 65.5 27.2
social alone
.08 .21
.16 .25
671.7 247.5
87.5 87.5
Activity
27
28
Section I: Measuring Well-Being in an International Context
TABLE 2.4 Comparison of Well-Being in Two Cities Columbus Structural Similarities (correlations) SAT, DIFMAX SAT, ACT ACT, DIFMAX ACT2 (quadratic), DIFMAX GFI, SAT GFI, DIFMAX GFI, ACT Time use, DIFMAX (multiple R) Time use, ACT (multiple R) Time use, SAT (multiple R)
Rennes
Difference significant?
.36 .13 .07 .28
.36 .21 .01 .32
ns ns ns ns
.43 .10 .21 .31
.50 .12 .18 .36
ns ns ns ns
.39
.35
ns
.25
.33
ns
18%
14%
.22 .05
.09 .22
< .01 < .01
.32
.33
ns
Some Content Differences Child Care Focal child care time (mothers only) Mean DIFMAX (z) Mean DIFMAX (z) Spouse present Mean DIFMAX (z) Spouse no kids
< .001
Time Use working/commuting walking reading praying/worshipping eating
5.4 hrs 7 min 33 min 19 min 52 min
6.3 hrs 19 min 48 min 3 min 117 min
< .05 < .001 < .001 < .001 < .001
On this background of general similarity, a few differences are worth noting. Marital status is somewhat more predictive of SAT in France, t(1447) = 2.09, p < .05. The interaction of marital status and country was in the same direction for DIFMAX, but weaker (t = 1.57). A rather larger difference was observed in the enjoyment of child care and of interactions with children. The American mothers spent more time focused on child care, but enjoyed it less (see bottom half of Table 2.4). This difference in the enjoyment of children spills over into the response to family occasions. Women in both samples enjoyed one-on-one interaction with their spouse
The Structure of Well-Being in Two Cities
29
about equally, but the corresponding values when children were present suggest that the presence of the spouse hardly makes American children less annoying to their mother. The relative unhappiness of American mothers obviously demands further study. Our French team member suggested that French children are simply better behaved, but other hypotheses should also be considered. The allocation of focal time is also broadly similar in the two cities, and the differences we do find confirm cultural stereotypes (Table 2.4). As expected, French women on average spend less time each day working and commuting, and more time walking and reading, while the Americans spend more time in praying/worshipping. The striking difference in the time spent eating is mainly due to differences in how attention is paid to that activity. Episodes in which eating was mentioned covered almost as much of the day in Columbus (17.1% vs. 20.8%), but eating was focal in those episodes almost twice as often for the French (56% vs. 30%). On the whole, the French women spent slightly more time on the more enjoyable activities—the importance of these differences is explored in another article (Krueger et al., 2009a). The conclusion of the present comparison is straightforward. The basic structure of well-being is the same for women in the two cities, but the content—the specific sources from which they draw happiness—is slightly different, reflecting differing cultural norms and social arrangements.
General Discussion
We began by raising familiar questions about the well-being of the rich and the married. To find an answer, we applied a methodology that provides complementary descriptions of people’s lives: by life circumstances (e.g., income, employment, marital status) that may be more or less conducive to satisfaction; and by the time spent in activities that may be more or less enjoyable or arousing. Our first conclusion was that the rich and the married are indeed somewhat more satisfied with their lives, but not much happier moment to moment. Favorable circumstances explained about 20% of the variance in Life Satisfaction but less than 2% of the variance in Experienced Happiness. Perhaps more important, the availability of a detailed description of people’s lives suggests a new approach to some standard questions. The classic puzzle involves the limited long-term hedonic effects of outcomes that are greatly desired in anticipation and evoke intense emotions when they occur. Winning the lottery and getting married are happy events, and it appears natural to ask why the happiness does not endure. Almost
30
Section I: Measuring Well-Being in an International Context
inevitably, this formulation invites the idea of a potent process of hedonic adaptation that eventually returns people to a set point determined by their personality (see Diener, Lucas, & Scollon, 2006; Headey & Wearing, 1989). Our findings suggest two revisions to this model. First, although personality surely matters, the claim that an individual’s experienced happiness must return to a set point that is independent of her local circumstances is probably false. For someone who enjoys socializing much more than commuting, a permanent reallocation of time from one of these activities to the other can be expected to have a permanent effect on happiness (Lyubomirsky, Sheldon, & Schkade, 2005). Second, there may be no need for a concept of hedonic adaptation to explain why life-changing events sometimes have little hedonic effect in the long run. The answer is simple: desirable circumstances such as marriage or high income have hedonic costs as well as benefits. Married women have more sex and more companionship, but they also spend much more time in chores and much less time with their friends than single women do. Does the fact that they are not much happier in steady state need another explanation? The duration bias—an unjustified expectation that intense affective states will endure—is the central mistake of affective forecasting (Gilbert, 2006). Perhaps we—the community of students of well-being—have been suffering all along from the same duration bias that we study in others. By invoking a process of adaptation to explain why extreme affective states do not persist, we reveal our own naive expectation that they should. Other questions become central in this framework. Why do people desire outcomes that will not make them happy in the long run? Why are they temporarily happy when desired life changes occur? The simple model we have discussed here must be enriched to answer these questions. Thus far we have explained people’s experienced happiness by the joint effects of their personality and the situations they encounter. But something else happens around major life changes such as marriage, a crippling accident, or a sudden bereavement. The normal influence of the immediate situation on affect is reduced, and in that temporary state one can be deliriously joyous while stuck in traffic, or miserable at a feast. In these special states, attention is constantly drawn back from the current situation to one’s preoccupations, and affect follows attention. However, the basic rule of attention is that it tends to be withdrawn from stimuli (or thoughts) as they become familiar. Although there are exceptions in which a recurrent thought can become self-sustaining, the normal outcome is a return of control to the immediate situation. The notions of hedonic treadmill and set point may not be the best way to explain the decline of passionate love, or the withdrawal of attention from life circumstances that that are no longer novel.
The Structure of Well-Being in Two Cities
31
Finally, we address a frequently raised and intuitively compelling objection to our approach: why should anyone care about duration-weighted experienced happiness? After all, it is generally agreed (and we join the consensus) that people care very much about the narrative of their life and the satisfaction and meaning they can find in it. The natural point of view for evaluating life is retrospective. In that view, what matters are the memorable and meaningful moments (Fredrickson, 2000)—it seems almost absurd to assign equal weight to such moments and to the time spent in routine activities. Time counts for little in narratives. If the quality of the story matters, achieving satisfaction with one’s life is the proper criterion of well-being. We share this intuition, but find another idea almost equally compelling: why do we think it so valuable to extend people’s lives, even when their life story will not be improved? Clearly, time of life is intrinsically valued. And if we care about extending people’s lives when they are old, we should care no less about the time that they spend commuting in middle age. Time is the ultimate finite resource, and the question of how well people spend it is a central issue in the study of well-being. Notes 1. For details of procedures and questionnaires, visit http://management.ucsd.edu/ faculty/directory/schkade/fa-study/. 2. About 300 participants in each country were recruited for Mondays to describe a weekend day. Half of them were instructed to describe the Saturday. 3. When a respondent mentioned multiple activities without choosing a focal one (14.9% of all episodes), a set of rules based on a hierarchy of attention was applied (e.g., love-making dominant, listening to music peripheral) to designate a focal choice. For details, visit http://management.ucsd.edu/faculty/directory/schkade/fastudy/. 4. ‘‘Impatient for it to end,’’ ‘‘Competent/confident,’’ ‘‘Tense/stressed,’’ ‘‘Happy,’’ ‘‘Depressed/blue,’’ ‘‘Interested/focused,’’ ‘‘Affectionate/friendly,’’ ‘‘Calm/relaxed,’’ ‘‘Irritated/angry,’’ ‘‘Tired.’’ 5. The other items are: ‘‘The conditions of my life are excellent’’; ‘‘I am satisfied with my life’’; ‘‘So far I have gotten the important things I want in life’’; ‘‘If I could live my life over, I would change almost nothing.’’ 6. Reported in Kahneman et al. (2006).
REFERENCES Argyle, M. (1999). Causes and correlates of happiness. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology. New York: Russell Sage Foundation. Belli, R. (1998). The structure of autobiographical memory and the event history calendar: Potential improvements in the quality of retrospective reports in surveys. Memory, 6, 383.
32
Section I: Measuring Well-Being in an International Context
Csikszentmihalyi, M., & Larsen, R. E. (1987). Journal of Nervous and Mental Disease, 175, 526. Diener, E. (1994). Assessing subjective well-being: Progress and opportunities. Social Indicators Research, 31, 103–157. Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The Satisfaction with Life scale. Journal of Personality Assessment, 49, 71–75. Diener, E., Lucas, R., & Scollon, C. N. (2006). Beyond the hedonic treadmill: Revising the adaptation theory of well-being. American Psychologist, 61, 305–314. Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. (1999). Subjective well-being: Three decades of progress. Psychological Bulletin, 125, 276–302. Fredrickson, B. L. (2000). Extracting meaning from past affective experiences: The importance of peaks, ends, and specific emotions. Cognition and Emotion, 14, 577–606. Gilbert, D. T. (2006). Stumbling on happiness. New York: Knopf. Headey, B., & Wearing, A. (1989). Personality, life events, and subjective well-being: Toward a dynamic equilibrium model. Journal of Personality and Social Psychology, 57, 731–739. Kahneman, D. (2003). A perspective on judgment and choice: Mapping bounded rationality. American Psychologist, 58, 697–720. Kahneman, D., Krueger, A., Schkade, D., Schwarz, N., & Stone, A. (2004). A survey method for characterizing daily life experience: The Day Reconstruction Method (DRM). Science, 306, 1776–1780. Kahneman, D., Krueger, A., Schkade, D., Schwarz, N., & Stone, A. (2006). Would you be happier if you were richer? A focusing illusion. Science, 312, 1908–1910. Krueger, A. B., Kahneman, D., Fischler, C., Schkade, D., Schwarz, N., and Stone, A. A. (2009a). Comparing time use and subjective well-being in France and the U.S. Social Indicators Research, 93, 7–18. Krueger, A. B., Kahneman, D., Schkade, D., Schwarz, N., & Stone, A. A. (2009b). National time accounting: The currency of life. In A. Krueger (Ed.), Measuring the subjective well-being of nations: National accounts of time use and well-being. Chicago: University of Chicago Press. http://www.krueger.princeton.edu/nta2.pdf Krueger, A. B., and Schkade, D. (2008). The reliability of subjective well-being measures. Journal of Political Economics, 92, 1833–1845. Lawton, M. P., Winter, L., Kleban, M. H., et al. (1999). Affect and quality of life— Objective and subjective. Journal of Aging and Health, 11, 169–198. Lykken, D. T. (1999). Happiness: What studies on twins show us about nature, nurture, and the happiness set-point. New York: Golden Books. Lyubomirsky, S., Sheldon, K., & Schkade, D. (2005). Pursuing happiness: The architecture of sustainable change. Review of General Psychology, 9, 111–131. Michelson, W. (2006). Time use: Expanding explanation in the social sciences. Paradigm Publishers: Boulder, CO. Ng, W., Diner, E., Arora, R., & Harter, J. (2009). Affluence, feelings of stress and wellbeing. Social Indicator Research, 94, 257–271. Robinson, M. D., & Clore, G. L. (2002). Belief and feeling: Evidence for an accessibility model of emotional self-reports. Psychological Bulletin, 128, 934. Robinson, J. P., & Godbey, G. (1999). Time for life: The surprising ways Americans use their time. University Park, PA: Penn State University Press. Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology, 39, 1161–1178.
The Structure of Well-Being in Two Cities
33
Russell, J. A. (2003). Core affect and the psychological construction of emotion. Psychological Review, 110, 145–172. Schkade, D. A., & Kahneman, D. (1998). Does living in California make people happy? A focusing illusion in judgments of life satisfaction. Psychological Science, 9, 340–346. Schwarz, N., & Strack, F. (1999). Reports of subjective well-being: Judgment processes and their methodological implications. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology. New York: Russell Sage. Stone, A. A., Shiffman, S. S., DeVries, M. W. (1999). Ecological momentary assessment. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology (pp. 61–84) . New York: Russell Sage.
Chapter 3 Culture and Well-Being: Conceptual and Methodological Issues Shigehiro Oishi University of Virginia
In this chapter, I will review major conceptual and measurement issues in culture and well-being research. I will use the term well-being in the most inclusive sense; namely, as the state of ‘‘being well’’ (Markus & Kitayama, 2000). Well-being, thus, includes objective indicators of being well such as being healthy and financially secure as well as concepts associated with eudaimonia (a well-lived life) such as meaning and purpose in life. In contrast, I will use the term subjective well-being when specifically referring to the cognitive evaluation of life as a whole or affective experiences (Diener, Suh, Lucas, & Smith, 1999; Veenhoven, 2009). Finally, I will use the term happiness in the way lay people use it to express and communicate favorable feelings or conditions. Although many scales have been developed to assess different aspects of well-being (see Andrews & Robinson, 1991; and Snyder & Lopez, 2002, for review), three types of single-item measures have dominated large-scale international surveys on well-being: (a) a general life-satisfaction item (e.g., ‘‘All things considered, how satisfied are you with your life as a whole these days?’’), (b) a Cantril’s (1965) Ladder item (‘‘Where on the ladder would you say you personally stand at the present time?’’), and (c) a happiness item (e.g., ‘‘Taking all things together, would you say you are 1 = very happy, 2 = quite happy, 3 = not very happy, 4 = not at all happy?’’). Many chapters in this volume (e.g., Diener, Kahneman, Tov, & Arora, 2009; Easterlin & Sawangfa, 2009; Helliwell, Barrington-Leigh, Harris, & Huang, 2009; Inglehart, 2009; and Layard, Mayraz, & Nickell, 2009) report the findings on these items (see Clark, 2009; and Harter & Arora, 2009, for 34
Culture and Well-Being: Conceptual and Methodological Issues
35
exceptions). The exclusive attention given to these three items in many of the chapters is understandable, considering that many extant large-scale international surveys included only some of these single-item scales. However, the exclusive attention paid to these three types of measures masks some important questions in cross-cultural subjective well-being research (e.g., conceptual equivalence, measurement equivalence, measurement error, response style, and online versus global reports). For instance, the reliability of single-item measures is typically lower than that of multiple-item measures, and yet measurement equivalence has not been formally established or measurement errors corrected in most research reported in this volume. I will discuss these broad issues important in the cross-cultural investigation of well-being. Specifically, I will first review historical and cultural variations in concepts of well-being, focusing on lay concepts of happiness. Then, I will discuss measurement issues in the context of national comparisons. Finally, I will summarize major findings on national differences in mean levels, correlates, and consequences of subjective well-being (see Diener, Oishi, & Lucas, 2003; Suh & Koo, 2008; Tov & Diener, 2007; and Mesquita & Leu, 2007, for comprehensive reviews on culture, well-being, and emotion). The three main theses of this chapter are (a) there are diverse concepts of well-being across times and cultures; (b) therefore, diverse measures that tap into these different concepts of well-being should be employed; and (c) national differences in mean levels, correlates, and consequences of well-being should be interpreted with the diversity of concepts and the limitations of different measures in mind.
The concept of well-being
Survey researchers often implicitly assume that the concept of well-being is apparent to people in most countries (e.g., Inglehart, 2009; Veenhoven, 2009). Indeed, survey researchers often translate items originally created in English and use them as if they had exactly the same meaning anywhere in the world. Most cultural psychologists and anthropologists are wary of this practice, as previous ethnographies revealed a surprising lack of some equivalent concepts in different societies (Kitayama & Markus, 2000; Shweder, 1991). Russell (1991) reviewed the literature on emotion and identified cultures that do not have corresponding words for the so-called basic emotions. According to Russell, there is no word for happiness in the Chewong language of peninsular Malaysia. There is no word for sadness in Tahitian and Chewong, no word for fear in Ifaluk, Utku, and Pintupi, and no word for surprise in Fore, Dani, Malay, and Ifaluk.
36
Section I: Measuring Well-Being in an International Context
Even when there is a corresponding concept, the way people think about well-being can be very different across cultures. For example, when Ed Diener travelled to India, he asked a middle-aged woman how satisfied she was with her life. To Ed’s surprise, this woman answered by saying, ‘‘Please ask my husband.’’ It was clear that she understood the concept of satisfaction. This woman just thought that her husband was the expert who could answer this question most accurately. In the United States and European countries, most people believe that they are the best judges of their own well-being. This assumption is not always tenable, however, and in some cultures, respondents might be answering well-being questions from the thirdperson perspective; namely, how significant others might answer for them (Cohen & Gunz, 2002). Furthermore, even when the corresponding word exists, conceptual comparability may not. For example, there is a great deal of agreement that the most appropriate Japanese translation of depression is 憂鬱 (yuutsu), and clearly there is a phenomenon labeled ‘‘depression’’ in Japan. Yet Tanaka-Matsumi & Marsella (1976) showed that the free associations with this concept given by English speakers were very different from those given by Japanese speakers. The most frequent spontaneous association with yuutsu by Japanese was rain, followed by dark, worries, gray, and cloud (i.e., four of the five most frequent associations were referring to external conditions).;, whereas the most frequent spontaneous association with ‘‘depression’’ by European Americans was sadness, followed by down, lonely, unhappy, and moody (i.e., all five of the most frequent associations were referring to internal feeling states). Because labeling is an important aspect of conscious awareness of a feeling state, these linguistic differences should result in predictable cultural differences in the frequency and intensity of these emotions. In addition, connotative differences of emotions might give rise to different co-occurrence patterns of emotions across cultures. Like other researchers in this volume (Diener et al., 2009; Inglehart, 2009; Veenhoven, 2009), I believe that most people in the world are indeed able to make subjective judgments of their living conditions and affective experiences. Nevertheless, it is instructive to consider a variety of concepts of well-being reported in various parts of the world and to examine rather than assume equivalence of concepts, desirability, and item functioning of survey items in cross-cultural research on well-being. For one thing, it allows us to evaluate whether the three types of subjective well-being measures often used in large-scale international surveys cover the diverse concepts of well-being in the world. First, it should be recognized that there are diverse views on the central aspects of well-being. Aristotle used the term eudaimonia (often translated as ‘‘happiness’’) to describe a well-lived life, and made a sharp distinction
Culture and Well-Being: Conceptual and Methodological Issues
TABLE 3.1
37
Components and Time-Frame of Well-Being Components of Well-Being
Time-Frame
Online Global
Affect
Cognitive
Eudaimonic
Happy now? Happy in general?
Satisfied now? Satisfied in general?
Meaningful now? Meaningful in general?
Note. Affective components include various positive emotions such as happy, joyful, excited, and relaxed. Cognitive components include global life satisfaction as well as domain satisfaction (e.g., job satisfaction). Eudaimonic components include various concepts such as meaning in life, purpose in life, positive relationships, autonomy, and self-acceptance. Online reports refer to the reports made on a brief time period such as ‘‘right now’’ or ‘‘yesterday.’’ In contrast, global reports cover a longer period of time such as ‘‘these days’’ and ‘‘in general.’’ Most international survey questions on well-being are limited to the affective and cognitive components of well-being that are assessed in a global term (‘‘these days’’ or ‘‘all things together’’).
between amusement and eudaimonia (Thomson, 1953). Researchers who are heavily influenced by Aristotle have advocated eudaimonic well-being and proposed that the concept of well-being should include not only life satisfaction and positive affect, but also purpose in life, a sense of autonomy, self-acceptance, connectedness, and a psychological sense of vitality (Ryff & Singer, 1998; Ryan & Deci, 2001). Some theorists argue that meaning in life (Steger, Kawabata, Shimai, & Otake, 2008) and meaningful work, or calling (Wrzesniewski, McCauley, Rozin, & Schwartz, 1997) are important ingredients of a well-lived life. Hedonist and utilitarian theorists, in contrast, conceive of pleasure and pain as the building blocks of the science of well-being (Kahneman, 1999), and view other constructs such as meaning in life as secondary to affective experiences. List theorists believe that well-being consists of several basic conditions, such as health and good relationships (Nussbaum & Sen, 1993). Other theorists believe that the satisfaction of one’s desires, wishes, and goals is a sine qua non of well-being (Griffin, 1986). Finally, some believe that the subjective evaluation of life as a whole is the essence of well-being (Diener, Sapyta, & Suh, 1998; Sumner, 1996). As seen in Table 3.1, then, it is interesting to note that nowhere in this volume did researchers consider international differences in eudaimonic well-being such as meaning in life, or the discrepancies between online and global reports of well-being.
Historical Changes in the Concept of Well-Being
In order to understand the concept of well-being, one approach that can be taken is a lexical approach. As personality psychologists have used
38
Section I: Measuring Well-Being in an International Context
dictionaries to gain an understanding of the structure of personality across cultures (Saucier & Goldberg, 2001), well-being researchers can use dictionary definitions of happiness to gain an understanding of the lay concept of happiness in different cultures and historical times (see McMahon, 2006, for historical changes in the meaning of the pursuit of happiness). The Oxford English Dictionary (OED) is useful in investigating historical changes in the meaning of happiness in English because it denotes the original use of each definition. According to the OED, when the term was first recorded in 1530, the primary meaning of happiness was ‘‘good fortune or luck in life or in a particular affair; success, prosperity.’’ The ‘‘fortune/luck’’ definition makes sense because the English term happiness came from the word hap, meaning luck and fortune. The second definition of happiness in the OED is ‘‘the state of pleasurable content of mind, which results from success or the attainment of what is considered good,’’ which entered the language in 1591. The third and final definition of happiness is ‘‘successful or felicitous aptitude, fitness, suitability, or appropriateness’’; this usage entered the language in 1599. My student and I (Hoobler & Oishi, 2008) have examined the Webster’s Unabridged Dictionary (WUD) from 1850 on, and found that the definitions of happiness in the nineteenth century were similar to the OED’s definitions from the sixteenth century. Namely, in its 1850, 1853, 1854, 1859, and 1861 editions, the WUD lists ‘‘the agreeable sensations which spring from the enjoyment of good’’ as the first definition, ‘‘good luck; good fortune’’ as the second, and ‘‘fortuitous elegance; unstudied grace’’ as the third definition. Interestingly, however, in the 1961 edition, the definition of happiness as ‘‘good fortune; good luck; prosperity’’ was deemed ‘‘archaic.’’ In other words, in the United States, some time after World War II, people stopped using ‘‘happiness’’ to refer to good luck or fortune, and happiness became a purely internal state of mind, or ‘‘a pleasurable or enjoyable experience.’’ Next, we examined whether the use of happiness changed over time in the United States (Graham & Oishi, 2008) by examining all the State of the Union addresses given since 1790. It is interesting that early presidents used the term happiness to refer to the favorable conditions of the United States rather than their own inner feeling state. For example, President George Washington in 1791 told Congress that ‘‘the intelligence from the army under the command of General Wayne is a happy presage to our military operations against the hostile Indians north of the Ohio.’’ President John Adams in 1799 told Congress the following: ‘‘the various and inestimable advantages, civil and religious, which, secured under our happy frame of government, are continued to us unimpaired, demand of the whole American people sincere thanks to a benevolent Deity for the merciful
Culture and Well-Being: Conceptual and Methodological Issues
39
dispensations of His providence . . . . As to myself, it is my anxious desire so to execute the trust reposed in me as to render the people of the United States prosperous and happy.’’ President James Madison said in 1812, ‘‘Such is the happy condition of our country, arising from the facility of subsistence and the high wages for every species of occupation.’’ In 1817, President Madison again stated, ‘‘At no period of our political existence had we so much cause to felicitate ourselves at the prosperous and happy condition of our country. . . . It is [in] contemplating the happy situation of the United States, [that] our attention is drawn with peculiar interest to surviving officers and soldiers of our Revolutionary army.’’ In 1820, President James Monroe said, ‘‘When, then, we take into view the prosperous and happy condition of our country in all the great circumstances which constitute the felicity of a nation. . . .’’ In 1823, President James Monroe stated, ‘‘From the present prosperous and happy state I derive a gratification which I cannot express.’’ In 1825, President James Quincy Adams said, ‘‘Perhaps of all the evidence of a prosperous and happy condition of human society the rapidity of the increase of population is the most unequivocal.’’ In 1830, President Andrew Jackson stated, ‘‘With a population unparalleled in its increase . . . we see in every section of our happy country a steady improvement in the means of social intercourse.’’ In 1833, President Jackson again stated, ‘‘it gives me pleasure to congratulate you upon the happy condition of our beloved country.’’ You get the picture. Earlier Presidents used the term happy when describing favorable conditions of the United States. In other words, the term happy was used to communicate good luck and fortune. Indeed, we rarely found the use of the term happy in reference to a president’s inner feelings of pleasure. President Thomas Jefferson was an exception in this regard. In 1801 he used the term in the exact way Americans do today, by stating, ‘‘I am happy to inform you that the continued efforts to introduce among them the implements and the practice of husbandry and the household arts have not been without success. . . . I am happy in this opportunity of committing the arduous affairs of our Government to the collected wisdom of the Union.’’ President Ronald Reagan used the term happy in an old-fashioned way in 1983 when he said, ‘‘I would like to talk with you this evening about what we can do together—not as Republicans and Democrats, but as Americans—to make tomorrow’s America happy and prosperous at home, strong and respected abroad, and at peace in the world.’’ Interestingly, this was the only time a president used the term happy along with prosperous (i.e., to mean a favorable condition of the collective) since the end of World War II. In short, in the United States both the dictionary definitions of happiness and the use of the term happy in State of the Union addresses show the shift
40
Section I: Measuring Well-Being in an International Context
in the meaning of happiness from one that emphasizes ‘‘lucky and fortunate conditions’’ to one meaning ‘‘favorable inner feelings and the satisfaction and accomplishment of one’s desires.’’ The historical changes in the use of the term happy in the United States might reflect the fact that people’s lives became over time much more controllable than in the earlier, frontier days (Turner, 1921).
Cross-Cultural Similarities and Differences in the Concept of Well-Being
The analyses of the dictionaries and the State of the Union addresses above reveal a historical shift in the United States in the definition of happiness from ‘‘good fortune and luck’’ to an internal feeling. In this section, my colleagues and I examine potential cultural variations in the definition and salient aspects of happiness. To this end, we investigated the dictionary definitions of happiness in various nations and languages. In Japan, Korea, and China, the following two Chinese characters are often used to refer to happiness: 幸福. The first character 幸 means goodness and the second character 福 indicates fortune and luck. Not surprisingly, the definition of 幸福 in Japanese is ‘‘luck, chance, and a stroke of good luck,’’ according to Kojien, the authoritative Japanese dictionary. Similarly, ‘‘luck and fortune’’ are dominant definitions of happiness in many other languages as well. The French word bonheur (happiness) means ‘‘happy event, favorable luck,’’ according to the Dictionary of the French Academy (9th edition) published in 2000. The German word for happiness, glu¨ck, literally means ‘‘luck.’’ Likewise, the Norwegian word for happiness, lykke, means ‘‘fortunate destiny and luck.’’ The Russian term for happiness is Sd‘Pelhe, which means ‘‘success/luck,’’ and ‘‘happy occasion, happy turn of events.’’ In contrast, the Italian word felicitA˜ means ‘‘the fulfillment of every desire.’’ The Portuguese felicitas is similar to Italian in that it is a ‘‘state of full/absolute inner satisfaction, well-being, in which all human beings’ aspirations are met/satisfied.’’ The Spanish term felicidad is defined as an ‘‘affective state involving satisfaction with the possession of something’’ and ‘‘satisfaction, pleasure, contentment’’ in the Diccionario de la Real Academia de la Lengua Espan˜ola with an enye (22nd edition). It is interesting, then, that different concepts of happiness are represented in most languages. Japanese, Chinese, Korean, French, German, Norwegian, Estonian, and Russian definitions of happiness capture the ‘‘good fortune, luck’’ concept of happiness. It should be noted that, according to Nussbaum (1986), Aristotle used the term eudaimonia (‘‘happiness,’’ or ‘‘well-lived life’’) interchangeably with markon (‘‘resource,’’ or ‘‘fortune’’). Italian,
Culture and Well-Being: Conceptual and Methodological Issues
41
Portuguese, and Spanish definitions of happiness capture the satisfaction of one’s desires, wishes, and goals. English definitions of happiness center on the satisfaction of one’s desires, wishes, and goals, and enjoyable experiences. As both McMahon (2006) and Nussbaum (1986) point out, if the concept of happiness is closely associated with luck, one’s happiness is vulnerable to external conditions. Ancient Greeks had the notion that happiness cannot be achieved until one is dead, because no matter how lucky one has been, one can always be subjected to a misfortune later in life. Until one’s life is over, it is hard to evaluate it as a whole. This might be a reason why more Japanese respond to the happiness question by saying ‘‘I don’t know’’ than do Americans or Canadians (e.g., 41 out of 1,362 Japanese respondents, or 3% of Japanese, said ‘‘I don’t know’’ to the happiness question, as opposed to 0 out of 1,931 Canadians, and 8 out of 1,200 Americans, or .7% of Americans in the World Values Survey). Finally, it should be noted that there are several empirical works on the lay concept of happiness. Lu & Gilmour (2004) asked participants to describe what happiness is to them personally. Whereas many American college students described happiness as ‘‘an intense feeling’’ and ‘‘excitement,’’ many Chinese described it as ‘‘a calm, peaceful feeling’’ and ‘‘a sense of equilibrium.’’ This suggests that the meaning and correlates of ‘‘happiness’’ might be quite different for Americans than for Chinese. Consistent with this analysis, Tsai, Louie, Chen, & Uchida (2007) found that American children’s picture book characters had wider smiles than those in Taiwanese books, suggesting that Americans value high-activation positive emotion (e.g., excitement) more than East Asians. Interestingly, Tsai, Miao, & Seppala (2007) found more high-activation positive emotions in Christian classical texts than in Buddhist texts, as well. Uchida & Kitayama (2007) also found that American college students spontaneously associated the word happiness with primarily positive words, whereas Japanese college students spontaneously associated the word happiness with both positive and negative words. It is not surprising, then, that American college students viewed ‘‘happiness’’ and ‘‘life satisfaction’’ as more desirable than did East Asian students (Diener, Suh, Smith, & Shao, 1995). It is further interesting to note that Diener (2000) found that happiness is thought to be more desirable in Spanish-, Italian-, and Portuguese-speaking countries in general where ‘‘happiness’’ is defined in terms of the satisfaction of one’s desires and goals than it is in East Asia, France, Germany, Russia, and other nations where happiness is defined in terms of luck and fortune. Furthermore, Diener, Suh, et al. (1995) showed that Chinese students thought that experiencing negative affect was more desirable and appropriate than did
42
Section I: Measuring Well-Being in an International Context
American students. Eid & Diener (2001) also showed that there was a class of Chinese college students who deemed positive emotions as neither desirable nor undesirable, and negative emotions as desirable. The existence of this subgroup in China suggests that correlates and consequences of happiness might be different between Chinese and Americans. Finally, the desirability of happiness is also positively correlated with self-reported mean life-satisfaction score at the nation level (r [40] = .36, p < 05 in Diener’s 2000 college student data). Colombians and Puerto Ricans were among the most satisfied with their lives in Diener’s (2000) data, and they also viewed happiness to be most desirable, whereas Chinese and Thais were among the least satisfied with their lives, and they also viewed happiness to be least desirable. Different conceptualizations of happiness could result in different expressions of happiness. If happiness reflects personal accomplishments, for instance, it is acceptable to express it. In contrast, if it reflects luck and fortune, it is not as desirable to express it in public. Indeed, Lyubomirsky (2000) found that Americans feel more comfortable expressing their happiness than Russians do. Similarly, Diener, Suh, et al. (1995) found that Americans viewed expressing positive emotions as more appropriate than did Koreans and Chinese. Lutz’s (1987) ethnography also showed that the Ifaluk condemn smiling because the expression of happiness is thought to lead to the neglect of one’s duties. Cultural variation in the desirability of happiness and the expression of happiness has an important implication for international comparisons on well-being. The recent Gallup World Poll asked participants whether they smiled a lot on the previous day (Diener et al., 2009). While this is a nice behavioral measure of happiness, Lutz’s ethnography suggests that this might not capture the experience of happiness in some cultures (e.g., the Ifaluk) as well as in other cultures (e.g., the U.S.). Furthermore, different conceptualizations of happiness manifest themselves as different beliefs regarding happiness. Consistent with the luck and fortune definition of happiness, Russians believe that happiness is less attainable, less controllable, and more fragile than do Americans (Lyubomirsky, 2000). Similarly, Koreans believe that the amount of happiness one can experience in a lifetime is fixed, and that if one is happy now, one is likely to be less happy in the future (Suh & Koo, 2008). In sum, the concept of happiness evolved over time from pure ‘‘fortune and good luck’’ to ‘‘satisfaction of one’s desires’’ and ‘‘pleasures and enjoyment’’ in the United States. The ‘‘good fortune’’ definition of happiness is still prevalent in East Asia and various European nations. Interestingly, the concept of happiness in Italian-, Spanish-, and Portuguese-speaking nations centers on the satisfaction of one’s desires. Happiness is believed to
Culture and Well-Being: Conceptual and Methodological Issues
43
be more desirable in Italian-, Spanish-, and Portuguese-speaking nations than in others. In addition, the expression of happiness is more desirable in some nations than in others. There is also a variety of beliefs regarding happiness across cultures. These conceptual differences among cultures could result in cross-cultural differences in mean levels and correlates of subjective well-being. Next, I will review measurement and methodological issues in the context of national comparisons.
Measurement and Methodological Issues
Consistent with the diverse definitions of well-being (Diener et al., 1999; Ryan & Deci, 2001; Ryff & Singer, 1998; Veenhoven, 2008) and happiness summarized above, researchers utilize various measures to capture different concepts of well-being (see Eid, 2008; Kahneman & Krueger, 2006; Pavot, 2008; and Schimmack, 2008, for reviews on the measurement of well-being). As summarized above, global life satisfaction, Cantril’s (1965) Ladder scale, and happiness items dominated large-scale international surveys. Although these three items probably represent a latent concept of well-being, there are some important qualifications. First, Helliwell, Barrington-Leigh, Harris, & Huang (2009) showed that Cantril’s Ladder scale is nearly normally distributed around 5 on the 0––10 scale, whereas a global lifesatisfaction item is highly negatively skewed (i.e., many more responses are above the central point than below the central point). Second, Diener et al. (2009) showed different patterns of correlations between these three types of well-being measures and material conditions. For instance, Cantril’s Ladder of Life item was more strongly associated with educational and financial achievements than were affective measures (e.g., enjoyment, smiled a lot). Similarly, Kahneman, Krueger, Schkade, Schwartz, & Stone (2006) found that household income was more strongly associated with life satisfaction than the percentage of time one is in a good mood (r = .32 vs. .20; see also Inglehart, Foa, Peterson, & Welzel, 2008). Because the Ladder item asks where respondents think they stand on the ‘‘ladder’’ of life, it probably primes people to think of how far they have come in life (e.g., their educational and financial accomplishments). In contrast, whether respondents enjoyed something a lot or smiled a lot on the previous day is unlikely to lead respondents to think of their material wealth or educational/career accomplishments. Thus, in general, the Ladder of Life item captures person’s educational, financial, and career achievements most, followed by a global life satisfaction item. Happiness and other affect items capture inner feeling states, which may or may not be linked closely to a
44
Section I: Measuring Well-Being in an International Context
person’s material conditions. Interestingly, however, Deaton, Fortson, & Tortora (2008) found that affect items were more strongly associated with a death in the family than the Ladder item. It appears then that, although affect items are less strongly associated with material conditions, they are more sensitive to interpersonal events and conditions than is the Ladder item. Researchers have just begun to unpack the differences among the three types of subjective well-being measures often used in large-scale international surveys (e.g., Deaton et al., 2008; Diener et al., 2009; Helliwell et al., 2009). In addition to examining differential patterns of correlations, it might also be helpful to obtain free-association data in response to these items to investigate whether the type of information respondents base their judgments on is different across the three types of questions. Divergence between cognitive and affective components of well-being is well documented (e.g., Lucas, Diener, & Suh, 1996). It is interesting to note that even within the same life satisfaction scale there are distinct concepts of well-being assessed. For example, an item in the Satisfaction with Life Scale (SWLS: Diener, Larsen, Emmons, & Griffin, 1985) assesses the ‘‘good fortune and luck’’ definition of happiness: ‘‘The conditions of my life are excellent.’’ In contrast, the satisfaction of one’s desires, wishes, and goals is assessed in another item of the SWLS: ‘‘So far I have achieved important things I want in my life.’’ It is possible that the ‘‘The conditions of my life . . .’’ item functions better in cultures where the primary definition of happiness is luck and fortune than elsewhere, whereas the ‘‘So far I have achieved . . .’’ item functions better in cultures where the primary definition of happiness is personal accomplishment than it does elsewhere. In addition to widely used life satisfaction scales such as SWLS (Diener et al., 1985) and affect scales such as the Subjective Happiness Scale (Lyubomirsky & Lepper, 1999) and Positive and Negative Affect Schedule (PANAS: Watson, Clark, & Tellegen, 1988), there are many other scales frequently used in psychological research on well-being, including the Meaning in Life Scale (e.g., Steger, Frazier, Oishi, & Kaler, 2006), the Vitality Scale (e.g., Ryan & Fredrick, 1997), and the Psychological Well-Being Scale, which consists of the environmental mastery, autonomy, self-acceptance, purpose in life, personal growth, and positive relationships subscales (Ryff, 1989). In addition to self-reports of well-being, some physiological measures are shown to correlate with self-reported well-being. For instance, Urry & colleagues (2004) showed that the relative left prefrontal brain activity captured by the EEG during the resting period was positively associated with the SWLS (Diener et al., 1985) and the psychological well-being score
Culture and Well-Being: Conceptual and Methodological Issues
45
(Ryff, 1989). Cortisol outputs were also shown to be negatively correlated with Ryff’s psychological well-being scale (Lindfors & Lundberg, 2002). Thus, it is important to note that there are physiological measures of wellbeing that have shown some promise. In addition to the diverse contents and measures of well-being, researchers use different time frames in the assessment of well-being. Because large-scale international surveys almost exclusively use global reports, there is little discussion on the discrepancy between global reports and other reports of well-being in this volume. However, this distinction is important because the cross-cultural differences that emerged from global reports could be quite different from those from other types of reports. For instance, in the experience sampling method, participants reported their affect when the signal went off at a random moment (e.g., Diener & Larsen, 1984; Riis et al., 2005; Stone & Schiffman, 1994). In a daily diary study, participants reported their emotions and satisfaction of the day (e.g., Diener & Emmons, 1985; Oishi, Diener, Choi, Kim-Prieto, & Choi, 2007). More recently, Kahneman, Krueger, Schkade, Schwarz, & Stone (2004) created the Day Reconstruction Method (DRM), in which participants review a previous day’s activities and moods in detail. In the present chapter, I will use the term online reports of well-being in reference to random-moment reports, event-contingent reports, and daily reports of well-being. As can be seen in Figure 3.1, both global and online reports of well-being reflect objective conditions of the respondents’ lives and life events (e.g., long commute, high income) and their beliefs and implicit theories about
Culture
Global Reports of Well-Being
Implicit Theory Self-Concept Belief
Personality Temperament
Objective Conditions Life Events
Online, Specific Reports of Well-Being
Figure 3.1 Two Sources of Global and Online Reports of Well-Being
46
Section I: Measuring Well-Being in an International Context
well-being (e.g., life is good; I am a happy person) to some extent. However, I propose that global reports of well-being reflect beliefs, implicit theories, and self-concepts to a greater degree than online reports of well-being, whereas online reports of well-being reflect objective reality to a greater degree than global reports, due to the different information available for each type of report (see Robinson & Clore, 2002, for a similar view). In addition, I hypothesize that in general there should be a larger effect of culture and personality on global reports of well-being than on online reports of well-being, because culture and personality affect one’s beliefs and self-concepts associated with well-being to a greater degree than one’s objective living conditions or life events.
Global Reports of Well-Being
I will first review the methodological issues associated with global reports of well-being. There are a number of problems raised by experimental social psychologists. For instance, Schwarz & Strack (1999) summarized the literature and argued that global reports of life satisfaction are vulnerable to the item-order, current mood, and other contextual effects (which are present at the time of the judgments). Eid & Diener (2004) showed, however, that the effect of current mood in a naturalistic testing condition is relatively small, or 16% of the variance, compared to 76% of the variance attributed to the stable trait of life satisfaction. Schimmack & Oishi (2005) also meta-analyzed the item-order effect on global reports of life satisfaction, and found only negligible item-order effects. In addition, if the context effects are so strong on global reports of life satisfaction, test-retest reliability of life satisfaction cannot be very high. However, test-retest stability of life satisfaction was substantial (rs = .40–.80; Schimmack & Oishi, 2005). Furthermore, self-reported life satisfaction converges with informant reports of life satisfaction fairly well (rs = .54–.58, Sandvik et al., 1991). Nevertheless, cultural comparisons based on global reports are difficult to interpret because there are additional factors that come into play. For instance, the scale-use might be different across cultural groups (e.g., one group likes to use high numbers, whereas the other group likes to use medium numbers). The item functioning might be different across groups (the same item might not mean the same thing in different cultures). In addition, there might be cultural differences in self-presentation, referencegroup and standard of comparison, or memory and positivity bias when making life-satisfaction judgments. I will discuss each of these issues below.
Culture and Well-Being: Conceptual and Methodological Issues
47
Number-Use
Previous research has shown that there are some cross-cultural variations in number-use in responding to the survey questions. For instance, Chen, Lee, & Stevenson (1995) found that Americans are more likely to use the extreme number (7 on the 1––7-point scale) than Taiwanese and Japanese, and Japanese are more likely than Americans to use the midpoint. It is therefore possible that some cultural differences in mean levels of subjective wellbeing could be due to cultural differences in number use. I have compared American college students’ and Japanese college students’ responses to the Satisfaction with Life Scale (SWLS: Diener, Larsen, Emmons, & Griffin, 1985; 5-item scale on a 1––7-point scale). According to the standard scoring methods (summing responses to the 5 items on the 7-point scale; range from 5 to 35), American college students are more satisfied with their lives than Japanese (M = 24.41, SD = 6.00 vs. M = 19.07, SD = 6.15), t (533) = 9.46, p < .001, d = .81. If these differences were due to numberuse, then the differences should disappear once the original responses were recoded into 3-point scale (1 to 3 = 1, 2 = 2, 5 to 7 = 3). That is, if satisfied Japanese are using numbers 5 or 6, whereas equally satisfied Americans are using numbers 6 or 7, when we treat them as the same, we should see no differences between groups. This is not the case, however. Indeed, when I converted the 7-point scale to a 3-point scale, the differences were nearly identical (M = 11.10, SD = 2.60 vs. M = 8.85, SD = 2.77), t (533) = 9.09, p < .001, d = .79. To make sure this is not specific to the U.S.–Japan comparison, I analyzed the differences between Chilean college students and Japanese college students on the SWLS as well. On the original 7-point scale, Chileans were more satisfied with their lives than Japanese: M = 26.38, SD = 5.13 vs. M = 19.07, SD = 6. 15, t (536) = 14.34, p < .001, d = 1.24. On the 3-point scale, the difference was nearly identical: M = 11.84, SD = 1.92 vs. M = 8.85, SD = 2.76), t (536) = 14.49, p < .001, d = 1.25. Although researchers should be concerned with cultural differences in number use, United States–Japan and Chile–Japan differences in global life satisfaction cannot be easily attributed to response sets. Item Functioning
There is a concern that some items function better in one culture than in another. For instance, one item from the SWLS, ‘‘If I live my life over, I would change almost nothing,’’ is a good indicator of general life satisfaction in one culture, but it might not be in another. It is possible to determine the measurement equivalence using Item Response Theory (IRT). IRT is a method in assessing measurement equivalence in multiple groups. I conducted several Differential Item Functioning (DIF) analyses (Oishi,
48
Section I: Measuring Well-Being in an International Context
2006) based on the SWLS (Diener et al., 1985) among American and Chinese college student respondents, using the Multilog 7.03 program (see Oishi, 2007, for the DIF analyses on Watson, Clark, & Tellegen’s 1988 PANAS scale). Based on the strict criteria, four of the five SWLS items showed significant DIF between American and Chinese students. Interestingly, the only non-DIF item was ‘‘The conditions of my life are excellent,’’ which directly assessed the ‘‘luck and fortune’’ of their current lives. The item that captured the satisfaction of one’s goals, ‘‘So far I have gotten important things I want in my life,’’ had the largest DIF. Most important, although the mean difference between Chinese and American college students decreased substantially from d = 1.18 to d = .71 by equating the more comparable items heavily than incomparable ones via IRT, American college students were still considerably more satisfied with their lives than were the Chinese students. Thus, I showed (2006) that the observed mean difference between Chinese and American college students on the SWLS is not due entirely to differential item functioning. Self-presentation
In addition to the number-use and the item functioning, there is another concern, which is self-presentation. If, in East Asia, it is not desirable to say that you are happy, and in the United States it is highly desirable to say that you are happy, then the mean difference between East Asians and Americans could be driven by cultural variations in self-presentation. Furthermore, if the self-presentation hypothesis is true, then we should not observe cultural differences in informant reports. It is therefore important to compare self-reports against informant reports. In one study (Oishi & Diener, 2008), we collected self-reports and informant reports of life satisfaction in Japan and the United States We found very similar magnitudes of mean differences in self-reports (M = 23.08 vs. 21.27 at Time 1, t [174] = 2.13, p < .05, d = .32; M = 23.78 vs. 20.23 at Time 2, t [174] = 3.74, p < .01, d = .57) and informant reports (M = 23.15 vs. 21.77, t [174] = 2.06, p < .05, d = .31). Thus, national differences in self-reported life satisfaction between Japanese and Americans might not be accounted for by modest responding of Japanese relative to Americans. However, it should be noted that recent research using implicit measures of self-esteem (the measure based on reaction time in categorizing ‘‘me’’ and positive versus ‘‘not me’’ and negative, and vice versa) has shown that national differences in selfreported self-esteem between Japanese and American college students might be due in part to self-presentation, as the researchers did not find any difference in implicit measures of self-esteem, despite the large differences in explicit measures of self-esteem (Yamaguchi et al., 2007).
Culture and Well-Being: Conceptual and Methodological Issues
49
Memory Bias
Another possible explanation for national variations in mean levels of life satisfaction is concerned with memory bias. It is possible that one group may pay a lopsided amount of attention to positive events that happened to them versus negative events, whereas another group may pay attention to positive and negative events equally. Then, national differences in global reports may reflect differences in how experiences are integrated in making life satisfaction judgments rather than how they feel day to day. In a first series of studies, I employed a daily diary and an experience sampling method (Oishi, 2002) to examine this possibility. Although Asian Americans and European Americans did not differ in their average daily satisfaction for seven days, European Americans evaluated the week as significantly more satisfying than did Asian Americans. Similarly, although Asian Americans and European Americans did not differ in the proportion of random moments in which they felt happy during the sevenday period, European Americans recalled having experienced happiness more frequently than did Asian Americans. In a related set of studies (Oishi & Diener, 2003), we also found that even when European American and Asian American participants solved a similar number of anagrams, European Americans remembered having solved more anagrams one week later than did Asian Americans. Similarly, European American participants remembered making more baskets than did Asian American participants, despite the fact that they made almost the same number of baskets one week earlier. These two papers, then, suggest that cultural variations in retrospective reports of well-being might not necessarily reflect differences in actual experiences. In the third study of this kind, Scollon, Diener, Oishi, & Biswas-Diener (2004) collected data from the United States, Japan, and India using an experience-sampling method. In this study, the authors did not find evidence of memory bias; that is, Hispanic Americans experienced more positive emotions at the random moments than did European Americans, Japanese, and Indians; and in retrospective reports, Hispanic Americans reported having experienced more positive emotions than other groups. Japanese and Indians experienced positive emotions less often at random moments, and reported having experienced positive emotions less often than European Americans. Thus, Scollon et al. failed to replicate my (2002) findings. However, in the fourth study of this kind, Wirtz’s experiencesampling study among European Americans and Asian Americans (2004) replicated my 2002 findings. Namely, although European Americans and Asian Americans did not differ in the proportion of time they felt happiness at random moments during one week. Later, European Americans reported
50
Section I: Measuring Well-Being in an International Context
having experienced happiness more often than did Asian Americans. In addition, my colleagues’ and my 2007 daily diary study in the United States, Japan, and Korea also replicated my 2002 findings in that there were no national differences in the average daily satisfaction over 21 days among European Americans, Asian Americans, Japanese, and Koreans; yet European Americans reported being more satisfied with their lives in general on the SWLS than did Asian Americans, Japanese, and Koreans. Finally, Whitchurch, Miao, Kurtz, and I (2009) conducted another study of this sort, this time using Kahneman et al.’s (2004) Day Reconstruction Method (DRM) in Korea and the United States. Again, there were no differences in positive affect experienced during the daily activities recorded between Korean and American participants, yet Americans reported being more satisfied with their lives in general on the SWLS than did Koreans. Positivity Bias
If there is a memory bias, there could be a ‘‘positivity’’ bias (or tendency to evaluate something more positively than expected from an objective criterion) in global reports of well-being as well. It is important, then, to examine the degree of satisfaction that respondents feel toward specific life domains, as well as life as a whole. Diener, Suh, Smith, & Shao (1995) examined mean differences in domain as well as global life satisfaction, first among Japanese, Korean, and American college students. American college students reported being more satisfied with their lives in general and experienced happiness more often than did Korean and Chinese college students. In addition, in 11 out of the 12 domains examined, American students reported being more satisfied than Korean and Chinese students. In their second study, Diener et al. also found similar national differences in both global reports of well-being and six of the 13 life domains. In another study, Diener and colleagues (Diener, Scollon, Oishi, Dzokoto, & Suh, 2000) explicitly tested the positivity bias hypothesis by computing positivity bias based on the discrepancy between specific and global domain satisfactions (e.g., satisfaction with professors, textbooks, and lectures versus satisfaction with education). Interestingly, in the nations where participants reported higher levels of global life satisfaction, positivity bias was also stronger. For instance, in Puerto Rico, Colombia, and Spain, respondents evaluated the global domains as much more satisfying than the corresponding specific domains, whereas in Japan, Korea, and China respondents evaluated the global domains as much less satisfying than the corresponding specific domains. Indeed, the correlation between the positivity bias score and the mean SWLS score of the nation was
Culture and Well-Being: Conceptual and Methodological Issues
51
substantial: r = .57, p < .001. These findings suggest that cross-national variations in specific domain satisfaction are smaller than cross-national differences in global reports of life satisfaction. One study (Oishi & Diener, 2008) examined the judgment bias across cultures. Participants in the United States and Japan completed a lengthy survey and were asked about their current moods and the degree to which they felt happy about their entire life. Because the testing conditions were very similar across cultures, there should not have been cultural differences in current moods at the time of the assessment. Indeed, American and Japanese participants were nearly identical in their current moods (M = 6.12, SD = 1.73 vs. M = 5.97, SD = 1.88 on the 1––9-point scale, t [530] = .87, p = .38, d = .08). However, Americans reported having been happier about their ‘‘entire life’’ than Japanese, M = 6.10, SD = 1.85 vs. M = 5.69, SD = 1.87, t = 2.35, p < .05, d = .20. These discrepancies between online and global reports of well-being in the magnitude of cultural differences are consistent with the model proposed in Figure 3.1, and present an alternative source of divergence between the affect items and Cantril’s Ladder of life items used in the Gallup World Poll (affect items refer to ‘‘yesterday’’ or a very narrow time frame, whereas the Cantril’s Ladder item refers to ‘‘life now,’’ which is probably interpreted as ‘‘life’’ these days). Reference-Group Effect
Finally, the interpretation of mean differences across cultures is difficult because different cultures might use different standards and reference groups for comparison. For instance, when rating how collectivist they are, Americans compare themselves with other Americans, and might conclude that they are quite collectivist. In contrast, Japanese respondents compare themselves with other Japanese, and might conclude that they are not that collectivist. Indeed, Heine, Lehman, Peng, & Greenholtz (2002) found support for this in comparisons between Japanese and American college students. Furthermore, Heine, Buchtel, & Norenzayan (2008) found that self-reported conscientiousness did not correlate with behavioral indicators of conscientiousness, such as walking speed. The reference-group effect, then, poses a serious threat to international comparisons on self-reports of well-being. Oishi & Roth (2009) investigated whether national means of self-reported life satisfaction and self-esteem would also show a lack of correlation with behavioral indicators. Unlike many of the personality trait measures (e.g., conscientiousness), however, national means of self-reported life satisfaction and self-esteem showed predicted relations to objective indicators such as the corruption index and
52
Section I: Measuring Well-Being in an International Context
the suicide rate (see also Helliwell, 2007, for the negative correlation between self-reported life satisfaction and suicide at the level of the nation). Thus, although the standard of comparison and the reference group pose some threats to the validity of self-reported values and personality traits, they do not appear to pose a similar threat to self-reports of wellbeing.
Online Reports of Well-Being
So far I have discussed methodological issues associated with global reports of well-being. I now turn to several methodological issues associated with online measures of well-being (see Scollon, Kim-Prieto, & Diener, 2003, for review). The number-use, item functioning, and self-presentation issues apply to most online measures of well-being. One major disadvantage of online measures is that they are much more time-consuming to collect data from than global reports of well-being. Life satisfaction, domain satisfaction, and affect questions can be answered in a matter of five to 10 minutes. In contrast, it takes at least a week to have good experience-sampling or daily diary data. The Day Reconstruction Method (Kahneman et al., 2004) is a great improvement in this regard, as it takes as little as 45 minutes to complete. Second, because of the intense nature of data collection, willing participants’ self-selection bias and relatively high attrition rate pose some additional challenges to the generalizability of the findings. For instance, depressed individuals are less likely to participate in and complete daily diary or experience-sampling studies than simple survey studies. Third, again because of the intense nature of data collection, it is very difficult to collect data from a large number of participants using online measures. This will further limit generalizability of the findings. Finally, online measures of well-being cover only a short period of one’s entire life; in the case of the DRM, it captures only one day of respondents’ lives. Thus, there is a concern regarding the representativeness of these data: namely, how representative are respondents’ online reports for one day or one week of their lives in general? In sum, both global and online measures of well-being have strengths and weaknesses. In a large international survey, it is unrealistic to include online measures of well-being. Thus, most national comparisons of wellbeing are limited to global self-reports of well-being. The interpretations of such data, therefore, require that some caution be taken, although as suggested below, the rank-order of national means of well-being corresponds fairly well with economic and social indicators.
Culture and Well-Being: Conceptual and Methodological Issues
53
Mean differences in subjective well-being across cultures
The previous research summarized above indicates that artifactual factors (e.g., number use, item functioning, and self-presentation) seem to play a relatively small role in explaining national differences in mean levels of life satisfaction. Previous research also indicates that mean differences between North Americans and East Asians are due in part to the ways in which individuals from different cultures make life satisfaction judgments. The discrepancies between specific/online reports and global reports of well-being seem to reflect the fact that subjective well-being is not a unitary construct, and that both specific/online reports and global reports of wellbeing should be considered important indicators of well-being. First, ranks of nations in subjective well-being are somewhat different, depending on the type of well-being measures used (Diener et al., 2009). For instance, Denmark had the highest score on Cantril’s Ladder scale, followed by Finland, the Netherlands, Switzerland, Norway, and New Zealand in the 2007 Gallup World Poll. According to the same data set, New Zealand had the highest score in terms of the ‘‘enjoyment’’ item, followed by Denmark, the Netherlands, Sweden, and Laos. In the most striking case, Laos was the eightieth out of 132 nations in terms of Cantril’s Ladder scale; yet it was the fifth in enjoyment. Although the rank-order of one well-being item with another is moderate to high, it is important to recognize some divergence. As noted by Diener (2008), it might be futile to look for the world’s happiest nation. Second, although previous research has shown memory and positivity bias in global reports of well-being (Schwarz & Strack, 1999, for review), national differences in global reports of well-being have shown systematic and predictable correlations with social and economic indicators. For instance, Diener, Diener, & Diener (1995) analyzed three large international surveys and one student survey, and found that the average subjective well-being score of these four surveys was highly correlated with individualism (r = .77, p < .01), GDP per capita (r = .58, p < .01), purchasing power (r = .61, p < .01), civil and political rights (r = .46), and the equality index (e.g., income equality, longevity equality; r = .48). Similarly, Veenhoven (1995) analyzed two international surveys and reached an almost identical conclusion: namely, income, social equality, political freedom, and access to knowledge/education were all significantly associated with the subjective well-being of nations. More recently, Veenhoven (2008) analyzed data from 90 nations and replicated the earlier findings. In addition, he showed that tolerance, trust, and safety are positive correlates of the well-being of nations (see also Veenhoven, 2009). Finally, Helliwell et al. (2009) showed that social contextual variables (e.g., whether there is
54
Section I: Measuring Well-Being in an International Context
someone respondents can count on when they need help) account for more variance than income in cross-national differences in mean levels of Cantril’s Ladder scale. Diener et al. (1995) showed that the rank-order of nations in global reports of well-being is highly stable. For instance, the rank-order of nations in one year and another was highly correlated (r = .67, p < .01). Cross-temporal stability of national means of life satisfaction, then, poses a question of sensitivity to external conditions of nations (see also Easterlin & Sawangfa, 2009; Graham, Chattopadhyay, & Picon, 2009; Layard et al., 2009, on this question). For instance, does a high-impact event (e.g., oil shock) affect the national mean of life satisfaction? According to Veenhoven (2008), the answer is yes. In reaction to the 1973 oil shock, Japanese life satisfaction showed a clear dip around 1974. Ronald Inglehart and his colleagues (Inglehart et al., 2008; see also Inglehart, 2009) have shown that an increase in perceived freedom was highly correlated with an increase in life satisfaction across nations. Namely, nations that experienced an increase in life satisfaction happened to be nations whose residents reported a greater sense of freedom than before. Previous research, then, suggests that nations with high GDP, political freedom, social equality, and access to education provide a fertile ground for individuals to thrive. These conditions do not guarantee high levels of subjective well-being for every citizen, as some individuals are predisposed to depression, or unfortunately experience various negative events (Bouchard, 2004). These are the personal and extraneous factors that social institutions cannot control. Thus, policies should aim to improve the economy for a greater number of citizens (though a rapid growth might have a negative consequence on well-being in the short run; Graham et al., 2009), in addition to increasing social equality, political freedom, and access to education, so that society may provide individuals with opportunities to thrive.
Correlates of subjective well-being across cultures
Besides mean differences across cultures, researchers discovered similarities and differences in correlates of subjective well-being across cultures. Several theorists (e.g., Ryan & Deci, 2001) have argued that the satisfaction of basic psychological needs such as relatedness and autonomy are the universal correlates of well-being, while others (e.g., Kitayama & Markus, 2000) have argued that correlates of well-being differ across cultures to the extent that ideal personhood is different in different cultures.
Culture and Well-Being: Conceptual and Methodological Issues
55
There are several findings consistent with Kitayama & Markus’s (2000) view. For instance, although self-esteem was positively associated with life satisfaction in the 31 nations studied, self-esteem was more strongly associated with life satisfaction in individualistic nations such as the United States and Canada than in collectivistic nations such as India and Cameroon (Diener & Diener, 1995; see also Oishi, Diener, Lucas, & Suh, 1999, for replications). Although financial satisfaction was positively associated with life satisfaction in virtually all nations surveyed, it was more strongly associated with life satisfaction in poor nations than in wealthy nations (Diener & Diener, 1995; Inglehart et al., 2008; Oishi et al., 1999). Also, although a sense of freedom was associated with life satisfaction in virtually all nations, this association was stronger in wealthy, individualistic nations than in less wealthy, collectivistic nations (Inglehart et al., 2008; Oishi et al., 1999). In addition to cultural variations in the link between domain satisfaction and global life satisfaction, researchers have found that the relationship between emotional experiences and life satisfaction varies systematically across cultures. For example, hedonic balance (the frequency of positive emotion versus negative emotion) was more strongly correlated with life satisfaction in individualistic nations than in collectivistic nations (Suh, Diener, Oishi, & Triandis, 1998). Among Americans, interpersonally disengaging emotions such as pride were strongly associated with happiness, whereas among Japanese interpersonally engaging emotions such as friendly feelings were strongly associated with happiness (Kitayama, Markus, & Kurokawa, 2000; Kitayama, Mesquita, & Karasawa, 2006). Finally, the type of motivation predictive of life satisfaction was also different across cultures. For instance, Japanese college students who were pursuing their goals to make their family and friends happy became more satisfied with their lives over time as they achieved their goals, whereas American college students who were pursuing their goals for themselves became more satisfied with their lives over time as they achieved their goals (Oishi & Diener, 2001). Similarly, among Americans, pursuing goals with the avoidant mindset was negatively associated with life satisfaction, whereas it was not negatively associated with life satisfaction among Koreans and Russians, where avoiding negative evaluations is deemed important (Elliot, Chirkov, Kim, & Sheldon, 2001). Other researchers found some similarities and differences across cultures, as well. For example, Schimmack et al. (2002) found that the relationship between extraversion and hedonic balance was similar among Americans, Mexicans, Japanese, Ghanaians, and Germans. However, replicating Suh et al. (1998), they also found that the relationship between hedonic balance and life satisfaction was stronger among Germans and Americans than
56
Section I: Measuring Well-Being in an International Context
among Mexicans, Japanese, and Ghanaians. Similarly, Kwan, Bond, & Singelis (1997) found that the correlations between big five traits (i.e., extraversion, neuroticism, openness to experience, conscientiousness, and agreeableness) and self-esteem were similar between Hong Kong and American college students. Interestingly, however, self-esteem was more strongly associated with life satisfaction among American college students than among Hong Kong students, whereas relationship harmony was more strongly associated with life satisfaction among Hong Kong than American students (see also Kang, Shaver, Sue, Min, & Jing, 2003). Finally, Uchida et al. (in press) showed that social support was equally associated with life satisfaction among Americans, Japanese, and Filipinos. However, these researchers also demonstrated that the link between social support and life satisfaction was fully mediated by self-esteem among Americans, whereas it was not among Japanese and Filipinos. Social support is associated with life satisfaction to the extent that it supports self-esteem among Americans, whereas it is associated with life satisfaction above and beyond self-esteem among Japanese and Filipinos. Finally, self-determination researchers have found more evidence for cultural similarities than differences in correlates of well-being. For instance, Chirkov, Ryan, Kim, & Kaplan (2003) showed that individuals high in internalized motivation as opposed to externalized (i.e., those who say they do what they do because they thoughtfully considered and fully chose to do so by themselves) report being more satisfied with their lives than individuals low in internalized motivation in the United States, Turkey, Russia, and South Korea. Similarly, Sheldon et al. (2004) showed that participants who say they are pursuing their goals for intrinsic and internalized reasons report having experienced more positive affect and less negative affect than those who say they are pursuing their goals for extrinsic reasons in the United States, South Korea, Taiwan, and China. To summarize, previous research has identified many cultural differences in the correlates of subjective well-being. Self-esteem and pride are the signatures of well-being in many individualistic nations, whereas they are not in less individualistic nations. Similarly, interpersonally disengaging positive emotions (e.g., pride, friendly) are strongly associated with wellbeing in individualistic nations, whereas interpersonally engaging positive emotions are associated with general well-being in collectivist cultures. In less affluent nations, the satisfaction of basic needs (e.g., financial satisfaction) is more strongly associated with life satisfaction than in affluent nations. At the most abstract level of motivation, however, researchers have identified cross-cultural similarities. Individuals high in well-being in any culture tend to engage in action from internalized motivation as opposed to externalized motivation.
Culture and Well-Being: Conceptual and Methodological Issues
57
Consequences of subjective well-being
Most researchers so far have focused on the correlates of subjective wellbeing, while few have investigated its consequences (Diener, et al., 1999, for review). Cross-cultural research has also focused on the correlates of subjective well-being so far. This is understandable, considering that happiness has long been thought of as an ultimate goal. Naturally, researchers did not question the consequences of the ultimate goal. Veenhoven (1988) and Lyubomirsky, King, & Diener (2005) are exceptions in this regard, as they have conducted meta-analyses on this topic. These researchers generally found positive outcomes from being happy or satisfied with life. For instance, happy individuals tend to live longer, have better social relationships, perform better at work than unhappy people, and make more money than unhappy people. More recently, researchers examined the optimal level of subjective well-being, and found that the moderate level of life satisfaction was associated with the highest level of later educational attainment and income, whereas the highest level of life satisfaction was associated with the highest level of relationship stability and voluntarism (Oishi, Diener, & Lucas, 2007). In terms of the distinction between global and online measures of wellbeing, it is interesting that most of the previous work on the consequences of subjective well-being used global reports of well-being. For instance, several researchers found that a decline in global reports of life satisfaction would predict later mortality among older adults (Gerstorf, Ram, Eastabrook, Schupp, Wagner, & Lindenberger, 2008; Gerstorf, Ram, Ro¨cke, Lindenberger, & Smith, 2008; Mroczek & Spiro, 2005). Among participants more than 50 years old at the beginning of the Midlife in the United States (MIDUS) project, life satisfaction at Time 1 predicted the likelihood of mortality in the following 10 years (Oishi & Schimmack, 2008). These findings indicate that global reports do predict important life outcomes such as mortality. Most longitudinal studies reviewed above regarding the consequences of subjective well-being were conducted in the United States and other Western, developed nations. Thus, the question remains whether the findings on the consequences and optimal levels of happiness will hold across cultures. Unfortunately, I was unable to locate any longitudinal data that directly speak to this issue. Instead I reviewed previous research on cultural differences in the relations between happiness and later choices and decision making. The desirability of happiness is important in the present context because happiness might not be associated with positive outcomes in a society where happiness is not as valued as it is in the United States For instance,
58
Section I: Measuring Well-Being in an International Context
if positive affect (PA) is not viewed as valuable and functional among East Asians as it is among European Americans (Diener, Suh, et al., 1995), then the actual experiences of PA might not predict East Asians’ future choices and decisions as well as those of European Americans. To test this idea, Diener and I (2003) had participants solve anagram tasks (Study 1) and play a basketball game (Study 2), while assessing their actual enjoyment of the task. After their completion of the task, participants were asked to indicate whether they would like to repeat the same task or perform another task. As predicted, actual enjoyment of the task predicted the choice of the task among European American, but not among Asian American participants (see Heine et al., 2001, for similar results). In other words, European Americans who enjoyed the anagram (basketball) chose the anagram (basketball) at Time 2, whereas European Americans who did not enjoy the anagram (basketball) chose the word fragment task (darts game) at Time 2. Among Asian American participants, we did not find such a straightforward relationship between actual enjoyment and subsequent choice. This might explain why, on average, East Asians are less satisfied with their lives than are Americans. Together, these findings suggest that PA has a more direct consequence for choice and decision making for European Americans than for Asian Americans. It is interesting to note, then, that European American findings are perfectly in line with the influential affect-as-information theory (Schwarz & Clore, 1988), which assumes that PA is an internal signal that things are going well and no change is required, whereas negative affect (NA) is a signal that things are not going well and some change is required. In contrast, Asian American findings are not consistent with the affect-as-information theory. In sum, these findings suggest that the link between happiness and important life outcomes might be more direct among European Americans than others, and that the optimal level of happiness for these life outcomes might be different across cultures.
Summary, Future Directions, and Conclusion
In this chapter, I have first summarized conceptual and methodological issues in culture and well-being research, and then presented representative findings on national differences in mean levels, correlates, and consequences of subjective well-being. Historically, the concept of happiness revolved around the notions of ‘‘fortune’’ and ‘‘good luck’’ in the United States. This concept is still prevalent in East Asia as well as various European nations. Interestingly, however, the concept of happiness centers on the satisfaction of one’s desires in Italian-, Spanish-, and Portuguese-speaking
Culture and Well-Being: Conceptual and Methodological Issues
59
nations. In addition, the ‘‘fortune’’ and ‘‘good luck’’ definitions were largely replaced by ‘‘pleasure’’ and ‘‘enjoyment’’ in American notions of happiness today. These diverse definitions of happiness are reflected in the way researchers assess subjective well-being. For instance, the item ‘‘the conditions of my life are excellent’’ in the SWLS (Diener et al., 1985) assesses the ‘‘fortune and good luck’’ definition of happiness. Another item of the SWLS, ‘‘I have gotten the most important things I want in my life,’’ captures the idea that happiness involves the satisfaction of one’s desires. Furthermore, researchers routinely assess the frequency with which individuals experience enjoyment and pleasure in their daily lives. Although large-scale international surveys have traditionally included only some of the three types of single-item measures of well-being (Cantril’s Ladder scale, a global life satisfaction item, and a happiness item), and therefore, the main focus of the current volume centers on the findings from these items, there are many well-validated scales of wellbeing (e.g., Meaning in Life scale). It will be critical in the future to map out various concepts of well-being and measurement tools that tap into different concepts of well-being in large-scale international surveys (see Diener et al., 2009; Inglehart, 2009; Inglehart et al., 2008, for an initial effort). With this map, the national differences in mean levels, correlates, and consequences of well-being become more interpretable. For instance, would the magnitude of national variations in subjective well-being decrease between East Asia and North America if the survey measured subjective well-being in terms of ‘‘luck’’ and ‘‘fortune,’’ as opposed to pleasure and enjoyment? Can national differences in correlates of subjective well-being be explained by conceptual differences in happiness? There are several important methodological issues when interpreting national differences in global reports of well-being, ranging from the number/scale-use (response style) and self-presentation motives to memory and positivity biases (see Table 3.2 for a checklist). Although online measures of well-being (e.g., experience-sampling method, daily diary method, Day Reconstruction Method) overcome some of the issues, such as memory and positivity biases, there are also several methodological issues uniquely associated with online reports of well-being, such as the difficulty of getting a representative sample, the time-consuming nature of data collection, and the representativeness of the time sampled. As theorized in Figure 3.1, different reports of well-being might reflect different sources of well-being; namely, self-concepts/beliefs associated with well-being (e.g., ‘‘I am a happy person’’; ‘‘life is good’’) and objective reality (e.g., death in the family). I believe that both types of reports capture ‘‘valid’’ variance in subjective well-being and predict different outcome measures. Global reports tend to predict one’s future decisions better than
60
Section I: Measuring Well-Being in an International Context
TABLE 3.2 A Checklist for Cross-Cultural Research on Well-Being: Things Researchers should Pay Attention to When Analyzing and Interpreting the Cross-Cultural Data on Well-Being 1. Conceptual Equivalence: Does the target concept (e.g., happiness, life satisfaction) have the same meaning across cultures? 2. Translation and Indigenous Words: Is the target concept accurately translated? Does it require an indigenous word? 3. Desirability of the Concept: Is the target concept equally desirable across cultures? 4. Response style: Are there notable differences in number use (e.g., tendency to use the middle point, extreme point) across cultures? 5. Item functioning: Do well-being items function the same across cultures? 6. Self-presentation concern: Are there notable differences in how people present themselves (e.g., modesty)? 7. Memory bias: Are there different kinds of memory biases across cultures (e.g., positivity bias)? 8. Validity criteria: Will the findings be replicated with different types of well-being questions (e.g., Cantril vs. happiness item)? Are there other ways to measure wellbeing besides global self-reports? Online measures? Behavioral measures? Physiological measures?
online reports (e.g., likelihood of returning to the same location for spring break—Wirtz, Kruger, Scollon, & Diener, 2003; future relationship stability—Oishi & Sullivan, 2006). Online reports, on the other hand, might better predict how individuals would actually feel moment-tomoment when they are in the same situation the next time than would global reports (Kahneman, 1999). With this distinction, it is important to consider national variations in online as well as global reports of well-being in the future. At this point, most international data rely on global reports of well-being, and are limited to three types of measures: Cantril’s (1965) Ladder item, a global life satisfaction item, and happiness and related affective scales. However, it is also important to recognize that global and online/specific reports of well-being capture distinct aspects of subjective well-being, and the picture that comes out of specific or online reports of well-being will probably be different from the one that comes out of global reports of well-being. In other words, although the current volume focuses on the distinction between cognitive and affective components of subjective well-being, different patterns of international differences might emerge, not only in terms of cognitive vs. affective components of well-being, but also in terms of global vs. online reports of well-being (see Table 3.1). As summarized above, national differences in global reports of well-being match up quite well with objective social and economic indicators. However, it is also true
Culture and Well-Being: Conceptual and Methodological Issues
61
that there are some nations that rank substantially higher or lower than expected in these social and economic indicators (e.g., Argentina, Japan). It is crucial to examine whether national differences in online reports of wellbeing will match up even better than global reports with objective social and economic indicators. If this is the case, national indicators of happiness should include online reports as well as global reports. As discussed above, mean levels of subjective well-being are higher in wealthier, safer, more democratic and egalitarian nations than in poorer, dangerous, less democratic nations with more social disparity (e.g., Diener et al., 1995; Inglehart et al., 2008; Veenhoven, 2008). These social and economic indicators are typically correlated with mean levels of subjective well-being at the range of .40 to .70. Recent research has also shown some within-nation changes in subjective well-being across time in response to major events such as the oil shock in the 1970s (Veenhoven, 2008) and changes in political freedom (Inglehart et al., 2008; see, however, Easterlin & Sawangfa, 2009). Yet the sensitivity of global reports of well-being to various changes in social policies has not been explicitly tested. It is important to document the degree of sensitivity of various measures of well-being. Then it might be possible to identify the factors that predict changes in subjective well-being in the future. There is an active scholarship examining cross-cultural similarities and differences in correlates of subjective well-being. Many studies showed that self-esteem and pride are better correlates of life satisfaction in individualistic nations than in collectivistic nations (Diener & Diener, 1995; Kitayama et al., 2000; Oishi et al., 1999). Hedonic balance is more strongly associated with life satisfaction in individualistic nations than in collectivistic nations (Schimmack et al., 2002; Suh et al., 1998). These findings suggest that the correlates of well-being differ across cultures, depending on cultural values, norms, and practices. Since most research in this area has relied on global reports of well-being, it will be critical in the future to examine whether correlates of well-being would be different across cultures, when online reports were employed (see Oishi, Diener, Scollon, & Biawas-Diener, 2004; and Oishi, Diener, Choi, et al., 2007, for an initial effort). Finally, I have summarized above that global reports of subjective wellbeing are associated with important life outcomes such as job performance, income, and mortality. Unlike on the topics of mean differences and correlates, there has not been a large-scale cross-national study on the consequences of subjective well-being. It is critical to investigate potential cultural variations in the consequences of subjective well-being. Like in the United States, would happy people be healthier, making more money, and living longer than less happy people in Japan, China, or Ghana?
62
Section I: Measuring Well-Being in an International Context
Conclusion
The science of well-being cannot fulfill its potential without solid measurement tools. The solid measurements of well-being in turn cannot be devised without proper understanding of the concepts of well-being. This chapter has reviewed cultural and historical variations in concepts of well-being and highlighted the importance of diverse measurement tools in capturing different conceptions of well-being. For practical reasons, large international surveys have relied exclusively on single-item global reports of well-being. It is important to interpret the results with the limitations associated with global reports of subjective well-being in mind (e.g., low reliability of these items). Simultaneously, it is critical to employ data-analytic techniques that test measurement equivalence across cultures and correct for measurement errors on one hand (e.g., a latent trait analysis in the structural equation model framework: see Oishi, 2007; Schimmack et al., 2002, for examples), and develop a new measurement tool of well-being that can be used in large international surveys on the other (Kahneman et al.’s 2004 DRM; and Harter & Arora, 2009, on time spent working and job fit, are some important first steps toward this goal). National differences in the concepts, mean levels, correlates, and consequences of well-being could be attributable to various factors, ranging from economic and political systems to sociohistorical and cultural traditions. Ultimately, culture and well-being researchers must understand the relative contributions of economic, political, socio-historical, and cultural factors, as well as complex interactions among these multiple factors.
Acknowledgments
I would like to thank Ed Diener, John Helliwell, Jamie Schiller, and an anonymous reviewer for invaluable comments on an earlier version of this paper.
References Andrews, F. M., & Robinson, J. P. (1991). Measures of subjective well-being. In J. P. Robinson, P. R. Shaver, & L. S. Wrightsman (Eds.), Measures of personality and social psychological attitudes (pp. 61–114). San Diego, CA: Academic Press. Bouchard, T. J., Jr. (2004). Genetic influence on human psychological traits. Current Directions in Psychological Science, 13, 148–151.
Culture and Well-Being: Conceptual and Methodological Issues
63
Cantril, H. (1965). The pattern of human concern. New Brunswick, NJ: Rutgers University Press. Chen, C., Lee, S-Y., & Stevenson, H. W. (1995). Response style and cross-cultural comparisons of rating scales among East Asian and North American students. Psychological Science, 6, 170–175. Chirkov, V., Ryan, R. M., Kim, Y., & Kaplan, U. (2003). Differentiating autonomy from individualism and independence: A self-determination theory perspective on internalization of cultural orientations and well-being. Journal of Personality and Social Psychology, 84, 97–110. Clark, A. (2009). Work, jobs, and well-being across the Millennium. In E. Diener, D. Kahneman, & J. F. Helliwell (Eds.), International differences in well-being. New York: Oxford University Press. Cohen, D., & Gunz, A. (2002). As seen by the other. . .: Perspectives on the self in the memories and emotional perceptions of Easterners and Westerners. Psychological Science, 13, 55–59. Deaton, A., Fortson, J., & Tortora, R. (2008). Life (evaluation), HIV/AIDS, and death in Africa. A working paper. Princeton, NJ: Princeton University. Diener, E. (2000). Subjective well-being: The science of happiness, and a proposal for a national index. American Psychologist, 55, 34–43. Diener, E., (2008). Myths in the science of happiness, and directions for future research. In M. Eid & R. J. Larsen (Eds.), The science of subjective well-being (pp. 493–514). New York: Guilford Press. Diener, E., & Diener, M. (1995). Cross-cultural correlates of life satisfaction and selfesteem. Journal of Personality and Social Psychology, 68, 653–663. Diener, E., Diener, M., & Diener, C. (1995). Factors predicting the subjective wellbeing of nations. Journal of Personality and Social Psychology, 69, 851–864. Diener, E., & Emmons, R. A. (1985). The independence of positive and negative affect. Journal of Personality and Social Psychology, 47, 1105–1117. Diener, E., Emmons, R. A., Larsen, R. L., & Griffin, S. (1985). The Satisfaction with Life Scale. Journal of Personality Assessment, 49, 71–75. Diener, E., Kahneman, D., Tov, W., & Arora, R. (2009). Income’s differential impact on judgments of life versus affective well-being. In E. Diener, D. Kahneman, & J. Helliwell (Eds.), International differences in well-being. New York: Oxford University Press. Diener, E., & Larsen, R. J. (1984). Temporal stability and cross-situational consistency of affective, behavioral, and cognitive responses. Journal of Personality and Social Psychology, 47, 871–883. Diener, E., Oishi, S., & Lucas, R. E. (2003). Personality, culture, and subjective wellbeing: Emotional and cognitive evaluations of life. Annual Review of Psychology, 54, 403–425. Diener, E., Sapyta, J. J., & Suh, E. (1998). Subjective well-being is essential to wellbeing. Psychological Inquiry, 9, 33–37. Diener, E., Scollon, C. K. N., Oishi, S., Dzokoto, V., & Suh, E. M. (2000). Positivity and the construction of life satisfaction judgments: Global happiness is not the sum of its parts. Journal of Happiness Studies, 1, 159–176. Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. (1999). Subjective well-being: Three decades of progress. Psychological Bulletin, 125, 276–302. Diener,E.,Suh,E.M.,Smith,H.,&Shao,L.(1995).Nationaldifferencesinreported subjective well-being: Why do they occur? Social Indicators Research, 34, 7–32.
64
Section I: Measuring Well-Being in an International Context
Eid, M. (2008). Measuring the immeasurable: Psychometric modeling of subjective well-being data. In M. Eid & R. J. Larsen (Eds.), The science of subjective well-being (pp. 141–167). New York: Guilford. Eid, M., & Diener, E. (2001). Norms for experiencing emotions in different cultures: Inter- and intranational differences. Journal of Personality and Social Psychology, 81, 869–885. Eid, M., & Diener, E. (2004). Global judgments of subjective well-being: Situational variability and long-term stability. Social Indicators Research, 65, 245–277. Elliot, A. J., Chirkov, V. I., Kim, Y., & Shelodn, K. M. (2001). A cross-cultural analysis of avoidance (relative to approach) personal goals. Psychological Science, 6, 505–510. Easterlin, R. A., & Sawangfa, O. (2009). Happiness and economic growth: Does the cross-section predict change over time? In E. Diener, D. Kahneman, & J. F. Helliwell (Eds.), International differences in well-being. New York: Oxford University Press. Gerstorf, D., Ram, N., Eastabrook, R., Schupp, J., Wagner, G. G., & Lindenberger, U. (2008). Life satisfaction shows terminal decline in old age: Longitudinal evidence from the German Socio-Economic Panel Study (SOEP). Developmental Psychology, 44, 1148–1159. Gerstorf, D., Ram, N., Ro¨cke, C., Lindenberger, U., & Smith, J. (2008). Decline in life satisfaction in old age: Longitudinal evidence for links to distance-to-death. Psychology and Aging, 23, 154–168. Graham, C., Chattopadhyay, S., & Picon, M. (2009). The Easterlin and other paradoxes: Why both sides of the debate may be correct. In E. Diener, D. Kahneman, & J. F. Helliwell (Eds.), International differences in well-being. New York: Oxford University Press. Graham, J., & Oishi, S. (2008). The concept of happiness across time: Whatever happened to ‘‘good fortune’’? A paper in preparation. Charlottesville, VA: University of Virginia. Griffin, J. (1986). Well-being: Its meaning, measurement, and moral importance. Oxford, England: Clarendon Press. Harter, J. K., & Arora, R. (2009). The impact of time spent working and job fit on wellbeing acround the world. In E. Diener, D. Kahneman, & J. F. Helliwell (Eds.), International differences in well-being. New York: Oxford University Press. Heine, S., Buchtel, E. E., & Norenzayan, A. (2008). What do cross-national comparisons of personality traits tells us? The case of conscientiousness. Psychological Science, 19, 309–313. Heine, S. J., Kitayama, S., Lehman, D. R., Takata, T., Ide, E., Leung, C., & Matsumoto, H. (2001). Divergent consequence of success and failure in Japan and North America: An investigation of self-improving motivations and malleable selves. Journal of Personality and Social Psychology, 81, 599–615. Heine, S., Lehman, D. R., Peng, K., & Greenholtz, J. (2002). What’s wrong with crosscultural comparisons of subjective Likert scales? The reference-group effect. Journal of Personality and Social Psychology, 82, 379–398. Helliwell, J. F. (2007). Well-being and social capital: Does suicide poses a puzzle? Social Indicators Research, 81, 455–496. Helliwell, J. F., Barrington-Leigh, C., Harris, A., & Huang, H. (2009). International evidence on the social context of well-being. In E. Diener, D. Kahneman, & J. Helliwell (Eds.), International differences in well-being. New York: Oxford University Press. Hoobler, E. A., & Oishi, S. (2008). The historical change in the definitions of happiness in American English. A paper in preparation. Charlottesville, VA: University of Virginia.
Culture and Well-Being: Conceptual and Methodological Issues
65
Inglehart, R. (2009). Faith and Freedom: Traditional and modern ways to happiness. In E. Diener, D. Kahneman, & J. Helliwell (Eds.), International differences in wellbeing. New York: Oxford University Press. Inglehart, R., Foa, R., Peterson, C., & Welzel, C. (2008). Development, freedom, and rising happiness: A global perspective (1981–2007). Perspectives on Psychological Science, 3, 264–285. Kahneman, D. (1999). Objective happiness. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology (pp. 3–25). New York: Russell Sage Foundation. Kahneman, D., & Krueger, A. B. (2006). Developments in the measurement of subjective well-being. Journal of Economic Perspectives, 20, 3–24. Kahneman, D., Krueger, A. B., Schkade, D. A., Schwarz, N., & Stone, A. A. (2004). A survey method for characterizing daily life experience: The day reconstruction method. Science, 306, 1776–1780. Kahneman, D., Krueger, A. B., Schkade, D. A., Schwarz, N., & Stone, A. A. (2006). Would you be happier if you were richer? A focusing illusion. Science, 312, 1908–1910. Kang, S-M., Shaver, P. R., Sue, S., Min, K-H., & Jing, H. (2003). Culture-specific patterns in the prediction of life satisfaction: Roles of emotion, relationship quality, and self-esteem. Personality and Social Psychology Bulletin, 29, 1596–1608. Kitayama, S., & Markus, H. R. (2000). The pursuit of happiness and the realization of sympathy: Cultural patterns of self, social relations, and well-being. In E. Diener & E. M. Suh (Eds.), Culture and subjective well-being. Cambridge, MA: MIT Press. Kitayama, S., Markus, H. R., & Kurokawa, M. (2000). Culture, emotion, and well-being: Good feelings in Japan and the United States. Cognition & Emotion, 14, 93–124. Kitayama, S., Mesquita, B., & Karasawa, M. (2006). Cultural affordances and emotional experience: Socially engaging and disengaging emotions in Japan and the United States. Journal of Personality and Social Psychology, 91, 890–903. Kwan, V. S. Y., Bond, M. H., & Singelis, T. M. (1997). Pancultural explanations for life satisfaction: Adding relationship harmony to self-esteem. Journal of Personality and Social Psychology, 73, 1038–1051. Laynard, R., Mayraz, G., & Nickell, S. (2009). Does relative income matter? Are the critics right? In E. Diener, D. Kahneman, & J. F. Helliwell (Eds.), International differences in well-being. New York: Oxford University Press. Lindfors, P., & Lundberg, U. (2002). Is low cortisol release an indicator of positive health? Stress and Health, 18, 153–160. Lu, L., & Gilmour, R. (2004). Culture and conceptions of happiness: Individual oriented and social oriented subjective well-being. Journal of Happiness Studies, 5, 269–291. Lucas, R. E., Diener, E., & Suh, E. (1996). Discriminant validity of well-being measure. Journal of Personality and Social Psychology, 71, 616–628. Lutz, C. (1987). Goals, events and understanding in Ifaluk emotion theory. In N. Quinn & D. Holland (Eds.), Cultural models in language and thought (pp. 290–312). Cambridge, England: Cambridge University Press. Lyubomirsky, S. (2000). In the pursuit of happiness: Comparing the United States and Russia. A paper presented at the annual meeting of the Society of Experimental Social Psychology. Atlanta, GA. Lyubomirsky, S., King, L., & Diener, E. (2005). The benefits of frequent positive affect: Does happiness lead to success? Psychological Bulletin, 131, 803–855. Lyubomirsky, S., & Lepper, H. (1999). A measure of subjective happiness: Preliminary reliability and construct validation. Social Indicators Research, 46, 137–155.
66
Section I: Measuring Well-Being in an International Context
McMahon, D. M. (2006). Happiness: A history. New York: Atlantic Monthly Press. Mesquita, B., & Leu, J. (2007). The cultural psychology of emotion. In S. Kitayama & D. Cohen (Eds.), Handbook of cultural psychology (pp. 734–759). New York: Guilford. Mroczek, D. K., & Spiro, A., III. (2005). Change in life satisfaction during adulthood: Findings from the Veterans Affairs Normative Aging Study. Journal of Personality and Social Psychology, 88, 189–202. Nussbaum, M. C. (1986). The fragility of goodness: Luck and ethics in Greek tragedy and philosophy. New York: Cambridge University. Nussbaum, M., & Sen, A (1993). The quality of life. New York: Oxford University. Oishi, S. (2002). Experiencing and remembering of well-being: A cross-cultural analysis. Personality and Social Psychology Bulletin, 28, 1398–1406. Oishi, S. (2006). The concept of life satisfaction across cultures: An IRT analysis. Journal of Research in Personality, 41, 411–423. Oishi, S. (2007). The application of structural equation modeling and item response theory to cross-cultural positive psychology research. In A.Ong & M. van Dulmen (Eds.), Oxford Handbook of Methods in Positive Psychology (pp. 126–138). New York: Oxford University Press. Oishi, S., & Diener, E. (2003). Culture and well-being: The cycle of action, evaluation and decision. Personality and Social Psychology Bulletin, 29, 939–949. Oishi, S., & Diener, E. (2008). Culture and well-being: Informant reports converge with self-reports. A paper in preparation. Charlottesville, VA: University of Virginia. Oishi, S., Diener, E., Choi, D. W., Kim-Prieto, C., & Choi, I. (2007). The Dynamics of daily events and well-being across cultures: When less is more. Journal of Personality and Social Psychology, 93, 685–698. Oishi, S., Diener, E., & Lucas, R. E. (2007). The optimal level of well-being: Can we be too happy? Perspectives on Psychological Science, 2, 346–360. Oishi, S., Diener, E., Lucas, R. E., & Suh, E. M. (1999). Cross-cultural variations in predictors of life satisfaction: Perspectives from needs and values. Personality and Social Psychology Bulletin, 25, 980–990. Oishi, S., Diener, E., Scollon, C. N., & Biswas-Diener, R. (2004). Cross-situational consistency of affective experiences across cultures. Journal of Personality and Social Psychology, 86, 460–472. Oishi, S., & Roth, D. P. (2009). The role of self-reports in culture and personality research: It is too early to give up on self-reports. Journal of Research in Personality, 43, 107–109. Oishi,S.,&Schimmack,U.(2008).Residentialmobility,well-being,andmortality:Residential moves hurt introverts. A paper under review. Charlottesville, VA: University of Virginia. Oishi, S., & Sullivan, H. W. (2006). The Predictive value of daily vs. retrospective wellbeing judgments in relationship stability. Journal of Experimental Social Psychology, 42, 460–470. Oishi, S., Whitchurch, E., Kurtz, J., & Miao, F. F. (2009). ‘‘Would I be happier if I moved?’’ Retirement status and cultural variations in the anticipated and actual levels of happiness. Journal of Positive Psychology, 4, 437–446. Pavot, W. (2008). The assessment of subjective well-being: Successes and shortfalls. In M. Eid & R. J. Larsen (Eds.), The science of subjective well-being (pp. 124–140). New York: Guilford. Riis, J., Loewenstein, G., Baron, J., Jepson, C., Fagerlin, A., & Ubel, P. A. (2005). Ignorance of hedonic adaptation to hemodialysis: A study using ecological momentary assessment. Journal of Experimental Psychology: General, 134, 3–9.
Culture and Well-Being: Conceptual and Methodological Issues
67
Robinson, M. D., & Clore, G. L. (2002). Belief and feeling: Evidence for an accessibility model of emotional self-report. Psychological Bulletin, 128, 934–960. Russell, J. A. (1991). Culture and the categorization of emotion. Psychological Bulletin, 110, 426–450. Ryan, R. M., & Deci, E. L. (2001). On happiness and human potentials: A review of research on hedonic and eudaimonic well-being. Annual Review of Psychology, 52, 141–166. Ryan, R. M., & Frederick, C. M. (1997). On energy, personality and health: Subjective vitality as a dynamic reflection of well-being. Journal of Personality, 65, 529–565. Ryff, C. D. (1989). Happiness is everything, or is it? Exploration on the meaning of psychological well-being. Journal of Personality and Social Psychology, 57, 1069–1081. Ryff, C. D., & Singer, B. (1998). The contours of positive human health. Psychological Inquiry, 9, 1–28. Sandvik, E., Diener, E., & Seidlitz, L. (1993). Subjective well-being: The convergence and stability of self-report and non–self-report measures. Journal of Personality, 61, 317–342. Saucier, G., & Goldberg, L. R. (2001). Lexical studies of indigenous personality factors: Premises, products, and prospects. Journal of Personality, 69, 847–880. Schimmack, U. (2008). The structure of subjective well-being. In M. Eid & R. J. Larsen (Eds.), The science of subjective well-being (pp. 97–123). New York: Guilford. Schimmack, U., & Oishi, S. (2005). The influence of chronically accessible and temporarily accessible information on life satisfaction judgments. Journal of Personality and Social Psychology, 89, 395–406. Schimmack, U., Radhakrishan, P., Oishi, S., Dzokoto, V., Ahadi, S. (2002). Culture, personality, and subjective well-being. Integrating process model of life satisfaction. Journal of Personality and Social Psychology, 82, 582–593. Schwarz, N., & Clore, G. L. (1988). How do I feel about it? Informative functions of affective states. In K. Fiedler & J. Forgas (Eds.), Affect, cognition, and social behavior (pp. 44–62). Toronto,: Hogrefe International. Schwarz, N., & Strack, F. (1999). Reports of subjective well-being: Judgmental processes and their methodological implications. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology (pp. 61–84). New York: Russell Sage Foundation. Scollon, C. N., Diener, E., Oishi, S., & Biswas-Diener, R. (2004). Emotions across cultures and methods. Journal of Cross-Cultural Psychology, 35, 304–326. Scollon, C. N., Kim-Prieto, C., & Diener, E. (2003). Experience sampling: Promises and pitfalls, strengths and weaknesses. Journal of Happiness Studies, 4, 5–34. Sheldon, K. M., Elliot, A. J., Ryan, R. M., Chirkov, V., Kim, Y., Wu, C., Demir, M., & Sun, Z. (2004). Self-concordance and subjective well-being in four cultures. Journal of Cross-Cultural Psychology, 35, 209–223. Shweder, R. (1991). Thinking through cultures. Cambridge, MA: Harvard University Press. Snyder, C. R., & Lopez, S. J. (2002). Handbook of positive psychology. New York: Oxford University Press. Steger, M. F., Frazier, P., Oishi, S., & Kaler, M. (2006). The meaning in life questionnaire: Assessing the presence of and search for meaning in life. Journal of Counseling Psychology, 53, 80–93. Steger, M. F., Kawabata, Y., Shimai, S., & Otake, K. (2008). The meaningful life in Japan and the United States: Levels and correlates of meaning in life. Journal of Research in Personality, 42, 660–678.
68
Section I: Measuring Well-Being in an International Context
Stone, A. A., & Shiffman, S. (1994). Ecological Momentary Assessment (EMA) in behavioral medicine. Annals of Behavioral Medicine, 16, 199–202. Suh, E., Diener, E., Oishi, S., & Triandis, H. C. (1998). The shifting basis of life satisfaction judgments across cultures: Emotions versus norms. Journal of Personality and Social Psychology, 74, 482–493. Suh, E. M., & Koo, J. (2008). Comparing subjective well-being across cultures and nations: The ‘‘what’’ and ‘‘why’’ questions. In M. Eid & R. J. Larsen (Eds.), The science of subjective well-being (pp. 414–427). New York: Guilford. Tanaka-Matsumi, J., & Marsella, A. J. (1976). Cross-cultural variations in the phenomenological experience of depression: I. Word association studies. Journal of Cross-Cultural Psychology, 7, 379–396. Thomson, J. A. K. (1953). The ethics of Aristotle: The Nicomachean ethics. New York: Penguin Books. Tov, W., & Diener, E. (2007). Culture and subjective well-being. In S. Kitayama & D. Cohen (Eds.), Handbook of cultural psychology (pp. 691–713). New York: Guilford. Tsai, J. L., Louie, J. Y., Chen, E. E., & Uchida, Y. (2007). Learning what feelings to desire: Socialization of ideal affect through children’s storybooks. Personality and Social Psychology Bulletin, 33, 17–30. Tsai, J. L., Miao, F. F., & Seppala, E. (2007). Good feelings in Christianity and Buddhism: Religious differences in ideal affect. Personality and Social Psychology Bulletin, 33, 409–421. Turner, F. J. (1921). The frontier in American history. New York: Henry Holt and Company. Uchida, Y., & Kitayama, S. (2007). Cultural constructions of happiness. A paper under review. The University of Kyoto, Kyoto, Japan. Uchida, Y., Kitayama, S., Mesquita, B., Reyes, J. A. S., Morling, B. (2008). Is perceived emotional support beneficial? Well-being and health in independent and interdependent cultures. Personality and Social Psychology Bulletin, 34, 741–754. Urry, H. L., Nitschke, J. B., Dolski, I., Jackson, D. C., Dalton, K. M., Muller, C. J., Rosenkranz, M. A., Ryff, C. D., Singer, B. H., & Davidson, R. J. (2004). Making a life worth living: Neural correlates of well-being. Psychological Science, 15, 367–372. Veenhoven, R. (1988). The utility of happiness. Social Indicators Research, 20, 333–354. Veenhoven, R. (2008). Measures of Gross National Happiness. Presentation at OECD World Forum on Statistics, Knowledge and Policy. Roundtable, ‘‘Measuring Happiness and Making Policy,’’ June 27–30, 2007, Istanbul, Turkey. Veenhoven, R. (2009). How universal is happiness? In E. Diener, D. Kahneman, & J. F. Helliwell (Eds.), International differences in well-being. New York: Oxford University Press. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of a brief measure of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. Wirtz, D. R. (2004). Focusing on the good versus focusing on the bad: An analysis of East-West differences in subjective well-being. Unpublished dissertation. The University of Illinois, Urbana-Champaign. Wirtz, D., Kruger, J., Scollon, C. N., & Diener, E. (2003). What to do on spring break? The role of predicted, online, and remembered experience in future choice. Psychological Science, 14, 520–524.
Culture and Well-Being: Conceptual and Methodological Issues
69
Wrzesniewski, A., McCauley, C., Rozin, P., & Schwartz, B. (1997). Jobs, careers, and callings: People’s relations to their work. Journal of Research in Personality, 31, 21–33. Yamaguchi, S., Greenwald, A. G., Banaji, M. R., Murakami, F., Chen, D., Shiomura, K., Kobayashi, C., Cai, H., & Krendl, A. (2007). Apparent universality of positive implicit self-esteem. Psychological Science, 18, 498–500.
Chapter 4 Life Satisfaction Arie Kapteyn, James P. Smith, and Arthur van Soest
Arthur van Soest, Netspar, Tilburg University, RAND & IZA1
This research was supported by grants from the National Institute on Aging to RAND. We are grateful to Ed Diener, John Helliwell, an anonymous reviewer, workshop participants in Princeton and seminar participants in Tilburg for useful comments.
Economists have discovered (or rediscovered) happiness, or at least research on subjective well-being and its economic correlates (see, e.g., Van Praag, Frijters, & Ferrer-i-Carbonell, 2003; Layard, 2005; or Clark, Frijters, & Shields, 2008). The rapidly growing research has touched on several important themes. These have included the so-called Easterlin paradox, whereby average happiness remains relatively constant over time in spite of large increases in income per capita (Easterlin, 1974, 1995; see also the chapter by Graham, Chattopadhyay, and Picon in this volume). In contrast, within-country cross-sectional and panel data almost always show that rising incomes ‘‘buy’’ additional satisfaction, although the magnitude of the within-country cross-sectional effect of income on satisfaction is under dispute (Blanchflower & Oswald, 2004; Di Tella et al., 2007; and Stevenson & Wolfers, 2008). Resolving this paradox, which is often interpreted as a fundamental challenge to the conventional economic theory of utility maximization, has generated a substantial amount of subsequent research attempting to reconcile the finding of a zero correlation between income and life satisfaction in aggregate time series evidence with the positive correlation in cross-section micro-estimates within a given country. 70
Life Satisfaction
71
This reconciliation has included adding relative incomes (of others or of oneself in the past) in the utility function (Van de Stadt et al., 1985; Clark et al., 2008), or a sometimes rapid process of adaptation to new circumstances (Di Tella et al., 2003) often labeled the ‘‘hedonic treadmill’’ (Di Tella et al., 2007). A recent contrary view is provided by Deaton (2008), who documents that if one considers a much wider range of countries arrayed by their level of economic development, the normally positive association of income with subjective life satisfaction reappears. His work also leads to the conclusion that the effect of income on life satisfaction according to cross-country regressions is, if anything, higher in the high income countries than in the low income countries. Stevenson and Wolfers (2008) revisit much of the earlier evidence and look at new data to reach similar conclusions. A considerable amount of research has focused on cross-country differences in subjective well-being, in particular comparing Europe and the United States where the United States appears to rank lower in satisfaction than many European countries with lower per capita incomes (Alesina et al., 2004; Di Tella et al., 2003; and Blanchflower & Oswald, 2004). For instance, Europeans apparently exhibit a stronger distaste for inequality than do Americans, which may be partly explained by a perception of greater mobility in the United States (Alesina et al., 2004). Blanchflower & Oswald (2004) study trends in well-being over time in the United Kingdom and the United States and find that reported levels of well-being have been dropping over time in the United States while they have been flat in the United Kingdom, despite the fact that in both countries average incomes have grown substantially over the last couple of decades. A fundamental problem in international comparisons and cross-sectional and time-series analyses of subjective well-being is that one has to assume that somehow response scales are the same across countries, across time, and across groups of respondents within a country. This critical and largely untested assumption becomes even more tenuous if question phrasings change or differ across surveys, as is often the case (see Stevenson & Wolfers, 2008). Here we address these problems head-on. In view of the specific interest of economists in the relationship between life satisfaction and income, we focus on the role of income. The population distribution of satisfaction in a country will depend on levels and distribution of incomes. Residents of different countries can, however, differ in the way they translate any given level of income into a subjective level of satisfaction. Moreover, residents of countries may differ in the subjective thresholds that they use in demarcating satisfaction into discrete categories, such as ‘‘very satisfied’’ or ‘‘not satisfied.’’ Income distributions, the translation from income to income satisfaction, and the
72
Section 1: Measuring Well-Being in an International Context
demarcation thresholds can all affect differences observed within and between countries in their distribution of stated levels of satisfaction. These distinct factors are often confused in the existing literature on life satisfaction and happiness. In our research, we have created unique data sources in two countries—the United States and the Netherlands—and developed a statistical methodology that allows us to separate out these distinct factors. This paper is divided into seven parts. The first describes the data sources that we developed and will rely on in this analysis. The second section summarizes responses of Dutch and American respondents to questions about their own life satisfaction in several key domains of their lives, while the third section describes the types of vignettes we developed and the responses to those vignettes by our Dutch and American respondents. In the next section, we summarize the vignette methodology that serves as the basis of our analysis and then sketch our statistical model, which corrects for response-scale differences across countries. The fifth section presents our main empirical results and their implications for interpreting observed differences in life satisfaction in the two countries. In the sixth section, simulations based on our estimated models are used to ascertain what Dutch distributions of life satisfaction would be if the Dutch had American parameters and thresholds rather than their own. The final section highlights our main conclusions.
Data Sources
Our analysis in this paper is based on information obtained from two Internet surveys, which we conducted in the Netherlands and the United States. For the Netherlands, we used CentERpanel, administered by CentERdata, affiliated with Tilburg University. CentERpanel includes about 2,250 households who have agreed to respond to questions every weekend over the Internet. The Dutch sample is not restricted to households with their own Internet access. Respondents are recruited by telephone. If they agree to participate and do not have Internet access, they are provided with Internet access (and if necessary, a set-top box). Thus, CentERpanel is representative of the adult Dutch population, except the institutionalized. Sampling weights provided by CentERdata and based upon comparing with a much larger survey by Statistics Netherlands are used to correct for unit non-response and attrition. The sample used for estimation has 2,244 respondents who participated in an interview with questions on life satisfaction (self-assessments as well as vignettes) in 2006. From multiple waves collected in the past, CentERpanel has a rich
Life Satisfaction
73
set of variables on the demographic, health, and economic characteristics of respondents. In our analysis, we use the most recent measurement of these background variables, reported a few months before our life satisfaction survey. The Internet infrastructure makes CentERpanel an extremely valuable tool for conducting experiments, with possibilities for randomization of content. Production lags are very short, with about one month between module design and data delivery. Our Internet survey for the United States is the RAND American Life Panel (ALP). This panel was initially recruited from respondents age 40plus in the Monthly Survey (MS) of Michigan’s Survey Research Center but has been subsequently supplemented with younger respondents.2 Background information collected for these respondents was similar to that available for Dutch respondents. The American sample used for estimation consists of 1,113 respondents interviewed during 2006–2007. The American data are weighted to match Current Population Survey demographic distributions in age, education, and gender.
Global Life Satisfaction by Domain
Respondents are asked to rate themselves on a five-point scale. They do so in the following specific life domains: income, family and social relations, job, and health. They are also asked a global question about their own life satisfaction. The scale that is used is the same for all domains: (very satisfied, satisfied, not satisfied or dissatisfied, not satisfied, and very dissatisfied). The exact self-assessment questions of life satisfaction are contained in Table 4.1. Table 4.1 summarizes responses obtained from the Dutch and American samples for the four domains of income, social contacts and family life, job or other daily activities, and health. The last panel in Table 4.1 presents responses to the question regarding global life satisfaction. For all domains and for global satisfaction, the distributions in the United States and the Netherlands are significantly different (see the tests reported at the bottom of each panel). Before turning to between-country differences, it is useful to first highlight some differences across the domains. Both the Dutch and Americans appear to be less satisfied with their incomes than with the other domains. The health domain is next, at least in terms of most negative responses; followed by job and daily activities, with respondents in both countries most satisfied with their lives in the family and social contacts domain. Differences among the Dutch in the three domains besides income are relatively small; sharper distinctions are present in these three domains
TABLE 4.1
Self Reports on Satisfaction with Domains of Life
Self report: How satisfied are you with the total income of your household? Very satisfied Satisfied Not satisfied or dissatisfied Not satisfied Very dissatisfied
NL 9.9 53.6 23.6 10.3 2.7
Country US 6.5 39.4 21.5 27.4 5.2
Test for independence: F(3.64, 12207.95) = 20.3117; p-value = 0.0000
Self report: How satisfied are you with your job or other daily activities? Very satisfied Satisfied Not satisfied or dissatisfied Not satisfied Very dissatisfied
NL 19.4 61.7 14.7 3.4 0.8
US 16.3 52.2 17.5 12.1 2.0
Test for independence: F(3.36, 11231.88) = 11.4447; p-value = 0.0000
Self report: How satisfied are you with your social contacts and family life? Very satisfied Satisfied Not satisfied or dissatisfied Not satisfied Very dissatisfied
NL 23.0 62.8 11.7 1.9 0.6
US 27.1 48.2 15.7 8.5 0.5
Test for independence: F(3.58, 11978.15) = 13.9798; p-value = 0.0000
Self report: How satisfied are you with your health? Very satisfied Satisfied Not satisfied or dissatisfied Not satisfied Very dissatisfied
NL 15.4 61.6 14.5 7.0 1.4
US 16.1 46.7 17.3 16.5 3.5
Test for independence: F(3.82, 12791.55) = 13.3638; p-value = 0.0000
Self report: How satisfied are you with your life in general? Very satisfied Satisfied Not satisfied or dissatisfied Not satisfied Very dissatisfied
NL 19.3 68.2 10.9 1.3 0.3
US 20.1 58.0 15.4 5.3 1.1
Test for independence: F(3.17, 10610.55) = 9.2306; p-value = 0.0000. Note: all frequencies are weighted with sampling weights. Tests are Pearson chi-squared tests for independence converted into F-statistics, accounting for the weighting (see Rao & Scott, 1984).
Life Satisfaction
75
among Americans. Finally, a much smaller proportion of both Dutch and American respondents appear to be dissatisfied with their lives when answering a global life satisfaction question than their answers in each of the four specific life domains would indicate. This appears to be due to relatively modest correlations in dissatisfaction across domains, so that dissatisfaction in one domain may be compensated for by satisfaction in a different domain. In fact, across the two samples, less than one percent of all respondents are ‘‘not satisfied’’ or ‘‘very dissatisfied’’ in all four domains, and among those, virtually all (94%) also report being ‘‘not satisfied’’ or ‘‘very dissatisfied’’ with their life in general. Turning to between-country differences in life satisfaction, consider first how satisfied respondents in the two countries are with their total household income. As we have analyzed in more detail elsewhere,3 Americans are much less satisfied with their incomes than the Dutch are, in spite of the fact that, on average, their incomes are considerably higher. Sixty-four percent of Dutch respondents say that they are either satisfied or very satisfied with their total household income. The comparable fraction for Americans is 46%—eighteen percentage points lower than the Dutch. Similarly, a much larger fraction of Americans respond that they are either not satisfied or very dissatisfied—a third of Americans compared to thirteen percent of the Dutch. The Dutch are also more satisfied with their jobs than Americans are, but these differences are smaller than those in the income domain. Fourteen percent of Americans are either not satisfied or very dissatisfied with their jobs, compared to four percent of the Dutch—a differential of ten percentage points, which is half as large as the differential in the income satisfaction domain. At the top of the scale, more than four in every five Dutch respondents are at least satisfied with their jobs, as are more than two-thirds of Americans. In both countries, respondents are much more satisfied with their job and other daily activities than they are with their incomes. There are actually more Americans very satisfied as well as very dissatisfied with the social aspects of their lives compared to the Dutch. Relatively few respondents in either country register displeasure (not satisfied or very dissatisfied) with this domain, although once again Americans are more likely to go to the bottom of the scale compared to the Dutch (9% compared to 2%). This avoidance of extremes is a common feature of Dutch responses to subjective scale questions and is similar to what we have found in prior work (see, for example, Kapteyn, Smith, & van Soest, 2007). This tendency may well have its origins in the Dutch culture. According to Wikipedia, ‘‘The Dutch typically see their countrymen as sober, practical, and down-to-earth people. Any form of
76
Section 1: Measuring Well-Being in an International Context
ostentation is likely to be criticized, and straightforwardness is generally appreciated.’’ The final specific domain on which we asked about life satisfaction was health. Based on objective health measures, the Dutch are a healthier population than the Americans (Kapteyn, Smith, & van Soest, 2007). In this case, that objective difference is reflected in their subjective answers to the question about their satisfaction with their health. Nineteen percent of Americans are either not satisfied or very dissatisfied with their health, compared to eight percent of the Dutch. The final panel in Table 4.1 displays the distribution of answers to questions evaluating global life satisfaction (satisfaction with life— SWL). Eighty-eight percent of the Dutch are either satisfied or very satisfied, compared to seventy-eight percent of the Americans. Similarly, while most respondents in both countries appear to be relatively satisfied with their lives, 6.4% of the Americans say that they are at a minimum not satisfied, compared to only 1.4% of the Dutch. Using models similar to those developed by Ferrer-i-Carbonell & van Praag (2002), Van Praag et al. (2003), and Van Praag & Ferrer-i-Carbonell (2008), Table 4.2 examines the relationship between responses to the ‘‘global life satisfaction’’ question to the level of the respondents’ satisfaction within the four specific life domains. It does so by listing coefficient estimates (and the associated ‘‘z’’ values) from an ordered probit model of global life satisfaction. The initial set of regressors in the first two columns are responses to life satisfaction in the four specific life domains, each indexed on a scale of one to five. Main effects are estimated Dutch coefficients, while the U.S. interactions test for differences between Americans and Dutch.4 As expected, these results show that satisfaction with life is positively associated with satisfaction within each of the four domains. As indicated by the estimated coefficients within each domain, income satisfaction received by far the lowest weight in global satisfaction, with the health domain in second-to-last place. The highest weight is in the family and social relations domain.5 While there is not much evidence of statistically significant differences between the two countries in the translation from satisfaction within a domain into global life satisfaction, there appears to be less weight in the United States assigned to the health domain. Remember that the coding goes from ‘‘very satisfied’’ to ‘‘very dissatisfied,’’ so the negative sign on the U.S. dummy means that U.S. respondents are happier, keeping satisfaction in each domain constant. That result, however, is not statistically significant. The second model in Table 4.2 adds a number of standard demographics to this model, including age, marital status, education, gender, and income,
Life Satisfaction
77
and once again allows all estimated effects to differ between the Dutch and Americans. All in all, the evidence for the need for demographics or interactions of these with the U.S. dummy is very weak. A test of the null that the effects of the demographics are equal to zero does not lead to rejection. Thus, it seems a model with just the domain-specific satisfaction variables is sufficient. There is a slightly different way of interpreting this outcome. If we state as a null hypothesis that global life satisfaction is a function of just the four domains we consider here, then the test would not reject that null. Of course the power of that test will depend on how much the possibly omitted satisfaction dimensions are correlated with the demographics included in Table 4.2.
TABLE 4.2 Ordered Probits for Global Life Satisfaction Against Satisfaction with Specific Domains
Income Domain Relations Domain Job Domain Health Domain US Income Domain US Relations Domain US Job Domain US Health Domain Dummy US Domain Married Age 40-50 ln family size Age 51-64 Age 65+ Ed med Ed high Working ln eq income Dummy US US female US married US ln family size US age 40-50 US age 51-64 US age 65+ US ed med US ed high US working US ln eq income
Coef.
z
Coef.
z
.225 .721 .625 .486 .052 .087 .020 .131 .145
6.85 12.22 12.27 11.02 0.85 1.11 0.26 1.98 0.82
.220 .708 .626 .497 .031 .101 .016 .143 .037 .128 .023 .024 .093 .023 .128 .117 .116 .009 .255 .084 .088 .226 .108 .198 .048 .038 .059 .077 .001
6.53 11.85 12.22 11.10 0.49 1.26 0.20 2.12 0.68 1.38 0.31 0.35 1.29 0.22 1.95 1.77 1.73 0.61 0.25 0.80 0.59 1.54 0.74 1.32 0.22 0.25 0.40 0.60 0.02
78
Section 1: Measuring Well-Being in an International Context
Description of Vignettes
In addition to their ratings of their own life satisfaction, respondents were given a set of vignettes covering the four life domains—income, family relations, job, and health. These domains were chosen because the current literature has documented them as key determinants of overall life satisfaction (see Easterlin, 2006). All vignettes were given with either a female or male name, which was randomized across respondents. Within each domain, vignettes were presented in random order to eliminate any possibility of order effects, whereby the initial vignette presented could affect the ranking of subsequent vignettes.6 Comparing rank ordering of vignette evaluations across respondents shows that different respondents tend to order vignettes in the same way. The scale that is used is the same for all domains: very satisfied, satisfied, not satisfied or dissatisfied, not satisfied, and very dissatisfied. The vignette questions in the income domain specify an income for the vignette person that is selected randomly, with the values being equal to either the median income in the Netherlands or the United States or a value that is half, twice, or four times the median income in each country. The six family-relation vignettes vary conditions in the vignette person’s family or friends’ life, including whether the vignette person is married or single, has children, has many or few friends, and the nature of the relationship with these important others. The work vignettes focus on whether the vignette person is working or retired, the amount of hours worked, the security of the job, and how stressful the job is. Finally, the four health domain vignettes vary conditions around the vignette person’s ability to engage in (light, vigorous) exercise, and possible problems with sleep, anxiety, and depression. In addition to these domain-specific life satisfaction questions, respondents are also given a subset of ten possible vignettes on global life satisfaction. These global life satisfaction vignettes succinctly describe the vignette person in a single vignette across the four sub-domains mentioned above— family relations, work, income, and health, combining the descriptions given in the domain-specific vignettes. This global vignette approach has the advantage of moving directly to an overall measure of life satisfaction, and for that reason, we will use them in the analysis in this paper. The domainspecific vignettes are not analyzed in the current paper; the satisfaction-withincome vignettes are analyzed in Kapteyn et al. (2008). The specific scenarios described in the global satisfaction vignettes are listed in the appendix. Table 4.3 presents a summary of the type of variation present in the ten global vignettes. The six ways that the global vignettes can vary are gender, age, family, income, work, and health. Gender of the vignette
Life Satisfaction
TABLE 4.3
79
Variation in Global Life Satisfaction Vignettes
Age
Family
Income
Work
Health
1
42
good
median +
stressful
2
50
moderate
half median +
3
65
bad
modest
ok with long hours retired
some pain good
4
25
no friends
half median +
5
25
good
6 7
57 75
bad good
8 9 10
62 70 50
good bad bad
half-median to twice median median + half median to twice median half median + half median + twice median +
no control or security no control but secure retired retired retired retired good
heart problems good good pain arthritis good moderate bad
person is randomly assigned to respondents through the feminine or masculine name of the vignette person. An exact age is always mentioned in each vignette, and these ages are listed in the second column of Table 4.3. With one exception (Vignette 3 where income is always ‘‘modest’’), income is also randomly assigned in the vignette with up to four values assigned (ranging from half the median to the median, twice the median, and four times the median). In most scenarios, only three of the four possible income values are used. The overall situation described in the family and social relations dimension relates to spouse, children, and/or the presence of close friends. In four of the vignettes, the social situation can reasonably be described as good; in one vignette as moderate; and in the other five it is problematic in some important way. The aspects of the work environment that are mentioned in the vignettes are stress, hours, control, security, and retirement status. In five vignettes, the vignette person has already retired. Finally, four of the vignettes describe a person in good health, while the remainder point to some type of health problem, ranging from moderate to serious. Responses to vignette questions on global life satisfaction
Table 4.4 lists the distribution of responses obtained in both countries for each global life satisfaction vignette. We divide these vignettes into three groups based on the age of the vignette person—those who have retired, the young, and middle-aged vignette persons. To eliminate the impact of
80
Section 1: Measuring Well-Being in an International Context
TABLE 4.4 Global Vignettes for the Retired Vignette Number
Family Health Age
3
6
7
8
9
Bad Heart bad 65
Bad Pain 57
Good Arthritis 75
Good Good 62
Bad Moderate 70
NL Very satisfied Satisfied Not satisfied or dissatisfied Not satisfied Very dissatisfied
US
NL
US
NL
US
NL
US
NL
US
0.1 1.0 1.6 0.0 13.3 18.1 70.7 77.4 1.4 5.4 25.0 12.7 19.6 10.5 67.4 69.2 26.1 18.8 7.5 4.4 48.9 27.1 48.6 40.6 16.3 9.0 3.2 3.0 39.5 19.6 24.7 51.7 28.6 48.1 1.3 7.6 1.6 0.8
3.0 0.0
3.8 0.0
Global Vignettes for the Young
0.0 0.0
0.8 49.0 63.0 0.0 2.7 7.6
Global Vignettes for the Middle Aged Vignette Number
4
5
1
2
10
Family Work
no friends no control no security
good no control good security
Good stressful
moderate ok-long hours
bad good
Health Age
good 25 NL US
good 25 NL US
some pain 42 NL US
good 50 NL US
bad 50
Very satisfied Satisfied Not satisfied or dissatisfied Not satisfied Very dissatisfied
NL
US
1.9 3.2 23.9 33.1 16.9 15.4 19.8 23.0 1.5 1.1 14.9 14.9 66.8 58.8 64.3 65.0 56.5 50.0 11.7 8.5 39.8 37.2 8.0 7.4 18.8 22.5 19.8 21.6 36.6 36.0 41.6 41.5 1.9 3.2
1.3 0.0
0.0 0.3
2.2 0.5
6.0 0.0
2.8 1.1
5.4 45.3 43.4 0.0 5.0 11.1
All vignettes evaluated at highest income level in the vignette.
income variation within the vignette, our comparisons in Table 4.4 apply to the highest income value mentioned in the vignette. The numbers in the rows at the top of the columns match the numbers of the vignette in Appendix A. The rows below describe the main attributes of the vignette person in terms of their age, family relations, health, and job situation. Let’s first examine the panel for the five global vignettes for retired persons. Besides a limited age variation within this group, the only meaningful variation in these vignette descriptions concerns the quality of family
Life Satisfaction
81
life and the health of the respondent. It is useful to start with the best case of the possible scenarios—Vignette 8, where both the social life and the health of the vignette respondent are good. In this case, vignette evaluations of both Dutch and American respondents mirror this best-case description, as three-quarters of respondents say that one should be very satisfied, and less than one percent respond as either ‘‘not’’ or ‘‘very’’ dissatisfied. When the life situation appears to be very good, as it certainly is in this case, a larger fraction of Americans answer ‘‘very satisfied’’—77% compared to 71% for the Dutch. Health status has a major impact in these evaluations of retired vignette persons. Even with Vignette 7 where the vignette person suffers only from arthritis but has many friends, there is a very sharp reduction in the percent of respondents who would be satisfied. Less than one in five of Dutch and American respondents apply the label ‘‘very satisfied,’’ compared to at least 70% for the healthy Vignette 8. As was the case with Vignette 7, somewhat more Americans say ‘‘very satisfied’’ with Vignette 8 compared to the Dutch. When the health situation of the vignette person gets even worse, as in Vignettes 3, 6, and 9, large fractions of Dutch and American respondents choose ‘‘not satisfied’’ or ‘‘very dissatisfied.’’ But this negative reaction to poor health appears to be more dramatic in the American sample. Using Vignette 3 to illustrate the point, we see that 59% of Americans respond as ‘‘not satisfied’’ or ‘‘very dissatisfied,’’ compared to only 26% of the Dutch. This result seems in contrast to Table 4.2, where we found that life satisfaction depends to a lesser extent on health satisfaction in the United States than in the Netherlands. Perhaps the result is misleading since we do not have an orthogonal design—variation in health is correlated with other variations in the vignette characteristics—and we need regressions to control for other differences in the vignette characteristics (see Table 4.5 below). Vignettes among the retired suggest the importance of health for life satisfaction among the retired population. Turn next to the two vignettes pertaining to the young. Age and even health are basically the same in these two vignettes, so that they differ only in how they describe the family and work domains. Compared to Vignette 5 where the social situation is good and the job situation is secure, the person in Vignette 4 has no friends and does not feel secure about his/her job. Those two problems take their toll of the evaluation of global life satisfaction. There is a tenfold reduction in the percent of respondents who are very satisfied, and the percent of respondents who are not satisfied or very dissatisfied increases from about one percent to over forty percent. Even among the young, where income and income growth is quite central to their lives, a negative situation in terms of either job security or friendships
TABLE 4.5 Effect of vignette descriptions on evaluation
Age Married Good relations Retired Modest income Half median income Twice median income Four times median income Health good job secure or under control at least 50 hrs Dummy US US Age US Married US Good relations US Retired US Modest income US Half median income US Twice median income US Four times median income US Health good US job secure or under control US at least 50 hrs
(1) Regression
(2) Ordered Probit
(3) Probit (very) dissatisfied
0.003 (1.29) 0.171 (3.97)** 1.211 (39.11)** 0.153 (2.11)* 0.947 (20.23)** 0.218 (8.22)** 0.145 (6.17)** 0.279 (9.35)** 0.477 (9.43)** 0.134 (3.66)** 0.693 (12.91)** 0.160 (1.00) 0.007 (2.41)* 0.227 (3.53)** 0.218 (4.65)** 0.269 (2.45)* 0.652 (8.75)** 0.100 (2.37)* 0.054 (1.52) 0.058 (1.27) 0.172 (2.25)* 0.200 (3.62)** 0.078
0.001 (0.39) 0.263 (4.12)** 1.636 (32.91)** 0.091 (0.83) 1.318 (18.77)** 0.334 (8.21)** 0.234 (6.43)** 0.448 (9.59)** 0.955 (11.47)** 0.290 (5.02)** 1.215 (14.38)** 0.216 (0.89) 0.010 (2.17)* 0.354 (3.63)** 0.321 (4.46)** 0.400 (2.38)* 0.925 (8.26)** 0.153 (2.36)* 0.093 (1.66) 0.109 (1.49) 0.200 (1.57) 0.284 (3.25)** 0.062
0.026 (5.21)** 0.259 (2.17)* 1.964 (24.24)** 0.765 (4.40)** 1.566 (12.27)** 0.406 (6.28)** 0.153 (2.46)* 0.294 (4.14)** 0.244 (1.43) 0.094 (0.88) 0.958 (6.09)** 0.164 (0.45) 0.009 (1.22) 0.510 (2.90)** 0.111 (0.95) 0.342 (1.31) 0.885 (4.52)** 0.020 (0.20) 0.172 (1.89) 0.158 (1.48) 0.418 (1.69) 0.399 (2.58)** 0.235 (continued )
82
Life Satisfaction
TABLE 4.5
83
(Continued)
Constant
(0.97) 3.265 (30.70)**
Observations R-squared
12051 0.55
(0.49)
12051
(1.02) 1.168 (4.87)** 12051
Notes: Robust t statistics in parentheses; * significant at 5% level; ** significant at 1% level Model (3): dependent variable: 1 if very dissatisfied or dissatisfied; 0 otherwise.
reduces overall life satisfaction a great deal; at least that is what our respondents believe. The final subset of global life satisfaction vignettes describes the middleaged. For them also, the age spread in the vignettes is quite limited, so that the principal variation across alternative vignettes relates to job, social relations, and health. In some respects, Vignette 10 may be the most interesting. In this case, the work situation is good and all vignettes are assigned an income either twice the median or four times the median. The answers summarized in Table 4.4 are for the four times median income in each country, so that the economic situation of the household is very good in the economic/work domains. However, both the family life and health are not good, and as a consequence many respondents evaluate the life situation described quite negatively. More than half of both Dutch and American respondents are not satisfied or are very dissatisfied with this vignette person’s life. Evaluating the independent impact of social relations, health, and job is difficult in the global vignettes as all three dimensions vary simultaneously across these vignettes. Table 4.5 presents an alternative approach for doing so, by showing the results of a regression of the rating of the vignettes on their main characteristics. The characterization chosen here is a little different from the one shown in Tables 3 and 4. Since there are only ten vignettes, we have to choose a parsimonious set of indicators to characterize the vignettes, in order to be able to identify their effect. The differences in comparison with Table 4.3 are as follows. We characterize family and social relations by means of two variables: whether the vignette person is married or not (Vignettes 1, 5, 6, 9, 10) and whether the relationship with family members or friends is good or not (Vignettes 1, 2, 3, 5, 7, 8). Work is characterized by whether one works long hours or not (50 hours per week or more, vignettes 2 and 4) and whether the vignette person has
84
Section 1: Measuring Well-Being in an International Context
control over his or her job or job security or not (Vignettes 2, 5, and 10). Health, finally, is coded as good or not (Vignettes 2, 4, 5, 8). Table 4.5 presents the results of an ordered probit, a regression, and a probit where we have combined ‘‘dissatisfied’’ and ‘‘very dissatisfied’’ into one category and the other three answers into another category. The latter has been included to investigate whether patterns are different at the lower end of the scale. As one can see, qualitatively the results of the regression and the ordered probit are very similar. We will only discuss the regression, as one can immediately interpret the size of the effects, whereas that is more complicated in the ordered probit case. At the end we will comment on cases where the probit deviates from the regression and the ordered probit. As before, the top panel presents the results for the Dutch sample, while the bottom panel presents the effects of interactions with a U.S. dummy. The U.S. dummy is negative, but insignificant. Thus the evidence for a uniform scale shift is limited. The difference in response patterns is more subtle than just a uniform shift in the scale used in both countries. One observes that in the United States there is some evidence that a higher age is assumed to be associated with lower life satisfaction (remember once again that the scale runs from 1 for very satisfied to 5 for very dissatisfied). For instance, in the United States an age difference of 40 years would be associated with a deterioration in life satisfaction of .40 on the five-point scale. At first sight, it looks like being married is not considered to be a good thing, particularly in the United States. Notice, however, that this has to be looked at in combination with the quality of the relationship with family and friends. Since good relations have an enormous effect on life satisfaction, being in a good marriage is a major source of happiness. Both countries attach a positive value to retirement, but the Americans more so than the Dutch. Working long hours is believed to have a negative effect on life satisfaction, but more so in the United States than in the Netherlands, suggesting that working conditions are more attractive in the Netherlands than in the United States. Having control over one’s job or having a secure job is valued more positively in the United States than in the Netherlands. Good health is worth about half a point in the Netherlands and about three tenths of a point in the United States, but the difference between the two countries is only marginally significant. The larger effect in the Netherlands is in line with Table 4.2 and shows that the differences in Table 4.4 are indeed somewhat misleading—they disappear when other vignette characteristics are controlled for.
Life Satisfaction
85
Income has a positive effect on life satisfaction in both countries, although the effect is not all that large. This is in line with the results presented in Table 4.2, which showed that, of the four domains, income is the least important one for global life satisfaction. The one exception is the dummy for modest income, which is quite large, but since this dummy uniquely identifies Vignette 3, its coefficient really identifies the evaluation of that vignette as a whole, not just the income level. We note that income is thought to be more important in the United States than in the Netherlands. The binary probit that explains the probability of rating a vignette as ‘‘dissatisfied’’ or ‘‘very dissatisfied’’ shows some minor deviations from the results in the other two columns. Both in the Netherlands and the United States, increasing age is associated with a greater probability of being rated dissatisfied or very dissatisfied. We also note that having a secure job or a job over which one exerts control reduces the chances of being rated dissatisfied or very dissatisfied.
The Theory of Vignettes
In this section, we provide an intuitive description of the use of vignettes for identifying response scale differences and then sketch our statistical approach.7 The basic idea is illustrated in Figure 4.1, which presents the
Figure 4.1 Comparing self-reported happiness across two countries in case of DIF
86
Section 1: Measuring Well-Being in an International Context
distribution of life satisfaction or happiness in two hypothetical countries. The density of the continuous happiness variable in Country A is to the left of that in Country B, implying that, on average, people in Country A are happier than in Country B. The people in the two countries, however, use very different response scales if asked to report their happiness on a fivepoint scale (very satisfied, satisfied, not satisfied or dissatisfied, not satisfied, and very dissatisfied). In the example in the figure, people in Country B attach much more positive labels to given points on the life satisfaction scale than do people in Country A. Someone in Country A with the life satisfaction indicated by the dashed line would report themselves not satisfied, while a person in Country B with the same actual satisfaction would report themselves satisfied. The frequency distribution of the self-reports in the two countries would suggest that people in Country B are more satisfied than those in Country A—the opposite of the true distribution. Correcting for the differences in the response scales (DIF, ‘‘differential item functioning,’’ in the terminology of King et al., 2004) is essential to compare the actual health distributions in the two countries. Vignettes can be used to do the correction. A vignette question describes the satisfaction of a hypothetical person and then asks the respondent to evaluate the satisfaction of that person on the same five-point scale that was used for the self-report of their satisfaction. Since the vignette descriptions are the same in the two countries, the vignette persons in the two countries have the same actual life satisfaction or happiness. For example, respondents can be asked to evaluate the life satisfaction of a person whose satisfaction is given by the dashed line. In Country B, this will be evaluated as ‘‘satisfied.’’ In Country A, the evaluation would be ‘‘not satisfied.’’ Since the actual level of satisfaction is the same in the two countries, the difference in the country evaluations must be due to DIF. Vignette evaluations thus help identify differences between the response scales. Using the scale in one of the two countries as the benchmark, the distribution of evaluations in the other country can be adjusted by evaluating them on the benchmark scale. The corrected distribution of the evaluations can then be compared to that in the benchmark country— they are now on the same scale. In the example in the figure, this will lead to the correct conclusion that people in Country A are more satisfied than those in Country B, on average. The underlying assumption is response consistency: a given respondent uses the same scale for selfreports and the vignette evaluations. We will apply the vignette approach to life satisfaction, using vignettes not only to obtain international comparisons corrected for DIF, but also for comparisons of different groups within a given country.
Life Satisfaction
87
Econometric Model
Our model explains respondents’ self-reports on satisfaction by themselves as well as their reports on the satisfaction of hypothetical vignette persons. Selfreports are modeled as a function of respondent characteristics Xi (including demographics, a country dummy, and interactions of all demographics with that dummy) and an error term Ei by the following ordered response equation: Yi ¼ Xi þ Ei ; Ei Nð0; 2 Þ; Ei independent of Xi
ð1:1Þ
< Yi ji ; j ¼ 1; : : : 5 Yi ¼ j if j1 i
ð1:2Þ
The thresholds ij between the categories are given by þ expð j Xi Þ; j ¼ 2; 3; 4 0i ¼ 1; 5i ¼ 1; 1i ¼ 1 Xi þ ui ; ji ¼ j1 i ð1:3Þ ui Nð0; 2u Þ; ui independent of Xi and the other error terms in the model: Since Xi includes a country dummy and interactions of all demographics with that dummy, this specification allows for completely different ways in which the response scales can vary with demographics in the two countries. The various cutoff points can also vary in different ways, which seems useful because of the observed tendency of the Dutch to avoid extremes, suggesting that a Dutch respondent will have a lower first cutoff point but a higher last cutoff point than a similar U.S. respondent. As noted before, the fact that different respondents i use different response scales ji is called ‘‘differential item functioning’’ (DIF). The term ui introduces an unobserved individual effect in the response scale. It will imply that reported evaluations of different vignettes (see equation (1.5) below) are positively correlated with each other and with self-reports (conditional on Xi), since some respondents will tend to use high thresholds and others will use low thresholds in all their reports. Since such positive correlations are observed in the data, incorporating ui helps to improve the model’s ability to predict the observed outcomes (the model fit). Define a benchmark respondent with characteristics Xi = X(B). The DIF adjustment involves comparing Yi to thresholds jB rather than ji , where jB is obtained in the same way as ji but using X(B) instead of Xi. A respondent’s reported satisfaction is computed using a benchmark scale instead of a respondent’s own scale. This does not give an adjusted score for each individual (since Yi is not observed), but it can be used to simulate adjusted distributions of Yi for the whole population or conditional upon some of the characteristics in Xi.
88
Section 1: Measuring Well-Being in an International Context
Using self-reports on ‘‘own life satisfaction’’ only, parameters and 1 are not separately identified, only the difference between and 1 . For example, consider country dummies: people in two different countries can have systematically different life satisfactions, but if the scales on which they report their life satisfaction can also differ across countries, then self-reports are not enough to identify the satisfaction difference between the countries. The vignettes will be used to identify and 1 separately. The evaluations Yli of vignettes l = 1,. . .,L = 10 are modeled using similar ordered response equations: Yli ¼ l þ Ili þ Eli
ð1:4Þ
Yli ¼ j if ij1 < Yli ji ; j ¼ 1; : : : 5
ð1:5Þ
Eli Nð0; 2 Þ; independent of each other; of Eri and of Xi
ð1:6Þ
Thus we include a dummy for each of the ten vignettes and allow the evaluations to depend on the log of the income assigned to the vignette (Ili ), which is randomized across respondents. The unobserved vignette evaluations Yli do not depend on respondent characteristics Xi (the assumption of vignette equivalence). The actually reported evaluations Yli do depend on Xi , but only through the thresholds. The maintained assumption here is that of ‘‘response consistency,’’ meaning that the thresholds ji are the same for self-reports and the vignettes. With these assumptions, it is clear how vignette evaluations can separately identify b and ð¼ 1 ; :::; 5 Þ: From the vignette evaluations alone, , ; 1 ; :::10 and can be identified (up to the usual normalization of scale and location). From self-reports, can then be identified in addition. Thus the vignettes can be used to solve the identification problem due to DIF. The two assumptions vignette equivalence and response consistency are crucial for solving the identification problem. Vignette equivalence may be problematic if life satisfaction is multidimensional and the weights are different in the two countries. The fact that in Table 4.2, the interactions between domain satisfactions and the U.S. dummy are jointly insignificant suggests that this is not a serious problem in our case. Response consistency may be violated if, for example, people make systematic mistakes in evaluating vignette persons but are much better able to evaluate their own satisfaction. Response consistency can be tested if an objective measure is available and such tests have typically supported the use of vignettes (see King et al., 2004, on vision; and Van Soest et al., 2007, on drinking behavior), but an objective measure of satisfaction with life seems hard to give.
Life Satisfaction
89
Empirical Results
This section highlights our main empirical findings. We discuss our main parameter estimates determining overall satisfaction with life and assess the consequences of different threshold parameters in both countries. The model presented above was estimated using the self-evaluations and vignettes in the Dutch CentERpanel and the RAND American Life Panel. The equations for global life satisfaction and for the response thresholds include a complete set of interactions with a country dummy for the United States. We also estimated the simpler model that does not allow for DIF. This amounts to a standard ordered probit for self-assessed satisfaction. Model of Global Life Satisfaction
Table 4.6 lists parameter estimates for two models explaining global life satisfaction, where the scale is from good to bad (1: very satisfied, . . ., 5: very TABLE 4.6
Self Assessment of Global Satisfaction Model with DIF
Model without DIF
b 1.005** 0.081 0.431** 0.205* 0.043 0.066 0.141 0.018 0.237** 0.160* 0.089*
s.e. 0.14 0.06 0.11 0.10 0.09 0.09 0.11 0.08 0.08 0.07 0.04
b 0.899** 0.052 0.361** 0.205* 0.119 0.054 0.169+ 0.054 0.163* 0.034 0.098**
s.e. 0.12 0.05 0.10 0.09 0.08 0.07 0.09 0.07 0.07 0.06 0.04
Interactions with dummy US Dummy 0.217 Female 0.008 Married 0.095 ln family size 0.211 Age 40-50 0.274+ Age 51-64 0.156 Age 65+ 0.346+ Ed med 0.107 Ed high 0.026 Working 0.008 ln eq income 0.415**
0.24 0.10 0.15 0.15 0.15 0.14 0.20 0.13 0.14 0.12 0.08
0.270 0.055 0.015 0.078 0.138 0.033 0.397** 0.018 0.027 0.127 0.327**
0.20 0.83 0.13 0.13 0.13 0.13 0.16 0.11 0.11 0.10 0.06
Constant Female Married ln family size Age 40-50 Age 51-64 Age 65+ Ed med Ed high Working ln eq income
** indicates significance at 1% level, * indicates significance at the 5% level, and + indicates significance at the 10% level.
90
Section 1: Measuring Well-Being in an International Context
dissatisfied). All regressors in these models (except the country dummy) are measured in deviations of their country-specific means, which makes it easier to interpret the constant term and, most importantly, the implications of the U.S. dummy. Demographic regressors include dummy variables for whether the respondent is female, married, age brackets 40–50, 51–64, 65+ (the left-out group is under 40 years old). Education is separated into three groups—low, medium, or high, with the low-education group the omitted category. Income is measured as log-equivalized family income where income is adjusted by the logarithm of family size. Log-family size is also a separate regressor, in part to test for the adequacy of this choice of functional form for the equivalence scale. Finally, a dummy variable is included indicating whether or not the respondent is working. For reasons outlined above, our preferred model is the model with DIF (adjusting for threshold differences). It is listed in the first two columns of Table 4.6. In the Dutch sample, there are no significant differences in satisfaction with life by gender or age. Higher income makes the Dutch more satisfied with their life. Conditional on income, higher education also makes the Dutch more satisfied. Since education is typically associated with higher income, this most likely reflects the fact that education is a reasonable proxy for permanent income of respondents. Finally, conditional on the equivalized income, married Dutch respondents and those with larger families are more satisfied with their lives. One interpretation of this finding is that marriage and family are, on average, a source of wellbeing for these households. Dutch respondents who work are more satisfied with their lives than those who do not. Turn next to our estimates of the differences in parameters between the two countries that implicitly set the U.S. parameters. Since regressors are measured in deviations from within-country means, the coefficient on the U.S. dummy gives the difference between the average American person and the average Dutch person, whose characteristics are different. This coefficient is positive but insignificant, suggesting that the average Dutch and American respondents have similar satisfaction with life, according to both the model with and the model without DIF. Similarly to the Dutch, there are no gender differences in life satisfaction among the Americans, but the estimated age patterns indicate that life satisfaction among Americans increases with age and that retired Americans are particularly satisfied with their lives. There is no differential impact of work among Americans. Here the contrast with some of the results in Table 4.5 is of interest. When evaluating vignettes, Americans seemed to think that getting older would reduce life satisfaction. Yet when it comes to their own satisfaction, getting older is a good thing. None of these effects is strongly significant, but they still would seem to cast some doubt on the assumption
Life Satisfaction
91
that Americans are able to evaluate the vignette persons in the same way as they evaluate themselves (response consistency) or that respondents of different ages evaluate the same vignette the same (vignette equivalence). To explore this further, we have included interactions between the respondents’ own age (coded as dummies as in Table 4.6) and the age of the vignette person in the regressions reported in Table 4.5. These interactions turn out to be totally insignificant, as they should be under the assumption of vignette equivalence. This implies that there is no evidence that respondents make systematic errors in evaluating vignettes describing persons of different ages than their own age. The most important variable for comparing the two countries is income. The impact of income in improving life satisfaction is much more pronounced in the United States than in the Netherlands (more than four times larger in the United States in the model with DIF).8 Since we estimate no intercept difference between the countries and the data are all demeaned within countries, the Dutch and Americans are about equally satisfied at their country-specific mean incomes. But Americans become more satisfied with life at high-income levels and much less satisfied than the Dutch at low-income levels. Another important question is how the corrections for threshold differences within and across countries affect our interpretation of the determinants of life satisfaction. This question is addressed by comparing the parameter estimates in the model without DIF to the model with DIF. Several estimated effects seem rather similar between the two models. We note, however, that for the Dutch the estimated effects of education and working are larger in the model with DIF than in the model without. Considering the interactions with the U.S. dummy, the effect of income on life satisfaction in the United States turns out to be more pronounced when we correct for DIF. These differences between the models with and without DIF are of course directly related to the estimated equations for the thresholds in the model with DIF. These estimates are presented in Table 4.7. For instance, consider the effect of log income interacted with the U.S. dummy. The negative coefficient for this variable in the first threshold equation means that the first threshold shifts to the left when log income increases in the U.S. As a result of that, a response is less likely to lie to the left of that threshold. Since this effect of log income on the first threshold explains part of the existing negative correlation between income and life satisfaction, incorporating the effect on the threshold reduces the negative effect of log income in the United States on self-rated global satisfaction. This explains the difference of the income effects in the United States on life satisfaction in the models with and without DIF. One should note however, that all thresholds play a role, not only the first one. Disentangling the effect of the threshold shifts may be a complicated matter. We prefer therefore to
92
Section 1: Measuring Well-Being in an International Context
TABLE 4.7 Thresholds of Estimated Equation for Global Life Satisfaction
Threshold 1 Constant Female Married ln family size Age 40-50 Age 51-64 Age 65+ Ed med Ed high Working ln eq income
b 0.00 0.04 0.04 0.00 0.07 0.03 0.08 0.06 0.10+ 0.21* 0.02
s.e. 0.00 0.05 0.07 0.07 0.07 0.06 0.08 0.06 0.06 0.06 0.04
ln (Threshold 2– Threshold 1) b 0.53 0.02 0.01 0.04 0.03 0.03 0.06 0.02 0.06* 0.09* 0.01
Interactions with dummy US Dummy 0.08 0.18 0.31 Female 0.08 0.07 0.03 Married 0.03 0.10 0.06 ln family size 0.17 0.11 0.03 Age 40-50 0.23* 0.11 0.06 Age 51-64 0.19+ 0.10 0.05 Age 65+ 0.04 0.15 0.02 Ed med 0.09 0.11 0.01 Ed high 0.03 0.11 0.02 Working 0.15 0.09 0.06 ln eq income 0.16* 0.05 0.02
ln (Threshold 3– Threshold 2)
ln (Threshold 4– Threshold 3)
s.e. 0.17 0.02 0.04 0.04 0.03 0.03 0.05 0.03 0.03 0.03 0.02
b 0.81* 0.06 0.11 0.12* 0.02 0.07 0.16* 0.01 0.02 0.03 0.07*
s.e. 0.20 0.04 0.07 0.06 0.05 0.05 0.07 0.04 0.04 0.04 0.02
b 0.34 0.00 0.15* 0.01 0.06 0.10 0.16+ 0.01 0.00 0.05 0.08*
s.e. 0.36 0.04 0.08 0.07 0.06 0.06 0.08 0.06 0.06 0.05 0.04
0.34 0.04 0.06 0.06 0.06 0.06 0.08 0.06 0.06 0.05 0.03
0.84+ 0.10+ 0.02 0.09 0.00 0.15+ 0.09 0.04 0.09 0.04 0.07
0.43 0.06 0.09 0.09 0.08 0.09 0.12 0.08 0.08 0.07 0.04
0.13 0.03 0.30* 0.13 0.05 0.13 0.17 0.04 0.08 0.07 0.01
0.55 0.06 0.10 0.10 0.09 0.09 0.12 0.09 0.09 0.07 0.05
* Indicates significance at the 5% level; + indicates significance at the 10% level. N = 2244 for NL and 1093 for US.
investigate the importance of threshold differences between countries and between demographic groups within countries by a series of simulations.
Model Simulations
A transparent way of understanding the implications of our approach is to simulate the distribution of life satisfaction in the two countries for different parameter values. Essentially we first simulate the Dutch distribution of self-reported life satisfaction and then replace various sets of parameters by the corresponding American values. Table 4.8 presents the results of these simulations by four age groups—those less than 40, 40–50 years old,
TABLE 4.8
Simulations from Model with DIF: Percent Distribution of Global Satisfaction by Age Group
93
Very Satisfied
Satisfied
Not Satisfied/ Dissatisfied
Dissatisfied
Very Dissatisfied
Age group younger than 40 Dutch sample using own parameters Dutch using US threshold parameters Dutch using US self-assessment parameters Dutch using all US parameters US sample using US parameters
21.3 24.2 17.5 19.4 17.3
66.2 59.8 64.8 59.0 56.9
11.2 12.9 14.8 16.1 18.2
1.3 3.0 2.7 5.1 7.1
0.0 0.0 0.2 0.4 0.5
Age group 40-50 Dutch sample using own parameters Dutch using US threshold parameters Dutch using US self-assessment parameters Dutch using all US parameters US sample using US parameters
19.1 28.1 10.8 16.8 17.1
66.0 55.8 63.1 55.2 55.0
13.2 13.2 21.0 20.5 20.3
1.8 2.9 4.8 7.0 7.2
0.0 0.0 0.3 0.5 0.4
Age group 50-64 Dutch sample using own parameters Dutch using US threshold parameters Dutch using US self-assessment parameters Dutch using all US parameters US sample using US parameters
20.0 27.5 14.5 20.1 19.6
64.0 55.9 61.9 55.6 54.0
14.3 13.0 19.9 17.5 18.0
1.6 3.6 3.5 6.5 7.8
0.0 0.1 0.2 0.4 0.5
Age group 65 and older Dutch sample using own parameters Dutch using US threshold parameters
26.5 27.6
58.9 55.3
13.4 14.3
1.1 2.7
0.0 0.0 (continued )
TABLE 4.8
(Continued)
94
Very Satisfied
Satisfied
Not Satisfied/ Dissatisfied
Dissatisfied
Very Dissatisfied
Dutch using US self-assessment parameters Dutch using all US parameters US sample using US parameters
35.5 36.3 32.2
53.6 51.3 52.1
10.1 10.5 13.0
0.8 1.9 2.6
0.0 0.1 0.1
All age groups Dutch sample using own parameters Dutch using US threshold parameters Dutch using US self-assessment parameters Dutch using all US parameters US sample using US parameters
21.5 26.7 18.7 22.4 19.6
64.1 56.9 61.4 55.6 55.1
12.9 13.3 16.7 16.4 17.9
1.5 3.1 3.0 5.3 6.6
0.0 0.0 0.2 0.3 0.4
N=2244 for NL and 1093 for US.
Life Satisfaction
95
50–64 years old, and at least 65 years old. The first row for each age group summarizes the distribution of satisfaction with income for the Dutch, using their own parameters. The second row replaces Dutch thresholds by American thresholds (cf. Table 4.7). The third row simulates the Dutch distribution if we replace the parameters in the Dutch satisfaction equation (i.e., Table 4.6 with DIF) by the American parameters. The fourth row replaces all Dutch parameters by American parameters. The fifth row simulates distributions for the American sample using American parameters. Table 4.9 lists similar simulations by income quartile instead of age. For each age group in Table 4.8, the first row approximately reproduces the distribution of self-reports in the Dutch sample, while the fifth row does the same for the U.S. sample. Comparing the first two rows in each panel shows that the Dutch self-reports would become more spread out when Dutch respondents evaluated their satisfaction with life using U.S. thresholds. Both the percentage very satisfied and the percentage dissatisfied go up. This corresponds to the notion that the Dutch tend to avoid extremes; giving them the U.S. thresholds makes them more likely to report the two extreme categories. Comparing rows 1 and 2 with row 5 then shows that correcting for response scale differences does not make the distribution of life satisfaction in the Netherlands and the United States more similar in all respects. For example, for all age groups combined (final panel), we find that, after the correction, a much larger fraction in the Dutch sample are very satisfied with their life than in the U.S. sample. The fraction not satisfied/dissatisfied or worse increases somewhat in the Dutch sample and comes somewhat closer to the U.S. fraction, but remains substantially smaller. Both before and after correction for response-scale differences, the Dutch population as a whole is more satisfied with their lives than the Americans. This does not apply to the oldest age group, however: Americans of 65 years and older are somewhat more satisfied with their lives than their Dutch counterparts, on average, irrespective of whether we give them the same scales or not. Rows 3 and 4 in each panel can be used to show how much of the remaining differences (keeping response scales constant across countries) is due to differences in observed characteristics, generalizing the traditional Oaxaca-Blinder decomposition to a non-linear model (cf., e.g., Yun, 2004). In particular, comparing rows 4 and 5 shows the differences explained by differences in background characteristics between the two countries, using U.S. evaluation standards (both in the self-assessment equation and for the thresholds). The results show that, although the differences are modest, the characteristics make the Dutch in all age groups more satisfied with their lives than the Americans. The most important characteristic driving this is partnership status: having a partner has a strong positive effect on
96
Section 1: Measuring Well-Being in an International Context
satisfaction with life, and the fraction of respondents with a partner is much higher in the Netherlands than in the United States (78% versus 64%). On the other hand, comparing rows 2 and 3 in each panel of Table 4.8 shows that giving the Dutch the U.S. parameters for the self-assessment (but keeping the Dutch thresholds) also brings about substantial shifts, where for younger ages the imposition of U.S. parameters on Dutch respondents leads to lower simulated satisfaction, while for higher ages it leads to more satisfaction. This is a direct reflection of the results in Table 4.7, which show rather strong interaction effects with the U.S. dummy for the age brackets 40–50 and 51–64. Next, let’s turn our attention to Table 4.9, which does the same thing as Table 4.8 but for income quartiles instead of age groups. The effect of assigning U.S. thresholds again leads to more dispersion in the responses (row 2 compared to row 1). For the highest income quartile, comparing row 2 and row 5 shows that the U.S. respondents are better off. This was not clear from the first row, due to the reluctance of the Dutch to classify themselves as dissatisfied or very dissatisfied. Assigning U.S. self-assessment parameters to the Dutch confirms the stronger effect of income on life satisfaction in the United States than in the Netherlands (row 3 versus row 2). We see that with the U.S. self-assessment parameters, Dutch respondents with low incomes would be considerably less satisfied. Conversely, with high incomes they would be more satisfied. When the Dutch are assigned both the U.S. self-assessment parameters and the U.S. thresholds, then the satisfaction distribution more closely resembles that of the United States (rows 4 and 5), and again shows that the differences in background characteristics somewhat favor the Dutch, mainly in the third income quartile.
Conclusions
We have analyzed the determinants of global life satisfaction, by using both self-reports and responses to a battery of vignette questions. Although more work needs to be done, some preliminary conclusions can be drawn. It appears that the four domains—job or daily activities, social contact and family, health, and income—provide a fairly complete description of global life satisfaction in both countries. Among the four domains, social contacts and family have the highest impact on global life satisfaction, followed by job and daily activities and health. Income has the lowest impact. As in other work, we find that American response styles differ from the Dutch in that Americans are more likely to use the extremes of the scale
TABLE 4.9
Simulations from Model with DIF: Percent Distribution of Global Satisfaction by Income Group
97
Very Satisfied
Satisfied
Not Satisfied/ Dissatisfied
Dissatisfied
Very Dissatisfied
Lowest Income Quartile Dutch sample using own parameters Dutch using US threshold parameters Dutch using US self-assessment parameters Dutch using all US parameters US sample using US parameters
20.9 28.5 11.3 14.9 14.2
65.4 56.4 59.4 53.7 51.8
12.4 12.4 22.9 21.4 22.8
1.3 2.7 5.8 8.9 10.3
0.0 0.1 0.6 1.1 0.9
Second Income Quartile Dutch sample using own parameters Dutch using US threshold parameters Dutch using US self-assessment parameters Dutch using all US parameters US sample using US parameters
22.3 27.6 17.6 21.7 22.7
64.4 57.0 64.0 57.5 58.4
12.0 12.6 16.0 16.3 14.9
1.3 2.7 2.3 4.5 3.9
0.0 0.0 0.0 0.1 0.1
Third Income Quartile Dutch sample using own parameters Dutch using US threshold parameters Dutch using US self-assessment parameters Dutch using all US parameters US sample using US parameters
22.1 26.3 22.1 25.7 22.7
63.0 56.8 61.2 55.8 56.2
13.3 13.6 14.5 14.5 15.9
1.6 3.2 2.1 4.0 5.0
0.0 0.0 0.0 0.1 0.1
Highest Income Quartile Dutch sample using own parameters Dutch using US threshold parameters
20.6 24.0
63.5 57.5
14.1 14.6
1.8 3.8
0.0 0.0 (continued )
TABLE 4.9
(Continued)
98
Very Satisfied
Satisfied
Not Satisfied/ Dissatisfied
Dissatisfied
Very Dissatisfied
Dutch using US self-assessment parameters Dutch using all US parameters US sample using US parameters
24.8 28.0 26.7
60.9 55.7 57.4
12.5 12.9 12.6
1.7 3.4 3.2
0.0 0.0 0.0
All Income groups Dutch sample using own parameters Dutch using US threshold parameters Dutch using US self-assessment parameters Dutch using all US parameters US sample using US parameters
21.5 26.7 18.7 22.4 19.6
64.1 56.9 61.4 55.6 55.1
12.9 13.3 16.7 16.4 17.9
1.5 3.1 3.0 5.3 6.6
0.0 0.0 0.2 0.3 0.4
N = 2244 for NL and 1093 for US; Income is equivalized income (per capita). Quartiles are country specific.
Life Satisfaction
99
(either very satisfied or very dissatisfied) than the Dutch, who are more inclined to stay in the middle of the scale. Although for both Americans and the Dutch, income is the least important determinant of global life satisfaction, it is more important in the United States than in the Netherlands. Indeed life satisfaction varies substantially more with income in the United States than in the Netherlands. There are some intriguing differences between the way respondents judge vignette persons and what turns out to influence their own satisfaction. Respondents in both the Netherlands and the United States appear to think that marriage does not contribute to life satisfaction when they judge vignettes. Yet their own satisfaction is positively influenced by being married. Similarly, respondents believe that, other things being equal, older persons should be less satisfied. Yet their own satisfaction goes up with age. The estimates of an econometric model are used to calculate counterfactual distributions of life satisfaction. Correcting for differences in response scales leads to some shifts, though the shifts are not very large. For most age and income groups, the conclusion that the Dutch are more satisfied with their lives than the Americans remains valid. For the oldest age group (65+) and the highest income group, however, the vignette corrections lead to different conclusions: giving Dutch respondents the American scales shows that these groups are somewhat less satisfied than their U.S. counterparts. This was not clear from the distributions using the respondents’ own country’s scales, mainly because of the Dutch reluctance to evaluate themselves as dissatisfied or very dissatisfied. Vignettes have been shown to bring objective and subjective measurements of health (in particular, vision) or drinking behavior closer in line with each other. An objective measurement for life satisfaction seems hard to give, so that other ways of validation need to be considered, perhaps looking at actual behaviors that are correlated with life satisfaction. This is one of the directions of future research.
Notes 1. Tilburg University, P.O. Box 90153, 5000 LE Tilburg, fax: +31-13-4663280; email
[email protected] 2. The MS, the leading consumer sentiments survey, produces the widely used Index of Consumer Attitudes. MS respondents are asked if they have Internet access and, if yes, if they are willing to participate in Internet surveys. Those who agree are added to our household panel to be interviewed regularly over the Internet. As with the CentERpanel, respondents who do not have Internet access are provided with a
100
3. 4.
5.
6.
7.
8.
Section 1: Measuring Well-Being in an International Context
set-top box (an MSN Web TV) that allows them to browse the Internet and send and receive email. We analyze answers to questions on income satisfaction in depth in Kapteyn, Smith, & van Soest (2008). To keep the specification parsimonious and following van Praag & Ferrer-i-Carbonell (2008), we include domain satisfactions as cardinal variables. Using dummies gives qualitatively similar results. Van Praag & Ferrer-i-Carbonell (2008, Chapter 4) perform similar regressions using panel data on Germany and the United Kingdom. They find a much larger role for satisfaction with the financial situation than we do for satisfaction with income. For the United Kingdom, they also find that social contacts are the most important factor; they do not have satisfaction with social contacts in the German data. In earlier work on health and work disability vignettes, we found some effects of the order of vignettes within each domain on the vignette evaluations. It might also be useful to randomize the order in which self-assessments by domain are presented, but we did not do this. Vignettes have been used earlier in economic research by Van Beek, Koopmans, & van Praag (1997), who analyze employer preferences by presenting hypothetical descriptions of job applicants. The coefficients of log income in the United States (0.504 in the model with DIF, 0.425 in the model without DIF) seem rather large compared to the coefficients reported in the chapter by Layard, Mayraz, and Nickell in this volume (between 0.33 and 0.58 on a ten-point SWB scale, where we use a five-point scale). The coefficients for the Netherlands are much smaller than what Layard et al. find for European countries.
References Alesina, A., Di Tella, R., & MacCulloch, R. (2004). Inequality and happiness: Are Europeans and Americans different? Journal of Public Economics, 88(9–10), 2009–2042. Blanchflower, D. G., & Oswald, A. J. (2004). Well-being over time in Britain and the USA. Journal of Public Economics, 88(7–8), 1359–1386. Clark, A. E., Frijters, P., & Shields, M. (2008). Relative income, happiness, and utility: An explanation for the Easterlin Paradox and other puzzles. Journal of Economic Literature, 46(1), 95–144. Deaton, A. (2008). Income, aging, health and well-being around the world: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22(2), 53–72. Di Tella, R., MacCulloch, R. J., & Blanchflower, D. G. (2003). The macroeconomics of happiness. Review of Economics and Statistics, 85(4), 809–827. Di Tella, R., Haisken-DeNew, J., & MacCulloch, R. (2007). Happiness adaptation to income and to status in an individual panel. NBER Working Paper 13159. Easterlin R.A. (1974). Does economic growth improve the human lot? Some empirical evidence. In R. David & M. Reder (Eds.), Nations and households in economic growth: Essays in honor of Moses Abramowitz (pp. 89–125). New York: Academic Press. Easterlin, R. A. (1995). Will raising the incomes of all increase the happiness of all? Journal of Economic Behavior and Organization, 27(1), 35–48. Easterlin, R. A. (2006). Life cycle happiness and its sources: Intersections of psychology, economics, and demography. Journal of Economic Psychology, 27, 463–482.
Life Satisfaction
101
Ferrer-i-Carbonell, A., & van Praag, B. M. S. (2002). The subjective costs of health losses due to chronic diseases. An alternative model for monetary appraisal. Health Economics, 11, 709–722. Kapteyn, A., Smith, J. P., & van Soest, A. (2007). Vignettes and self-reports of work disability in the U.S. and the Netherlands. American Economic Review, 97(1), 461–473. Kapteyn, A., Smith, J. P., & van Soest, A. (2008). Are Americans really less happy with their incomes? RAND Working Paper, WR-591. King, G., Murray, C., Salomon, J., & Tandon, A. (2004). Enhancing the validity and cross-cultural comparability of measurement in survey research. American Political Science Review, 98(1), 567–583. Layard, R. (2005). Happiness: Lessons from a new science. London: Penguin Books. Rao, J. N. K., & Alastair, J. S. (1984). On chi-squared tests for multiway contingency tables with cell proportions estimated from survey data. Annals of Statistics, 12, 46–60. Stevenson, B., & Wolfers, J. (2008). Economic growth and subjective well-being: Reassessing the Easterlin Paradox. Working Paper, Wharton School, University of Pennsylvania, prepared for Brookings Papers on Economic Activity, Spring 2008. Van Beek, K. W. H., Koopmans, C. C., & and van Praag, B. M. S. (1997). Shopping at the labour market: A real tale of fiction. European Economic Review, 41(2), 295–317. Van Praag, B. M. S., & Ferrer-i-Carbonell, A. (2008). Happiness quantified—A satisfaction calculus approach. Oxford: Oxford University Press. Van Praag, B. M. S., Frijters, P., & Ferrer-i-Carbonell, A. (2003). The anatomy of subjective well-being. Journal of Economic Behavior and Organization, 51, 29–49. Van de Stadt, H., Kapteyn, A., & van de Geer, S. (1985). The relativity of utility: Evidence from panel data. The Review of Economics and Statistics, 67, 179–187. van Soest, A., Delaney, L., Harmon, C., Kapteyn, A., & Smith, J. P. (2007). Validating the use of vignettes for subjective threshold scales. RAND Labor and Population Working Paper WR-501. Yun, M-S. (2004). Decomposing differences in the first moment. Economics Letters, 82(2), 275–280.
APPENDIX – Life Satisfaction Vignettes
Global Life Satisfaction
Global 1: (Name) is 42 years old, happily married, with two children who are doing well at school and generally get on well with their parents. His/her family income is about xxx (median, twice median, four times median). He/she likes his work although some days it is somewhat stressful. (Name) suffers from rather serious back pain that keeps him/her awake at night about once a week, but has no other serious health problems. Global 2: (Name) is 50 years and divorced. He/she has one daughter of 22 with whom he/she gets on well, although he/she sees her only once a year. (Name) works about 60 hours per week, and feels he/she has a very secure job over which he/she has a lot of control. He/she makes about xxx (half the median, median, twice the median, four times the median) per year. He/she has no serious health problems. Global 3: (Name) is 65 years old. His/her wife/husband died 3 years ago and he/she still spends most of his time thinking about his/her and the good times they had together. He/she has 4 children and 10 grandchildren who visit him/her regularly. (Name) has a small pension and receives social security; he/she can make ends meet but has no money for extras such as expensive gifts to his/her grandchildren. He/she has had to stop working recently due to heart problems. He/she gets tired easily and, for example, 102
Life Satisfaction
103
cannot walk more than one block without taking a pause. Otherwise, he/she has no serious health conditions. Global 4: (Name) is 25 years old and single. He/she does not have many friends. He/she works about 50 hours a week and makes xxx (half the median, median, twice the median). He/she feels he/she has little control over his/her job and worries about loosing it. He/she has no health problems, but feels a little stressed sometimes. He/she does not exercise. Global 5: (Name) is 25 years old and recently married, no children. He/she works about 35 hours per week and makes xxx (half the median, median, twice the median). He/she works out regularly and on vacations he/she makes long hikes in the mountains with his/her husband. His/her job is satisfying, though a bit dull sometimes. He/she feels he/she does not have a lot of control over his/her job, but it is a very secure job. Global 6: (Name) is 57 years old and recently married his/her second wife/ husband. He/she has two children from his/her first marriage, but has little contact with them. He/she draws DI, because he/she has serious back pains. He/she often has trouble sleeping. His/her DI benefits are xxx (half the median, median, twice the median). Global 7: (Name) is 75 years old and a widow. His/her pension benefits are xxx (half the median, median, twice the median). He/she owns the house he/she lives in and has a large circle of friends. He/she plays bridge twice a week and goes on vacation regularly with some friends. Lately he/she has been suffering from arthritis, which makes work in the house and garden painful. Global 8: (Name) is 62 years old and has been retired for five years. He/she quit his job as soon as he/she could. He/she has never regretted his/her decision to retire. His/her pension is xxx (median, twice median; four times the median) He/she is physically very active and makes long bicycle trips in Southern Europe. He/she is single, but usually makes the trips with friends his/her age. Global 9: (Name) is 70 years old and has been retired for five years. His/her pension is xxx (median, twice median; four times the median). He/she still misses the contacts with his/her colleagues and would have liked to keep working part time. He/she and his wife/her husband take a few small vacations every year. For the rest they each lead their own lives and don’t do many things together. They have two children but rarely see them. He/ she is overweight and gets tired when walking more than a few blocks. He/ she has been a smoker all his/her life.
104
Section 1: Measuring Well-Being in an International Context
Global 10: (Name) is 50 years old and does not exercise. He/she cannot climb stairs or do other physical activities because he/she is obese. He/she has pain in his/her knees, elbows, wrists, and fingers, and the pain is present almost all the time. He/she has an executive job in a big firm and feels that he/she has a lot of control over his/her job. He/she makes xxx (twice the median, four times the median). He/she has been married for a long time, but he/she and his/her wife spend very little time together.
Chapter 5 Life (Evaluation), HIV/AIDS, and Death in Africa Angus Deatona, Jane Fortsonb, and Robert Tortorac a
Princeton University Mathematica Policy Research c Gallup Organization b
Angus Deaton is at the Center for Health and Wellbeing and the Research Program in Development Studies at the Woodrow Wilson School of Princeton University and is a senior research scientist at Gallup. Jane Fortson is at Mathematica Policy Research, and Robert Tortora is Chief Methodologist and Regional Research Director for sub-Saharan Africa at Gallup. We are grateful to Anne Case and Danny Kahneman for helpful discussions. Angus Deaton acknowledges financial support from NIA through Grant P30 AG024361 to Princeton University and Grant P01 AG05842–14 to the National Bureau of Economic Research.
The HIV/AIDS epidemic has brought large increases in morbidity and mortality to many countries in sub-Saharan Africa. For some countries, the epidemic has eliminated the large gains in life expectancy that took place between 1950 and 1990. More than 20 million Africans are estimated to be HIV positive, and between one and a half and two million die from AIDS every year. In the South African province of KwaZulu-Natal, it has been estimated that more than half of women aged 25 to 29 are HIV positive (Welz et al., 2007), and according to data from the Demographic and Health Surveys (DHS), national infection rates in Zimbabwe are more than 30 percent for women aged 30 to 39 and men aged 35 to 44. According to data from the 2006 wave of the Gallup World Poll, more than 80 percent of people in Kenya, Malawi, Rwanda, Uganda, Zambia, and Zimbabwe reported knowing someone who had died of AIDS. In the 2007 wave, more than a third of respondents in Uganda and Tanzania reported having lost an immediate family member to AIDS in the last year. 105
106
Section I: Measuring Well-Being in an International Context
In the context of this epidemic, we use the African data from the Gallup World Poll, supplemented with information from the African Demographic and Health Surveys, to look at the links between disease and self-reported well-being. Among the measures we examine is an overall life-evaluation measure, Cantril’s ‘‘self-anchoring striving scale’’ (1965), also known as the Ladder of Life, as well as measures of emotion or affect, including sadness, depression, smiling, and enjoyment. These data provide direct evidence on the emotional and hedonic impact of one of history’s most serious epidemics as reported by those who are directly experiencing it. The World Poll also contains a set of questions about the perceived importance of HIV/AIDS relative to other factors restricting well-being, and we use these, together with the well-being questions, to investigate the importance that Africans place on HIV/AIDS compared with other factors in their lives, such as other diseases, income, poverty, employment, and education. We use this information to address two distinct questions. The first is the value of life in sub-Saharan Africa, a topic that has long been controversial. We use self-reported well-being measures to calculate the change in income required to compensate people for the reduction in well-being associated with the death of an immediate family member. The second question concerns the self-reported well-being measures themselves. Is it legitimate to use them as a basis for calculating compensation? Beyond that, can self-reported well-being, or ‘‘happiness,’’ the blanket term often used in the literature, serve as an adequate guide to well-being in designing policies for public health and social welfare? The use of self-reported well-being measures to calculate compensation for the death of relatives has previously been recommended by Oswald & Powdthavee (2008). Their paper is part of a literature that uses self-reported well-being measures to calculate the income compensation associated with non-monetized factors, such as the value of informal care (van den Berg & Ferrer-i-Carbonell, 2007), airport noise (van Praag & Baarsma, 2005), and urban renewal (Dolan & Metcalfe, 2008, who argue that these measures will often be superior to direct assessment of willingness-to-pay). In earlier work using the Gallup World Poll, one of us found a strong positive—and approximately linear—relationship between average national life-evaluation and the logarithm of national income, but, conditional on income, could find no effect of life expectancy, or of the national prevalence of HIV infection (Deaton, 2008). If this finding is correct—and the more comprehensive analysis here will suggest that it is, at least if we confine ourselves to the life-evaluation measure—it would appear that Africans require little compensation for the consequences of the epidemic and attach relatively little priority to dealing with it. Such a finding has implications for a
Life (Evaluation), HIV/AIDS, and Death in Africa
107
number of important topics, including the design of foreign assistance for Africa, and more broadly, for the measurement of the level and distribution of international well-being using measures that incorporate both health and income. We shall address these issues in the final section of the chapter. Our findings are also important for the ‘‘happiness’’ literature, about the meaning of measures of self-reported well-being, and about how and whether they can be used in assessing welfare and in directing public policy. The African results show that the death of immediate family members has little effect on life evaluation, but a substantial effect on measures of negative affect, such as depression and sadness. These results show that different measures of well-being, although correlated, are by no means the same thing; measures of life evaluation capture different aspects of experience than do measures of affect. The sums of money required to compensate affect for the death of a family member are much larger than those required to compensate life evaluation. In consequence, it is not legitimate to subsume both under a blanket measure of ‘‘happiness,’’ let alone to use them more or less interchangeably as the practical counterparts of Benthamite utility and as a guide to utilitarian public policy. In the final section we argue that while measures of both life evaluation and affect are relevant for assessing well-being, each would be seriously compromised as an exclusive guide. The chapter is laid out as follows. Section 1 uses data on HIV prevalence from the DHS, together with data on self-reported well-being from the Gallup World Poll. Within countries in both data sets, we match means by sex and age group. These calculations are designed to reexamine and extend to sub-national aggregates the aggregate results in Deaton (2008), who found no effect of national health measures on average life evaluation across countries. In section 2, we move to micro-data from the World Poll, and look at the consequences for subjective well-being (SWB) of knowing someone who has died of HIV/AIDS (2006 survey), or of having a family member who died in the previous year (2007 survey). We compare these effects to the effects of higher income, placing a monetary value on the health outcomes. Following Tortora (2008), we also summarize the results of questions about the importance that people attach to dealing with various problems—such as joblessness, poverty, lack of education, and disease— and compare these with the results of the hedonic regressions. Section 3 discusses the implications of our findings. Our analysis is based on the African data from the 2006 and 2007 waves of the Gallup World Poll, which is a representative survey of adults from countries around the world. Samples of around 1,000 adults are drawn from each country: 140 countries in 2006 and 150 in 2007. In developing countries, including all of those we cover here, data are collected in
108
Section I: Measuring Well-Being in an International Context
face-to-face interviews. In both years the surveys almost always covered more than 95 percent of the population aged 15 or over in each country, with some exceptions, such as Angola, where areas with landmines were not surveyed. Typically, the frame consists of a list of Enumeration Areas used by the country’s central statistical office, from which primary sampling units (PSUs) were selected by Gallup. In each country, the PSUs are sorted into six strata, from those in cities of more than a million people, to the very few rural areas with a population of less than 10,000, and are selected with probability proportional to population within each stratum. One hundred and twenty-five PSUs are selected within each country, and eight interviews are obtained in each. A random route procedure is used to select households, and the Kish Grid is used to select one respondent, aged 15 or older, at random from each selected household. The Gallup World Poll covers more than 90 percent of the population of sub-Saharan Africa as a whole. A core set of questions, including questions on well-being, education, and income, is asked in every country. There are also questions that are different in each region, and the questions about death from HIV/ AIDS and other diseases are asked only in sub-Saharan Africa, where 32 countries were covered by either or both of these two waves.
1. HIV infection and well-being in Africa
Deaton (2008) found that, conditional on the logarithm of national income, national HIV prevalence is uncorrelated with mean life evaluation. In this section, we use better data on HIV prevalence from the DHS to investigate in more detail the link between HIV and life evaluation within some of the most highly affected countries. We also broaden the set of outcomes to include measures of affect, including enjoyment, smiling, sadness, and depression. We use HIV-related data drawn from the DHS for fourteen countries in sub-Saharan Africa: Burkina Faso, Cameroon, Ethiopia, Ghana, Guinea, Kenya, Malawi, Mali, Niger, Rwanda, Senegal, Tanzania, Zambia, and Zimbabwe. In these fourteen countries, a recent wave of the DHS has included the collection of blood samples for HIV testing. These blood samples, as well as responses to the individual questionnaire, yield information about HIV prevalence, HIV knowledge, and perceived HIV risk. HIV prevalence estimates from these data are arguably the best estimates available. The data from the DHS are described in the data notes at the end of the chapter. Table 5.1 shows, separately for each country, three measures of HIV from the DHS—HIV prevalence, the fraction of respondents who say that they know someone who has AIDS or has died of AIDS, and the fraction of
Life (Evaluation), HIV/AIDS, and Death in Africa
TABLE 5.1
109
HIV and Subjective Well-being, Means by Country
Country
HIV HIV HIV Prev. Knows Risk Ladder Enjoyment Smiling Sadness Depression
Burkina Faso Cameroon Ethiopia Ghana Guinea Kenya Malawi Mali Niger Rwanda Senegal Tanzania Zambia Zimbabwe
0.02 0.06 0.01 0.02 0.02 0.07 0.12 0.02 0.01 0.03 0.01 0.07 0.16 0.18
0.47 0.46 0.10 0.38 0.13 0.75 0.66 0.25 0.18 0.78 0.08 – 0.79 0.29
– – – – – 0.22 – – – – – 0.23 0.50 0.27
3.91 4.18 4.03 4.93 4.32 4.24 4.41 4.05 4.01 4.23 4.65 4.14 4.50 3.64
0.68 0.63 0.58 0.65 0.71 0.72 0.70 0.80 0.78 0.68 0.78 0.73 0.68 0.68
0.69 0.63 0.54 0.74 0.67 0.70 0.72 0.75 0.73 0.82 0.75 0.76 0.72 0.73
0.19 0.24 0.22 0.17 0.25 0.15 0.17 0.14 0.13 0.25 0.16 0.19 0.20 0.25
0.08 0.21 0.46 0.14 0.14 0.09 0.13 0.10 0.08 0.25 0.04 0.18 0.11 0.20
Notes: In all cells, the sample is restricted to adult respondents aged 15–49 and results are weighted using provided sample weights. HIV Prevalence, the mean of HIV Knows, and the mean of HIV Risk are calculated using DHS data. Means of the Ladder, enjoyment, smiling, sadness, and depression are calculated using Gallup data. HIV Knows is the fraction of DHS respondents who say that they know someone who has AIDS or has died of AIDS. HIV Risk is the fraction of DHS respondents who say that they are at moderate or higher risk of being infected with HIV. Ladder is the Cantril Ladder of Life on a scale from 0 (‘‘the worst possible life’’) to 10 (‘‘the best possible life’’). Enjoyment, sadness, and depression are indicators for whether, on the previous day, the respondent experienced these emotions a lot of the day. Smiling is an indicator for whether, on the previous day, the respondent smiled and laughed a lot of the day.
respondents who say that they are at moderate or higher risk of being infected with HIV. The last five columns show national average levels of life evaluation, enjoyment, smiling, sadness, and depression, calculated using data from the Gallup World Poll. Life evaluation is measured using the Cantril Ladder, which ranges from 0, ‘‘the worst possible life,’’ to 10, ‘‘the best possible life’’; we shall refer to points on this evaluative Ladder as ‘‘rungs’’ or ‘‘steps,’’ of which there are eleven, from 0 to 10. Enjoyment, sadness, and depression are indicators for whether, on the previous day, the respondent experienced these emotions a lot of the day. Smiling is an indicator for whether the respondent reported smiling and laughing a lot on the previous day. The national averages of these outcomes in Table 5.1 show no significant correlations across countries between HIV prevalence and average levels of subjective well-being. Our approach here is to look within countries at the relationship between HIV and measures of subjective well-being. In particular, we use individual-level HIV testing data from the DHS for each country to calculate
110
Section I: Measuring Well-Being in an International Context
prevalence separately by sex for each of seven five-year age groups (spanning ages 15–49). Figure 5.1 provides an overview of what drives our more detailed results. The first panel shows DHS estimates of HIV prevalence for each age group and sex, separately for high- and low-HIV prevalence countries. High-prevalence countries are those with prevalence above six percent (Kenya, Malawi, Tanzania, Zambia, and Zimbabwe). The figure shows the high degree of variability in HIV prevalence across countries and shows that, within countries, HIV infection is strongly related to age and sex, with prevalence in high-HIV countries peaking among women ages 30–34 and men ages 40–44. The pattern by sex and age of averages of well-being measures— including life evaluation, enjoyment, smiling, sadness, and depression— do not mirror the age profile in HIV prevalence in any obvious way. The difference in HIV infection for a sex/age group between high- and low-HIV countries at the top of the third column of Figure 5.1 bears little relation to the sex/age group pattern in mean well-being in other rows. The figures show that adults in high-HIV countries are more likely to report smiling and less likely to report depression; women in high-HIV countries report more enjoyment than women in low-HIV countries. Of course, the differences illustrated in Figure 5.1 could well be driven by the fact that high-HIV countries have other characteristics, such as better economic conditions, that raise well-being. In Table 5.2, we estimate the relationship between life evaluation and HIV, controlling for other individual and country characteristics. Here (as elsewhere), we do not weight by country population so that, for example, a sample person in Ghana gets the same weight as a sample person in Nigeria. Standard errors are clustered at the country, sex, and age group level. Controlling for log GDP per capita, HIV prevalence in a country, sex, and age group is associated with lower life evaluation (Table 5.2, column 2). The coefficient is about a third larger (more negative) in column 3, which also includes country fixed effects. The addition of the sex dummy in column 4 does very little, as does the replacement of national income by individual household income in column 5, although income itself has a large positive effect on the Ladder. These magnitudes imply that adults in a country, sex, and age group with prevalence of ten percent report life evaluation values about a tenth of a rung lower than those in a country, sex, and age group without HIV. One problem with these estimates is that the previous literature has established the existence of pronounced age-patterns in life evaluation; see, for example, Helliwell (2003) or Blanchflower & Oswald (2008), who argue that there is a universal U-shape in life evaluation, with a minimum in middle-age. This is approximately the mirror image of the
High HIV
High HIV-Low HIV
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
15
20
25
30
35
40
45
Difference –.4 –.2 0
0
Difference
0
–.1 –.08–.06 –.04 –.02 0
Depression .05 .1 .15 .2
Sadness 0 .05 .1 .15 .2
Difference –.1 –.05 0 .05
0
.2
Smiling .4 .6
Difference 0 .02.04.06.08 .1
.8
0
–.05
Difference 0 .05 .1
Enjoyment .2 .4 .6 .8
0 Enjoyment .2 .4 .6 .8 0 .8 Smiling .4 .6 .2 0 Sadness 0 .05 .1 .15 .2 Depression .05 .1 .15 .2 0
111
15 Life Evaluation 1 2 3 4 5
20
Life Evaluation 1 2 3 4 5
15
.2
Women
Difference 0 .05 .1 .15 .2
HIV Prevalence 0 .05 .1 .15 .2 .25
HIV Prevalence 0 .05 .1 .15 .2 .25
Low HIV Men
Figure 5.1 HIV Prevalence, Life Evaluation, Enjoyment, Smiling, Sadness, and Depression, by Age Group and Sex (separately for low- and high-HIV countries). Notes: ‘‘High HIV’’ countries are Kenya, Malawi, Tanzania, Zambia, and Zimbabwe. ‘‘Low HIV’’ countries are Burkina Faso, Cameroon, Ethiopia, Ghana, Guinea, Mali, Niger, Rwanda, and Senegal. In the left and middle columns, each bar represents the weighted mean in a particular age group, separately by sex. In the right column, each bar represents the difference, within an age group/sex category, between high- and low-HIV countries (i.e., middle columnleft column). HIV prevalence data are drawn from the DHS. Life evaluation, enjoyment, smiling, sadness, and depression are drawn from the Gallup World Poll. Enjoyment, sadness, and depression are indicators for whether, on the previous day, the respondent experienced these emotions a lot of the day. Smiling is an indicator for whether, on the previous day, the respondent smiled and laughed a lot of the day. Five-year age groups are as follows: 15–19 (‘‘15’’ in the figure), 20–24 (‘‘20’’ in the figure), 25–29 (‘‘25’’ in the figure), 30–34 (‘‘30’’ in the figure), 35–39 (‘‘35’’ in the figure), 40–44 (‘‘40’’ in the figure), and 45–49 (‘‘45’’ in the figure).
112
Section I: Measuring Well-Being in an International Context
TABLE 5.2 Life Evaluation and HIV Prevalence Ladder HIV ln GDP
(1)
(2)
(3)
(4)
(5)
(6)
0.483 (0.322) 0.401 (1.834)
0.986* (2.512) 0.388* (5.834)
1.392* (3.305)
1.348* (3.167)
1.323* (2.648)
0.860 (1.426)
0.020 (0.527)
0.001 (0.015) 0.414* (14.984)
0.011 (0.254) 0.415* (14.967)
Yes No 0.035 21663 14
Yes No 0.081 14210 12
Yes Yes 0.082 14210 12
Female ln y Country FEs? Age group FEs? R2 obs countries
No No 0.264 14 14
No No 0.010 21663 14
Yes No 0.035 21663 14
Notes: In all columns, the sample is restricted to adult respondents aged 15–49 and results are weighted using provided sample weights. The dependent variable is the Ladder. In column 1, the dependent variable is the country-level weighted average of the Ladder. In columns 2–6, the dependent variable is the individual Ladder. HIV is country-level HIV prevalence among adults 15–49 in column 1 and country/sex/age group-level prevalence in columns 2–6. ln GDP is the log of country-level GDP per capita in 2005, as measured by the International Comparison Program, World Bank (2008). ln y is the log of family income, using individual responses from the Gallup survey. In columns 2–6, standard errors are clustered at the country/sex/age group-level. Absolute values of t-statistics are in parentheses. * p < 0.05.
age pattern in HIV prevalence (see Figure 5.1), so that it is possible that the results in columns 2 through 5 of Table 5.2 are driven purely by a correlation in these age profiles. Since life satisfaction is U-shaped in countries where HIV prevalence is low or zero, we cannot safely use these patterns in countries in Africa where the infection rate is high. To deal with these concerns, we add controls for age groups. Our preferred specification is in column 6 of Table 5.2, where we control for sex, log family income, age group, and country. Here the estimated coefficient on HIV prevalence, although negative, is no longer statistically significant. In a check to see whether this finding might come from the small sample size, we have repeated (but do not show) the regressions adding data from Cambodia, the Dominican Republic, Haiti, and India—for which there are also DHS HIV-testing data—and have found the same insignificant result. Table 5.3 repeats the specification in the final column of Table 5.2, but with measures of affect replacing the Ladder measure of life evaluation. Controlling for sex, log income, age group, and country, HIV is not significantly associated with enjoyment, smiling, or sadness, but is
Life (Evaluation), HIV/AIDS, and Death in Africa
TABLE 5.3
113
Emotions and HIV Prevalence
HIV Female ln y
(1) Enjoyment
(2) Smiling
(3) Sadness
0.151 (1.060) 0.001 (0.136) 0.056* (10.992)
0.087 (0.598) 0.013 (1.249) 0.050* (9.446)
0.053 (0.413) 0.009 (0.919) 0.029* (6.264)
(4) Depression 0.198* (2.083) 0.005 (0.716) 0.020* (4.902)
Country FEs? Age group FEs?
Yes Yes
Yes Yes
Yes Yes
Yes Yes
R2 obs countries
0.036 14172 12
0.030 14048 12
0.017 14157 12
0.091 14127 12
Notes: In all columns, the sample is restricted to adult respondents aged 15–49 and results are weighted using provided sample weights. In column 1, the dependent variable is an indicator for whether, on previous day, the respondent experienced enjoyment a lot of the day. In column 2, the dependent variable is an indicator for whether, on previous day, the respondent smiled and laughed a lot of the day. In column 3, the dependent variable is an indicator for whether, on previous day, the respondent experienced sadness a lot of the day. In column 4, the dependent variable is an indicator for whether, on previous day, the respondent experienced depression a lot of the day. HIV is country/sex/age group-level HIV prevalence. ln y is the log of family income, using individual responses from the Gallup survey. In all columns, standard errors are clustered at the country/sex/age group-level. Absolute values of t–statistics are in parentheses. * p < 0.05.
significantly associated with depression. Adults in a country, sex, and age group with HIV prevalence of ten percent, compared to adults in a country, sex, and age group without HIV, are about two percentage points more likely to report experiencing depression much of the day on the previous day. In Appendix Tables 5.A.1 and 5.A.2, we estimate these regressions substituting for HIV prevalence the fraction of adults who report knowing someone who has AIDS or has died of AIDS. The results are broadly consistent with the results in Tables 5.2 and 5.3, and the preferred specification (Table 5.A.1, column 6) shows no significant relationship between life evaluation and HIV knowledge. Likewise, and with the exception of smiling, HIV knowledge is unrelated to the outcomes in Table 5.A.2. Adults in country/sex/age groups in which a higher fraction know someone affected by AIDS are less likely to report smiling a lot on the previous day. Appendix Tables 5.B.1 and 5.B.2, for which the sample sizes are admittedly much smaller, likewise show little relationship between HIV—now measured as the fraction of adults who report being at moderate or higher risk of being infected with HIV—and these outcomes.
114
Section I: Measuring Well-Being in an International Context
Overall, these results are indecisive, but they provide no clear evidence of strong effects of HIV on any of the measures of self-reported well-being. But they can hardly be taken to establish that HIV/AIDS has little effect on wellbeing: being HIV positive is not the same as having AIDS, and people who are infected may have no knowledge of the fact. High rates of infection in the population may not imply an increase in morbidity for respondents, especially if, in the absence of antiretroviral therapy, survival times with fullblown AIDS are short. Mortality from AIDS may also be attributed to other causes, especially in populations where background adult mortality is high, so that group variation in infection rates may not have a perceptible effect on group variation in life evaluation or affect, so that we may be looking at the wrong variables. Moreover, the use of the two different data sources means that we can only merge at the country, age group, and sex level. The lack of individual-specific data is likely to reduce the precision of our estimates, and we do not know for sure that the prevalence rate in one’s own sex and age group is the one that people are aware of or care about.
2. Well-being and mortality among individuals
We now take a more direct approach by looking at the effects on respondents of knowing someone who has died. In the 2006 sub-Saharan Africa module of the Gallup World Poll, respondents were asked ‘‘Do you personally know anyone that has died from X?’’ where X includes tuberculosis (TB), malaria, HIV/AIDS, smallpox, polio, hepatitis, and cholera. In the 2007 round, with an overlapping group of countries, the question was changed to ‘‘Please tell me if any one in your immediate family has died from X in the past 12 months?’’ where X includes the same diseases as before, plus death from chronic (more than six months) diarrhea and deaths of women in childbirth. In countries where people often have little contact with doctors or clinics, some of these diagnoses are manifestly unreliable, but at the least they provide an indication of how people perceive the effects of these diseases. In interpreting the usefulness of these answers, it should also be kept in mind that reliable data on adult mortality are almost completely absent in many of the countries covered here, and even official estimates of mortality from HIV/AIDS are little more than intelligent guesses based on small surveillance sites or projections from infection rates from surveys or antenatal clinics. Note also that, even where qualified personnel are in attendance, cause of death is not always easily ascertained, especially when the decedent suffered from multiple diseases. In the current context, this is particularly important for HIV/AIDS, which opens the way to opportunistic infections, particularly TB, with which a large fraction of the population has long been
Life (Evaluation), HIV/AIDS, and Death in Africa
115
TABLE 5.4 Fractions of people who report knowing someone who died of various conditions
Angola Benin Botswana Burkina Faso Burundi Cameroon Chad Ethiopia Ghana Kenya Madagascar Malawi Mali Mauritania Mozambique Niger Nigeria Rwanda Senegal Sierra Leone Tanzania Togo Uganda Zambia Zimbabwe
Malaria
HIV/ AIDS
TB
TB or HIV/AIDS
Any of seven
UNAIDS mortality
0.69 0.40 0.13 0.71 0.84 0.56 0.85 0.74 0.49 0.81 0.39 0.83 0.77 0.52 0.78 0.87 0.40 0.74 0.73 0.78 0.64 0.58 0.88 0.78 0.62
0.26 0.18 0.58 0.54 0.85 0.58 0.87 0.71 0.33 0.83 0.02 0.85 0.21 0.04 0.61 0.29 0.32 0.93 0.06 0.13 0.60 0.47 0.93 0.82 0.88
0.54 0.16 0.29 0.31 0.53 0.49 0.66 0.51 0.24 0.50 0.19 0.79 0.33 0.38 0.41 0.43 0.22 0.46 0.28 0.48 0.40 0.24 0.43 0.66 0.73
0.60 0.28 0.64 0.60 0.88 0.71 0.90 0.82 0.43 0.88 0.20 0.93 0.43 0.39 0.71 0.50 0.40 0.95 0.31 0.52 0.67 0.54 0.94 0.89 0.94
0.84 0.54 0.69 0.79 0.99 0.79 0.96 0.95 0.64 0.93 0.53 0.98 0.84 0.65 0.95 0.91 0.59 0.98 0.78 0.91 0.78 0.75 0.99 0.97 0.96
0.69 0.39 6.23 0.70 1.46 2.39 1.44 0.94 0.95 2.91 0.03 5.28 0.43 0.16 4.09 0.29 1.29 0.86 0.15 0.60 2.50 1.48 2.67 4.80 10.76
Notes: All but the final column report fractions of respondents in the 2006 wave of the Gallup World Poll, weighted by sampling weights, who answer positively to the question ‘‘Do you know someone who has died from X?’’ where the column heads show X. The seven diseases are those listed, plus cholera, hepatitis, polio, and smallpox. The final column is an estimate of AIDS mortality per thousand based on the estimate of AIDS deaths taken from UNAIDS (2008), divided by a thousand times 2005 population taken from the 2007 World Development Indicators.
asymptomatically infected. A substantial fraction of what are recorded as deaths from TB are probably attributable to HIV infection. Table 5.4 lists the fractions of people in the 2006 wave who reported that they knew someone who died of malaria, HIV/AIDS, and TB. In columns 4 and 5, we report the fractions of people who know someone who died either of HIV/AIDS or of TB as well as who died of any of the seven listed diseases, including polio, hepatitis, smallpox, and cholera. There are close to 1,000 respondents in each country. Clearly, these numbers should be treated with great caution as indicators of mortality rates. For example, only people with long memories could have known people who died of smallpox, yet in one or
116
Section I: Measuring Well-Being in an International Context
two countries, smallpox is frequently listed, for example by 41 percent of respondents in Chad, 20 percent in Sierra Leone, and 13 percent in Niger (not shown in the tables). These figures are most likely indicative of the quality of these data from those countries. Even so, the figures in Table 5.4 are broadly sensible; the correlations between column 2 or column 4 on one hand, and on the other, column 6, the UNAIDS (2008) estimates of mortality rates—themselves subject to error—are 0.57 and 0.56. The combined TB and HIV/AIDS numbers are probably the more accurate for AIDS deaths in the countries where the epidemic is severe, but this would not be the case where HIV/AIDS prevalence is low, as it is in most of West Africa. One important feature of Table 5.4 is that respondents typically know more people who died of malaria than who died of HIV/AIDS. This is not only true where it is to be expected, in the countries of West Africa where HIV/AIDS is relatively rare, such as Benin, Burkina Faso, Ghana, Mali, Mauritania, Niger, Nigeria, Senegal, Sierra Leone, and Togo, but it is also true, or within a point or two of being true, in some of the countries where the HIV/AIDS epidemic is the most severe, such as Burundi, Ethiopia, Kenya, Malawi, Mozambique, Tanzania, and Zambia. There is clearly substantial background mortality from disease in Africa, even before the advent of HIV/AIDS. Table 5.5 presents data from the 2007 wave of the Gallup World Poll on an overlapping group of countries, using the more focused question of whether respondents have lost a member of their immediate family in the last twelve months. These data also include two causes of death that were not asked about in 2006: death of a family member in childbirth, and death of a family member from chronic diarrheal disease. The latter did not generate many positive responses, and is not included in the table. The numbers in Table 5.5 are much smaller than those in Table 5.4, as must be the case, and a few remain implausible, such as the very high numbers for HIV/AIDS in Chad and the Central African Republic. The correlation between column 2 and the UNAIDS-based estimates of mortality rates is now only 0.45. However, perhaps the most important numbers are those in the fifth column for women in the immediate family who have died in childbirth. For half of the countries, particularly those in West Africa, these numbers are higher than the numbers of family members dying from HIV/AIDS, though typically not higher than those dying from malaria. Once again, there is a major cause of (at least perceived) mortality that is currently present, and was present (and presumably more severe than now) long before the advent of HIV/AIDS. Table 5.6 presents the first evidence on the effects of having recently lost an immediate family member on five measures of self-reported well-being.
Life (Evaluation), HIV/AIDS, and Death in Africa
117
TABLE 5.5 Fractions of people who report losing an immediate family member in the last twelve months
Angola Benin Burkina Faso Cameroon Central African R. Chad D. R. of the Congo Ethiopia Ghana Guinea Kenya Liberia Madagascar Malawi Mali Mauritania Mozambique Namibia Niger Nigeria Senegal Sierra Leone South Africa Sudan Tanzania Uganda Zambia Zimbabwe
Malaria
AIDS
TB
TB or AIDS
Childbirth
Any
0.19 0.09 0.25 0.29 0.26 0.52 0.19 0.11 0.06 0.24 0.18 0.15 0.10 0.26 0.18 0.13 0.15 0.05 0.39 0.06 0.20 0.33 0.00 0.09 0.33 0.27 0.21 0.14
0.07 0.05 0.10 0.17 0.44 0.47 0.12 0.08 0.01 0.04 0.17 0.01 0.00 0.15 0.03 0.03 0.08 0.19 0.02 0.02 0.00 0.03 0.11 0.03 0.35 0.38 0.13 0.28
0.07 0.02 0.05 0.17 0.26 0.32 0.11 0.10 0.03 0.10 0.06 0.05 0.06 0.18 0.05 0.04 0.09 0.08 0.09 0.03 0.02 0.10 0.05 0.02 0.19 0.03 0.10 0.18
0.11 0.07 0.13 0.27 0.51 0.51 0.20 0.16 0.04 0.12 0.19 0.05 0.06 0.26 0.07 0.06 0.14 0.21 0.11 0.04 0.02 0.11 0.13 0.03 0.39 0.39 0.19 0.34
0.10 0.22 0.11 0.18 0.17 0.36 0.16 0.07 0.11 0.11 0.10 0.09 0.10 0.11 0.09 0.04 0.05 0.01 0.21 0.07 0.08 0.21 0.02 0.08 0.25 0.11 0.07 0.10
0.30 0.32 0.33 0.48 0.62 0.69 0.39 0.29 0.16 0.38 0.35 0.23 0.18 0.45 0.27 0.17 0.27 0.23 0.45 0.13 0.24 0.47 0.14 0.17 0.60 0.52 0.35 0.38
Notes: Fractions of respondents in the 2007 wave of the Gallup World Poll, weighted by sampling weights, who answer positively to the question ‘‘Please tell me if any one in your immediate family has died from X in the past 12 months?’’ where the column heads show X. The correlation between the HIV/AIDS fractions in column 2 and the UNAIDS mortality numbers in the final column of Table 5.4 for the overlapping countries is 0.45.
The measure of loss we have used is whether the respondent has lost an immediate family member in the last twelve months to one of (a) malaria, (b) TB, (c) HIV/AIDS, or (d) death in childbirth. The World Poll does not have a question on all-cause mortality, and we have ignored the other reported causes (hepatitis, cholera, polio, smallpox, and chronic diarrhea) because the fractions reporting are very small, and because there are substantial numbers of missing values from ‘‘don’t knows’’ or refusals. We look at five different measures of well-being (described in section 1): (a) the Cantril Ladder of Life, (b) enjoyment, (c) smiling, (d) sadness, and
118
Section I: Measuring Well-Being in an International Context
TABLE 5.6 Differences in well-being measures between those who report losing an immediate family member in the last 12 months and those who do not (differences in bold are more than twice their estimated standard errors).
Angola Benin Burkina Faso Cameroon Central African R Chad D.R. Congo Ethiopia Ghana Guinea Kenya Liberia Madagascar Malawi Mali Mauritania Mozambique Namibia Niger Nigeria Senegal Sierra Leone South Africa Sudan Tanzania Uganda Zambia Zimbabwe Average FE regression
Ladder
Enjoyment
Smiling
Sadness
Depression
0.03 0.15 0.08 0.24 0.11 0.19 0.11 –0.44 0.06 0.18 0.07 –0.48 0.08 0.22 0.23 0.04 –0.36 0.04 –0.40 0.00 –0.38 0.07 0.22 0.39 0.32 0.23 0.16 0.36 –0.11 –0.12
–0.16 0.03 0.02 0.01 0.05 0.03 –0.08 0.06 0.13 0.05 0.11 0.01 –0.12 0.05 –0.06 –0.18 0.03 0.03 0.08 –0.14 0.06 0.04 0.00 0.05 0.01 0.04 0.04 –0.16 –0.04 –0.04
0.08 0.08 0.04 0.03 0.00 0.08 0.05 0.06 0.01 0.08 0.05 0.05 0.04 –0.11 –0.13 –0.17 0.01 0.05 –0.12 –0.10 0.01 0.05 0.07 0.02 0.01 0.13 0.04 –0.13 –0.03 –0.02
0.10 0.05 0.05 0.05 0.12 0.04 0.03 0.09 0.05 0.06 0.00 0.00 0.05 0.00 0.11 0.09 0.06 0.13 0.07 0.12 0.03 0.07 0.14 0.01 0.08 0.01 0.09 0.14 0.06 0.06
0.06 0.02 0.08 0.08 0.04 0.02 0.06 0.10 0.10 0.08 0.01 0.03 0.18 0.02 0.03 0.09 0.01 0.08 0.04 0.15 0.00 0.01 0.13 0.02 0.07 0.01 0.03 0.11 0.05 0.05
Notes: The average is a simple average over countries (calculated using the within country weights) without population weighting. Each column shows the difference in means between those reporting to have lost, in the last twelve months, an immediate family member to malaria, HIV/AIDS, TB, or childbirth and those who did not so report. Standard errors are corrected for design effects. The FE regression in the last row comes from regressing the dependent variable on a dummy for mortality from any cause and a set of country fixed effects, with standard errors corrected for survey design.
(e) depression. Each of these measures captures a potentially different aspect of feelings and of life assessment, and we have no prior expectation that they will respond in the same way to the deaths of family members. Table 5.6 shows the differences in the SWB measures between people who report having lost a family member and those who do not. Figures in
Life (Evaluation), HIV/AIDS, and Death in Africa
119
bold are more than twice their standard errors, computed taking into account the survey design. The results are generally in the direction that would be expected: deaths of immediate family members reduce the Ladder value, reduce the probability of having smiled, laughed, or enjoyed oneself yesterday, and increase the likelihood of experiencing sadness or depression. But there are exceptions, a few of which are significantly different from zero (e.g., the positive effect of a death on the Ladder in Sudan, or on enjoyment in Kenya), and only about a quarter of the differences are significantly different from zero. Sadness and depression are the two most consistent of the indicators, with only two negative signs for sadness (Senegal and Sudan) and four for depression (Benin, Kenya, Mozambique, and Sudan), and these six differences are small and insignificantly different from zero. We deal with the heterogeneity by pooling across countries, and by either running regressions with country fixed effects, or by averaging the results over countries, taking simple averages with each country as a data point. (Weighting by population would give most of the weight to Nigeria.) In the fixed-effect regression, we regress each SWB measure on a dummy for a death and a set of country dummies. This regression also yields an average of the country effects, but where each country difference is weighted by the inverse of its estimated variance, so that more precisely estimated differences get higher weights. Both numbers are shown in the bottom panel of Table 5.6. As it turns out, the two sets of estimates are almost the same. Over sub-Saharan Africa as a whole, the death of an immediate family member reduces the Ladder values by 0.11 or 0.12 of a step, the probability of enjoyment or smiling/laughter by 2 to 4 percentage points, and increases the probability of experiencing sadness and depression by 5 to 6 percentage points. These estimates come from simple differences of averages in each country, with no controls for other differences across people who have and have not lost a family member in the last year. SWB is generally sensitive to demographic status, such as age and sex, as well as to education and income, all of which could potentially confound the effects of a death. As we shall see, education is positively related to SWB, and more educated people are more likely to be HIV positive in several of these countries (Fortson, 2008). Tables 5.7 and 5.8 move to a multivariate analysis, using the same outcomes as Table 5.6 (an 11-point scale for the Ladder, and linear probability models for the other SWB measures) and including age, sex, education, and income as controls. As before, we present estimates based on pooled regressions with country fixed effects (Table 5.7) and estimates that come from estimating regressions for each country, and then averaging the coefficients (Table 5.8). Given that the World Poll data have
TABLE 5.7
Country fixed-effect regressions on SWB of knowing a family member who died and other controls, including income Ladder
Malaria TB AIDS Childbirth Age
120
Age2 Female Education ln y obs countries F(equal) F(zero)
0.083 (2.0) 0.104 (1.9) 0.104 (1.9) 0.035 (0.7) 0.00452 (1.1) 0.00006 (1.3) 0.013 (0.5) 0.458 (14.4) .. 26,232 28 2.80 2.76
0.082 (1.8) 0.128 (2.2) 0.100 (1.8) 0.042 (0.8) 0.00153 (0.3) 0.00002 (0.3) 0.011 (0.4) 0.273 (7.6) 0.366 (19.2) 19,459 26 3.36 2.70
Enjoyment 0.025 (2.3) 0.032 (2.1) 0.009 (0.7) 0.010 (0.7) 0.00218 (1.9) 0.00001 (0.7) 0.004 (0.6) 0.098 (11.9) .. 26,172 28 1.55 3.03
0.029 (2.3) 0.024 (1.4) 0.002 (0.1) 0.005 (0.3) 0.00225 (1.6) 0.00001 (0.4) 0.000 (0.0) 0.070 (7.1) 0.057 (11.7) 19,396 26 1.17 2.14
Smiling 0.004 (0.4) 0.038 (2.5) 0.006 (0.4) 0.016 (1.2) 0.00543 (5.1) 0.00005 (3.8) 0.018 (2.5) 0.053 (6.7) .. 26,060 28 1.45 2.23
0.011 (0.9) 0.028 (1.6) 0.001 (0.1) 0.014 (0.9) 0.00488 (3.7) 0.00004 (2.8) 0.008 (1.0) 0.024 (2.6) 0.046 (10.1) 19,320 26 0.41 1.42
Sadness 0.027 (2.9) 0.038 (2.8) 0.050 (4.0) 0.029 (2.5) 0.00051 (0.5) 0.00000 (0.3) 0.012 (1.9) 0.053 (7.6) .. 26,295 28 0.75 15.31
0.034 (3.1) 0.038 (2.5) 0.050 (4.0) 0.013 (1.0) 0.00044 (0.3) 0.00001 (0.5) 0.021 (2.8) 0.032 (4.0) 0.035 (8.6) 19,439 26 1.42 13.14
Depression 0.026 (3.1) 0.045 (3.4) 0.017 (1.6) 0.034 (3.2) 0.00249 (2.6) 0.00003 (2.5) 0.002 (0.4) 0.041 (6.5) .. 26,255 28 0.80 13.37
0.030 (3.2) 0.045 (3.1) 0.020 (1.8) 0.028 (2.3) 0.00153 (1.3) 0.00002 (1.3) 0.004 (0.6) 0.028 (3.6) 0.024 (6.2) 19,414 26 0.48 10.88
Notes: Each pair of columns shows two country fixed-effect regressions with the dependent variable as column head, the first excluding the log of income, the second including it. Education is a dummy that is 1 if the person has completed eight or more years of education. Obs is the number of observations in each regression; note the drop when income is included. The two F-statistics are F-tests for the coefficients on the four causes of death being equal, and for all effects being zero.
TABLE 5.8 Averaged coefficients from country regressions on SWB of knowing a family member who died and other controls, including income Ladder Malaria TB AIDS Childbirth
121
Age Age2 Female Education ln y obs countries F(equal) F(zero)
0.110 0.102 (2.2) (2.0) 0.116 0.129 (2.0) (2.0) 0.202 0.166 (2.9) (2.0) 0.035 0.094 (0.6) (1.6) 0.00052 0.00471 (0.1) (0.9) 0.00000 0.00006 (0.0) (1.0) 0.009 0.020 (0.4) (0.6) 0.456 0.273 (14.1) (7.1) .. 0.374 (16.8) 26,232 19,459 28 26 4.48 3.96 3.73 2.98
Enjoyment 0.023 (1.8) 0.036 (2.1) 0.006 (0.4) 0.016 (1.1) 0.00226 (1.9) 0.00001 (0.8) 0.005 (0.7) 0.102 (12.3) .. 26,172 28 0.88 3.13
0.030 (2.0) 0.033 (1.5) 0.005 (0.2) 0.001 (0.1) 0.00275 (1.9) 0.00001 (0.8) 0.002 (0.2) 0.071 (7.1) 0.064 (12.0) 19,396 26 0.69 2.35
Smiling 0.003 (0.2) 0.053 (3.1) 0.022 (1.2) 0.011 (0.8) 0.00536 (4.8) 0.00005 (3.7) 0.019 (2.6) 0.057 (7.0) .. 26,060 28 2.31 3.92
0.009 (0.6) 0.053 (2.4) 0.010 (0.5) 0.003 (0.2) 0.00439 (3.2) 0.00004 (2.3) 0.004 (0.5) 0.029 (3.0) 0.055 (10.7) 19,320 26 1.21 2.37
Sadness 0.034 (3.0) 0.021 (1.5) 0.085 (3.7) 0.033 (2.4) 0.00032 (0.3) 0.00000 (0.0) 0.012 (1.9) 0.053 (7.5) .. 26,295 28 1.77 10.71
0.034 (2.5) 0.013 (0.8) 0.127 (5.0) 0.021 (1.3) 0.00091 (0.7) 0.00002 (0.9) 0.025 (3.3) 0.028 (3.3) 0.042 (9.1) 19,439 26 4.67 11.08
Notes: See Table 5.7 above. Here the regressions are run country by country, and the coefficients averaged over them.
Depression 0.029 (2.5) 0.049 (3.3) 0.016 (0.9) 0.039 (3.4) 0.00282 (3.0) 0.00003 (3.0) 0.004 (0.7) 0.040 (6.4) .. 26,255 28 0.66 10.10
0.035 (2.7) 0.063 (3.7) 0.025 (1.1) 0.033 (2.2) 0.00240 (2.0) 0.00003 (2.1) 0.007 (1.1) 0.029 (3.7) 0.028 (6.8) 19,414 26 0.79 9.43
122
Section I: Measuring Well-Being in an International Context
about 1,000 observations for each country, the estimation of separate regressions for each country is entirely feasible, and the averaging recognizes that fact. Nor is it clear that when we average across countries, we wish to give more weight to the more precisely estimated coefficients rather than counting each country as a unit. Even so, Tables 5.7 and 5.8, while differing in detail, are remarkably similar, so that little depends on which we use, and we have the comfort of knowing that our results are not affected by which of the two methods we use. For each measure of SWB, we present two regressions, one with and one without income. We want to condition on income, which has consistently proved to be one of the most powerful predictors of the Ladder and of the affect measures in the World Poll, but there are two countries here (Guinea and Mali) where we do not have useable income data, and there are many missing observations within countries that do have income data, so that we lose about a quarter of the sample size when income is introduced. By showing regressions with and without income, we can check that the restriction of the sample does not have a major effect on the results. We also note that the income measure comes from a single question in which the individual respondent is asked to choose an income bracket for family income. In sub-Saharan Africa—as in other poor, largely agricultural areas of the world—such questions are unlikely to elicit more than an extremely imprecise estimate of the usual concept of income. As a result, there is likely to be substantial attenuation bias in the estimates of the effects of income. We work with the logarithm of income; previous work by Deaton (2008) and by Stevenson & Wolfers (2008) has shown that the logarithmic form works well both within and between countries, at least as far as the Ladder is concerned. Table 5.7 presents the fixed-effect regression results for the five SWB measures. All coefficients and their t–values are shown, other than the country fixed effects. In each case, the first column shows the estimates without income, and the second column shows the estimates including the logarithm of income; the second column always has two fewer countries. Of the non-mortality variables, education—whether the respondent has eight or more years of completed schooling—and income have consistently positive effects on life satisfaction and on reported emotions. More educated and higher income people report higher values on the Ladder of Life, they are more likely to remember laughing and smiling and enjoying themselves yesterday, and they are less likely to remember being sad or depressed. The effects of being better educated are similar in size to an increase of one unit in the logarithm of income, which corresponds to a 172 percent increase. Given the errors of measurement in the income variable, it would be a mistake to interpret the tables as showing the separate effects of
Life (Evaluation), HIV/AIDS, and Death in Africa
123
income and education, since the latter is likely to pick up at least some effects of the former. The effects of gender and age are inconsistent and usually weak. In the World Poll in general, women are more likely than men to report both more positive and more negative emotions, but these results are not clearly apparent in these sub-Saharan African countries, and the significance of estimates depends on whether or not income is included. We do not replicate the standard finding that life evaluation is U-shaped over the life-cycle, though there is a U-shape for smiling and laughing (reaching the minimum around age 50), and (more weakly) for being depressed, with a minimum around age 40. Our main results are in the first four rows, which report the effects on well-being of having lost an immediate family member in the last twelve months. The pattern of these estimates differs sharply across the SWB measures as well as across causes of death within the measures. The Ladder of Life is close to the life satisfaction or ‘‘happiness’’ measure that is used by Layard (2005) and others to measure overall well-being. Yet the loss of immediate family members has only a modestly negative effect on the Ladder. Deaths from TB, malaria, and childbirth have negative effects, while a death from HIV/AIDS has an apparent positive effect. The t–values for all of the estimates are unimpressive, and we can barely reject the hypothesis that all are zero (final row, F statistic is 2.76 with p–value of 0.027) or that a death has the same effect no matter what its cause (F-statistic is 2.80 with a p–value of 0.039). In Table 5.8, where the country regressions are averaged, the t–values are somewhat larger and the tests of no effects and of equality are larger and more significant. The anomalous effects of HIV/AIDS may have some real basis—for example, there is some evidence that some deaths, such as deaths from cancer or other ‘‘dread diseases,’’ are feared more than others (Sunstein, 2004). Different causes of death will typically involve different family members of different ages— we do not know who died in these data—but it seems likely that in most of Africa, even now, HIV/AIDS deaths are more common among those who are relatively well off, which is consistent with the reduction of the positive effect when income is introduced, albeit imperfectly, given the measurement error in income. Note also that the change in the estimates with income also comes from the change in sample size; for example, the upward revision of the effect of an AIDS death on sadness in Table 5.8 can be replicated without income but with the regression confined to the sample for which income is not missing. Even if we take the largest negative coefficient, which is for the loss of a family member to TB, the estimate in Table 5.8 is 0.129, compared with 0.374 for log income, so that the income equivalent of losing an immediate
124
Section I: Measuring Well-Being in an International Context
family member to TB is a change in log income of 0.129 divided by 0.374, or 0.345, equivalent to a 29 percent reduction in income. Alternatively, the compensation for the loss would be 41 percent of income for as many years as the effect lasts. By the same token, the compensation for a loss to HIV/AIDS or to childbirth (at least in Table 5.8) is negative. These numbers seem absurd on their own terms, even before we consider comparing those monetary values to similar monetary values from rich countries. And they are almost certainly gross overestimates, given attenuation bias in the estimates of income through measurement error. The estimated effects of mortality on the two positive emotions, smiling and enjoyment, are qualitatively similar to the estimated effects of mortality on the Ladder. Deaths from TB and malaria inhibit smiling or laughing and inhibit enjoyment, but deaths from HIV/AIDS or from childbirth have sometimes positive and sometimes negative effects, depending on whether income is included. The coefficients are sufficiently far from zero that we can reject the null of no effect from any death in most cases, but we can usually accept the hypothesis that all deaths have the same effect on the experience of these emotions. The results (in Table 5.7) for the last two measures—sadness and depression—are closely in line with what we might expect, and at variance with the results for the Ladder. People are sharply and significantly more likely (up to five percentage points) to report feeling sad or depressed if they have lost a family member. All four causes of death have similar effects, and in spite of their individual and joint significance, we cannot reject the hypothesis that the four estimates are identical. The anomalous effect of an HIV/AIDS death—and in some cases of a death of a woman in childbirth—on the Ladder or on positive feelings does not recur for sadness or depression. A death is a death and leads to sadness and depression. The evident coherence and sensibleness of these results contrasts with those for the Ladder, and make it much harder to attribute the latter to the effects of poorly designed questions, lack of understanding by respondents, or general measurement error. If we were to choose to express the effects on sadness and depression in terms of the effects of income, the results would be much larger than for the Ladder of Life. For example, if we were to take 0.04 as a representative estimate of the effects of disease, and 0.025 as a representative estimate of the effect of log income, the ratio is 1.6, so that the effects of the death on sadness or depression would be reproduced by an 80 percent reduction in income, or offset by a fivefold increase. Of course, if we were to decide to use these numbers to calculate the monetary equivalent of the death of a family member, we would have to explain why they are to be preferred over
Life (Evaluation), HIV/AIDS, and Death in Africa
125
the much smaller (and barely significantly different from zero) numbers that come from looking at the Ladder of Life, particularly given that the Ladder is much closer to the life-satisfaction measures that have been used in the previous literature. Before addressing that question, it is worth considering another measure, previously reported by Tortora (2008). The World Poll asked respondents in Africa to rank in importance twelve objectives based on the Millennium Development Goals. The objectives were (1) providing more jobs for youth, (2) achieving primary education for all, (3) reducing the spread of malaria and TB, (4) improving access to safe drinking water, (5) reducing the death rate among children under five, (6) reducing poverty, (7) reducing the number of women dying during childbirth, (8) reducing the spread of HIV/AIDS, (9) achieving gender equality and empowering women, (10) improving access to sanitation facilities, (11) providing access to new technology, and (12) reducing hunger. Each respondent was given a random selection of six of the twelve objectives, and asked to rank them from one (most important) to six (least important). Tortora’s Table 1 shows that reducing poverty and reducing hunger handily win this race, with average ranks of 2.41 and 2.48 respectively. Next, but with a considerably lower rank, comes reducing the spread of HIV/AIDS, with an average rank of 3.05, followed by jobs for youth (3.17), reducing the death rate from children under five (3.34), reducing deaths in childbirth (3.38), achieving primary education for all (3.62), reducing the spread of malaria and TB (3.64), and improving access to safe drinking water (3.75). There is then another substantial gap in the rankings before we come to improved sanitation (4.09), gender equality (4.38), and providing access to new technology (4.65). Kharas (2008) reports similar findings from the Afrobarometer surveys from Kenya, Mozambique, Nigeria, South Africa, Tanzania, Uganda, and Zambia, where respondents listed their top priorities as jobs, income, support for agriculture, and improvement in infrastructure, with health issues, including HIV/AIDS, attracting much lower rankings. These results are consistent with some, although not all, of the findings from the SWB analysis. The high rank for reducing poverty and hunger is consistent with the (dominant) importance of income and education as determinants of life evaluation. That HIV/AIDS comes next, and is ranked higher than TB and malaria, in spite of the higher prevalence of both of the latter (Tables 5.4 and 5.5) is perhaps attributable to the current attention given to HIV/AIDS relative to the more long-established and familiar diseases. TB and malaria are ranked well behind deaths of children and deaths of mothers in childbirth; the former might have been captured by the World Poll question on deaths from chronic diarrhea, but evidently
126
Section I: Measuring Well-Being in an International Context
were not. Note also that only some of the ranking questions refer explicitly to mortality, and the ‘‘spread’’ questions presumably elicit responses about morbidity or other consequences of the diseases. So perhaps differences are to be expected. But the important point about all of the results is the much greater importance attached to income (poverty and hunger) than attached to disease.
3. Discussion: value of life and subjective well-being
Consider first our findings on the value of life in sub-Saharan Africa, and suppose for the moment that it is appropriate to use the life evaluation measures in this way, an issue to which we will return. Given this, we find very small numbers. The largest estimates are 30 to 40 percent of income, and even those estimates are biased upwards by errors of measurement in income. These compensations refer to annual income for the death of an immediate family member in the past twelve months; we have no information on the required compensation in subsequent years. In a comparable exercise for Britain, using data from 1992 to 2002, Oswald and Powdthavee (2008) estimate compensation for the loss of a family member to be between £200,000 (upper-end estimate for loss of a partner) and £16,000 (lower-end estimate for loss of a sibling) with monetary amounts in 1996 prices. Median earnings in 1997 were approximately £12,500. Viscusi and Aldy (2003) review estimates of the value of a statistical life; these are based on the now-standard methodology, dating back to Rosen’s (1988) formulation, in which a value of life is inferred from the earnings premium that workers receive in riskier jobs. For the United States, their central estimate for the value of a statistical life is $6.8 million for a prime-age worker earning $26,000 a year, or more than 250 times annual earnings. They review comparably based estimates from around the world—though none from Africa—and estimate that the international income elasticity of the value of life is 0.6 to 0.8, which would imply that the ratio of the value of life to income will be higher in lower-income countries. The theoretical concept underlying these estimates is the value of a person’s own life, which is arguably higher than the value of the life of a family member, but they nevertheless provide an indication of the magnitude that is used in the literature and by various government agencies. There is additional, albeit less precise, evidence for a low value of life in Africa. We have already cited the findings on policy priorities from the World Poll and the Afrobarometer surveys. Related evidence comes from the high price-elasticity of demand in Africa for healthcare, with large negative responses to user fees or to small charges for medicine or
Life (Evaluation), HIV/AIDS, and Death in Africa
127
preventative measures (see Easterly, 2009, for a review and discussion). These findings may reflect a lack of understanding of the benefits of Western medicine, or they may reflect a more fundamental adaptation to and acceptance of the high levels of morbidity and mortality that have long been a feature of African history (Iliffe, 1995). Certainly, the World Poll’s findings on mortality rates in childbirth, or from malaria and TB, show that HIV/AIDS is not usually the leading cause of disease compared to other, long-standing scourges. African households have many mechanisms that might help them deal with the consequences of losing family members. Households are large, and there is a great deal of coming and going, particularly in economies that depend heavily on migration, which, not coincidentally, are those most heavily afflicted by the HIV/AIDS epidemic. In such places, even before this latest epidemic, it is not unusual for people to depart for long periods, sometimes never to return. These arguments all support the belief that the value of life in Africa is very low. Yet it is important to clarify exactly what this means. It does mean that Africans are prepared to give up relatively little money in order to prolong the life of immediate relatives, if not their own lives. It does not mean that African lives are worth less than American or European lives, that international health policy should be predicated on that supposition, or that we can assess the level and distribution of international well-being based on these low values. The belief in the relatively low worth of African lives was a feature of imperialism, as documented by historians such as Davis (2001) and Watts (1997), the latter of whom begins his book with an 1835 statement about how little the plague meant to the Egyptians. It is even incorporated into the UNDP’s Human Development Index, which, by adding life expectancy to the logarithm of income, values an additional year of life expectancy in the United States as worth 20 times an additional year in India and nearly 50 times an additional year in Tanzania (Ravallion, 1997). Similarly, the analysis of the global convergence of full income by Becker, Philipson, & Soares (2005) accepts willingness to pay as relevant for international comparisons, though the convergence they document would be much reduced or eliminated if African lives were valued at the low income equivalents found here. Even if we accept our and the other estimates of the value of life as representing people’s own trade-offs between health and income, that does not imply that we must attach the same social value to additional money to all people in the world. Indeed, international agencies, by prioritizing aid to the poor, particularly in Africa, certainly believe that money is worth more to poorer people. One way to think about this has been developed by Fleurbaey (2005), who works in terms of money-metric utility; the approach has been applied to international comparisons by Fleurbaey &
128
Section I: Measuring Well-Being in an International Context
Gaulier (2007). If each individual has a utility function ðh; yÞ, where h is a list of health conditions, and y is income, we can define the quantity y by
ðh ; y Þ ¼ ðh; yÞ
ð1Þ
where h is the list of health states corresponding to perfect health. The difference between y and y is the amount that the person would reduce income to be restored to perfect health, and y is the money-metric utility that captures both current health and current income. Social welfare—here international, or cosmopolitan social welfare—is defined over the indivi dual levels of y , and we can use this social welfare function to calculate, not only the priority in income that should be given to the poor, but also the social value of health interventions directed towards them. Suppose that we write W for the social welfare function, and there is a health innovation , the effect of changing on social welfare takes the form " # X @W X @y @W i @hj ¼ @hj @ @ @yi i j
ð2Þ
where i indexes individuals, and j indexes the health states. The term outside the square brackets on the right-hand side is the social value of money to individual i, which is higher the poorer the individual is—the standard argument for foreign aid—while the weights applied to the derivatives of the health states inside the brackets are each individual’s own willingness to pay for health. The point of this formulation is that each individual’s own monetary evaluation of health is respected in making social judgments, but the overall value of the intervention also depends on the marginal social value of income to each person. So it is entirely rational for international agencies to attach great value to improving health in Africa, even if Africans themselves are prepared to give up relatively little money to do so. This analysis leaves unresolved a number of difficult issues. For example, giving money to Africa would be even better than giving health care, and the assistance intended for health care is likely to be subverted towards poverty reduction and income enhancement by local politicians, even those who are acting in the interests of their constituents. So the low value placed on life by Africans still poses problems for the current refocusing of foreign aid away from support for growth towards support for health care, perhaps because it is more difficult to reach people with cash, or because aid agencies value lives differently than individuals do, or because the methods based on self-reported well-being do not tell us what we want to know, an issue to which we now turn.
Life (Evaluation), HIV/AIDS, and Death in Africa
129
The immediate issue is that we have two different measures of wellbeing: a life-evaluation measure for which the monetary compensation for a death is small, and affect measures, for which the monetary compensation for a death is large. The Ladder question requests an overall evaluation of life; this, or the related question about life satisfaction, is often loosely referred to as ‘‘happiness’’ and has a more plausible claim than momentary feelings or emotions to being a comprehensive measure of individual well-being. Yet the affect measures yield more plausible measures of compensation. If we are to decide between them, or possibly rule both to be incorrect, we need a better understanding of what these measures tell us. The Ladder is an evaluation of life as a whole, affected by momentary experiences and feelings, but distinct from them (Kahneman & Riis, 2005). One interpretation is that the Ladder is a measure of life achievement, in which material success, education, and social standing are the key ingredients. If so, it is easy to imagine why someone who has lost a parent, for example, could be sad and depressed, but would not necessarily downgrade his or her sense of achievement in life, though we would hardly expect this to be true for the death of a partner or of a child. It is even possible that a sense of dealing well with the misfortune might lead to an improvement in life evaluation; it is possible to wake up in the morning feeling depressed, but still believe that one’s life as a whole is going well (Annas, 2004). If this argument is accepted, neither life evaluation nor life satisfaction measures, informative though they may be, are useful for calculating the compensation for emotional distress. To quote Annas, ‘‘if you rush to look for empirical measures of an unanalyzed ‘subjective’ phenomenon, the result will be confusion and banality.’’ Here the ‘‘banality’’ is our finding that the loss of an immediate family member makes people sad, while the ‘‘confusion’’ is that this sort of unhappiness is the same thing as ‘‘happiness’’ measured as life evaluation. A reasonable position is one in which both life evaluation and affect are both components of well-being, without having an exclusive claim either separately or together; it is good to have a sense of achievement, and it is good not to be depressed, but other things—such as health—matter, too, even if they are not fully captured in either a sense of achievement or in a lack of depression. Any argument for focusing on either affect or life evaluation would also need to deal with the imperfections of each. Affect measures are subject to adaptation and are easily influenced by trivial features of the situation, while life evaluations often misremember the affective content of past episodes (Kahneman, Wakker, & Sarin, 1997). We are surely on safer ground if we take a capability approach, through which we value health, or income, or other things by the opportunities for freedom that they provide
130
Section I: Measuring Well-Being in an International Context
(Sen, 2001). Improving health extends capabilities, even if those capabilities are not adequately captured by self-reported well-being.
Data Notes
This analysis uses data from the 2006 and 2007 waves of the Gallup World Poll in conjunction with data from the Demographic and Health Surveys (DHS). The DHS data come from the following country-years: Burkina Faso (2003), Cameroon (2004), Ethiopia (2005), Ghana (2003), Guinea (2005), Kenya (2003), Malawi (2004), Mali (2001), Niger (2006), Rwanda (2005), Senegal (2005), Tanzania (2003), Zambia (2001/2002), and Zimbabwe (2005/2006). In some robustness checks, we also use data from Cambodia (2005), the Dominican Republic (2002), Haiti (2005), and India (2005/2006). The data from the following countries are from preliminary releases of the data: Cambodia, Ethiopia, Haiti, India, Niger, Senegal, Tanzania, and Zimbabwe. The DHS are nationally representative household surveys that are available from ORC Macro (http://www.measuredhs.com). DHS surveys include a household questionnaire, a women’s questionnaire, and a men’s questionnaire. An exception is the 2003 DHS for Tanzania (also referred to as the HIV/AIDS Indicator Survey [AIS], which covers only mainland Tanzania); this survey has an individual questionnaire (administered to both men and women), rather than separate men’s and women’s questionnaires. In the fourteen cross-sections in our sample, the survey also includes an HIV test. The household questionnaire and women’s questionnaire are administered to all households responding to the survey; the men’s questionnaire and HIV test are, in some countries, administered to only a sub-sample of households. Results of HIV testing can be linked to individual survey responses, except in Mali and Zambia. However, the HIV testing component in these countries includes some information about respondents tested, including basic demographic characteristics. Our analysis using DHS data uses three measures of HIV. Though in a few cases these are calculated at the national level, for the most part they are calculated within each country for each five-year age group, separately by sex. Five-year age groups are as follows: 15–19, 20–24, 25–29, 30–34, 35–39, 40–44, 45–49. In some countries, the DHS collects HIV test results from men (but not women) aged 50–59, but for consistency across countries (as well as across genders), we restrict the analysis to adults 15–49. ‘‘HIV Prevalence’’ is the fraction of adults infected with HIV (among those tested), using results from the HIV testing component. ‘‘HIV Knowledge’’ is the fraction of DHS respondents who say that they know someone who has AIDS or has died of AIDS. This question was asked only of those who said that they had heard of AIDS; the fraction
Life (Evaluation), HIV/AIDS, and Death in Africa
131
is calculated among those who had heard of AIDS. Responses are drawn from the women’s and men’s questionnaires; this question was not asked in Tanzania. ‘‘HIV Risk’’ is the fraction of DHS respondents who said that they were at moderate or greater risk of getting AIDS (including those who said that they have AIDS). This question was asked only of those who said that they had heard of AIDS; the fraction is calculated among those who had heard of AIDS. Responses are drawn from the women’s and men’s questionnaires; this question was asked only in Kenya, Tanzania, Zambia, and Zimbabwe. When calculating the prevalence of HIV infection, we weight results using DHS-provided HIV sample weights. When calculating the fractions with HIV Knowledge and HIV Risk, we weight results using DHS-provided individual sample weights. Data on GDP per capita are for 2005, and come from the latest round of the International Comparison Program, World Bank (2008). References Annas, J. (2004). Happiness as achievement. Daedalus, 133(2), 44–51. Becker, G. S., Philipson, T. J., & Soares, R. R. (2005). The quantity and quality of life and the evolution of world inequality. American Economic Review, 95(1), 277–291. Blanchflower, D. G., & Oswald, A. J. (2008). Is well-being U-shaped over the lifecycle? Social Science and Medicine, 66, 1733–1749. Cantril, H. (1965). The pattern of human concern. New Brunswick, NJ: Rutgers University Press. Davis, M. (2001). Late Victorian holocausts: El Nin˜o famines and the making of the third world. New York: Verso. Deaton, A. (2008). Income, health, and well-being around the world: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22(2), 53–72. Dolan, P., & Metcalfe, R. (2008). Comparing willingness-to-pay and subjective wellbeing in the context of non-market goods. CEP Discussion Paper 890, LSE, London, mimeograph (October). Easterly, W. (2009). Can the West save Africa? Journal of Economic Literature, 47 (2), 373–443. Fleurbaey, M. (2005). Health, wealth, and fairness. Journal of Public Economic Theory, 7(2), 253–284. Fleurbaey, M., & Gaulier, G. (2007). International comparisons of living standards by equivalent incomes. CEPII Working Paper No. 2007–03, January. Fortson, J. G. (2008). The gradient in sub-Saharan Africa: Socioeconomic status and HIV/AIDS. Demography, 45(2), 303–322. Helliwell, J. F. (2003). How’s life? Combining individual and national variables to explain subjective well-being. Economic Modelling, 20(2), 331–360. Iliffe, J. (1995). Africans: The history of a continent. Cambridge, UK: Cambridge University Press. Kahneman, D., Wakker, P. P., & Sarin, R. (1997). Back to Bentham? Explorations of experienced utility. Quarterly Journal of Economics, 112(2), 375–405.
132
Section I: Measuring Well-Being in an International Context
Kahneman, D., & Riis, J. (2005). Living, and thinking about it: Two perspectives on life. In F. A. Huppert, N. Baylis, & B. Keverne (Eds.), The science of well-being (pp. 285–304). London: Oxford University Press. Kharas, H. (2008). A reality check on African aid. The Brookings Institution, February 20, http://www.brookings.edu/opinions/2008/0220_african_aid_kharas.aspx, accessed on November 17, 2008. Layard, R. (2005). Happiness: Lessons from a new science. New York: Penguin. Oswald, A. J., & Powdthavee, N. (2008). Death, happiness, and the calculation of compensatory damages. Journal of Legal Studies, 37(S2), S217-S251. Ravallion, M. (1997). Good and bad growth: The human development reports. World Development, 25(5), 631–638. Rosen, S. (1988). The value of changes in life expectancy. Journal of Risk and Uncertainty, 1(3), 285–304. Sen, A. (1999). Development as freedom. New York: Knopf. Stevenson, B., & Wolfers, J. (2008). Economic growth and subjective well-being: Reassessing the Easterlin paradox. Brookings Papers on Economic Activity, No. 1, Spring 2008, 1–87. Sunstein, C. R. (2004). Valuing life: A plea for disaggregation. Duke Law Journal, 54, 385–445. Tortora, R. D. (2008). Sub-Saharan Africans Rank the Millennium Development Goals (MDGs). Gallup. http://www.gallup.com/poll/116431/Research-Reports.aspx. UNAIDS. (2008). 2008 Report on the Global AIDS Epidemic. Geneva, Switzerland: UNAIDS and WHO. van den Berg, B., & Ferrer-i-Carbonell, A. (2007). Monetary valuation of informal care: The well-being valuation method. Health Economics, 16, 1227–1244. van Praag, B. M. S., & Baarsma, B. E. (2005). Using happiness surveys to value intangibles: The case of airport noise. Economic Journal, 115, 224–246. Viscusi, W. K, & Aldy, J. E. (2003). The value of a statistical life: A critical review of market estimates throughout the world. Journal of Risk and Uncertainty, 27(1), 5–76. Watts, S. (1997). Epidemics and history: Disease, power and imperialism. New Haven: Yale University Press. Welz, T., Hosegood, V., Jaffar, S., Ba¨tzing-Feigenbaum, J., Herbst, K., & Newell, M. (2007). Continued very high prevalence of HIV infection in rural KwaZulu-Natal, South Africa: A population-based longitudinal study. AIDS, 21(11), 1467–1472. World Bank. (2008). Global purchasing power parities and real expenditures. Washington, DC.
Appendix A
TABLE 5.A.1
Life Evaluation and HIV Knowledge
Ladder HIV Knows ln GDP
(1) 0.079 (0.225) 0.410 (1.815)
(2)
(3)
0.012 0.958* (0.130) (3.524) 0.421* (5.740)
Female
(4)
No No
No No
Yes No
R2 obs countries
0.264 13 13
0.010 0.037 19,973 19,973 13 13
(6)
1.264* (4.818)
1.226* (4.189)
0.544 (1.606)
0.119* (3.361)
0.110* (2.565) 0.426* (15.211)
0.074 (1.706) 0.426* (15.083)
ln y Country FEs? Age group FEs?
(5)
Yes No
Yes No
Yes Yes
0.038 19,973 13
0.084 13,056 11
0.085 13,056 11
Notes: In all columns, the sample is restricted to adult respondents aged 15–49 and results are weighted using provided sample weights. The dependent variable is the Ladder on a scale from 0 (‘‘the worst possible life’’) to 10 (‘‘the best possible life’’). In column 1, the dependent variable is the country–level weighted average of the Ladder. In columns 2–6, the dependent variable is the individual Ladder. HIV Knows is country–level HIV knowledge among adults 15–49 in column 1 and country/sex/age group-level knowledge in columns 2–6. Knowledge is the fraction of DHS respondents who say that they know someone who has AIDS or has died of AIDS. ln GDP is the log of country-level GDP per capita in 2005, as measured by the International Comparison Program, World Bank (2008). ln y is the log of family income, using individual responses from the Gallup survey. In columns 2–6, standard errors are clustered at the country/sex/age group-level. Absolute values of t–statistics are in parentheses. * p < 0.05.
133
TABLE 5.A.2
Emotions and HIV Knowledge
HIV Knows Female ln y
(1) Enjoyment
(2) Smiling
(3) Sadness
0.002 (0.017) 0.005 (0.401) 0.060* (10.770)
0.207* (2.033) 0.002 (0.122) 0.051* (8.845)
0.005 (0.058) 0.007 (0.645) 0.032* (6.788)
(4) Depression 0.065 (1.023) 0.005 (0.568) 0.020* (4.566)
Country FEs? Age group FEs?
Yes Yes
Yes Yes
Yes Yes
Yes Yes
R2 obs countries
0.037 13,019 11
0.029 12,918 11
0.019 13,015 11
0.099 12,986 11
Notes: In all columns, the sample is restricted to adult respondents aged 15–49 and results are weighted using provided sample weights. In column 1, the dependent variable is an indicator for whether, on previous day, the respondent experienced enjoyment a lot of the day. In column 2, the dependent variable is an indicator for whether, on previous day, the respondent smiled and laughed a lot of the day. In column 3, the dependent variable is an indicator for whether, on previous day, the respondent experienced sadness a lot of the day. In column 4, the dependent variable is an indicator for whether, on previous day, the respondent experienced depression a lot of the day. HIV Knows is country/sex/age grouplevel knowledge in all columns. Knowledge is the fraction of DHS respondents who say that they know someone who has AIDS or has died of AIDS. ln y is the log of family income, using individual responses from the Gallup survey. In all columns, standard errors are clustered at the country/sex/age group-level. Absolute values of t–statistics are in parentheses. * p < 0.05.
134
Appendix B
TABLE 5.B.1
Life Evaluation and HIV Risk
Ladder
(1)
(2)
(3)
(4)
(5)
(6)
HIV Risk
1.155 (3.574) 0.746 (7.132)
0.467 (1.698) 0.791* (6.355)
0.960* (2.245)
1.124* (2.658)
1.242 (1.951)
1.321 (1.135)
0.065 (0.815)
0.099 (1.064) 0.374* (7.849)
0.106 (0.869) 0.376* (7.898)
ln GDP Female ln y Country FEs? Age group FEs?
No No
No No
Yes No
Yes No
Yes No
Yes Yes
R2 obs countries
0.986 4 4
0.024 6,894 4
0.030 6,894 4
0.030 6,894 4
0.079 5,264 4
0.082 5,264 4
Notes: In all columns, the sample is restricted to adult respondents aged 15–49 and results are weighted using provided sample weights. The dependent variable is the Ladder, on a scale from 0 (‘‘the worst possible life’’) to 10 (‘‘the best possible life’’). In column 1, the dependent variable is the country–level weighted average of the Ladder. In columns 2–6, the dependent variable is the individual Ladder. HIV Risk is country–level HIV risk among adults 15–49 in column 1 and country/sex/age group-level risk in columns 2–6. Risk is the fraction of DHS respondents who say that they are at moderate or higher risk of being infected with HIV. ln GDP is the log of country-level GDP per capita in 2005, as measured by the International Comparison Program, World Bank (2008). ln y is the log of family income, using individual responses from the Gallup survey. In columns 2–6, standard errors are clustered at the country/sex/age group-level. Absolute values of t–statistics are in parentheses. * p < 0.05.
135
TABLE 5.B.2
Emotions and HIV Risk (1) Enjoyment
HIV Risk Female ln y
0.346 (1.774) 0.015 (0.728) 0.049* (6.497)
(2) Smiling
(3) Sadness
0.206 (0.958) 0.016 (0.659) 0.050* (6.897)
0.072 (0.476) 0.014 (0.803) 0.024* (3.274)
(4) Depression 0.147 (0.995) 0.013 (0.726) 0.023* (3.899)
Country FEs? Age group FEs?
Yes Yes
Yes Yes
Yes Yes
Yes Yes
R2 obs countries
0.024 5,253 4
0.023 5,206 4
0.019 5,242 4
0.026 5,229 4
Notes: In all columns, the sample is restricted to adult respondents aged 15–49 and results are weighted using provided sample weights. In column 1, the dependent variable is an indicator for whether, on previous day, the respondent experienced enjoyment a lot of the day. In column 2, the dependent variable is an indicator for whether, on previous day, the respondent smiled and laughed a lot of the day. In column 3, the dependent variable is an indicator for whether, on previous day, the respondent experienced sadness a lot of the day. In column 4, the dependent variable is an indicator for whether, on previous day, the respondent experienced depression a lot of the day. HIV Risk is country/sex/age grouplevel risk in all columns. Risk is the fraction of DHS respondents who say that they are at moderate or higher risk of being infected with HIV. ln y is the log of family income, using individual responses from the Gallup survey. In all columns, standard errors are clustered at the country/sex/age group-level. Absolute values of t–statistics are in parentheses. * p < 0.05.
136
Section II
International Comparisons of Income and Well-Being Through Time
This page intentionally left blank
Chapter 6 Does Relative Income Matter? Are the Critics Right? R. Layard, G. Mayraz, and S. Nickell1
In the United States, happiness has been roughly constant since the early 1950s, despite massive income growth. The same is true in Europe’s leading economy, (West) Germany, since their records began in the 1970s. At the same time, evidence has accumulated that people in advanced countries care a lot about how their income compares to that of other people. This has led to the following hypothesis about advanced countries: (i) Higher individual income produces a substantial increase in individual happiness, other things equal. (ii) Higher average income in a country produces much less of an increase in average happiness, and (iii) Individuals care greatly about their income relative to others’ (which makes [i] and [ii] consistent with each other). One could well call this the Easterlin hypothesis,2 but we shall not do so in order to emphasize that we are only concerned with the advanced countries of North America and Western Europe. What people in these countries want to know is: How far is general income growth (beyond income levels already achieved) likely to increase average happiness? This is a question about time series relationships. Even with time series evidence it is not easy to isolate the causal effects of income growth. But with cross-sectional evidence it is even more difficult to do so, since income differences across countries are so highly correlated with 139
140
Section II: Comparisons of Income and Well-Being Through Time
differences in institutions and cultures.3 If we use time-series data, it is much easier to control for these other influences on happiness. So this chapter is about the time series in advanced countries. Fortunately, there is enough evidence about these countries for us to make well-founded statements about the income–happiness relationship there, without bringing in evidence about poorer countries where the relationship may be very different. Despite this, the hypothesis we began with has been sharply criticized in two interesting and influential papers by Deaton (2008) and Stevenson/Wolfers (2008), in which the main weight of evidence comes from countries poorer than those in North America and Western Europe and in which much of the evidence is cross-sectional. The conclusion of these studies is summed up by Stevenson and Wolfers as follows: ‘‘Our evidence is consistent with the view that only absolute income matters to happiness. Indeed, whereas previous analyses of the link between income and happiness suggested a prima facie case for relative income playing a dominant role, our updated analysis finds no such case.’’4 The purpose of this chapter is to present the evidence (mainly timeseries) that relative income matters in advanced countries, and finally, to ask whether the additional evidence presented by Deaton and Stevenson/ Wolfers is sufficient to outweigh the other evidence. In what follows we therefore present our own analysis of the United States, West Germany and other West European countries. We also review other evidence in favor of the relative income hypothesis. Finally, we turn to Deaton and Stevenson/Wolfers, and we conclude that their evidence is too weak to overthrow the weight of other evidence. Throughout this chapter we standardize the measures of happiness/lifesatisfaction, so that they run from 0 to 10 (with intermediate replies equally spaced in between). This is the procedure we adopted in our previous crosssectional study of the effects of individual income on individual happiness within countries. From a wide range of population surveys we found very similar coefficients, with a unit rise in log income raising happiness by 0.6 units on average.5 This is the standard against which we can compare country-level time-series effects. For advanced countries they are very much lower, as we shall see.
United States
We begin with the United States. The United States is not of course just ‘‘one data point’’ in a large number of countries.6 It includes 50 states, many of them larger than separate countries. It is a large proportion of the
Does Relative Income Matter? Are the Critics Right?
141
advanced world, and people in other countries look to it as important evidence about their own futures. In social science, U.S. evidence alone is regularly treated as sufficient for developing major propositions.
Aggregate data, 1950 to –now
Happiness in the United States has not risen since 1950, despite massive income growth. From 1950 to 1973 (a period not covered by Stevenson/ Wolfers), income growth was especially rapid (Figure 6.1). Yet happiness failed to rise over that period, and has not risen since.7 The happiness data are shown in Figure 6.2. From 1950 to 1970, we have nine observations from the American Institute of Public Opinion (Gallup), which asked the question: ‘‘In general, how happy would you say you are—very happy, fairly happy, or not too happy?’’ From 1963 to 1976, we have seven observations from the National Opinion Research Center (NORC) which asked the question: ‘‘Taken all together, how would you say things are these days? Would you say that you are very happy, pretty happy, or not too happy?’’
9
Log real GDP per capita 9.5 10
10.5
Real GDP Per Capita in the United States
1950
1960
1970
1980 Year
1990
2000
Series shows the log of real GDP per capita (chain series) in 2000 constant prices. Source: Penn World Tables 6.2
Figure 6.1 US GDP per capita over time.
Section II: Comparisons of Income and Well-Being Through Time
6
6.5
7
7.5
Reported happiness over time in the United States
5.5
Mean happiness in survey (0 10 scale)
142
1950
1960
1970
1980 Year
AIPO (1950–1970) GSS (1972–2006)
1990
2000
NORC (1963–1976)
AIPO is the American Institute of Public Opinion (later Gallup). NORC is the National Opinion Research Center, and administrator of GSS. GSS is the General Social Survey. Original questions asked on a 3-point scale.
Figure 6.2 Reported happiness in the United States: 1950–2006. The surveys shown were chosen to cover the entire period with no gaps. Other surveys with a long time series, such as Gallup Ladder of Life person to person interviews and Gallup Ladder of Life telephone interviews show a comparable trend.
And from 1972 to 2006, we have 26 observations from the General Social Survey (GSS) which asked the same question as NORC.8 As throughout this paper, we score the data by treating the highest answer as 10, the lowest as 0, and the others as equally spaced in between. The figure shows the average score in each year. For all three surveys there is a downward time trend. When considering the period since 1973, Stevenson/Wolfers quite rightly draw attention to the growing inequality in that period (see Figure 6.3) which means that the average of log income grows more slowly than the log of average income. But this is not sufficient to explain why average happiness is declining. To examine this, the obvious strategy is to look at the richest two quintiles of income where income rose substantially: even they experienced no rise in happiness (see Table 6.1 and Figure 6.4). Moreover, in the earlier period (1950–1970), inequality was falling, and yet happiness failed to rise. From this analysis one concludes that growing national income is not a sufficient condition for happiness to grow. One cannot say that growing
Does Relative Income Matter? Are the Critics Right?
143
.42 .4 .38 .36 .34
Gini coefficient for families
.44
Family income inequality over time
1950
1960
1970
1980
1990
2000
Year Source: U.S. Census Bureau. Table F-4.
Figure 6.3 US Gini coefficient over time.
TABLE 6.1 Time trend of happiness and of log real income in the United States, by income quintile (from high to low). Quintile
Avg. happiness trend
Log real income trend
1 2 3 4 5
.0003 (.0035) .0010 (.0034) .0100 (.0033) .0069 (.0035) .0165 (.0037)
.0140 (.0004) .0079 (.0002) .0059 (.0001) .0057 (.0002) .0012 (.0007)
Combined
.0059 (.0016)
.0082 (.0004)
Source: General Social Survey, 1973–2006. Trends reported are the coefficient on the year in a regression of average reported happiness (log income) against year with no other controls. Income is adjusted for the number of adults in the household. Standard errors are in parentheses. Sample size: 40,012
income had no positive effect on happiness—it might have been outweighed by other negatives9. But, in reflecting on why growing income had so little effect, it is reasonable to ask if one reason was the importance of relative income.
5
Average happiness 5.5 6 6.5
7
Average happiness by income quintile
1970
1980
1990 2000 gss year for this respondent Quintile 1 Quintile 3 Quintile 5
2010
Quintile 2 Quintile 4
11 10.5 10 9.5 9
Average log income
11.5
Average log income by income quintile
1970
1980
1990 2000 gss year for this respondent Quintile 1 Quintile 3 Quintile 5
2010
Quintile 2 Quintile 4
Source: GSS 1972–2006. Income is adjusted for the number of adults in the household.
Figure 6.4 Average happiness and average log real income in the United States by income quintile (from high to low). Source: General Social Survey, 1972–2006. Income is adjusted for the number of adults in the household.
144
Does Relative Income Matter? Are the Critics Right?
145
Individual data, 1972 to now
To investigate this hypothesis, we can use the individual data from the General Social Survey over the years since 1972.10 For each individual we have one observation that tells us, inter alia: happiness (very happy ¼ 10, pretty happy ¼ 5, not too happy ¼ 0) household real income per adult11 average Y in same household type in same year relative income as perceived by the individual12 (well above average ¼ 10, above average ¼ 7.5, average ¼ 5, below average ¼ 2.5, well below average ¼ 0) X age, age2, sex, marital status, employment status
H Y Y R
To clarify the analysis, we confine it to whites aged 30–55; and, to reduce the effects of measurement error, we omit observations where reported income is in the top and bottom 5% of the range.13 Controls (X) are always included. We then use these data to explain individual happiness (see Table 6.2). First we include only the log of income (plus controls). It attracts a coefficient of 0.58. As we have said, this is the typical size of coefficient that emerges in the average country from a cross-sectional regression of happiness or
TABLE 6.2 Regressions for explaining individual happiness in the United States using the General Social Survey (N ¼ 14,836). The dependent variable is reported happiness scaled to a 0–10 range. Years: 1972–2006. Y is real income per adult,Y average real income per adult in the year in observations matched for household type, R is financial situation relative to others (see text) scaled to the 0–10 range. Robust standard errors in parentheses. Regressions restricted to whites aged 30–55. Observations with extreme incomes were removed (see text for details). Controls include sex, quadratic in age and education, and marital and work status dummies. In column (3) the beta coefficients are 0.06 on Y, 0.10 on R, and 0.02 on Y
logY
(1)
(2)
(3)
0.565 (0.055)
0.578 (0.055) 0.690 (0.243)
0.330 (0.062) 0.486 (0.244) 0.150 (0.016)
0.075
0.075
0.082
logY R
Rsq
146
Section II: Comparisons of Income and Well-Being Through Time
life-satisfaction upon log income (when happiness/life satisfaction is scaled to the range 0–10). This is a big effect. But what happens if we now introduce comparator income, measured by the average income in the same year in the same type of household? Log comparator income has an effect of 0.69. In other words relative income has an effect of 0.69 on life satisfaction, and the estimated effect of absolute income is negative. Even if a time trend is introduced into the equation, the effect of absolute income is barely above zero.14 Regression results can of course be a fluke, and we know little about exactly how people estimate comparator income. But fortunately the GSS asks respondents another key question—about how they perceive their position in the income distribution (well above average, above average, average, below average, or well below average). This adds enormously to the plausibility of the analysis because it directly measures cognition. When this variable (R) is introduced (column 3), it proves highly significant. This is important evidence in favor of the view that individual happiness really does depend on relative income as people experience it. And at the same time as it is introduced, the estimated effect of absolute income becomes even more negative.15
Financial satisfaction
This gives us a coherent understanding of the overall U.S. time series (even though we do not have time series for individuals).16 This understanding is enhanced when we look at specifically economic sources of life-satisfaction. In the GSS, respondents are asked ‘‘Would you say you are pretty well satisfied with your present financial situation, more or less satisfied, or not satisfied at all?’’ If people cared only about absolute income, they would have become more satisfied with their financial position as they became richer. But, in fact, the level of financial satisfaction has fallen somewhat (see Figure 6.5). Regression analysis shows that one reason for this is, again, the powerful influence of perceived relative income. As Table 6.3 shows, the respondents’ own income has a strong influence on their financial satisfaction. But when comparator income Y is introduced in Column (2), it becomes clear that comparator income is a major part of the story. If we go further and introduce perceived relative income (R), the position becomes even more clear. In this analysis the beta-coefficient on R is as large as that on absolute income. This is remarkable when one considers how crude the 5-point measure of perceived relative income is, compared with the measurement of absolute income.
4.5
5
5.5
6
6.5
Happiness and Financial Satisfaction in the United States
1970
1980
1990 Year
Happiness
2000
2010
Financial satisfaction
Series shows the mean of (1) reported happiness and (2) financial satisfaction, both on a 0–10 scale. Source: General Social Survey 1972–2006.
Figure 6.5 Happiness and Financial satisfaction in the United States.
TABLE 6.3 Regressions for explaining individual financial satisfaction in the United States using the General Social Survey (N ¼ 14,756). The dependent variable is reported financial satisfaction scaled to a 0–10 range. Years: 1972–2006. Y is real income per adult, Y average real income per adult in the year in observations matched for household type, R is financial situation relative to others (see text) scaled to the 0–10 range. Robust standard errors in parentheses. Regressions restricted to whites aged 30–55. Observations with extreme incomes were removed (see text for details). Controls include sex, quadratic in age and education, and marital and work status dummies. In column (3) the beta coefficients are 0.23 on Y, 0.26 on R, and 0.03 on Y
logY
(1)
(2)
(3)
2.17 (0.062)
2.20 (0.062) 1.552 (0.279)
1.403 (0.071) 0.965 (0.271) 0.480 (0.019)
0.157
0.159
0.207
logY R
Rsq
147
148
Section II: Comparisons of Income and Well-Being Through Time
100
120
140
160
Actual and required real income
1950
1960 Actual real income
1970 year
1980
1990
Required real income
Figure 6.6 Actual and required real income per head in the United States (1952=100).
We can end the U.S. story with a famous Gallup finding. For many years Gallup asked their respondents, ‘‘What is the smallest amount of money a family of four needs to get along in this community?’’ As Figure 6.6 shows, the answers moved clearly in line with average real incomes in the community. This supports the idea that people have a mental concept of an income norm, which rises with general levels of living standards.
West Germany
We can turn now to the leading European economy, which also has the advantage of a longitudinal panel beginning in 1984 and providing our study with 81,000 person-year observations.17 Figure 6.7 shows average life satisfaction among those aged 30–55 in the panel, together with average life satisfaction for all adults in Eurobarometer. As can be seen, there has been no rise in average life satisfaction in either survey. In the meantime, real income rose sharply (Figure 6.8). To estimate the determinants of life satisfaction at the individual level, we first proceeded exactly as we did with the GSS, ignoring the panel features of the data. The comparator income was taken as that in the same age/sex/education group.18 The results are in Table 6.4. In columns 1 and 2,
7 6.5 6
Mean reported life satisfaction
7.5
Reported life satisfaction in West Germany
1970
1980
1990 Year
Eurobarometer 1973–2007
2000
2010
GSOEP 1985–2006
Source: Eurobarometer and German Socio–Economic Panel. Mean life satisfaction reported on a 0–10 scale.
Figure 6.7 Average life satisfaction in West Germany.
10 9.8 9.6 9.4
Log real GDP per capita
10.2
Log real GDP per capita in Germany
1970
1980
1990 Year
Source: OECD.
Figure 6.8 Log real GDP per capita in Germany. 149
2000
2010
150
Section II: Comparisons of Income and Well-Being Through Time
we repeated the analyses in the first two columns of the U.S. table, with remarkably similar results. The effect of absolute log income taken on its own was once again around 0.5, but it became negative when relative income was introduced. Even if a time trend was introduced, the effect of relative income remained twice as large as that of absolute income.19 But these estimates are vulnerable to other possible unmeasured characteristics of individuals that are correlated with income and with life satisfaction. We can deal with this problem by exploiting the panel features of the data and including both an individual fixed effect and state/year dummies to explain each observation. The results are shown in columns 3 and 4 of Table 6.4. When log income is included on its own, it attracts a reduced coefficient of 0.33—lower than the original cross-sectional estimate but still quite high. However, when comparator income is introduced, relative income has a coefficient of 0.29, while the effect of absolute
TABLE 6.4 Regressions for explaining individual life satisfaction in West Germany using the German Socio-Economic Panel (N ¼ 90279 for the first 4 regressions, 56240 for the last two). The dependent variable is reported life satisfaction on a 0–10 scale. Y is real net household income, Y is average real household income in year in observations matched for sex, age (–5 years), and education (3 groups). Robust standard erorrs in parentheses. Regressions restricted to non-immigrant West Germans aged 30–55. Controls for regular regressions include sex, quadratic in age and education, marital and work status, and state dummies. Controls for panel regressions include person fixed effects, a quadratic in age and education, marital and work status, and state/year combination dummies. No fixed effects
logY
Fixed effects
(1)
(2)
(3)
0.501 (0.014)
0.527 (0.015)
0.326 (0.022)
0.648 (0.072)
logY
(4) 0.329 (0.022)
log Y t 2 log Y t 3 0.077
0.078
(6)
0.321 (0.029)
0.324 (0.029)
0.291 (0.134)
log Y t 1
Rsq
(5)
0.572
0.572
0.331 (0.169) 0.012
0.011
(0.028)
(0.028)
0.005
0.006
(0.028)
(0.028)
0.059
0.057
(0.027)
(0.027)
0.594
0.594
Does Relative Income Matter? Are the Critics Right?
151
income falls to 0.04 (0.33 minus 0.29). This seems to provide a good explanation of why average happiness has not risen in response to rises in absolute income.
Adaptation versus social comparisons
There is another possible explanation, however—adaptation. Psychologists in particular tend to favor this explanation. We can use the German panel data to examine the size of the effects of adaptation and social comparison. This is done in columns 5 and 6 of the table. Column 5 ignores the possible effects of social comparisons. It shows that, if we include lagged income over the previous three years, the three lagged income terms have a total effect of –0.07, implying that the long-run effect of own income is about four fifths of the short-term effect.20 This provides a small contribution to explaining why happiness has not risen as income has risen. But when we also introduce comparator income (in column 6) we get a complete explanation. The negative effect of comparator income is equal in size to the positive effect of own income. Adaptation provides a small additional contribution, but that is all.
Other Evidence on the Importance of Relative Income in Advanced Countries
Thus, for the two largest economies of North America and Western Europe, there is strong support for the importance of relative income. Such findings are not of course new.21 In studying the United States, Luttmer (2005) used a different survey, the National Survey of Families and Households, to estimate a happiness equation for U.S. couples. The explanatory variables were household income and the average earnings in the local area (localities averaging around 100,000 population). The equation when estimated in cross-section was H ¼ 0:12 log Y 0:24 log Y þ controls ð0:02Þ
ð:07Þ
Like our Table 6.2 results for the United States, this shows the key importance of relative income.22 Luttmer also showed that a unit increase in log Y raised hours of work by around two hours a week—consistent with the view that higher Y reduces utility. In Canada a similar cross-sectional micro-study included as the comparator the average income in the same census tract. It found that relative
152
Section II: Comparisons of Income and Well-Being Through Time
income was the only significant income variable affecting life-satisfaction.23 Likewise, in Germany there have been at least two studies of the German Socio-Economic Panel (GSOEP) similar to ours, with similar findings of relative income effects.24 Required income
One would like to be sure that these findings were not statistical flukes, but corresponded to some inner psychological process. The GSOEP is very helpful here because it asks people about what income they consider sufficient. The question is: ‘‘Whether you feel an income is good, or not so good, depends on your present life circumstances and expectations. In your case, what net household income is just sufficient income?’’ Let us call this the individual’s required income. Using fixed effects, Frey & Stutzer (2005) showed that life satisfaction depended on the ratio of income to required income, with no independent role for absolute income. A similar result was obtained in Switzerland, using cross-sectional data (Stutzer, 2004). This raises the question of what causes these income requirements. This is a well-established field going back to the work of van Praag, who pioneered the study of what individuals consider to be a sufficient income.25 In Figure 6.5, we showed Gallup estimates of required income that followed quite closely the course of absolute income. But Kapteyn and his colleagues were the first to show econometrically that required income (thus defined) is powerfully affected by generally prevailing levels of income.26 Their work used a Dutch panel survey of households. Stutzer (2004) found similar effects from the Swiss Poverty Study (SPS)—with required income responding to average income with an elasticity of one half. We can then combine this Swiss evidence with the happiness equation quoted earlier, in order to obtain the reduced form effect of average income upon individual happiness. By doing this, we can see once more that relative income is having a powerful impact on happiness. Hypothetical questions
A final type of evidence comes from asking people hypothetical questions. Solnick & Hemenway (1998) asked a group of Harvard graduate students to choose between living in two imaginary worlds in which prices were the same: A. You get $50k a year, while others get $25k. B. You get $100k and others get $200k.
Does Relative Income Matter? Are the Critics Right?
153
A majority of respondents preferred the first type of world.27 This only makes sense if relative income is more important to them than absolute income.
The Critics
Given all this evidence, it is not surprising that there has been widespread support for the hypothesis of this chapter. But the two recent studies by Deaton and Stevenson/Wolfers are thought by some to have disproved the hypothesis. How far should they make one change one’s mind? Deaton’s analysis is fascinating but purely cross-sectional. It covers 132 countries, from the poorest to the richest, and finds a strong relationship between average life satisfaction and log average incomes, with a coefficient of 0.84, which is at least as high as the within-country effect of log income at the individual level.28 However, if the analysis is confined to advanced countries (with income per head above $20,000), the estimated coefficient is 0.38, and the coefficient is much smaller than its standard error (0.78). As Deaton says, this ‘‘is consistent both with a true slope of zero and a slope that is the same or higher than the low-income countries’.’’ One might think that was as much as could be said, but Deaton argues that ‘‘the latter is the natural conclusion. These results support a finding that the relationship between the log of income and life satisfaction offers a reasonable fit for all countries, whether high-income or low-income, and if there is any evidence for deviation, it is small and probably in the direction of the slope being higher among the high-income countries.’’ It is difficult to see how this could be the natural conclusion unless one had a strong prior assumption that all countries work the same way, whether rich or poor, and that any evidence from outside the study should be disregarded, such as that just quoted showing the difference (in advanced countries) between effects across individuals and across country aggregates. But the weakest feature of the Deaton evidence from a policy point of view is its cross-sectional nature. Institutional factors correlated with income play a major role in these cross-sectional country differences.29 So we get little evidence from the cross-section about causality, which is what the public debate requires. By contrast, the exhaustive study by Stevenson/Wolfers covers time-series as well as cross-sectional data, again for countries from the poorest to the richest. They conclude that ‘‘the relationship between subjective well-being and income within a country (that is, contrasting the happiness of rich and poor members of a society) is similar to that seen across countries (contrasting rich
154
Section II: Comparisons of Income and Well-Being Through Time
and poor countries), which in turn is similar to the time series relationship (comparing the happiness of a country as it gets richer or poorer).’’ This finding, if true, implies that relative income has no effect on happiness. But does this finding apply to the advanced countries taken on their own? Most of the analyses in Stevenson/Wolfers are for all countries, from the poorest to the richest.30 But they do give separate time-series analyses for the United States, Japan, and Europe. They described the United States as ‘‘a data point supporting the Easterlin Paradox,’’ but argued that ‘‘it should be regarded as an interesting exception warranting further scrutiny.’’ In order to help explain U.S. experience since 1972, they invoked increased inequality, and, once they had done this, they concluded that ‘‘the U.S. experience could be roughly consistent with the accumulated evidence of a robust happiness–income link.’’ As we have seen, this explanation of the period after 1972 does not work, and it works even worse in the period from 1950 to 1972, when economic growth was strongest and inequality fell. For Japan, their data showed a strong positive impact of income on happiness until income reached around $20,000 a head, at which point the impact became negative.31 But, for Europe, they have an important point. The Eurobarometer data beginning in 1972 show rising happiness in a number of West European countries, though falling in a few others. In their statistical analysis of the average effect of income in Europe, they included the countries of the former Warsaw Pact, where it is clear that other factors besides income are changing rapidly. To understand what is happening, we should confine ourselves to countries that were never Communist. And to understand the impact of income we should distinguish between the cyclical and long-run changes in income. This we do in the next section.32
Western Europe
Our analysis relates to 1973–2007 and covers 16 countries. Figure 6.9 show the basic data from Eurobarometer.33 The dependent variable is average life satisfaction in each country and year.34 Average income in each year we split into two components—a cyclical component (YC ) and a trend component (YT ), using the Hodrik/Prescott filter. We also include unemployment and inflation, which are well-documented factors affecting the time-series behavior of life satisfaction.35 There is a fixed effect for each country. The main effect we are interested in is that of trend income, which tells us the long-run effect of higher living standards.
5
6
7
8
9
Austria: 0–.039 (0.015)
1970
1980
1990
2000
2010
2000
2010
2000
2010
5
6
7
8
9
Belgium: 0–.016 (0.006)
1970
1980
1990
5
6
7
8
9
Denmark: 0.017 (0.002)
1970
1980
1990
Figure 6.9 Happiness over time in Western Europe. Titles include country name and regression slope (standard error in parentheses). Source: Eurobarometer 1973–2007. 155
5
6
7
8
9
Finland: 0.046 (0.016)
1970
1980
1990
2000
2010
2000
2010
2000
2010
5
6
7
8
9
France: 0.018 (0.003)
1970
1980
1990
5
6
7
8
9
Germany: 0.001 (0.004)
1970
1980
1990
Figure 6.9 (Continued). 156
5
6
7
8
9
Greece: 0.006 (0.009)
1970
1980
1990
2000
2010
2000
2010
2000
2010
5
6
7
8
9
Ireland: 0.005 (0.006)
1970
1980
1990
5
6
7
8
9
Italy: 0.032 (0.004)
1970
1980
1990
Figure 6.9 (Continued).
157
5
6
7
8
9
Luxembourg: 0.012 (0.003)
1970
1980
1990
2000
2010
5
6
7
8
9
Netherlands: 0.007 (0.003)
1970
1980
1990
2000
2010
2000
2010
5
6
7
8
9
Norway: 0.042 (0.055)
1970
1980
1990
Figure 6.9 (Continued). 158
5
6
7
8
9
Portugal: 0–.012 (0.009)
1970
1980
1990
2000
2010
2000
2010
2000
2010
5
6
7
8
9
Spain: 0.025 (0.009)
1970
1980
1990
5
6
7
8
9
Sweden: 0.023 (0.013)
1970
1980
1990
Figure 6.9 (Continued).
159
160
Section II: Comparisons of Income and Well-Being Through Time
5
6
7
8
9
United Kingdom: 0.007 (0.002)
1970
1980
1990
2000
2010
Figure 6.9 (Continued).
The results are shown in Table 6.5. The coefficient on the cyclical component of income is very large. But the coefficient on trend income is only 0.20—well below the within-country effect of over 0.50. And it is extremely ill-defined, with a standard error of 0.17, reflecting the different TABLE 6.5 Panel fixed effects regressions for explaining annual country-level mean life satisfaction in West Europe. The dependent variable is mean life satisfaction. T is time trend, Y T is trend GDP per head, Y C is cyclical GDP per head, u is the unemployment rate and p is the inflation rate. Country fixed effects included. Regressor
(1)
(2)
T
0.002 (0.004)
Year Dummies
log YT
0.201 (0.174)
0.180 (0.167)
log YC
3.625 (1.014)
1.855 (1.406)
u
0.023 (0.006)
0.033 (0.007)
p
0.006 (0.002)
0.010 (0.004)
0.930
0.943
Rsq
Source: Eurobarometer: 1973–2007. Mean life satisfaction calculated using the Eurobarometer sampling weights. GDP trend and cyclical components computed using Hodrick-Prescott filter with a parameter of 9.5 (a parameter of 6.25 produces similar results).
Does Relative Income Matter? Are the Critics Right?
161
effects of income in different countries. (Unemployment and inflation have the usual effects, with a percentage point of unemployment being far more harmful than a percentage point of inflation.) These findings confirm that, even in the typical country in Western Europe, the long-run effect of higher average income at country level is quite small—and much less than the effect on individual happiness when one person’s income rises while everyone else’s stays the same.
Conclusions
Our main conclusions are these. 1. In the United States, happiness has not risen since the 1950s, despite large increases in income at all points in the income distribution. Nor has life satisfaction risen in Western Germany since these records began, though it has in some other Western European countries. 2. In micro-data for the United States and West Germany, a rise in individual log income (all else constant) raises individual happiness/ life satisfaction substantially. In our metric (scaled 0–10) the effect is of the order of 0.5 for a one unit rise in log income, in both the United States and Germany. 3. We can split this effect into an effect of relative income and an effect of absolute income. In these data the whole effect is of relative income. This remains true if we include individual fixed effects. When adaptation is allowed for, it provides little additional explanation for why happiness has been so stable in Germany while income has risen. 4. The importance of income comparisons is confirmed by including psychological perceptions as explanatory variables. If ‘‘perceived relative income’’ is included, it has powerful effects. Similarly, if ‘‘required income’’ is included, life satisfaction depends simply on income relative to ‘‘required income’’ (which in turn depends on the average prevailing level of living). 5. If we turn to the time series for 20 West European countries, we can estimate the effect of a long-run change in the average income of the whole population. This is much less than the effect when one individual becomes richer (all else constant). For the average West European country, the country-level time series effect is around 0.2, which compares with the figure of 0.5–0.6 for individuals—again consistent with an important role for relative income.36
162
Section II: Comparisons of Income and Well-Being Through Time
6. The findings of this chapter are consistent with a substantial body of earlier research on the United States, Germany, Switzerland, and the Netherlands. They are also consistent with psychological research where individuals are asked to choose between one world in which they are relatively rich and absolutely poor, and another where they are relatively poor and absolutely rich. They are also consistent with honest introspection by most of us—who can say that they are completely unaffected by what others get? 7. Yet two recent papers dispute the role of relative income in affecting life-satisfaction/happiness. These papers cover the whole range of countries from poor to rich, while our argument is about rich countries only. For rich countries, these recent papers provide no reason to question the importance of relative income as an influence on life-satisfaction. Notes 1. Paper prepared for the Conference on Well-Being at Princeton University, October 13–14, 2008. We are grateful to the Esmee´ Fairbairn Foundation for financial support. 2. Easterlin (1974). 3. For the effects of institutional and cultural factors on the average happiness levels in different countries, see, for example, Helliwell (2003, 2008, and this volume); Inglehart & Klingemann (2000); and Inglehart et al. (2008). 4. Stevenson & Wolfers (2008), p. 29. 5. Layard et al. (2008). We found coefficients like 0.61 for the World Values Survey and 0.60 for the European Social Survey. Similar results are reported below for the U.S. General Social Survey and the German Socio-Economic Panel. 6. Stevenson & Wolfers (2008), p. 24 7. Smith (1979) surveys in detail the data up to 1977 and concludes that, ‘‘with the effects of variant wordings, seasons and contexts taken into consideration, it appears that happiness rose from the late Forties to the late Fifties, then fell until the early Seventies.’’ 8. The average sample size each year was 1,900 for AIPO; 1,400 for NORC; and 1,950 for the GSS. 9. See, for example, Layard (2005, pp. 34–35), which guesses that higher income brought some benefits that were outweighed by deteriorating human relationships. See also Bartolini et al. (2008), who trace how declining social capital reduced happiness. 10. This is an updated version of Layard (2005), Annex 4.1 (1). The data are corrected in the ways described by Stevenson/Wolfers, though these corrections do not substantially affect the results. 11. We ignore children (since they are a choice variable) and treat one adult as 1 unit, two adults as 1.6 units, and three adults as 2.1 units. We exclude households with over three people. ‘‘Income’’ is gross income.
Does Relative Income Matter? Are the Critics Right?
163
12. The question is: ‘‘Compared with other American families in general, would you say your family income is far below average, below average, average, above average, or far above average?’’ 13. In the GSS, income is reported in bands, to each of which we assign values by the method used by Stevenson/Wolfers. To determine the 10% of outliers, we regress income on the normal explanatory variables and select the outliers on the basis of the residuals. However, including them produces similar but less well-defined results. In our later analysis of Germany, we have precise income data and include all respondents. Excluding the 10% of outliers in Germany produces similar but less well-defined results. 14. The coefficients on log Y and log Y are 0.58 (0.05) and 0.57 (0.59). In this regression the time trend is completely insignificant at 0.001 (0.007). 15. If, to allow for dynamics, we introduce dummy variables to allow for whether income grew, was constant, or fell, then the coefficient on absolute income becomes –0.17 while that on relative income is unaffected. 16. See note 13. 17. The life-satisfaction question was: ‘‘In conclusion we would like to ask you about your satisfaction with your life in general. Please answer according to the following scale: 0 means completely dissatisfied and 10 means completely satisfied. How satisfied are you with your life, all things considered?’’ 18. A more detailed comparator group is possible than in the GSS, due to larger sample size. 19. The coefficients on log Y and log Y are 0.54 (.01) and –0.35 (.08). In this regression the trend has a coefficient of –0.009 (0.001). 20. In a similar analysis of the German data, di Tella et al. (2007) use four lags, even though the fourth lag is insignificant. When we do this on our sample, the total effect of the four lagged terms is –0.09. 21. For a survey, see Clark et al. (2008). 22. Happiness was measured from 1–7. In fixed-effects panel estimation on two data points for each respondent: Hi ¼ :05 log Yi :37 log Y ð:05Þ
23. 24. 25. 26. 27.
ð:16Þ
Helliwell & Huang (2005). See also Barrington-Leigh & Helliwell (2008). Ferrer-i-Carbonell (2005), Vendrik & Woltjer (2007) See for example van Praag & Frijters (1999) de Stadt et al. (1985) By contrast, in a similar question about leisure, respondents preferred more absolute leisure even if it was less relative to others’ leisure. This is highly relevant to tax policy: see Layard (2006). 28. The life-satisfaction question was as follows: ‘‘Please imagine a ladder with steps numbered from zero at the bottom to ten at the top. Suppose we say that the top of the ladder represents the best possible life for you and the bottom of the ladder represents the worst possible life for you. On which step of the ladder would you say you personally feel you stand at this time, assuming that the higher the step the better you feel about your life, and the lower the step the worse you feel about it? Which step comes closest to the way you feel?’’ 29. See note 3.
164
Section II: Comparisons of Income and Well-Being Through Time
30. At one point (p. 11) they test whether the cross-section inter-country effect is flatter for countries above $15,000 per head of household and conclude that it is not. But this cutoff includes many countries that are far from the experience of advanced countries, and a more appropriate cutoff for the public debate is $20,000, as used by Layard (2005) and Deaton (2008). 31. Their Figure 18. 32. We are grateful to Stevenson/Wolfers for providing us with their Eurobarometer data file. 33. Typical sample sizes vary between one and two thousand per year. 34. The question is ‘‘On the whole are you very satisfied, fairly satisfied, not very satisfied or not at all satisfied with the life you lead?’’ 35. di Tella et al. (2001) 36. The cross-sectional estimates of the effect of higher incomes already allow substantively for the role of adaptation.
References Barrington-Leigh, C., & Helliwell, J. (2008). Empathy and emulation: Life satisfaction and the urban geography of comparison groups. NBER Working Paper 14593, December. Bartolini, S., Bilanchini, E., & Pugno, M. (2008). Did the decline in social capital depress Americans’ happiness? Quaderni del Dipartimento di Economia Politica, University of Siena, N.540. Clark, A. E., Frijters, P., & Shields, M. A. (2008). Relative income, happiness and utility: An explanation for the Easterlin paradox and other puzzles. Journal of Economic Literature, 46(1), 95–144. de Stadt, H., Kapteyn, A., & de Geer, S. (1985). The Relativity of Utility: Evidence from Panel Data. Review of Economics and Statistics, 67(2), 179–187. di Tella, R., MacCulloch, J., & Oswald, A. J. (2001). Preferences over inflation and unemployment: Evidence from surveys of happiness. American Economic Review, 91(1), 335–341. di Tella, R., Haisken-De New, J., & MacCulloch, R. (2007). Happiness adaptation to income and to status in an individual panel. NBER Working Paper 13159. Deaton, A. (2008). Income, health and well-being around the world: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22(2), 53–72. Easterlin, R. (1974). Does income growth improve the human lot? Some empirical evidence. Reprinted in R. Easterlin (Ed.) (2002), Happiness in economics. Cheltenham, UK: Edward Elgar. Ferrer-i-Carbonell, A. (2005). Income and well-being: An empirical analysis of the comparison income effect. Journal of Public Economics, 89, 997–1019. Frey, B. (2008). Happiness. A revolution in economics. Cambridge, MA: MIT Press. Frey, B., & Stutzer, A. (2005). Testing theories of happiness. In L. Bruni & P. Porta (Eds.), Economics and happiness: Framing the analysis. Oxford and New York: Oxford University Press. Helliwell, J. (2003). How’s life? Combining individual and national variables to explain subjective well-being. Economic Modelling, 20, 331–360. Helliwell, J. (2008). Well-being and social capital: Does suicide pose a puzzle? Social Indicators Research, 81, 455–496.
Does Relative Income Matter? Are the Critics Right?
165
Helliwell, J., & Huang, H. (2005). How’s the job? Well-being and social capital in the workplace. NBER Working Paper 11759, November; forthcoming Industrial and Labour Relations Review. Inglehart, R., & Klingemann, H.-D. (2000). Genes, culture, democracy and happiness. In E. Diener & E. Suh (Eds.), Culture and subjective well-being. Cambridge, MA: MIT Press. Inglehart, R., Foa, R., Peterson, C., & Welzel, C. (2008). Development, freedom and rising happiness: A global perspective (1981–2007). Perspectives on Psychological Science, 3(4), 264–285. Layard, R. (2005). Happiness—Lessons from a new science. London: Penguin. Layard, R. (2006). Happiness and public policy: A challenge to the profession. Economic Journal, 90, 737–750. Layard, R., Mayraz, G., & Nickell, S. (2008). The marginal utility of income. Journal of Public Economics, 92, August, Special Issue on Happiness and Public Economics. Luttmer, E. F. P. (2005). Neighbors as negatives: Relative earnings and well-being. Quarterly Journal of Economics, 120(3), 923–1002. Smith, T. (1979). Happiness: time trends, seasonal variations, intersurvey differences and other mysteries. Social Psychology Quarterly, 42(1), 18–30. Solnick, S., & Hemenway, D. (1998). Is more always better? A survey on positional concerns, Journal of Economic Behaviour and Organisation, 37, 373–383. Stevenson, B., & Wolfers, J. (2008). Economic growth and subjective well-being: Reassessing the Easterlin Paradox. Brookings Papers on Economic Activity, Spring. Page numbers refer to their manuscript of August 19, 2008. Stutzer, A. (2004). The role of income aspirations in individual happiness. Journal of Economic Behavior and Organization, 54(1), 89–109. van Praag, B., & Frijters, P. (1999). The measurement of welfare and well-being: The Leyden approach. In D. Kahneman, E. Diener, & N. Schwarz, Well-being: The foundations of hedonic psychology. New York: Russell Sage Foundation. Vendrik, M. C. M., & Woltjer, G. B. (2007). Happiness and loss aversion: Is utility concave or convex in relative income? Journal of Public Economics, 91, 1423–1448.
Chapter 7 Happiness and Economic Growth: Does the Cross Section Predict Time Trends? Evidence from Developing Countries Richard A. Easterlin and Onnicha Sawangfa University of Southern California
Little is known about happiness trends in the developing world. As a result, point-of-time comparisons between richer and poorer countries have typically been used to infer the likely course of happiness as GDP per capita rises. In this chapter, we bring together some of the limited evidence available on happiness in the developing world to examine whether trends over time are consistent with expectations based on cross-sectional data. The developing countries are of special interest, because, according to the cross-sectional comparisons, this is where economic growth would be expected to have its biggest impact on happiness. As it turns out, for the limited number of countries studied here, we find that the actual trends do not conform to those predicted by the cross-sectional relationship. We examine three specific questions: (1) What is the nature of happiness trends in the developing countries? (2) Does the commonly observed international cross-sectional pattern of diminishing marginal utility of income predict the time trend in happiness? (3) Are higher rates of economic growth typically accompanied by more positive trends in happiness? We start with no preconceptions as to what the answers should be. Our interest is factual: to find out what the evidence is on these questions. Studies of time trends in happiness in developing countries are virtually nonexistent due to the limited and fragmentary nature of the available data. In the economics literature, the most notable exception is a recent paper by Stevenson &Wolfers (2008). This study is commendable for its effort to mobilize a wide range of data to assess the happiness–growth relation, not only cross-sectionally, but in time series as well. The time series analysis suffers, however, from, among other things, a failure to distinguish 166
Happiness and Economic Growth
167
short- from long-term relationships between happiness and GDP per capita, as will be discussed in the Methods section below. Outside of economics, there is a time series study by two quality-of-life specialists, Michael Hagerty & Ruut Veenhoven (2003), which, like the Stevenson and Wolfers paper, claims to find that happiness is positively associated with economic growth. This study has been critiqued by Easterlin (2005). (See also the reply by Hagerty & Veenhoven, 2006.) Finally, political scientist Ronald Inglehart and his collaborators in an analysis of World Values Survey data for recent decades report that ‘‘happiness rose in 45 of 52 countries for which substantial time series data were available’’ (Inglehart, Foa, Peterson, & Welzel, 2008, p. 264). As will be explained in the next section, this inference appears to result from an upward bias in the happiness measure on which the authors’ conclusion relies.1 Our criterion for including a country in this study of happiness trends is that there be at least three comparable observations on subjective wellbeing spanning at least ten years; the average period spanned is actually about 16 years. This is a short time series for studying happiness. In the original time series study of happiness and economic growth, Easterlin (1974, p. 110) found that, when comparing identical happiness questions, there was an increase in happiness in the United States from 1946 to 1956–1957, followed by a decline to 1970, with a negligible net change over the entire period. The rise and fall observed in the United States over more than two decades suggests that even a ten- or fifteen-year period may fail to give a valid indication of the long-term trend. But even when we set minima as low as a ten-year period and comparable observations at three, we are left with fairly few developing countries—a total of only thirteen. This small number is somewhat compensated for by the fact that most are quite populous, and several have very high rates of economic growth. Indeed, four (Brazil, China, Japan, and South Korea) have the distinction of being among the thirteen ‘‘success stories’’ of economic growth recently featured in a World Bank Report (Commission on Growth and Development, 2008). Happiness depends on many factors (Bruni & Porta, 2005; Frey & Stutzer, 2002; Layard, 2005). Generalizations about growth and happiness in developing countries, however, have typically been based on bivariate cross-sectional comparisons of national measures of subjective well-being and per capita income. A succinct example of inferring time trends of subjective well-being for developing countries from a bivariate multicountry cross section appears in the valuable survey volume by two of the leading economists in the study of the economics of happiness: Comparing across countries, it is true that income and happiness are positively related and that the marginal utility falls with higher income.
168
Section II: Comparisons of Income and Well-Being Through Time
Higher income clearly raises happiness in developing countries, while the effect is only small, if it exists at all, in rich countries (Frey & Stutzer, 2002, p. 90). Hence, in testing here such generalizations against time series evidence, we, too, employ a bivariate methodology. Some analysts look at happiness in relation to the absolute amount of change in income; others, the percentage change. In the analysis below, we look at both. In keeping with much of the literature the term happiness has been used to this point as a proxy for measures of subjective well-being (SWB) in general. The following analysis is based specifically on two measures of SWB: overall life satisfaction and satisfaction with finances. Although satisfaction with finances is less comprehensive than overall life satisfaction, one might expect it to be even more closely related to economic growth and thus an even better measure to test the happiness–growth relationship. Most importantly, by using two measures we are able to gauge whether they provide consistent answers to our three questions. In what follows, the term subjective well-being is used when referring to the two measures either separately or together.
Data and Measures
The principal data set is the World Values Survey (WVS), conducted in an increasing number of countries throughout the world in five waves: 1981–84, 1989–93, 1994–99, 1999–2004, and 2005–2007 (World and European Values Surveys Four-Wave Integrated Data File, 2006; World Values Survey 2005 Official Data File V.20081015, 2008). Most of the developing countries included here were first surveyed in wave 2, but four (Argentina, Japan, South Korea, and Mexico) were covered in wave 1. A second major source is the Latinobarometer, conducted almost annually since 1995. For the six Latin American countries studied here—Argentina, Brazil, Chile, Mexico, Peru, and Venezuela—we rely on the Latinobarometer once it is available, rather than the WVS, because of its fuller time series coverage. Two African countries, Nigeria and South Africa, are included. For South Africa, besides the WVS, there is a separate survey, the South African Quality of Life Trends Study commissioned to MarkData (hereafter SA MarkData) that provides a check on the WVS data.2 Five Asian countries are studied—China, India, Japan, South Korea, and Turkey. For China two other survey sources, Gallup and the Asiabarometer, are used to check the WVS. For Japan the primary series comes from the ‘‘Life in Nation’’ surveys. These surveys started in 1958 when Japan’s GDP per capita, at
Happiness and Economic Growth
169
11 percent of the U.S. level in 2000, put it well within the developing bloc. The series extends through 2007, and is complemented by the WVS and a survey conducted by the Cabinet Office of Japan and kindly provided by Takayoshi Kusago, covering the period from 1978–2005 (Kusago, 2007).3 The three surveys for Japan all terminate at a point when Japan’s GDP per capita is upwards of 80 percent of the U.S. 2000 level, and thus span the widest range of growth experience covered here. In the Japanese ‘‘Life in Nation’’ surveys there are several changes in the survey question between 1958 and 2007 (Hirata, 2001; Stevenson & Wolfers, 2008). Stevenson & Wolfers (2008, Table 5) give a valuable digest of these data. We therefore divided the series into three segments that we analyze separately, Japan 1 (1958–1969), Japan 2 (1970–1991), and Japan 3 (1992–2007). The 11-year Japan 1 series comprises two shorter segments that we pool in order to satisfy our ten-year-minimum criterion, using a dummy variable to account for the difference in response level attributable to the change in survey question. The WVS data for Japan (1981–2005), used in the analysis of satisfaction with finances, are labeled Japan 4. The GDP data are those of the World Bank from 1975 onwards (World Development Indicators Online, 2008). Those for Japan prior to 1975 are based on a backward extrapolation of the World Bank series using the Penn World Table (Heston, Summers, & Aten, 2006). We date the observations on subjective well-being here, not at the actual survey dates, but to match the annual GDP observations that they most likely reflect. The GDP data are for calendar years, while the SWB surveys typically relate to a single month or span only a few months. If a survey were conducted early in a year, it would clearly be meaningless to link it to a GDP estimate covering the entire twelve months of the same year. Our procedure, therefore, is as follows. An SWB survey conducted within the first four months of a year—say, January through April 1991—is linked to 1990 GDP; a mid-year survey, conducted in the period from May through August 1991, is compared to the GDP average of 1990 and 1991 and dated 1990.5; and a survey conducted in the latter part of the year, September through December 1991, is compared with 1991 GDP and dated 1991. In the WVS, life satisfaction and satisfaction with finances are measured on a response scale ranging in integer values from 1 (dissatisfied) to 10 (satisfied). The specific question for each subjective well-being measure used here is given in Appendix 7.A. We do not use the four-category happiness measure from the WVS, except for India from wave 3 onward. The 10-response categories of the other two measures have the obvious advantage of greater sensitivity, but
170
Section II: Comparisons of Income and Well-Being Through Time
there is another reason for not using the happiness measure. In wave 2, the happiness question that was asked differs from that in later waves, and as a result, happiness diverges markedly from the other two SWB measures in its direction of change between waves 2 and 3. This is particularly noticeable for the transitional countries of central and eastern Europe for which a number of studies concur in the finding that the economic collapse in these countries was accompanied by a substantial decrease in subjective wellbeing (Easterlin, 2009; EBRD, 2007; Guriev & Zhuravskaya, 2008; Llekes, 2006; Sanfey & Teksoz, 2007; Saris & Andreenkova, 2001; Veenhoven, 2001). In most of these transitional countries, the average happiness answers rise from wave 2 to wave 3 of the WVS despite marked declines in life and financial satisfaction. Similarly, in all but one of the developing countries studied here, the happiness measure increases between waves 2 and 3, while the other two measures are almost always consistent with each other in direction of change, sometimes moving up together, sometimes down together. The reason for the increase typically shown in the WVS happiness measure between waves 2 and 3 appears to be a ‘‘primacy’’ bias resulting from a change in the instruction accompanying the happiness question. In wave 2, interviewers were instructed to alternate the order of response choices from one respondent to the next. Thus respondent 1 would be presented with choices ranging from ‘‘very happy’’ down to ‘‘not at all happy,’’ while respondent 2 would be presented with ‘‘not at all happy’’ first. There are a number of survey studies demonstrating a tendency for respondents to favor earlier over later choices (Belson, 1966; Chan, 1991; Schuman & Presser, 1981, pp. 56–77). In wave 2, therefore, half the respondents would have been more inclined toward less happy choices by virtue of being presented with the more negative options first. In wave 3, the ‘‘very happy’’ option appears first, and the instruction to alternate response options no longer appears. Hence happiness responses in wave 3 would tend to be biased upward relative to wave 2, a movement consistent with the positive between-wave changes almost universally observed in the happiness measure in both the transitional and developing countries. For overall life satisfaction there is also a bias problem, one that would tend to reduce life satisfaction in waves 3 and 4 compared to the earlier and later waves. In this case it is a ‘‘focusing bias,’’ due to placing the question on financial satisfaction before that on overall life satisfaction in waves 3 and 4 (cf. Stevenson & Wolfers, 2008). However, the question on financial, as opposed to life, satisfaction appears to be comparable across waves in terms of both content and context. Both life and financial satisfaction typically exhibit similar directions of change
Happiness and Economic Growth
171
between waves 2 and 3, suggesting that the bias in the question on life satisfaction, to the extent it exists, is not as serious as that in the happiness measure. A paper by Inglehart, Foa, Peterson & Welzel (2008) proposes a substantive reason for the disparate time series movements in happiness and life satisfaction in the WVS.4 The authors suggest that the two measures reflect different determinants—the former, political conditions, and the latter, economic circumstances—citing specifically the experience of the transitional countries of central and eastern Europe. In their interpretation ‘‘many ex-communist countries experienced democratization accompanied by economic collapse, resulting in rising happiness and falling satisfaction’’ (Inglehart, Foa, Peterson, & Welzel, 2008, p. 277). Aside from the issue of statistical bias in the happiness measure, two pieces of evidence may be noted that call into question the argument that happiness moves differently from life satisfaction, because it is affected by democratization and life satisfaction is not. As has been mentioned, in the transitional countries of central and eastern Europe, happiness typically increases between waves 2 and 3 of the WVS, but life satisfaction declines. Thanks to a question included in the Eurobarometer surveys in these countries, it is possible to test directly whether democratization explains the rise in happiness. The question is: ‘‘On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not satisfied at all with the way democracy works in (your country)?’’5 There are twelve transitional countries included in both waves 2 and 3 of the WVS, and happiness increases in nine of the twelve. If the rise in happiness is due to democratization, a regression for all twelve countries of the change in happiness on the change in satisfaction with democracy should yield a positive slope coefficient. Actually, the regression yields a negative, though not statistically significant, coefficient. For two of the three countries that have the highest increase in happiness between waves 2 and 3, satisfaction with democracy declines. For two of the three countries in which happiness declines between waves 2 and 3, satisfaction with democracy increases. These results do not support the hypothesis that the movement in happiness is a result of democratization. Democratization may, however, temporarily affect happiness, but contrary to the view in the Inglehart et al. article, it affects life satisfaction similarly. There are almost no surveys like the WVS that ask both happiness and life satisfaction questions, but the SA MarkData survey for South Africa does. In May 1994, one month after the country’s first full-suffrage democratic election, a survey was conducted that included questions about both happiness and life satisfaction. Here, for both measures, is the percentage of the black population in the top two (out of five) categories at that
172
Section II: Comparisons of Income and Well-Being Through Time
time and the corresponding percentages at the two adjacent dates when similar surveys were conducted:
Happiness Life satisfaction
1988
1994
1995
32 37
80 86
39 45
Note how by both measures the well-being of blacks soared at the time of the election. But as Valerie Møller (2007, p. 248) observes: ‘‘[P]ost-election euphoria was short-lived. Satisfaction levels have since returned to ones reminiscent of those under the former regime.’’ This return is registered by both SWB measures. Moreover, the magnitude of rise and fall is virtually identical for the two measures. This evidence suggests that democratization may temporarily affect subjective well-being, but this effect appears not only in happiness but in life satisfaction as well. The Latinobarometer, used here together with the WVS for all six Latin American countries, also has comparability problems. The response categories for life satisfaction in the first two surveys (1997 and 2000) differ from those in subsequent surveys; hence the observations for the first two years are not used here. Also, the 2006 Latinobarometer data on financial satisfaction are omitted. In this case, a focusing bias occurred in 2006, because of the placement before the ‘‘financial satisfaction’’ question of a new question involving comparison with one’s parents’ situation, which tended to bias upward the subsequent response on financial satisfaction.6
Methods
We compute the long-term growth rate of SWB by regressing it on time, taking as our period of analysis for each country the longest time span available. The long-term growth rate of GDP per capita is computed from the GDP per capita values at the start and end of the period covered by the SWB observations. When more than one data set is available for SWB, as in the case of the Latin American countries, we compute a pooled regression with a dummy variable to identify the different data set. Growth rates for both SWB and GDP are per year; the change in SWB is measured in absolute terms, and that in GDP per capita, in percentage terms. In taking long periods for analysis the purpose is specifically to distinguish the longer- from the shorter-term relationship between SWB and GDP per capita. There is ample evidence that short-term fluctuations in SWB are positively correlated with macroeconomic conditions. This is
Happiness and Economic Growth
173
nowhere more apparent than in the massive economic collapse and subsequent recovery of the transitional countries, but it is also evident in developed countries (Di Tella, MacCulloch, & Oswald, 2001; Easterlin, 2009; Guriev & Zhuravskaya, 2008; Sanfey & Teksoz, 2008). This shorterterm positive association between fluctuations in SWB and GDP per capita should not be mistaken for the longer-term relationship. Imagine, for example, two series: one of SWB and one of GDP per capita, exhibiting synchronous sawtooth movements, but those in SWB are around a horizontal trend line, while those in GDP per capita are about a positive trend. The short-term relationship between the growth rate of SWB and that of GDP per capita is positive, but the longer-term one is not. If data with shorter and longer time spans for different countries are pooled, the shorter-term positive relationship will tend to dominate a regression of SWB on GDP per capita. This is because short-term growth rates of GDP per capita, both positive and negative, are larger (disregarding sign) than long-term growth rates; hence the short-term rates are more likely to be the outlying observations and disproportionately affect the regression results. An example of the failure to distinguish the short-term from the longterm relationship is the widely-publicized Stevenson & Wolfers (2008) paper, which seeks to establish a positive relationship between SWB and economic growth. In the part of their time series analysis that uses WVS data, they estimate regression relationships like those done here between the change in life satisfaction and that in GDP per capita. They report the results of three ‘‘short first differences’’ and three ‘‘long first differences’’ regressions after eliminating WVS countries whose data they consider noncomparable over time (Stevenson & Wolfers, 2008, Figure 15). The time spans of the ‘‘short first differences’’ regressions—typically five to six years—are too brief to identify the long-term relationship between SWB and GDP per capita. Of their three ‘‘long first differences’’ regressions, only two have a significant positive coefficient. The significant positive coefficient for the waves II to IV regression (based on observations for 32 countries) is due to the inclusion of eleven transitional countries, whose data particularly reflect the concurrent collapse and recovery of SWB and GDP per capita in these countries. If the transitional countries are omitted from the regression, the slope coefficient is no longer significant.7 The significant positive coefficient for the waves I to IV analysis, based on seventeen countries, is due to the inclusion of one transitional country, Hungary, with low growth in GDP per capita and a negative change in life satisfaction, and one developing country, South Korea, with very high growth in GDP per capita (it is off the scale in their diagram) and high growth of life satisfaction. Among the other 15 countries, all of which are
174
Section II: Comparisons of Income and Well-Being Through Time
developed, there is no significant relationship between the change in life satisfaction and that in GDP per capita. Thus, the positive association between the change in life satisfaction and that in GDP per capita reported by Stevenson and Wolfers rests almost entirely on the positively correlated V-shaped movement of the two variables during the post–1990 collapse and recovery in the transitional countries. The slope of a regression line of life satisfaction on time passing through both the contraction and expansion phases in these countries is not significant in 10 of the 11 countries for which data are available (Easterlin, 2009). As mentioned, the WVS response scale for the two SWB questions used here ranges in integer values from 1 (= dissatisfied) to 10 (= satisfied). In other surveys included here, the response options are typically categorical, not numerical. In keeping with the usual practice in the literature, we have assigned integer values to each category from 1 (= the worst response option) up to the number corresponding to the best response option (e.g., 4 if there are four response categories). We then compute the mean SWB from the integer responses for all respondents for each year that data are available. Because the WVS satisfaction questions are on a 1–10 scale, while the non-WVS questions are on a 1–4 or 1–5 scale, we rescale the WVS responses to conform to the non-WVS scale. In rescaling the WVS to a 1–4 response scale, for example, we assume that a WVS response of 10 corresponds to a non-WVS response of 4, and a WVS response of 1 to a non-WVS response of 1, and make a linear transformation using the formula: y ¼ 0:333x þ 0:667; where y ¼ SWB on the non-WVS scale; and x ¼ SWB on the WVS scale: In the regression results in the analysis below, the slope coefficients are those based on the rescaled WVS responses. The WVS surveys are imperfect. In the preceding section we noted that the wording or context of the SWB questions sometimes changes over time, and concluded that the disparate change between waves 2 and 3 in happiness compared with life and financial satisfaction was due to a primacy bias. For this reason, we chose to focus our analysis on life and financial satisfaction. Another problem is that, for some countries, the geographic coverage of the surveys changes over time. Sometimes sample weights are provided to adjust for variations in sample coverage, but weights are not given regularly enough to yield time series comparable for our purpose.
Happiness and Economic Growth
175
A valuable summary based on WVS documentation of shifts in survey coverage is given by Stevenson & Wolfers (2008, Appendix B). Their criterion for including a WVS survey in their analysis appears to be that the survey must be nationally representative. As a result they discard a number of surveys for seven of the countries studied here. Although a nationally representative survey is clearly preferable, it seems premature to dismiss other surveys as unusable. Invariably, when the WVS surveys are not nationally representative, they cover the more literate and urbanized segments of the population. These are precisely the groups that are most likely to be experiencing the income benefits of economic growth. Hence, if economic growth is raising subjective well-being, it is in these population groups that one would expect to see the improvement in SWB most clearly. The real problem for comparability over time arises when the survey coverage changes, typically to a nationally representative survey. We try to minimize the effects of such shifts in several ways, depending on the nature of the available data. For two countries, Argentina and Chile, we use only the earlier WVS surveys that appear to be comparable over time in geographic coverage; these surveys cover about 70 percent of the population. For two other countries, Mexico and Nigeria, where the survey coverage is shifting, especially in regard to coverage of the rural population, we base our analysis on the SWB means for the population living in places of 100,000 population or more, those likely to be covered fairly consistently. Finally, for two other countries, China and South Africa, we have independent surveys by other organizations that provide support for the time series change in SWB reported in the WVS. Of the seven countries included here whose surveys are largely or wholly discarded by Stevenson and Wolfers, we are left with only India, a country that appears here solely in the analysis of life satisfaction, and whose exclusion would not alter the results. The basic data for all thirteen countries studied here and the rationale for the data chosen are given in Appendix 7.B. We do not show there the results of an additional test that we ran—an estimate of the SWB mean at each date after controlling to the extent possible for shifts in population composition by gender, age, education, and size of place of residence. This analysis yielded patterns of change much like those in the data selected. For countries for which we use two surveys to establish the trend in SWB, we compute the ordinary least squares (OLS) time trend for each series separately, and then a pooled regression that includes a dummy variable for one of the series. Because the wording of the SWB question and/or response categories changes from one series to another, the dummy variable tells one whether the change has a significant effect on the level of the series. The detailed regression results for each country are given in Appendix 7.B.
176
Section II: Comparisons of Income and Well-Being Through Time
Results Trends in subjective well-being
Within a country, life satisfaction and financial satisfaction almost always trend in the same direction, providing mutually consistent evidence of the longer-term movement in subjective well-being. In nine of the twelve countries for which the two measures are available, they both move upward, and in two, downward (compare the upper right and lower left quadrants of Figure 7.1). Peru is a borderline exception. Life satisfaction is often seen as the net outcome of satisfaction with various domains of life—finances, family, health, work, friends, and so on (Easterlin & Sawangfa, 2007; Saris, Veenhoven, Scherpenzeel, & Bunting, 1995; van Praag & Ferrer-i-Carbonell, 2004). The significant relationship between financial and life satisfaction found here provides new evidence consistent with the view that satisfaction with economic conditions is an important determinant of the time series movement in life satisfaction (cf. Easterlin & Plagnol, 2008). This relationship holds in Figure 7.1 despite the fact that the number of years spanned by the financial satisfaction series is
Change per year in Life Satisfaction (absolute amount x 10–2)
5 VEN
4 3 2
MEX NIG
TUR ARC KOR
1
JAP4 PER
SAF
BRA
CHN
0
CHL
–1 –2 –2
–1
0
1
2
3
4
5
Change per year in Financial Satisfaction (absolute amount x 10–2)
Figure 7.1 Annual Rate of Change in Life Satisfaction and Financial Satisfaction. Source: Column 4 of Tables 7.1 and 7.2. The fitted OLS regression is y = 0.47081+0.35442x; n = 12; adjusted R2 = 0.24; t-statistics in parentheses. (1.19) (2.12)
Happiness and Economic Growth
177
sometimes different from that for life satisfaction. In four countries the difference is about one year, but in one (South Africa) it is around nine years (see col. 2 of Tables 7.1 and 7.2). In most countries, the rates of change in subjective well-being are not very great, and, in all but Mexico and Venezuela, not significantly different from zero (Tables 7.1 and 7.2, column 4; the Japan 4 series, available for both life and financial satisfaction, is used in the Figure 7.1 analysis). It is noteworthy, however, that in nine of twelve countries, the regression coefficients for both life and financial satisfaction are positive, signifying some improvement over time, if not enough to achieve statistical TABLE 7.1 Annual Growth Rates of Life Satisfaction and GDP per Capita in Specified Country during Specified Period (countries arrayed from high to low by growth rate of GDP per capita) (1)
(2)
(3)
(4)
Life satisfaction
Country Japan 1 China South Korea India Chile Japan 2 Turkey Peru Mexico Brazil Argentina Japan 3 Venezuela Nigeria South Africa Addendum Japan Japan 4
Change per year *102 (1– 4) scale
(5) GDP per capita % change per year (PPP 2000 international dollars)
Years
Number of observations
1958–1969 1995–2007 1980–2005
11 12 25
12 3 4
1.37* 0.16 1.04
9.40 8.61 5.40
1995–2006 1989.5–2006 1970–1991 1990–2007 1995.5–2006 1989.5–2006 1991–2006 1984–2006 1992–2007 1995–2006 1989.5–2000 1981–2007
11 16.5 21 17 10.5 16.5 15 22 15 11 10.5 26
3 7 25 4 7 8 7 8 14 7 3 5
0.28 0.94 0.56** 1.60 0.15 2.10* 0.22 1.13 1.22** 3.90* 1.80 0.37
4.95 4.04 3.29 2.33 2.18 1.63 1.28 1.15 1.12 0.53 0.19 0.17
1958–2007 1981–2005
49 24
51 5
0.13 0.42
4.03 1.97
Period
+ significant at 10%; * significant at 5%; ** significant at 1% Source: Life satisfaction, Appendixes 7B and 7C. GDP, World Bank 2007. The World Bank series for Japan was extrapolated from 1975 back to 1958 using the annual rate of change in the Penn World Table 6.2.
178
Section II: Comparisons of Income and Well-Being Through Time
TABLE 7.2 Annual Growth Rates of Financial Satisfaction and GDP per Capita in Specified Country during Specified Period (countries arrayed from high to low by growth rate of GDP per capita) (1)
(2)
(3)
(4)
Financial satisfaction
Country China South Korea Chile Turkey Japan 4 Peru Mexico Brazil South Africa Argentina Nigeria Venezuela
Change per year *102 (1–5) scale
(5) GDP per capita % change per year (PPP 2000 international dollars)
Years
Number of observations
19952007 1980–2005
12 25
3 5
0.58 0.92
8.61 5.40
1989.5–2005 1990–2007 1981–2005 1994.5–2005 1989.5–2005 1991–2005 1990–2007
15.5 17 24 10.5 15.5 14 17
12 4 5 10 12 12 4
0.27 1.47 0.13 1.34 4.91* 0.93 3.55
4.11 2.33 1.97 1.91 1.50 1.19 0.98
1984–2005 1989.5–2000 1994.5–2005
21 10.5 10.5
13 3 10
0.65 4.53 1.98
Period
0.86 0.19 0.14
+ significant at 10%; * significant at 5%; ** significant at 1% Source: See Table 7.1.
significance. Possibly the lack of significance for six of the twelve countries is partly because the number of observations is only 3 to 5. If, however, we regress log GDP instead of SWB on time, using only the dates for which there are SWB values, the trend coefficient is significant at the five percent level or better for three of the same six countries despite the few observations. Moreover, the issue of significance, or lack thereof, does not affect the findings on the two main questions below, with which this chapter is primarily concerned. The principal issue is whether countries with better economic performance exhibit more positive trends in SWB. To answer this, we turn to two specific questions: (1) Do the actual trends in SWB conform to what might be expected based on the cross-sectional relation of SWB and GDP per capita? (2) Are the trends in SWB positively associated with the rate of economic growth?
Happiness and Economic Growth
179
Predicted and actual trends in SWB
The predicted trends in SWB are estimated here by the practice common in the literature of using a point-of-time comparison between richer and poorer countries to infer the likely course of happiness as GDP per capita rises (Frey & Stutzer, 2002; Layard, 2005; Stevenson & Wolfers, 2008). We then compare the actual trends in SWB in our thirteen countries with the predicted trends to see to what extent the international cross-sectional relation of SWB to GDP per capita predicts the observed trends. Our finding is that the actual trends are not significantly related to the predicted trends. Put more strongly: Knowing the actual change over time in a country’s GDP per capita and the multi-country cross-sectional relation of SWB to GDP per capita adds nothing, on average, to one’s ability to predict the actual time series change in SWB in a country. Our procedure in general is as follows. We first estimate the crosssectional relation between SWB and log GDP per capita by pooling all of the observations from all countries included in waves 1–4 of the WVS (the regression equations are reported in the source notes to Figures 7.2 and 7.3).
3.5
Life satisfaction (1–4 scale)
3.3 3.1
0.07
2.9 2.7
0.29
2.5 2.3 2.1 5,000
5,000
1.9 0
10000
20000
30000
40000
50000
Per capita GDP (PPP 2000 international dollars)
Figure 7.2 Regression of Life Satisfaction on GDP per capita, WVS Cross Section (GDP per capita on absolute scale; 195 observations for 89 countries). Source: WVS, Waves 1-4. The fitted regression is y ¼ 0.405 þ 0.270ln(x) (n = 195, adjusted 2 R = 0.452); t-statistics in parentheses. (2.05) (12.68)
180
Section II: Comparisons of Income and Well-Being Through Time
3.5 0.08
Financial satisfaction (1–5 scale)
3.3 3.1 2.9
0.33
2.7 2.5 2.3 2.1
5,000
5,000
1.9 0
10000
20000
30000
40000
50000
Per capita GDP (PPP 2000 international dollars)
Figure 7.3 Regression of Financial Satisfaction on GDP per capita, WVS Cross Section (GDP per capita on absolute scale; 136 observations for 54 countries). Source: WVS, Waves 1–4. The fitted regression is y ¼ 0.338 þ 0.305 ln(x) (n = 136, adjusted R2 = 0.31); t-statistics in parentheses. (0.96) (7.85)
We then estimate from the regression equations the predicted change in SWB by entering the change in log GDP per capita between the initial and terminal dates of observation for each country. Finally, we compare the predicted change in SWB with the actual change between these dates as estimated from a regression on time of the observed values of SWB. The cross-sectional relation of both life and financial satisfaction to absolute GDP per capita exhibits the typical pattern of diminishing marginal utility of income (Figures 7.2 and 7.3). A $5,000 increment of GDP per capita from an initial level of $2,500, for example, raises life satisfaction by almost 0.3 points on a 1–4 scale; in contrast, from an initial level of $17,500, the increment in life satisfaction for the same absolute change in GDP per capita is less than one-fourth as much, 0.07 points (Figure 7.2). The increments in financial satisfaction for corresponding changes in GDP per capita are respectively 0.33 and 0.08 on a 1–5 scale (Figure 7.3). Thus, the cross-sectional regressions in Figures 7.2 and 7.3 demonstrate the assertion commonly made that a given absolute increase in GDP per capita has a greater impact on SWB in a poor than a rich country.
Happiness and Economic Growth
181
The calculation of the time series changes in life satisfaction predicted by the WVS cross section of Figure 7.2 can be illustrated by comparing the prediction for China, which comes close to the poorer country’s situation in Figure 7.2, with that for Japan 1992–2007 (the Japan 3 segment), which illustrates the richer country’s situation. In the period under study, China’s absolute growth in GDP per capita is $4,631; Japan’s, not much different, $4,409 (Table 7.3, col. 4). But China’s mean GDP per capita is about one fifth of Japan’s—$5,047 compared with $26,372 (column 2). Consequently, the slope of the regression curve, which indicates the increment in life satisfaction associated with a given increment in the dollar amount of GDP per capita, is considerably greater at China’s GDP level than Japan’s (column 3). To calculate the change in life satisfaction that one would predict for China based on the observed change in GDP per capita, we express China’s GDP per capita at the beginning and end of the period in log terms, compute the difference (0.992; column 5), and multiply it by the slope coefficient (0.270) of the cross-sectional regression equation underlying Figure 7.2 (given in the source note for Figure 7.2). The result is a predicted change in China’s life satisfaction of 0.268 for its GDP change of $4,631. The same procedure for Japan yields a predicted change of 0.045—about one-sixth of that for China—for its GDP change of $4,409. The actual change in life satisfaction for each country is estimated from a regression on time of the life satisfaction values actually observed. For China, the slope coefficient of the regression, the change per year in life satisfaction is –0.0016 on a 1–4 scale (Table 7.1, column 4). Multiplying this by the number of years in the period under study, 12, yields the actual change in life satisfaction, –0.019, as estimated from the time trend. For Japan, the slope coefficient of observed life satisfaction on time is –0.0122; the period, 15 years; and the estimated actual change in life satisfaction, –0.183. Thus we obtain for China a predicted change in life satisfaction from the WVS cross section of 0.268, and an actual change, based on the observed life satisfaction time trend, of –0.019. For Japan, the corresponding values are: predicted, 0.045; actual, –0.183 (these are the values for China and Japan that appear in columns 6 and 7 of Table 7.3). For life satisfaction there are 13 other cases—yielding a total of 15 altogether—for which we can estimate the actual change in life satisfaction and compare it with that predicted from the WVS cross section. These 15 paired observations (listed in columns 6 and 7 of Table 7.3) are plotted in Figure 7.4. If the actual changes corresponded to those predicted from the cross-sectional relationship, the 15 observations would fall along the positively sloped dotted line in Figure 7.4, a line with a slope of 1.0. For 9 of the 15 life satisfaction observations, the actual change is less than the predicted change (the points falling below the dotted line), and in four of these cases
TABLE 7.3 Full-Period Change in Life Satisfaction, Actual and That Predicted from Figure 2 Cross-section (Countries arrayed from low to high by mean GDP per capita in specified period) (1)
(2)
(3)
(4)
(5)
(6)
(7)
D GDP per capita
Country
Period
Mean GDP per capita (PPP 2000 international dollars)
Predicted D LS per $1000 D GDP per capita
PPP 2000 international dollars
Natural log
Predicted D LS, full period
Actual D LS, full period
182
Nigeria India China Peru Venezuela Turkey Brazil Japan 1 Chile Mexico South Africa Argentina South Korea Japan 2 Japan 3
1989.5–2000 1995–2006 1995–2007 1995.5–2006 1995–2006 1990–2007 1991–2006 1958–1969 1989.5–2006 1989.5–2006 1981–2007 1984–2006 1980–2005 1970–1991 1992–2007
838.1 2,626.0 5,047.0 5,145.1 6,301.7 6,778.3 7,148.5 7,591.5 8,314.9 8,800.6 10,562.4 12,135.1 12,461.5 18,079.2 26,372.0
0.322 0.105 0.058 0.053 0.043 0.040 0.038 0.038 0.034 0.031 0.026 0.022 0.025 0.016 0.010
16.4 1,363.8 4,631.1 1,160.0 367.2 2,623.7 1,354.5 6,946.9 5,247.3 2,333.3 453.9 3,034.6 14,364.1 11,830.5 4,408.9
0.020 0.532 0.992 0.226 0.058 0.392 0.190 0.988 0.653 0.267 0.043 0.251 1.314 0.679 0.168
0.005 0.144 0.268 0.061 0.016 0.106 0.051 0.267 0.177 0.072 0.012 0.068 0.355 0.184 0.045
0.189 0.030 0.019 0.016 0.429 0.271 0.033 0.151 0.154 0.346 0.096 0.248 0.261 0.117 0.183
Addendum Japan Japan 4
1958–2007 1981–2005
16,347.2 22,222.8
0.021 0.012
24,458.4 10,243.5
1.937 0.469
0.524 0.127
0.062 0.100
Happiness and Economic Growth
183
.5
Expected relation of actual to predicted
VEN
.4
MEX Actual change
.3 TUR ARG .2
KOR
NIG JAP1
Observed relation of actual to predicted
.1 0
JAP2
SAF BRA PER IND
CHN
–.1 JAP3
–.2 –.2
–.1
0
CHL
.1 .2 Predicted change
.3
.4
.5
Figure 7.4 Actual Change in Life Satisfaction and That Predicted from Change in GDP per capita Using WVS Cross Sectional Regression in Figure 2. Source: Table 7.3, columns 6 and 7; regression statistics, Table 7.5. For each country the change in life satisfaction is measured over the full period spanned by the life satisfaction data.
(China, Chile, India, Japan 3), the actual change is negative while the predicted change is positive. Six countries (Venezuela, Mexico, Turkey, Argentina, Nigeria, and South Africa) have changes in life satisfaction considerably in excess of what one would expect based on the crosssectional relation. For financial satisfaction, the distribution of countries above and below the dotted line is quite similar to that for life satisfaction, except that Brazil shifts from slightly below to slightly above the dotted line (Figure 7.5). A regression line fitted to the 15 observations in Figure 7.4 —the solid line plotted in the figure—actually has a somewhat negative slope, –0.152, though it is not statistically significant (the regression statistics are reported in Table 7.5). Moreover, the dotted line slope coefficient of 1.0 lies outside a two-standard-error range of the solid line slope coefficient of –0.152. The same conclusions hold for the coefficients in the analysis of financial satisfaction in Figure 7.5. We conclude, therefore, that for the thirteen countries studied here, the actual changes in SWB typically have no relation to the cross-sectional pattern. The practice common in the
184
Section II: Comparisons of Income and Well-Being Through Time
.8
MEX
Expected relation of actual to predicted
SAF
.6
Actual change
NIG .4
VEN ARG BRA
.2
TUR
KOR Observed relation of actual to predicted
JAP4
0
CHL
CHN
PER –.2 –.2
–.1
0
.1
.2
.3
.4
.5
.6
.7
.8
Predicted change
Figure 7.5 Actual Change in Financial Satisfaction and That Predicted from Change in GDP per capita Using WVS Cross Sectional Regression in Figure 3. Source: Table 7.4, columns 6 and 7; regression statistics, Table 7.5. For each country the change in financial satisfaction is measured over the full period spanned by the financial satisfaction data.
literature, to infer time trends in SWB from the cross-sectional relationship of SWB to GDP per capita, finds no support in this analysis. Japan is of special interest, because it traverses such a broad range of GDP per capita. Compared to the United States GDP per capita level in 2000, Japan moved from 11 percent in 1958 to about 80 percent in 2006. Based on the international cross-sectional pattern in Figure 7.2, this impressive growth in GDP per capita should have raised life satisfaction by 0.52 points, about one-sixth of the way along a 1–4 scale. The actual change in life satisfaction over this 48-year period, as estimated from a regression of life satisfaction on time, is slightly (but not significantly) positive, 0.06 points. (The regression is given in Appendix 7.B.) Subjective well-being and economic growth
When plotted against the natural log, rather than the absolute value of GDP per capita, the cross-sectional regressions underlying Figures 7.2 and 7.3 above imply that higher rates of economic growth are accompanied by
TABLE 7.4 Full-Period Change in Financial Satisfaction, Actual and That Predicted from Figure 3 Cross Section (Countries arrayed from low to high by mean GDP per capita in specified period) (1)
(2)
(3)
(4)
(5)
(6)
(7)
Actual D FS, full period
D GDP per capita
Country
185
Nigeria Peru China Venezuela Turkey Brazil Chile Mexico South Africa Argentina South Korea Japan 4
Period 1989.5–2000 1994.5–2005 1995–2007 1994.5–2005 1990–2007 1991–2005 1989.5–2005 1989.5–2005 1990–2007 1984–2005 1980–2005 1981–2005
Mean GDP per capita (PPP 2000 international dollars) 838.1 4,890.3 5,047.0 6,020.4 6,778.3 7,055.1 8,156.9 8,625.9 9,968.3 11,663.5 12,461.5 22,222.8
Predicted D FS per $1000 D GDP per capita 0.364 0.063 0.065 0.051 0.046 0.043 0.039 0.036 0.031 0.026 0.028 0.014
PPP 2000 international dollars
Natural log
Predicted D FS, full period
16.4 970.0 4,631.1 88.5 2,623.7 1,167.8 4,931.2 1,983.9 1,642.2 2,091.5 14,364.1 10,243.5
0.020 0.199 0.992 0.015 0.392 0.166 0.624 0.231 0.165 0.180 1.314 0.469
0.006 0.061 0.303 0.004 0.120 0.051 0.190 0.071 0.050 0.055 0.401 0.143
0.475 0.141 0.070 0.208 0.250 0.130 0.042 0.762 0.604 0.137 0.231 0.032
186
Section II: Comparisons of Income and Well-Being Through Time
TABLE 7.5 Regression Relationships between Full-Period Actual Change in Subjective Well-Being and Change Predicted from WVS Cross-Section Actual Change
Predicted change Constant Observations R-squared
Life satisfaction
Financial satisfaction
0.15191 [0.744] 0.13662 [0.084]+ 15 0.008
0.72643 [0.305] 0.30221 [0.024]* 12 0.104
p values in brackets + significant at 10%; * significant at 5%; ** significant at 1%. Source: Life satisfaction, Table 7.3, cols. 6 and 7; financial satisfaction, Table 7.4, cols. 6 and 7. For each country the change in SWB is measured over the full period spanned by the SWB data.
greater improvements in SWB. In the Figure 7.6 cross section, for example, where GDP is now on a logarithmic scale, an increase in GDP per capita from $1,000 to $2,000 (an increase in log GDP per capita from 6.91 to 7.60) raises life satisfaction by 0.18 points. Doubling the rate of economic
3.5
Life satisfaction (1–4 scale)
3.3 3.1 2.9 2.7 0.36
2.5 0.18
2.3
X2/X1 = 4
2.1
X2/X1 = 2
1.9 400 6
4000 8000 16000 32000 1000 2000 GDP per capita (PPP 2000 international dollars) 7
8 9 Log GDP per capita
10
64000 11
Figure 7.6 Life Satisfaction and GDP per capita, WVS Cross-section (GDP per capita on logarithmic scale). Source: The regression equation is the same as that for Figure 7.2.
Happiness and Economic Growth
187
Life satisfaction (1–4 scale)
3.5 3.3 3.1 2.9 2.7 0.21
2.5
0.42 X2/X1 = 4
2.3 2.1
X2/X1 = 2
1.9 400 6
1000 2000 4000 8000 16000 32000 GDP per capita (PPP 2000 international dollars) 7
8 9 Log GDP per capita
10
64000 11
Figure 7.7 Financial Satisfaction and GDP per capita, WVS Cross-section (GDP per capita on logarithmic scale). Source: The regression equation is the same as that for Figure 7.3.
growth; i.e., raising GDP per capita from $1,000 to $4,000 (log GDP per capita, from 6.91 to 8.29), doubles the increase in life satisfaction to 0.36 points. Figure 7.7 indicates for financial satisfaction a similar relationship, a doubling of the rate of economic growth accompanied by a doubling of the rate of subjective well-being. The relationship that one would predict from the cross-sectional regression between the rate of change per year in GDP per capita (in percentage terms) and that in SWB (in absolute amounts) is shown by the dotted lines marked ‘‘Predicted Relation’’ in Figures 7.8 and 7.9. The points in these lines are obtained by using the regression equations given at the bottom of Figures 7.2 and 7.3 to estimate the improvement in SWB that would result from annual growth rates of GDP per capita of 1, 3, 5 percent, and so on. The dotted lines demonstrate that higher growth rates of SWB are expected to be associated with higher rates of economic growth. Is a higher growth rate of GDP per capita in fact accompanied by a higher growth rate of SWB? To assess this, we plot in Figures 7.8 and 7.9 the values for each country of the actual rates of change per year in SWB and GDP per capita (given in columns 4 and 5 of Tables 7.1 and 7.2). Without exception, countries with quite high rates of growth in GDP per capita, 4 percent per year or more, fall below the dotted line, indicating that the improvement in SWB is
Change per Year in Life Satisfaction (absolute amount x 10–2)
5 4
VEN
3
Predicted Relation MEX
2
NIG
TUR
JAP1
ARG
1 SAF 0
KOR
BRA PER
Actual Relation
JAP2
CHN
IND
–1
CHL
JAP3
–2 2 4 6 8 Change in GDP Per Capita (percent per year)
0
10
Figure 7.8 Annual Rate of Change in Life Satisfaction and in GDP per capita, Actual and That Predicted from WVS Cross Section. Source: Table 7.1, cols. 4 and 5; regression statistics, Table 7.6.
5
MEX
Change per Year in Life Satisfaction (absolute amount x 10–2)
NIG 4 SAF 3
Predicted Relation
2
VEN TUR
1
BRA ARG
KOR JAP4
0
CHL
–1
PER
CHN Actual Relation
–2 0
2 4 6 8 Change in GDP Per Capita (percent per year)
10
Figure 7.9 Annual Rate of Change in Financial Satisfaction and in GDP per capita, Actual and That Predicted from WVS Cross Section. Source: Table 7.2, cols. 4 and 5; regression statistics, Table 7.6.
Happiness and Economic Growth
189
TABLE 7.6 Regression Relationships between Annual Growth Rates of Subjective Well-Being and GDP per capita
D GDP per capita Constant Observations R-squared
(1)
(2)
D Life satisfaction
D Financial satisfaction
0.10084 [0.419] 1.08787 [0.050]* 15 0.051
0.4007 [0.096]+ 2.37295 [0.010]** 12 0.252
p values in brackets + significant at 10%;* significant at 5%;** significant at 1%. Source: Col. 1 from Table 7.1, cols. 4 and 5; col. 2 from Table 7.2, cols. 4 and 5. Growth rates of SWB are in absolute amount per year; of GDP per capita, percent per year.
less than what one would expect based on the cross section. In contrast, countries with growth rates of GDP per capita lower than 4 percent fall on either side of the dotted line—some have a greater improvement in SWB than would be predicted from the cross section; others less. Regression lines fitted to the country data on growth rates of SWB and GDP per capita—the solid lines in Figures 7.8 and 7.9 marked ‘‘Actual Relation’’—are, in fact, negatively, not positively inclined. For one of them, ‘‘financial satisfaction,’’ the slope coefficient is statistically significant at the 10-percent level (Table 7.6). In both figures, the dotted line (‘‘Predicted Relation’’) slope coefficient lies outside a two-standard-error range of the solid line (‘‘Actual Relation’’) slope coefficient. A conservative conclusion would seem to be that for the thirteen developing countries studied here there is no evidence that more rapid economic growth is accompanied by a greater improvement in happiness.
Conclusion
The experience of thirteen developing countries, scattered across three continents, fails to show any consistent relationship between long-term economic growth and the growth rate of subjective well-being. This is true for two measures of subjective well-being that were separately analyzed, life satisfaction and financial satisfaction. The absence of any relationship between financial satisfaction and GDP per capita is especially noteworthy, because this is the SWB measure that one would expect to be most strongly affected by economic growth. However, the two measures of subjective well-being themselves typically trend similarly within a country,
190
Section II: Comparisons of Income and Well-Being Through Time
providing mutually supporting evidence of the movement in subjective well-being. Point-of-time cross-sectional regressions of SWB on GDP per capita are often used as a basis for inferring that economic growth raises subjective well-being, especially in poorer countries. For the developing countries studied here there is no relationship between expectations based on crosssectional data and actual time series experience—one more bit of evidence that cross-sectional relationships are a questionable basis for inferring change over time (cf. Easterly, 1999). Of course, we are examining here only the simple bivariate relation between the growth rates of SWB and GDP per capita, but it is the bivariate relation between SWB and GDP per capita that has provided the cross-sectional evidence underlying generalizations that economic growth would raise subjective well-being. Our conclusions are subject to a number of qualifications. We considered only thirteen of the world’s developing countries, though a number of them are the most populous in the world and several have quite high rates of economic growth. Our time series for subjective well-being are short, averaging 16 years. We have tried to minimize shorter-term disturbances by computing growth rates over the full time span covered by the data for each country, but we cannot rule out the possibility that shorter-term influences still affect our estimates of the long-term trends. To maximize the time span covered, we used for some countries data from two different sources. We have tried to screen and test all of the data for comparability over time, but there is no assurance that our procedures are foolproof. Despite our efforts, one can undoubtedly find reason to fault the SWB data. But for those so inclined, it is perhaps worth pointing out the consistency in the present results. Consider, for example, three countries with very high recent growth rates of GDP per capita—China, Chile, and South Korea. China’s growth rate implies a doubling of real income in less than 10 years; South Korea’s, in 13 years; and Chile’s in 18 years. With the per capita amount of goods multiplying so rapidly in a fraction of a lifetime, one might think many of the people in these countries would be so happy they’d be dancing in the streets. Yet both China and Chile show mild (not statistically significant) declines in SWB—China in surveys conducted by three separate statistical organizations. South Korea—none of whose surveys has been faulted—shows a (not statistically significant) increase, but all of the increase results from the low value reported in the 1980 survey, one that was conducted a few months after the assassination of the country’s president. Thereafter, in four surveys from 1990 to 2005, a period when GDP per capita continued to grow rapidly, averaging 5 percent per year, subjective well-being is constant or declining slightly. With incomes increasing so greatly in three different countries, it is
Happiness and Economic Growth
191
surprising, to say the least, that there are no surveys that register the marked improvement in subjective well-being that one might expect based on the cross-sectional relationship. We do find that countries that have higher growth rates of satisfaction with finances typically have higher growth rates of overall life satisfaction. One might reasonably ask, therefore, if financial satisfaction is closely linked to life satisfaction, as this result indicates, why doesn’t subjective well-being improve with economic growth? One possible answer to this question is that there are offsetting changes in other happiness domains such as family life. Another, which has some support in the growing body of happiness research, is that while economic growth raises objective living conditions, it also raises the standards by which people judge their living conditions (cf., e.g., Clark, Frijters, & Shields, 2008). There is nothing to suggest that those in developing countries are immune to this rise in material aspirations, or that this aspiration mechanism operates only beyond some unspecified ‘‘basic needs’’ point. This rise in standards would undercut the positive impact on wellbeing of objectively improved living conditions, as measured by GDP per capita. As a result, perceptions of one’s financial situation, which is what the survey questions here capture, do not rise commensurately with the objective improvement in living conditions. Clearly, the results here point to the need for deeper research into the happiness–growth relationship—to the need for consideration of the various effects of economic growth—not only via the accumulation of material goods, but also through aspirations, and work, health, and family concerns. They suggest, too, the urgency of time series studies to test the current plethora of SWB generalizations based only on point-of-time cross sections.
Acknowledgments
All those who have used the WVS data must be grateful for this impressive survey undertaking by Ronald Inglehart and his collaborators that has placed in the public domain information on subjective attitudes and well-being for so many countries throughout the world over the past two to three decades. Without these data, the present study would not have been possible. We have benefited from suggestions made by Timothy Biblarz, Ed Diener, John Ham, John Helliwell, Betsey Stevenson, and Justin Wolfers, and the generous assistance of Jacqueline Smith and Laura Angelescu. For their help in providing data and comments thereon, special thanks are due to Valerie Møller (South Africa) and Takayoshi Kusago (Japan). Financial support was provided by the University of Southern California.
192
Section II: Comparisons of Income and Well-Being Through Time
Notes 1. Analysts sometimes try to infer time series change by comparing the responses to ‘‘ladder-of-life’’ questions of the type asked in the recent Gallup World Poll to Hadley Cantril’s (1965) results. To assume the recent responses are comparable to Cantril’s is questionable. Before presenting respondents with the ladder-of-life question, Cantril’s interviewers conducted a lengthy in-depth interview probing the respondents’ concerns about the best and worst of all possible worlds (see Cantril, 1965, pp. 22–24; and Easterlin, 1974, pp. 91–92). The recent ladder-oflife questions have no counterpart to this lengthy preamble. 2. Special thanks go to Professor Valerie Møller of Rhodes University, Grahamstown, South Africa, for providing tabulations of these data (which are not in the public domain) and valuable help regarding the comparability over time of the survey questions. Professor Møller has published extensively on quality of life in South Africa (see, e.g., Møller, 1998; 2001; 2007). 3. The Kusago series is used here in Appendix 7.B as a check on the ‘‘Life in Nation’’ series. 4. Helliwell & Putnam (2004) find somewhat different results for happiness versus life satisfaction in relation to a variety of social capital variables, but their analysis, unlike that of Inglehart and his collaborators, is cross-sectional and multivariate. 5. For five countries, the period spanned by the question is the same as that for the WVS happiness question; for the other seven the period covered by the question differs by one year from that to which the happiness question refers. 6. In the Latinobarometer, the questions preceding the family economic situation question in 2005 and 2006 were: 2005. In the following twelve months, do you think that, in general, the economic situation of the country will be much better, a little better, the same, a little worse, or much worse than now? 2006. Do you think you have a job much better, a little better, the same as, a little worse, or much worse than your father did?
The response categories are identical for the two questions, so we computed for each country the mean response to each of the questions on a scale from 5 (much better) to 1 (much worse). We hypothesize that the shift to the 2006 question invoking a comparison of one’s personal economic situation (one’s job) with one’s father’s situation would put the respondent in a more favorable context to respond to the next question on one’s current family economic situation than the 2005 question about the country’s general economic outlook. The evidence confirmed this. In every one of 17 Latinobarometer countries, the 2006 question requiring comparison with one’s father elicited a mean response much higher than did the 2005 question on the country’s economic outlook; the answer averaging 0.9 points on the 1–5 scale. Moreover, the more favorable context invoked by the 2006 comparison with one’s father resulted in more favorable response to the following question on one’s family’s economic situation. In 14 of the 17 countries there was an increase from 2005 to 2006 in the assessment of one’s own economic situation, the average increment being about 0.1 points on a 1–5 scale. If one compares the 17 countries, one finds that the magnitude of the 2005–2006 increase in one’s family’s economic situation was significantly positively correlated with the magnitude
Happiness and Economic Growth
193
of the 2005–2006 increase between the immediately preceding questions. We are grateful to Malgorzata Switek for carrying out this analysis. 7. In redoing the Stevenson and Wolfers’ analysis we have converted both variables to a per-year basis, because the interval between two given waves sometimes differs among countries. The effect on the results is negligible.
References Belson, W. A. (1966). The effects of reversing the presentation order of verbal rating scales. Journal of Advertising Research, 6, 30–37. Bruni, L., & Pier, L. P. (2005). Economics and happiness: Framing the analysis. Oxford, UK: Oxford University Press. Cantril, Hadley. (1965). The pattern of human concerns. New Brunswick, NJ: Rutgers University Press. Chan, J. C. (1991). Response-order effects in Likert-type scales. Educational and Psychological Measurement, 51, 531–540. Clark, A. E., Frijters, P., & Shields, M. A. (2008). Relative income, happiness, and utility: An explanation for the Easterlin paradox and other puzzles. Journal of Economic Literature, 46(1), 95–144. Commission on Growth and Development (2008). The growth report: Strategies for sustained growth and inclusive development. Washington, DC: The World Bank. Di Tella, R., MacCulloch, R. J., & Oswald, A. J. (2001). Preferences over inflation and unemployment: Evidence from surveys of happiness. American Economic Review, 91, 335–341. Easterlin, R. A. (1974). Does economic growth improve the human lot? In Paul A. David & Melvin W. Reder (Eds.), Nations and households in economic growth: Essays in honor of Moses Abramovitz. New York: Academic Press. Easterlin, R. A. (2005). Feeding the illusion of growth and happiness: A reply to Hagerty and Veenhoven. Social Indicators Research, 74(3), 429–443. Easterlin, R. A. (2009). Lost in transition: Life satisfaction on the road to capitalism. Journal of Economic Behavior and Organization, 71(2), 130–145. Easterlin, R. A., & Plagnol, A. C. (2008). Life satisfaction and economic conditions in East and West Germany pre- and post-unification. Journal of Economic Behavior and Organization, 68, 433–444. Easterlin, R. A., & Sawangfa, O. (2009). Happiness and domain satisfaction: Theory and evidence. In Amitava Krishna Dutt & Benjamin Radcliff (Eds.), Happiness, economics, and politics: New lessons for old problems. Northhampton, MA: Edwards Elgar Publishers. An earlier version appeared as IZA Discussion Paper 2584. Easterly, W. (1999). Life during growth. Journal of Economic Growth, 4(3), 239–275. European Bank for Reconstruction and Development (2007). People in transition. Transition Report 2007. London: European Bank for Redevelopment. Frey, B. S., & Stutzer, A. (2002). Happiness and economics: How the economy and institutions affect well-being. Princeton, NJ: Princeton University Press. Hagerty, M. R., & Veenhoven, R. (2003). Wealth and happiness revisited—growing national income does go with greater happiness. Social Indicators Research, 64, 1–27.
194
Section II: Comparisons of Income and Well-Being Through Time
Hagerty, M. R., & Veenhoven, R. (2006). Rising happiness in nations 1946–2004: A reply to Easterlin. Social Indicators Research, 76, 421–436. Helliwell, J. F., & Putnam, R. D. (2004). The social context of well-being. Philosophy Transactions of the Royal Society of London Biological Science, 359, 1435–1446. Heston, A., Summers, R., & Aten, B. (2006, September). Penn World Table Version 6.2. Center for International Comparisons of Production, Income and Prices, University of Pennsylvania. (URL: http://pwt.econ.upenn.edu/php_site/pwt_index.php, accessed on February 14, 2008.) Hirata, J. (2001). Happiness and economics: Enriching economic theory with empirical psychology. Unpublished master’s thesis, Maastricht, Netherlands: Maastricht University. Kusago, T. (2007). Rethinking of economic growth and life satisfaction in post-WWII Japan—A fresh approach. Social Indicators Research, 81, 79–102. Layard, R. (2005). Happiness: Lessons from a new science. New York: Penguin Press. Lelkes, O. (2006). Tasting freedom: Happiness, religion and economic transition. Journal of Economic Behavior and Organization, 59, 173–194. Møller, V. (1998). Quality of life in South Africa: Post-apartheid trends. Social Indicators Research, 43, 27–68. Møller, V. (2001). Happiness trends under democracy: Where will the new South African set-level come to rest? Journal of Happiness Studies, 2, 33–53. Møller, V. (2007). Researching quality of life in a developing country: Lessons from the South Africa case. In I. Gough & J. A. McGregor (Eds.), Wellbeing in developing countries: From theory to research (pp. 242–258). Cambridge, UK: Cambridge University Press. Saris, W. E., & Andreenkova, A. (2001). Following changes in living conditions and happiness in post-Communist Russia: the Russet panel. Journal of Happiness Studies 2(2), 95–109. Saris, W. E., Veenhoven, R., Scherpenzeel, A. C., & Bunting, B. (Eds.) (1995). A comparative study of satisfaction with life in Europe. Budapest, Hungary: Eo¨tvo¨s University Press. Schuman, H., & Presser, S. (1981). Questions and answers in attitude surveys: Experiments on question form, wording and context. New York: Academic Press. Stevenson, B., & Wolfers, J. (2008). Economic growth and subjective well-being: Reassessing the Easterlin paradox. IZA Discussion Paper 3654, Bonn, Germany: Institute for the Study of Labor. van Praag, B. M. S., & Ferrer-i-Carbonell, A. (2004) Happiness quantified: A satisfaction calculus approach. Oxford, UK: Oxford University Press. Veenhoven, R. (2001). Are the Russians as unhappy as they say they are? Journal of Happiness Studies, 2(2), 111–136. World and European Values Surveys, Four Wave Integrated File, 1981–2004, v.20060423, 2006. World Value Survey Association (www.worldvaluessurvey.org) and European Values Study Foundation (www.europeanvalues.nl), accessed on May 8, 2007. World Development Indicators. World Bank. Retrieved from February 26, 2008, to June 11, 2008, from http://go.worldbank.org/IW6ZUUHUZ0. World Values Survey 2005 Official Data File, v.20081015, 2008. World Values Survey Association (www.worldvaluessurvey.org). Aggregate File Producer: ASEP/JDS, Madrid.
Appendix 7.A Survey Questions
World Values Survey
Life satisfaction: All things considered, how satisfied are you with your life as a whole these days? Please use this card to help with your answer. 1 ‘‘Dissatisfied’’ 2
3
4
5
6
7
8
9
10 ‘‘Satisfied’’
Financial satisfaction: How satisfied are you with the financial situation of your household? If ‘‘1’’ means you are completely dissatisfied on this scale and ‘‘10’’ means you are completely satisfied, where would you put your satisfaction with your household’s financial situation? 1 ‘‘Dissatisfied’’ 2
3
4
5
6
7
8
9
10 ‘‘Satisfied’’
Happiness: Taking all things together, would you say you are: 1 ‘‘Very happy’’
2 ‘‘Quite happy’’
3 ‘‘Not very happy’’
4 ‘‘Not at all happy’’
Latinobarometer
Life satisfaction: In general terms, would you say that you are satisfied with life? 1 = Very satisfied; 2 = Pretty satisfied; 3 = Not very satisfied; 4 = Not satisfied at all 195
196
Section II: Comparisons of Income and Well-Being Through Time
Financial satisfaction: How would you define, in general, the current economic situation of yourself and your family? Would you say that it is . . . 1 = Very good; 2 = Good; 3 = Regular; 4 = Bad; 5 = Very bad
Gallup Survey (China)
Life satisfaction: Overall, how satisfied or dissatisfied are you with the way things are going in your life today? Would you say you are: Very satisfied (4), Somewhat satisfied (3), Somewhat dissatisfied (2), or Very dissatisfied (1)?
Life in Nation Survey (Japan)
1958–1963: How do you feel about your circumstances at home? Please choose one of the following: Satisfied, Not satisfied / not dissatisfied, Somewhat dissatisfied, or Extremely dissatisfied. 1964–1969: How do you feel about your life at home? Please choose one of the following: Completely satisfied, Satisfied, Somewhat dissatisfied, or Completely dissatisfied. 1970–1991: How do you feel about your life now? Please choose one of the following: Completely satisfied, Satisfied, Somewhat dissatisfied, or Completely dissatisfied. 1992–2007: Overall, to what degree are you satisfied with your life now? Please choose one of the following: Satisfied, Somewhat satisfied, Somewhat dissatisfied, or Dissatisfied.
Cabinet Office of Japan
Life satisfaction: Are you happy with your life overall? 1 = Very satisfied; 2 = Satisfied; 3 = Not satisfied or unsatisfied; 4 = Unsatisfied; 5 = Never satisfied
Eurobarometer
Life satisfaction: On the whole, how satisfied are you with the life you lead? 1 ‘‘Not at all satisfied’’
2
3
4
5
6
7
8
9
10
‘‘Absolutely satisfied’’
Happiness and Economic Growth
197
South African Quality of Life Trends Study (Mark Data)
Life satisfaction: [1983, 1999 wording]; (revised phrasing) Taking all things together in your life, how satisfied are you with your life as a whole these days? [On the whole] (Generally speaking) would you say you are very satisfied, satisfied, dissatisfied, or very dissatisfied? (neither/nor, don’t know) (At other dates, ‘‘neither’’ is the middle item.) Happiness: 5 = Very happy; 4 = Fairly happy; 3 = Neither happy nor unhappy; 2 = Fairly unhappy; 1 = Very unhappy
Appendix 7.B Data and Regressions, by Country
The data sources for the subjective well-being data used here, specific survey dates, and date of the GDP observations with which each SWB survey is paired are given in Appendix C. In this appendix we give for each country the World Values Survey (WVS) data on mean subjective wellbeing used in the regressions (in boldface) and the regression results. For the six Latin American countries we also give mean subjective well-being for the initial and terminal years of the Latinobarometer series that we use. The Latinobarometer data often overlap the WVS series and thus provide a test of the consistency between the two surveys in the change in subjective well-being. For six of the remaining countries (all but Japan), the regressions are based entirely on the WVS data, and for three of these (China, South Africa, and Turkey) we give not only the WVS means but also the means from other surveys as a check of the WVS data. We also give means for subgroups of the population by education or size of place of residence, where these data are relevant to evaluating the WVS series. A brief comment is provided for each country to explain the data-selection decisions. All data are presented based on their original scale (1–10, 1–4, or 1–5). The regressions are based on rescaled values of the WVS to 1–4 or 1–5 scale, as appropriate. Where the regression results are based on two component series, we present regression results separately for each segment as well as those for the combined regression. The time variable in the regressions is set equal to zero at 1980. Latin America
Argentina: The first three WVS surveys cover the higher-income, more urbanized central portion of the country with about 70 percent of the 198
Happiness and Economic Growth
199
population; thereafter the surveys are nationally representative. Comparing the five WVS surveys, one finds that differences among educational groups are quite consistent, and so too are the trends by level of education. In the periods of overlap between the WVS and Latinobarometer surveys, the directions of change are consistent for life satisfaction, 1998–2006, and financial satisfaction, 1995–2006. We use the first three WVS surveys, when the geographic coverage was constant. Brazil: The WVS surveys are designed to be nationally representative. Across the three WVS surveys below, differentials between places with populations of 100,000 and over and less than 100,000 are quite consistent, as are trends for the two size-of-place groups. Between the last two dates below, the directions of change in the Latinobarometer and WVS survey are consistent for both life and financial satisfaction. Chile: The first two WVS surveys cover the higher-income central portion of the country and include about two-thirds of the population. We use these two surveys with consistent geographic coverage together with the Latinobarometer surveys in our regressions below. The third WVS survey is based on a sample of 29 cities; the fourth appears to be nationally representative. Where the WVS and Latinobarometer surveys overlap, they are consistent in the direction of change between dates. Mexico: The 1990 WVS survey was confined to cities with populations of 50,000 or more; thereafter, the survey was nationally representative. Across four WVS surveys, differentials in subjective well-being were usually small between places with populations of 100,000 plus and those under 100,000. The Latinobarometer and WVS trends are fairly consistent, but the 1995 WVS value for financial satisfaction seems out of line, appearing to be on the high side. To minimize the effect of the shift in geographic coverage after 1990, the analysis uses the WVS observations through 1999 for places with populations of 100,000 or more. Peru: For financial satisfaction, the Latinobarometer is used throughout. For life satisfaction, the first two WVS values for the total population are used, together with the Latinobarometer data starting in 2000. Where the WVS and Latinobarometer surveys overlap, the changes in subjective wellbeing are consistent between them. Across three WVS surveys, differences by level of education in subjective well-being are consistent. The trends by education for the three surveys are also consistent. We therefore use the WVS values for the total population. Venezuela: The wave-5 WVS survey is not yet available. For financial satisfaction we use the Latinobarometer surveys from 1995 on. For life satisfaction we use the 1995 and 2000 WVS surveys complemented by the Latinobarometer 2000 on. To judge from the WVS sample distribution, there is a substantial increase between the years 1995 and 2000 in the
200
Section II: Comparisons of Income and Well-Being Through Time
representation of places with populations of 100,000 or more. However, the change in life satisfaction between 1995 and 2000 is much the same by size of place of residence. Hence, we use the WVS life-satisfaction values for the total population in 1995 and 2000. Between 1995 and 2000, the direction of change in WVS life satisfaction is the same as that in financial satisfaction in both the WVS and Latinobarometer.
Asia
China: There are four WVS surveys of China, but we exclude the first survey, which was restricted almost entirely to the urban population. The 1995 survey covered central China, home to almost 68 percent of the population, while the two most recent surveys appear to be nearly nationally representative. If one compares the 1995 and 2007 WVS surveys, differentials by level of education are similar. The mild decline in subjective well-being from 1995 to 2007 in the WVS surveys is consistent with declines reported in Gallup surveys conducted in 1997, 2001, and 2004, and Asiabarometer surveys in 2003 and 2006.1 Because of the consistency in trends across three different surveys, we treat the 1995 WVS survey as comparable to those in 2000.5 and 2007. India: There are four WVS surveys of India. In the two most recent surveys, the responses for both life and financial satisfaction are confined almost entirely to five categories (1, 3, 5, 7, and 10), whereas in the first two surveys they are distributed among all ten options. As a result there is a substantial downward bias in the means for life and financial satisfaction in the last two surveys compared with the first two, making them non-comparable over time. We chose therefore to use the responses on happiness to analyze the trend in subjective well-being. We used the surveys from 1995 on, because, as explained in the text, the happiness question changed between 1990 and 1995, making the 1990 mean not comparable with the later surveys. The 2001 and 2006 surveys appear to be fairly representative of the population generally, but the 1995 survey was in Hindi. Differentials by level of education in all these surveys, however, are quite similar in magnitude. Also, the direction of change between waves is the same at each level of education. We used, therefore, the WVS values for happiness for the total population in the last three surveys as comparable over time. Japan: The primary series for life satisfaction is from the ‘‘Life in Nation’’ surveys, 1958–2007 (Stevenson & Wolfers 2008, Table 5). Financial satisfaction is from the five WVS surveys, 1981–2005. Life
Happiness and Economic Growth
201
satisfaction from these WVS surveys is included below, because it is used in the analysis in Figure 7.1 of the text. South Korea: We use the five nationally representative WVS surveys since 1980. In the 1996 survey the question on life satisfaction was not asked, but that on financial satisfaction was. Turkey: The four surveys appear to be nationally representative (except that the 1990 survey omitted the southeastern, predominantly Kurdish, region). The sizeable increase in subjective well-being between 2001 and 2007 in the WVS is consistent with that in the Eurobarometer.
Africa
Nigeria: The first WVS survey was designed to be carried out in the urban and literate segments of the population; thereafter the surveys appeared to be more nationally representative. To improve comparability, we use for all three dates data for the population in places of 100,000 population or more. South Africa: The WVS surveys other than that in 2007 substantially underrepresent the black population. To obtain a more accurate estimate for the total population we have weighted the means for the four population subgroups—blacks, colored, Indian, and white—by the population distribution reported by the national statistics agency. The resulting trends for the total population are generally consistent in direction of change with those reported in the SA MarkData surveys. In making comparisons with the WVS, it is important to note that the ordering of response categories differs in some of the SA MarkData surveys. When the surveys are arranged in terms of comparable response options, two comparisons with the WVS are possible. The first is between the early 1980s and around 2000; both the SA MarkData and WVS surveys show a decline in life satisfaction for the population as a whole between the two dates. The second comparison is from around 1990 to 2007; both surveys report an increase in life satisfaction. Because of the consistency between the two surveys, we use the WVS survey means for the total population after reweighting the component population groups by the appropriate population distribution.
Note 1. We omit the 1994 Gallup survey, which had five rather than the four response categories of the later surveys.
Appendix 7.B (continued) Mean SWB
Latin America
Argentina Life satisfaction LB (1–4 scale) WVS Total Pop ACEb
1984
1990
1995
1998
2006
18+ 13–17 £ 12
6.77 6.83 6.69 6.52
7.25 7.05 7.43 7.01
6.93 6.92 6.99 6.74
2.82 a 7.30 7.39 7.33 7.15
Financial satisfaction LB (1–5 scale) WVS Total 18+ Pop ACEb 13–17 £ 12
5.52 5.70 5.24 4.71
5.31 5.40 5.12 4.67
2.80 c 4.96 5.09 4.97 4.49
2.79 5.49 5.84 5.19 5.38
3.16 d 6.46 6.77 6.43 5.82
WVS Sample size ‘‘n’’ education % ACE 18+ % ACE 13–17 % ACE £ 12
974 746 30 50 20
992 672 28 48 25
1076 995 48 36 16
1268 1210 48 35 17
995 1002 54 29 17
a. 2000.5; b. ACE = age completed education; c. 1995.5; d. 2005
202
3.02 7.70 7.92 7.80 7.42
Brazil
1991
1996.5
2006
7.37 7.28 7.57
2.71 a 7.15 6.99 7.37
2.81 7.64 7.49 7.86
Financial satisfaction LB (1–5 scale) WVS Total Pop 100k+ <100k
5.51 5.28 5.85
3.05 b 5.48 5.39 5.60
3.19 c 5.87 5.78 6.04
WVS Sample size % 100k+ % <100k
1782 60 40
1149 58 42
1500 57 43
Life satisfaction LB (1–4 scale) WVS Total Pop 100k+ <100k
a. 2000; b. 1997; c. 2005
Chile Life satisfaction LB (1–4 scale) WVS Total Pop 100k+ <100k Financial satisfaction LB (1–5 scale) WVS Total Pop 100k+ <100k WVS Sample size % 100k+ % <100k
1989.5
1996
2000
2005
7.55 7.56 7.53
6.92 6.87 7.21
2.77 a 7.12 7.14 6.70
2.86 7.24 7.04 7.76
5.91 5.93 5.83
2.98 b 5.91 5.83 6.39
2.78 a 5.64 5.68 5.36
3.09 5.80 5.61 5.94
1500
1000
1200
1000
82 18
88 12
94 6
84 16
a. 2000.5; b. 1995.5
203
Mexico
1981
1989.5
1995
1999
2005 3.05 b 8.23 8.33 8.11
Life satisfaction LB (1–4 scale) WVS Total Pop. 100k+ <100k
7.97 n.a. n.a.
7.41 7.41 7.41
7.54 7.52 7.57
2.97a 8.13 8.12 8.17
Financial satisfaction LB (1–5 scale) WVS Total Pop. 100k+ <100k
n.a. n.a. n.a.
6.15 6.14 6.18
2.68 c 7.18 7.28 7.21
2.98 6.63 6.70 6.48
3.16 b 7.13 7.17 7.09
WVS Sample size % 100k+ % < 100k
n.a. n.a. n.a.
1531 79 21
2043 53 47
1292 45 55
1560 52 48
a. 2000.5; b. 2005; c. 1995.5
Peru
1995.5
2000.5
2006
Life satisfaction LB (1–4 scale) WVS Total Education High school+ Vocational, incomplete secondary None, primary
6.36 6.65 6.28 6.02
2.48 6.44 6.53 6.34 6.31
2.50 7.02 7.19 7.06 6.78
Financial satisfaction LB (1–5 scale) WVS Total Education High school+ Vocational, incomplete secondary None, primary
2.80 5.12 5.44 5.01 4.74
2.66 5.11 5.30 4.97 4.65
2.86 5.68 6.19 5.53 5.15
WVS Sample size % High school+ % Vocational, Incomplete secondary % None, primary
1191 36 45 19
1490 53 34 13
1490 34 42 24
204
Venezuela Life satisfaction LB (1–4 scale) WVS Total Pop.
100k+ <100k
Financial satisfaction LB (1–5 scale) WVS Total Pop. 100k+ <100k WVS Sample size % 100k+ % <100k
1995
2000
2005
6.72 6.79 6.65
3.25 7.52 7.55 7.43
3.44 n.a. n.a. n.a.
2.80 a 5.00 4.93 5.07 1200 53 47
3.01 6.19 6.20 6.17 1200 71 29
3.27 n.a. n.a. n.a. n.a. n.a. n.a.
a. 1995.5
Asia
China
1990
1995
2000.5
2007
2.82 a
2.67 c 3.68 e 6.76 7.08 6.61 6.32
Life satisfaction Gallup (1–4 scale) Asiabarometer (1–5 scale) WVS Total Education Vocational+ Primary No education
7.29 n.a. n.a. n.a.
6.83 7.01 6.80 6.49
2.78 b 3.73 d 6.53 6.53 6.42 6.74
Financial satisfaction WVS Total Education Vocational+ Primary No education
6.12 n.a. n.a. n.a.
6.11 6.28 6.09 5.78
5.65 5.63 5.72 5.56
5.94 6.23 5.89 5.39
996
1495 47 32 21
991 58 28 14
1959 48 26 26
WVS Sample size % Vocational+ % Primary % No education
a. 1997; b. 1999; c. 2004; d. 2003; e. 2006
205
India Happiness WVS Total Education
College Middle school or high school None, primary
WVS Sample size % College % Middle school or high school % None, primary
1995
2001
2006
3.04 3.24 3.14 2.86
2.95 3.14 3.06 2.83
3.02 3.27 3.18 2.85
2010 19 40 42
1968 14 34 52
1995 12 38 50
Japan
1981
1990
1995
2000.5
2005
Life satisfaction WVS Financial satisfaction WVS
6.58 6.14
6.53 6.03
6.61 6.33
6.48 6.17
6.99 6.16
South Korea
1980
1989.5
1996
2001
2005
Life satisfaction WVS Financial satisfaction WVS
5.33 5.17
6.69 5.75
n.a. 5.69
6.21 5.79
6.39 5.71
Turkey
1990
1996
2001
2007
Life satisfaction Eurobarometer (1–4 scale) WVS Financial satisfaction WVS
6.41 5.09
6.18 5.26
2.25 5.62 4.29
3.00 7.46 5.99
Africa Nigeria
1989.5
1995
2000
Life satisfaction WVS Total Pop Pop. 100k+ Pop. <100k
6.59 6.61 6.56
6.59 6.76 6.37
6.87 7.18 6.60
Financial satisfaction WVS Total Pop Pop. 100k+ Pop. <100k
5.51 5.51 5.51
5.68 5.87 5.48
6.28 6.59 6.01
WVS Sample size % 100k+ % <100k
997 58 42
1989 59 41
2022 48 52
206
South Africa
1990
1996
2001
2007
2.97d 6.20 5.86j
3.07e 5.59 5.48
3.07c 3.02f 5.81 5.68
3.13 7.03 7.02
2.67 5.40
2.88 4.99
2.94 5.25
3.01 6.80
3.75 n.a.
3.57 6.83
3.10 6.86
3.37 7.56
3.80 7.52
3.38 7.08
3.39 7.66
3.09 7.62
3.98 7.63
3.50 7.47
3.42 7.37
3.65 8.27
Financial satisfaction WVS Original Reweightedf
5.46 4.34
4.78 4.09
5.09 4.52
6.00 5.66
WVS Sample size % Black % Colored % Indian % White
2696 40 7 7 45
2927 55 13 7 25
2995 43 17 10 30
2977 69 10 4 17
Life satisfaction Total Pop. SA MarkDataa (1–5 scale) Q1 Q2 WVS Original WVS Reweightedg Black SA MarkData (1–5 scale) WVS Colored SA MarkData (1–5 scale) WVS Indian SA MarkData (1–5 scale) WVS White SA MarkData (1–5 scale) WVS
1981
3.13b 6.79 6.45h
a. In Q1 the neutral category, neither satisfied nor dissatisfied, is the last of five options; in Q2, the neutral category is the third out of five; b. 1983; c. 1999; d. 1988; e. 1995; f. 2002; g. The mean weights are black, 77 percent; colored, 9 percent; Indian, 3 percent; and white, 12 percent. The actual weights used here differ somewhat from these means, because we used the population distribution appropriate for each date; h. Assumes same difference (0.34) from WVS raw mean as in 1990; j. Reweighted value for total population includes an assumed value for colored population of 7.00, about the same difference from the white population as in 1996.
207
Appendix 7.B (continued) Regression results
Latin America
Life satisfaction (1984–2006) Argentina time
WVS only
LB only
Combined WVS only
LB only
Combined
0.00547 0.03027 [0.761] [0.028]*
0.01127 0.02237 [0.204] [0.120] 0.23406 [0.104] 2.65123 3.11133 [0.000]** [0.009]** 8 3 0.477 0.965
0.02252 [0.242]
0.00652 [0.663] 0.09953 [0.615] 2.73248 [0.000]** 13 0.027
wvs_dum Constant Observations R-squared
2.94131 [0.032]* 3 0.135
2.20465 [0.001]** 5 0.843
Life satisfaction (1991–2006) Brazil
WVS only
time
0.01311 0.01286 [.] [0.300]
LB only
wvs_dum Constant
Financial satisfaction (1984–2005)
3.26682 [.] Observations 2 R-squared 1
2.42714 [0.002]** 5 0.342
Financial satisfaction (1991–2005)
Combined WVS only 0.00219 [0.826] 0.37965 [0.020]* 2.67675 [0.000]** 7 0.934
208
2.41739 [0.000]** 10 0.166
LB only
Combined
0.002 [.]
0.01089 [0.178]
3.02441 [.] 2 1
2.82097 [0.000]** 10 0.214
0.00932 [0.196] 0.01693 [0.814] 2.85183 [0.000]** 12 0.21
Financial satisfaction (1989.5–2005)
Life satisfaction (1989.5–2006) Chile
WVS only
LB only
time
0.03268 0.0173 [.] [0.365]
Combined WVS only 0.00935 [0.585] 0.13778 [0.505] 3.05969 [0.001]** 7 0.683
wvs_dum Constant
3.49499 [.] Observations 2 R-squared 1
2.43351 [0.008]** 5 0.274
time
WVS only a
LB only
0.02376 [0.299]
0.01405 [0.416]
2.88317 [0.040] * 3 0.796
2.71973 [0.004] ** 5 0.228
wvs_dum Constant Observations R-squared
Combined
0.00011 0.00322 0.00272 [.] [0.781] [0.785] 0.2131 [0.082]+ 3.18242 3.01245 3.00255 [.] [0.000]** [0.000]** 2 10 12 1 0.01 0.451
Life satisfaction (1989.5–2006) Mexico
LB only
Financial satisfaction (1989.5–2005)
Combined WVS only 0.02096 [0.046] * 0.36632 [0.008] ** 2.55754 [0.000] ** 8 0.812
LB only
Combined
0.02993 [0.618]
0.05713 [0.016] *
3.10143 [0.133] 3 0.318
1.78783 [0.002] ** 9 0.587
0.04913 [0.014] * 0.87698 [0.000] ** 1.94619 [0.000] ** 12 0.773
a. WVS for residents of places with population 100,000 or more
Life satisfaction (1995.5–2006) Peru
WVS only
LB only
time
0.00523 [.]
0.0006 [0.985]
2.70612 [.] 2 1
2.56802 [0.035]* 5 0
wvs_dum Constant Observations R-squared
209
Combined 0.00153 [0.944] 0.25444 [0.166] 2.5182 [0.007]** 7 0.598
Financial (1995.5–2005) LB only 0.01344 [0.247]
3.02113 [0.000]** 10 0.163
Financial (1995–2005)
Life satisfaction (1995–2006) Venezuela
WVS only
LB only
Combined
LB only
time
0.05297 [.]
0.03097 [0.140]
0.01983 [0.248]
2.1128 [.] 2 1
2.61935 [0.006]** 5 0.57
0.03901 [0.031]* 0.07398 [0.465] 2.43114 [0.001]** 7 0.904
wvs_dum Constant Observations R-squared
2.63233 [0.000]** 10 0.163
Asia Financial Life satisfaction Life in Nation Surveys
Japan time 1958–1963 1964–1969
Kusago
China time Constant Observations R-squared
WVS
1970–
1992–
1958–
1978–
1981–
1981–
1969
1991
2007
2007
2002
2005
2005
0.01371 [0.046]* reference 0.19562 [0.001]**
0.00557 [0.004]**
0.01221 [0.002]**
0.01149 [0.000]**
0.00417 [0.321]
0.00135 [0.677]
2.70033 [0.000]** 12 0.803
2.52122 [0.000]** 25 0.302
3.186406 [0.000]** 14 0.571
0.00126 [0.492] reference 0.12089 [0.004]** 0.12398 [0.010]* 0.13113 [0.100]+ 2.74392 [0.000]** 51 0.216
3.12771 [0.000]** 10 0.801
2.72454 [0.000]** 5 0.319
3.24616 [0.000]** 5 0.066
1992–2007
Observations R-squared
WVS
1958–
1970–1991
Constant
satisfaction
Life (1995–2007)
Financial (1995–2007)
0.00155 [0.888] 2.93496 [0.040]* 3 0.031
0.0058 [0.785] 3.29797 [0.068]+ 3 0.11
South Korea time Constant Observations R-squared
210
Life (1980–2005)
Financial (1980–2005)
0.01044 [0.389] 2.57374 [0.004]** 4 0.374
0.00924 [0.104] 2.92133 [0.000]** 5 0.641
India time Constant Observations R-squared
Happiness (1995–2006) 0.00276 [0.790] 3.06245 [0.035]* 3 0.105
Turkey time Constant Observations R-squared
Life (1990–2007)
Financial (1990–2007)
0.01596 [0.550] 2.50908 [0.029]* 4 0.202
0.01468 [0.657] 2.5759 [0.044]* 4 0.117
Africa Nigeria time Constant Observations R-squared
South Africa time Constant Observations R-squared
Life (1989.5–2000)
Financial (1989.5–2000)
0.01802 [0.197] 2.68181 [0.021]* 3 0.908
0.04527 [0.166] 2.54075 [0.047]* 3 0.934
Life (1981–2007)
Financial (1990–2007)
0.00371 [0.777] 2.6439 [0.001]** 5 0.031
0.03551 [0.174] 1.97063 [0.027]* 4 0.683
Note: p-values in bracket; + significant at 10%; * significant at 5%; ** significant at 1%
211
Appendix 7.C Data sources, survey dates, and date of GDP observations with which each survey is paired
Life satisfaction
Financial satisfaction
Survey dates Latin America Argentina WVS 10
LB 4
start
end
Survey dates Paired with GDP for:
start
end
Paired with GDP for:
1984 1984 199102 199104 199508 199509 199901 199902 2006 2006
1984 1990 1995 1998 2006
WVS 10
1984 1984 199102 199104 199508 199509 199901 199902 2006 2006
1984 1990 1995 1998 2006
200104 200307 200405 200508 200610
2000.5 2002.5 2003.5 2005 2006
LB 5
199505 199606 199711 199811 200001 200104 200204 200307 200405 200508
1994.5 1995.5 1997 1998 1999 2000.5 2001.5 2002.5 2003.5 2005
200105 200308 200406 200509 200611
199506 199607 199712 199811 200002 200105 200205 200308 200406 200509
(continued )
212
(continued) Life satisfaction
Financial satisfaction
Survey dates Latin America Brazil WVS 10
LB 4
Chile WVS 10
LB 4
Mexico WVS 10
LB 4
start
end
Survey dates end
Paired with GDP for:
199201 199706 2006 199506 199607 199712 199811 200002 200104 200205 200308 200406 200509
1991 1996.5 2006 1994.5 1995.5 1997 1998 1999 2000 2001.5 2002.5 2003.5 2005
199005 199005 1996 1996 200011 200011 2005 2005
1989.5 WVS 10 199005 199005 1996 1996 1996 2000 200011 200011 2005 2005 2005
1989.5 1996 2000 2005
200104 200307 200405 200508 200610
2000.5 LB 5 2002.5 2003.5 2005 2006
199505 199606 199712 199811 200003 200105 200205 200308 200406 200509
1994.5 1995.5 1997 1998 1999 2000.5 2001.5 2002.5 2003.5 2005
199005 199005 199509 199603 200001 200002 2005 2005
1989.5 WVS 10 199005 199005 1995 199509 199603 1999 200001 200002 2005 2005 2005
1989.5 1995 1999 2005
200104 200105 200307 200308 200405 200406
2000.5 LB 5 2002.5 2003.5
1994.5 1995.5 1997
199110 199201 199706 199706 2006 2006 200104 200104 200308 200405 200406 200508 200509 200610 200610
200105 200308 200406 200509 200610
Paired with GDP for: 1991 1996.5 2006 2000 2002.5 2003.5 2005 2006
start
WVS 10 199110 199706 2006 LB 5 199506 199606 199712 199811 200001 200104 200205 200308 200405 200508
199504 199606 199712 199811 200003 200104 200204 200307 200405 200508
199505 199505 199606 199606 199711 199801
(continued)
213
(continued) Life satisfaction Survey dates start
Peru WVS 10
LB 4
end
Paired with GDP for:
Survey dates start
end 199812 200002 200105 200205 200308 200406 200509
Paired with GDP for:
200508 200509 200610 200610
2005 2006
199812 200001 200104 200204 200307 200405 200508
199605 199605 200107 200107 2005 2005
1995.5 2000.5 2005
WVS 10 199605 199605 200107 200107 2005 2005
1995.5 2000.5 2005
200104 200307 200405 200508 200610
2000 2002.5 2003.5 2005 2006
LB 5
1994.5 1995.5 1997 1998 1999 2000 2001.5 2002.5 2003.5 2005
1995 2000
WVS 10 199603 199604 200011 200012
1995 2000
2000 2002.5 2003.5 2005 2006
LB 5
1994.5 1995.5 1997 1998 1999 2000 2001.5 2002.5 2003.5 2005
200104 200308 200405 200509 200610
Venezuela WVS 10 199603 199604 200011 200012 LB 4
Financial satisfaction
200104 200307 200405 200508 200610
200104 200308 200406 200509 200610
214
199505 199606 199712 199811 200002 200104 200204 200307 200405 200508
199505 199606 199712 199811 200002 200104 200204 200307 200405 200508
199506 199606 199712 199811 200002 200104 200205 200308 200405 200509
199506 199607 199712 199812 200002 200104 200205 200308 200406 200509
1998 1999 2000.5 2001.5 2002.5 2003.5 2005
Life satisfaction Survey dates Asia China WVS 10
G4
India WVS 10
WVS 4 (Happy)
Japan WVS 10
K5
SW 4(a) SW 4(b) SW 4(c) SW 4(d)
Financial satisfaction
start
end
199007 1995 200103 2007 1997 199905 200411
199012 1995 200106 2007 1997 199905 200411
1990 1995 2000.5 2007 1997 1998.5 2004
199007 1995 2006 1995 200108 2006
199012 1995 2006 1995 200110 2006
1990 1995 2006 1995 2001 2006
1981 1981 199009 199009 1995 1995 200007 200007 2005 2005 1978, 1981, 1984, 1987, 1990, 1993, 1996, 1999, 2002, 2005
Survey dates
Paired with GDP for:
1981 1990 1995 2000.5 2005 1978, 1981, 1984, 1987, 1990, 1993, 1996, 1999, 2002, 2005 1958–1963 1964–1969 1970–1991 1992–2007
215
start
end
Paired with GDP for:
WVS 10 199007 199012 1995 1995 200103 200106 2007 2007
1990 1995 2000.5 2007
WVS 10
1981 1990 1995 2000.5 2005
1981 1981 199009 199009 1995 1995 200007 200007 2005 2005
Life satisfaction Survey dates start
end
Financial satisfaction
Paired with GDP for:
Survey dates start
South Korea WVS 10 198103 198103 1980 199006 199007 1989.5
WVS 10 198103 198103 1980 199006 199007 1989.5 1996 1996 1996 200111 200111 2001 2005 2005 2005
200111 200111 2001 2005 2005 2005 Turkey WVS 10
EB 10 Africa Nigeria WVS 10
199011 199612 200109 2007 200501
199005 1995 200010 South Africa WVS 10 198110 199010 1996 200103 2007 M
199101 199701 200201 2007 200502
end
Paired with GDP for:
1990 1996 2001 2007 2004
WVS 10 199011 199101 1990 199612 199701 1996 200109 200201 2001 2007 2007 2007
199006 1989.5 1995 1995 200011 2000
WVS 10 199005 199006 1989.5 1995 1995 1995 200010 200011 2000
198110 199011 1996 200105 2007
WVS 10 199010 199011 1990 1996 1996 1996 200103 200105 2000.5 2007 2007 2007
1981 1990 1996 2000.5 2007 1983, 1988, 1994–95
Key WVS = World Values Survey LB = Latinobarometer G = Gallup SW = ‘‘Life in Nation’’ survey as reported in Stevenson & Wolfers (2008) EB = Eurobarometer M = Mark Data Numbers following data source are number of response categories. Letters a, b, c, and d indicate change in wording of response options. We did not include ‘‘life satisfaction’’ in the Latinobarometer for 1997 and 2000 because the question differs from that for later dates.
216
Chapter 8 Happiness Adaptation to Income Beyond ‘‘Basic Needs’’ Rafael Di Tellaa and Robert MacCullochb a
Harvard Business School, NBER and CIfAR Imperial College London and CEPR
b
We thank John Helliwell, as well as seminar participants at the Princeton Conference ‘‘Understanding National Differences in Well-Being,’’ October 2008.
In a 1974 paper, Richard Easterlin found that for the United States, a measure of subjective well-being (like happiness) did not increase appreciably in the post–World War II period, in spite of large increases in per capita income. This finding was observed for other periods and other nations in Europe and even Japan, and has become a ‘‘stylized fact’’ of happiness research. And researchers who do detect positive trends in some of the happiness time-series have to face up to the fact that these tend to be small.1 This stylized finding is consistent with happiness adaptation to income in the long run in these nations. A problem with this general conclusion, however, is that we do not have a long time-series of happiness in a poor country. But if a flat time-series was present in poor nations it should also lead to no positive relationship between happiness and income across nations, given that most international income differences are long established. But the evidence is not supportive of this conjecture. People in poor countries, like Nigeria, report lower levels of happiness than in rich countries, like the United States (e.g., Veenhoven, 1991; Diener et al., 1995; Inglehart, 1997; and Deaton, 2008). Figure 8.1 shows the pattern reported in Inglehart & Klingemann (2000). Such work suggests that a positive cross-country relationship should be a second ‘‘stylized fact’’ of happiness 217
218
Section II: Comparisons of Income and Well-Being Through Time
Mean of percent Happy and percent Satisfied with life as a whole
100 95 90 85 80 75 70 65 60
N. Nether- Iceland Ireland Denmark Switzerland lands Ireland Norway Finbind Sweden U.S.A. Australia New Belgium Puert Britain Canada Zealand Italy Taiean South Kones West Germany France Colombia Philippines Brazil Japan Venezuel Austria Spain Mexico Ghania East Nigeria Germany China Dom.Rep. ArgentinaChile Portugal BangPoland Crech Pakistan badesh India Turkey Slovenia S. Africa Croatia Slovakia YugoHungar slavia Macedonia Peru Azerbaija Latvia
55
Estonia
50 45
Romaina Georgia Lithuania Aurmenia Bulgaria
40 Russia
35 30
Ukraine Belarus Moldova
1000
5000
9000
13000
17000
21000
25000
GNP / capita (World Bank purchasing power parity estimates, 1995 U.S.)
Figure 8.1 Subjective Well-being (from World Values Survey) versus GNP per capita. Source: Inglehart and Klingemann, ‘Genes, Culture and Happiness’, 2000.
research.2 The explanation presented in Veenhoven (1991) is that more income improves happiness only until basic needs are met. Beyond the point where there is enough income so that people are no longer hungry and absolute poverty has been eliminated, income does not matter for happiness. That is, once wealthy countries have satisfied basic needs, they are on the ‘‘‘flat of the curve,’ with additional income buying little if any extra happiness’’ (see Clark, Frijters, & Shields, 2008). Fitting these two ‘‘facts’’ (flat time-series and positive slope in the crosssection of countries) has become an important challenge for happiness researchers.3 Note that, within the ‘‘second stylized fact’’ (the cross-sectional finding) there is at least some evidence that the relationship is not linear (although the extent of this is not agreed upon). More recent evidence in
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
219
favor of the proposition that income depends on (the logarithm of) the level of income across countries is due to Deaton (2008), using the latest Gallup World Poll. The purpose of this chapter is to find out whether higher income has long-lasting (historical) impacts on happiness or whether these effects dissipate over time. We use two different strategies. In the first, we focus on panels (of countries and of people). We estimate regressions of the form, Happinessg ¼ ð0 log yg þ 1 log y 1 g þ :: T g log y T g Þ þ eg , where there are ‘‘T’’ lags of (each individual’s or country’s) income, y. The possibility that, once basic needs are satisfied, income has no long-lasting effects on happiness, can be tested by partitioning our data by wealth level, g = poor or g = rich. If the levels and lags of income sum to a positive number for the ‘‘poor’’ and to zero for the ‘‘rich,’’ then we will be able to explain both the observed time-series and cross-sectional patterns between happiness and income. (For a continuum of wealth levels, we just interact each of the levels and lags of income with each level of wealth). The individual panel data from the German Socioeconomic Panel shows that home-owners, who are presumably wealthier than renters, adapt fully to the effects of higher income (after around seven years), whereas renters do not. We also find evidence consistent with full adaptation to (the logarithm of) GDP per capita in the wealthy countries within Europe, though not in the poor. However, even in the wealthy developed countries adaptation may take at least five years, so the happiness gains that they experience from higher income levels, whilst not permanent, can still be very long-lasting. In the second approach, we focus on the cross-section of countries included in the 2005 Gallup World Poll. Using the well-being question from 2005, but data for GDP per capita for 1960–2005, we can evaluate different theories connecting income to well-being. For example, we can estimate the correlation between happiness (measured by the Cantril ‘‘Ladder of Life’’ question) and the growth rate, conditional on initial levels of GDP per capita. We find that the past 45 years of economic growth (from 1960 to 2005) in the rich nations of the world have not brought happiness gains above those that were already in place once the 1960s standard of living had been achieved. However, in the poorest nations, we cannot reject the null hypothesis that the happiness gains they have experienced from the past half century of economic growth have been the same as the gains from growth prior to the 1960s. In other words, for these nations, it is still only the absolute level of (the logarithm of) income that matters for happiness. A small literature in economics has emerged following Easterlin’s paper.4 Of particular interest, given our focus on adaptation effects, are the models of Pollak (1970), Wathieu (2004), and Rayo & Becker (2007), inter alia. Although habituation has been studied in the context of
220
Section II: Comparisons of Income and Well-Being Through Time
physical disability, marriage, divorce, unemployment, and other life circumstances, our interest is primarily in habituation to income changes. An influential paper in psychology by Brickman, Coates, and JanoffBullman (1978) showed that individuals who had won between $50,000 and $1,000,000 at the lottery the previous year reported comparable life satisfaction levels as those who did not.5 Frederick & Loewenstein (1999) and Diener & Diener (2002) present reviews of the evidence available, gathered largely by psychologists. Recent studies of habituation using happiness data include Clark (1999) on how (job) satisfaction adapts to changes in wages, Di Tella et al. (2003), who estimate the effect of income lags on happiness in a panel of 12 OECD countries, and Gardner & Oswald (2001) who use data on a panel of individuals who receive windfalls (by winning a lottery or receiving an inheritance). Di Tella, Haisken-De New, & MacCulloch (2005) regress life satisfaction on current and several lags of personal income (and on current and several lags of status) and find that full adaptation occurs to income after about four years (but not to higher levels of status). Our explanation is related to the work on satisfaction with income (rather than with life) by van Praag & Kapteyn (1973) showing that income aspirations rise in proportion to income (sometimes called ‘‘preference drift’’). Indeed, van de Stadt, Kapteyn, & van de Geer (1985) find that the hypothesis of one-for-one changes in income aspirations and income cannot be rejected (see also van Praag & Ferrer-i-Carbonell, 2004; & Stutzer, 2004). More recently, Easterlin (2003) argued that family aspirations do not change as marital status and family size change, but that material aspirations increase commensurately with household wealth. Two caveats are in order. We do not devote too much space to the development of a careful definition of ‘‘basic needs,’’ although we do think that an exact empirical definition would be of considerable value; in particular, one that clarifies the extent to which basic needs are socially determined. Instead, our approach is to split the sample across relatively rich and poor sub-samples in each of the data sets that we use. Given the quality of wealth data, we use broad categories, such as top and bottom half of the samples using withinsample income data or home ownership status. Of course, with the sample split this way, it is imprecise to say that the bottom half have not met their ‘‘basic needs’’ (for more on this see, Di Tella & MacCulloch, 2008b). The second caveat is that we will focus on three different data sets, and the wellbeing questions available in each of them vary somewhat. The precise wording of the questions is important because it is possible that they tap into different emotions. For economists who think these emotions aggregate into a summary measure of utility, this is less problematic (although for evidence that the measures available for contentment and happiness correlate
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
221
differently with macroeconomic variables and age, see Di Tella & MacCulloch, 2008a; and Deaton, 2008; see also Konow & Early, 2008, for evidence on material motivation using well-being data). The rest of the chapter is organized as follows. Section 2 discusses the data and empirical strategy used to quantify the behavioral effects (at least in the sub-samples where they are found to exist). Section 3 presents the results. The final section concludes. 2. Data and Empirical Strategy
Each of the data sets, and the empirical strategy that we use, is described in the following sub-sections. a. Individual-Level Panel Data: the German Socio-Economic Panel, 1984–2000
We use the German Socio-Economic Panel, a longitudinal data set begun in 1984 that randomly samples households living in the western states of the Federal Republic of Germany. In 1990 the eastern states were added to provide a representative sample of the (reunited) Germany, although in this paper we concentrate only on the West German sample. The survey contains the following ‘‘happiness’’ question: ‘‘In conclusion, we would like to ask you about your satisfaction with your life in general: please answer according to the following scale: 0 means completely dissatisfied and 10 means completely satisfied: How satisfied are you with your life, all things considered?’’ The possible answers appear on a scale showing the numbers ‘‘0 1 2 . . . 9 10.’’ The words ‘‘Completely dissatisfied’’ correspond to ‘‘0’’ and ‘‘Completely satisfied’’ correspond to ‘‘10.’’ The second key variable that we use is a measure of each individual’s income, y = Personal Income. There are several different income-related questions in the survey that are relevant to this measurement. We use ‘‘Real Household Post-Government Income’’ from the Cross-National Equivalent File. Table 8.A shows the summary statistics and the Appendix 8.A provides a richer description of the German Socio-Economic Panel’s sampling methods. We run a series of regression specifications that are based on the following general form: Happinessi ; t g ¼ ð0 g log yi ; t g þ 1 g log yi ; t 1 g þ :: T g log yi ; t T g Þ ð1Þ þ g X i ; t g þ f i g þ t g þ ei ; t g where lags on (the logarithm of) income, log yitg are used to explain (current) life satisfaction levels, Happinessi,tg, of individual, i, at year, t.6
222
Section II: Comparisons of Income and Well-Being Through Time
TABLE 8.A Summary Statistics for the German Socio-Economic Panel: 1985–2000 Variable
Units
No. of Obs.
Mean
Std dev
Happiness – between – within Personal income – between – within log(Personal income) – between – within
0–10 scale
Total = 29,852 n = 4,987 t = 5.9
7.13
1.67 1.40 1.05
DM 1995
Total = 29,852 n = 4,987 t = 5.9
61,974
32,958 150 30,462 176 13,377 143,155
724,403 438,790 347,586
Total = 29,852 n = 4,987 t = 5.9
10.90
0.55 0.53 0.24
13.49 12.93 13.62
log(1995 DM)
Min. 0 0 0.86
5.01 5.17 6.46
Max. 10 10 13.91
Note: All variable definitions are in the appendix.
The level of income is measured by the logarithm of real (net) household income from all sources during the current year. Consequently, the equation measures the degree to which people’s happiness adapts to income over time. We also include an unobserved individual and year fixed effect, fi g and t g, respectively. The maximum number of lags used is T = 7, and the equation is estimated for different groups, g, of people. The (i.i.d.) random noise term is ei,tg. The vector Xi,tg consists of individual characteristics: Marital state (a set of dummies depending on whether the respondent is married, divorced, separated, or widowed), Employment state (a set of dummies depending on whether the respondent is retired, at school, at home, in the military, selfemployed, or a public servant) and Education (a set of dummies measuring their level of high school achievement, vocational training, or college degree). Data on all of these variables exist for a sample of 4,987 West Germans from 1985 to 2000. Estimation is done using an Ordinary Least Squares (fixed-effects) model, although similar conclusions emerge when a more flexible cardinalization is used (see Ferrer-i-Carbonell & Frijters, 2004, for a discussion, and the results in Kohler, Behrman, & Skytte, 2005; see also the approach in Di Tella & MacCulloch, 2005). The formal hypothesis that we use to test for adaptation effects is: Ho :
T X i¼ 1
gi ¼ 0
versus
H1 :
T X
gi 6¼ 0
ð2Þ
i¼ 1
If individuals adapt to the effects of the higher income levels, then we would expect the sum of the lagged income coefficients (above) to be negative.
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
223
The formal hypothesis that we use to test for long-run income effects is: Ho :
T X i ¼0
gi ¼ 0
versus
H1 :
T X i ¼0
gi 6¼ 0
ð3Þ
If there are long-lasting effects of income on happiness, then we would expect the sum of the present and lagged income coefficients (above) to be positive. The two groups, g, of particular interest to us are the wealthy and the poor. Due to the scarcity of wealth data, the proxy that we use in the German Socio-Economic Panel is whether or not the individual owns their own home. The survey question that was used to generate this variable asks: ‘‘Are you a tenant or an owner? 1. Tenant or 2. Owner.’’ That is, g = Tenant or g = Owner, depending on the person’s response.
b. Pooled Cross Country-Time Series Data: The Eurobarometer Cumulative Surveys, 1975–2002
We also use pooled cross-country time series data from the Eurobarometer Survey Series. These surveys interview a random sample of Europeans during the 28-year period from 1975 to 2002, and ask a series of socioeconomic questions. The main question of interest asks: ‘‘On the whole, are you very satisfied, fairly satisfied, not very satisfied or not at all satisfied with the life you lead?’’ (The small ‘‘Don’t know’’ and ‘‘No answer’’ categories are not studied here). Data are available on this question for 605,020 people in 16 countries for whom a complete set of data on a large number of personal characteristics is also available. Table 8.B shows the summary statistics, and the Appendix 8.A provides a richer description of the Eurobarometer’s sampling methods. We run a series of regression specifications that are based on the following general form: Happinessi ; c ; t g ¼ ð0 g log yc ; t g þ 1 g log yc ; t 1 g þ :: T g log yc ; t T g Þ þ g Z i ; c ; t g þ kc g þ YEARc g þ ft g þ "i ; c ; t g ð4Þ where lags on (the logarithm of) real GDP per capita, yc,tg, are used to explain the (current) life satisfaction level, Happinessi,c,tg, of individual, i, in country, c, and year, t. We also include an unobserved country and year fixed effect, kcg, and ftg, respectively. Finally YEARcg denotes a countryspecific time trend. The maximum number of lags used is T = 7 and the equation is estimated for different groups, g, of people. The (i.i.d.) random
224
Section II: Comparisons of Income and Well-Being Through Time
TABLE 8.B Summary Statistics for the European Pooled Cross-Section Time Series: 1975–2002 Std dev.
Min.
Max.
3.07
0.28 0.27 0.08
1 2.63 2.84
4 3.57 3.29
Total = 307 n = 16 t = 19.2
19,012
6,913 6,605 3,466
6,424 9,059 6,050
48,655 34,011 33,656
Total = 307 n = 16 t = 19.2
9.79
0.35 0.35 0.17
8.77 9.10 9.34
10.79 10.40 10.37
Variable
Units
No. of Obs.
Mean
Happiness – between –within GDP per capita – between – within log(GDP per capita) – between – within
1–4 scale
Total = 605,020 n = 31,511 t = 19.2
2000 US $
Log(2000 US $)
Note: All variable definitions are in the appendix.
noise term is Ei,c,tg. The vector Zi,c,tg consists of individual characteristics: Marital state (a set of dummies depending on whether the respondent is married, divorced, separated, or widowed); Employment state (a set of dummies depending on whether the respondent is retired, at school, at home, in the military, self-employed, or a public servant); and Education (a set of dummies measuring the respondent’s level of low, middle, or higher education). Estimation is done using an Ordered Probit Model. If individuals adapt to the effects of higher levels of GDP per capita, then we would expect the sum of the lagged income coefficients to be negative. The formal hypothesis test that we use is: Ho :
T X
gi ¼ 0
versus
H1 :
T X i ¼1
i¼ 1
gi 6¼ 0
ð5Þ
If there are long-lasting effects of income on happiness, then we would expect the sum of the present and lagged income coefficients to be positive. The formal hypothesis test that we use is: Ho :
T X i¼ 0
gi ¼ 0
versus
H1 :
T X i ¼0
gi 6¼ 0
ð6Þ
The two groups of particular interest to us are the wealthy and the poor. As we do not have wealth data for individuals in the Eurobarometer Surveys, we use the initial level of GDP per capita at the start of the sample period in 1975 as a proxy for the average level of wealth for individuals in their respective countries. We divide the sample into two
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
225
halves. That is, g = Bottom half of individuals as ranked by their country’s GDP per capita in 1975, or g = Top half of individuals as ranked by their country’s GDP per capita in 1975. c. Cross Country Data: The World Gallup Poll, 2005
The final test we use to identify whether higher levels of GDP per capita lead to long-lasting impacts on happiness, or whether these effects disappear over time once basic needs have been satisfied, uses cross-sectional data. These come from the cross-country surveys of subjective well-being in the 2005 Gallup World Poll, measured consistently across 132 countries. Similar questions were asked in all countries, and the survey contains data for each country that are nationally representative of people aged 15 and older (with sample size close to 1,000 in each country). The survey asks a variety of subjective well-being questions. The one we use is the Cantril ‘‘Ladder of Life’’ question that asks ‘‘Please imagine a ladder with steps numbered from zero at the bottom to ten at the top. Suppose we say that the top of the ladder represents the best possible life for you and the bottom of the ladder represents the worst possible life for you. If the top step is 10 and the bottom step is 0, on which step of the ladder do you feel you personally stand at the present time?’’ We use country averages of the well-being question. Table 8.C shows the summary statistics and Appendix 8.A provides a richer description of the Gallup World Poll’s sampling methods. We can indirectly test for the importance of adaptation to income versus basic needs using the Gallup Poll by regressing well-being ‘‘now’’ on TABLE 8.C
Summary Statistics for the Gallup World Poll: 2006
Variable
Units
Cantril Ladder of Life
0–10 scale 2000 US $
GDP per capita in 1960 log (GDP per capita in 1960) GDP per capita in 2005 log (GDP per capita in 2005) log (1 + growth rate of GDP per cap 1960–2005)
2000 US $
No. of Obs.
Mean
Std dev.
Min.
Max.
76
5.60
1.25
3.24
8.00
76
2909
3874
98
18,580
76
7.05
1.45
4.58
9.83
76
9128
11,818
100
40,597
76
7.91
1.78
4.61
10.61
76
0.86
0.73
0.70
Note: All variable definitions are in the appendix.
2.89
226
Section II: Comparisons of Income and Well-Being Through Time
historical past levels of GDP almost half a century ago—and also the growth in income over the past half century. That way we can see whether the latter has bought extra happiness.7 First note that:
log GDP2005
DGDP2005 T ¼ log GDPT ð1 þ Þ GDPT
ð7Þ
where GDPT is GDP per capita in year, T; and DGDP2005T = GDP2005GDPT. We then estimate the following OLS, setting T = 1960: Happinessg2005;c ¼ log GDPg1960;c þ log ð1 þ GDPGrowthRateg20051960;c Þ þ "gc
ð8Þ
where Happiness2005,cg is the average happiness (on the 0–10 scale) across individuals living in country, c, in year, 2005. We use measures of real GDP per capita (in year 2000 U.S. dollars) from the World Development Indicators as our proxy for average income in each country.8 The GDP Growth Rate2005–1960 is defined as DGDP2005–1960/GDP1960. The (i.i.d.) random noise term is ecg, and the equation is estimated for different groups, g, of countries. We test the null hypothesis that current happiness depends solely on the logarithm of the current level of GDP per capita (i.e., the Deaton, 2008, hypothesis): H0 :
¼
versus
H1 : 6¼
ð9Þ
Rejection of the above null hypothesis in favour of a>b would imply that people may not be experiencing happiness gains from the more recent growth of their living standards (since the 1960s) much above the happiness gains that were already in place in 1960. In this case, basic needs would already have been (partially) satisfied in 1960, so that further increases in income since the 1960s may have been subject to the adaptation effect. We divide the sample into two halves, g = poor and g = rich countries, to check whether this is the case for the rich countries (i.e., a>b for g = rich), whereas in the poor countries people may still be trying to satisfy basic needs (i.e., a = b for g = poor).9 3. Main Results: Adaptation to Income and Basic Needs a. Individual-Level Panel Data: German Socio-Economic Panel
The first column in Table 8.1 presents a benchmark estimate of the effect of the level of the logarithm of current income on the current level of life
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
TABLE 8.1
Happiness, West Germany, 1985–2000: Adaptation to Income All
Dependent Variable: Happiness (Life Satisfaction 0–10) Current level of real income log (Personal Income in year t)
Wealth Proxy
(1)
(2)
Tenant (3)
Owner (4)
0.21 (0.02)
0.26 (0.03)
0.27 (0.05)
0.25 (0.05)
0.01 (0.04) 0.10 (0.04) 0.02 (0.04) 0.02 (0.04) 0.02 (0.04) 0.07* (0.04) 0.03 (0.03)
0.07 (0.06) 0.16 (0.06) 0.08 (0.06) 0.08 (0.06) 0.12 (0.06) 0.15 (0.06) 0.03 (0.05)
0.06 (0.06) 0.03 (0.06) 0.13 (0.06) 0.05 (0.06) 0.12* (0.07) 0.05 (0.06) 0.12 (0.05)
0.21 0 0.05 0.33
0.09 0.28 0.18 0.02
0.18 0 0.07 0.26
0.03
0.04
0.01
Past levels of real income log (Personal Income t-1) log (Personal Income t-2) log (Personal Income t-3) log (Personal Income t-4) log (Personal Income t-5) log (Personal Income t-6) log (Personal Income t-7) Results of F tests Income Lags Prob ( Lags > F) Current & Lagged Income Prob ( Current & Lagged Income>F) R2 overall
227
0.03
Note: [1] All regressions are robust OLS and include individual and year fixed effects. Starred tests are significant at the 10 percent level. Tests in bold face are significant at the 5 percent level. Total no. of observations is 29,852, individuals is 4,987 and mean years is 6.0 for col. (1); 29,852 observations, 4,987 individuals and 6.0 mean years for col. (2); 14,951 observations, 2,603 individuals and 5.7 mean years for col. (3); 14,901 observations, 2,384 individuals and 6.3 mean years for col. (4) [2] Dependent variable is the individual responses to the question: ‘‘Please answer according to the following scale: 0 means completely dissatisfied and 10 means completely satisfied: How satisfied are you with your life, all things considered?’’. Personal Income is real household net income. [3] Wealth is proxied by whether you own your own house or have been renting during the sample period. All regressions include controls for employment status, personal income, education and marital status.
satisfaction in Germany, together with individual and year fixed effects, as well as a set of personal characteristics. That is, it estimates equation (1) but restricting all lagged coefficients, a-t, to be zero. The coefficient is positive and highly significant—suggesting that the log of income is a significant determinant of happiness. However, the size of the effect is not large: a doubling of income would move one up just 0.15 units on the 0–10
228
Section II: Comparisons of Income and Well-Being Through Time
happiness scale. Note that the summary statistics reported in Table 8.A show that happiness has a total standard deviation equal to 1.67 (the between- equals 1.40 and the within- equals 1.05). Thus, a one-standarddeviation increase in log income accounts for 6.9% of a standard-deviation increase in happiness (0.21*0.55/1.67). In column 2, we add an arbitrary number of lags of each individual’s income. To keep it general, we include seven (but see the discussion below).10 The coefficient on current income is still positive and significant. One measure of the amount of adaptation to income in the sample is captured by the sum of the coefficients on the lags (from one to seven). The sum is negative, and an F-test shows that it is significant at the 1 percent level, which allows us to reject the hypothesis of no adaptation (see equation 1 with T = 7). The sum of the lagged coefficients is equal to 0.21 (i.e., 0.01 0.10 0.02 0.02 + 0.02 0.07 0.03). Consequently, of the initial impact of income, 80.8% is lost over the ensuing seven years (i.e., 0.21/0.26) leaving a long-run effect of 0.05. Put another way, although the current effect of income from this specification suggests that a rise in average real incomes of 12% (from 56,429 DM in 1986 to 63,042 DM in 2000) would have added 0.03 units onto happiness scores (i.e., 0.26*log(1.12)) adaptation effects reduce the size of the effect to only 0.01 units (i.e., 0.05*log(1.12)). We are also not able to reject the conclusion that adaptation to income after seven years of time is total. An F-test of whether the sum of all eight coefficients on income (i.e., current and seven lags) is equal to zero cannot be rejected at the 1 percent level of significance.11 Note that in Di Tella, Haisken-De New, & MacCulloch (2005), where each individual’s job status is also used as an explanatory variable, the period of time over which we cannot reject full adaptation reduces to four years. Our approach allows us to provide estimates across different subgroups.12 For example, we can estimate and compare adaptation to income amongst the poor and the wealthy. As accurate wealth data are not easily available, we use whether a person is a tenant or a home-owner as a proxy for their wealth level. In column 3 the hypothesis that renters do not adapt to higher levels of (the logarithm of) income after seven years cannot be rejected at conventional levels. And the hypothesis that they stop enjoying higher happiness levels after seven years due to an increase in the level of their personal income can be rejected at the 2 percent level. In other words, we can reject full adaptation to the effects of higher income for the tenants in our sample (i.e., the sum of the coefficients on the current and seven lags of income is 0.18, which the F-test shows is significant at the 2 percent level). Turning to column 4, the hypothesis that home-owners do not adapt to higher levels of (the logarithm of) income after seven years can now be rejected at the 1 percent level. But the hypothesis that they stop
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
229
enjoying higher happiness levels after seven years from an increase in the level of their personal income cannot (in contrast to the result for tenants) be rejected at conventional levels. In other words, we cannot reject the hypothesis of full adaptation to the effects of higher income for the homeowners in our sample—but we can for the tenants. Causality
Part of the interest of the results in the above section (estimating differential adaptation across the rich and poor in a panel of individuals) is that the classic concern in this type of setting is whether personality traits may be driving the connection between happiness and income. One answer is to employ a panel of individuals, so that the inclusion of individual fixed effects can deal with the special case of fixed traits, such as ability, preferences, personality, or family background.13 However, there is still the possibility that time-varying shocks to happiness may later change an individual’s income.14 Such shocks are closer to measures of happiness and positive affect than to those of overall life satisfaction and Cantril’s ‘‘Ladder of Life’’ that we employ in this paper. Work by Lucas et al. (1996) suggests that overall life satisfaction is not influenced by the affective state of the person at the time of the interview.15 Permanent (versus Transitory) Income
Another possible caveat regarding the interpretation of our results concerns the question of whether people are consuming on the basis of their lifetime (or ‘‘permanent’’) income. In this case we may expect to see approximately flat life satisfaction over time due simply to income smoothing (i.e., people consuming based on their permanent income). An income shock may yield some rise in satisfaction but just to the extent to which it is perceived as being long-lasting (versus transitory). This relationship may also be affected by the extent to which individual responses to the German Socio-Economic Panel ‘‘life satisfaction’’ question are forward-looking (versus instantaneous assessments of well-being). However, these kinds of considerations would not necessarily explain why there exist differences across the rich and poor subgroups within Germany, as they should affect both of these groups similarly.
b. Pooled Cross Country-Time Series Data: The Eurobarometer Cumulative Surveys
The first column in Table 8.2 presents a benchmark estimate of the effect of the level of (the logarithm of) current income on the current level of life
230
Section II: Comparisons of Income and Well-Being Through Time
TABLE 8.2 Happiness, 16 European Countries, 1975–2002: Adaptation to Income Wealth Proxy: 1975 GDP Dependent Variable: Happiness (Life Satisfaction 1–4) Current level of real income log (GDP per capita t)
All Countries
Top half (4)
1.28 (0.40)
1.82 (0.50)
1.35 (0.54)
0.29 (0.70) 0.92 (0.44) 0.61 (0.60) 0.55 (0.48) 0.38 (0.32)
0.63 (0.81) 1.66 (0.46) 0.79 (0.72) 1.82 (0.36) 1.12* (0.61)
0.77 (0.79) 0.06 (0.44) 0.46 (0.44) 0.002 (0.47) 0.24 (0.41)
1.07 0.28 0.21 0.31
1.13 0.0 0.69 0
1.53* 0.1 0.18 0.56
0.09
0.06
(2)
0.65 (0.52)
Part levels of real income log (GDP per capita t-1) log (GDP per capita t-2) log (GDP per capita t-3) log (GDP per capita t-4) log (GDP per capita t-5) Results of F tests Income Lags Prob ( Lags > F) Current & Lagged Income Prob ( Current & Lagged Income>F) R2 overall
Bottom half (3)
(1)
0.09
0.09
Note: [1] All regressions are ordered probits and include controls for age, sex, employment status, personal income, education and marital status at the micro level, and for unemployment and inflation rates at the macro level. A complete set of country and year fixed effects is also included, as well as country-specific time trends. Starred tests are significant at the 10 percent level. Tests in bold face are significant at the 5 percent level. Total no. of observations is 605,020 individuals for col. (1), 605,020 individuals for col. (2), 323,815 individuals for col. (3) and 281,205 individuals for col. (4). [2] Dependent variable is the individual responses to the Euro-Barometer Survey question that reads: ‘‘On the whole, are you very satisfied, fairly satisfied, not very satisfied or not at all satisfied with the life you lead?’’. Accordingly, four ordered categories were created. GDP per capita is real Gross Domestic Product per capita in 2000 US dollars. [3] Wealth is proxied by whether an individual is in the bottom or top half of the income distribution, ranked on the basis of GDP per capita in their nation 1975.
satisfaction for 16 nations in Europe over the past 28 years. The regression controls for a set of country and year effects, country-specific time trends, as well as a large set of personal characteristics. That is, it estimates equation (4) but restricts all lagged coefficients, b-t, to be zero. The coefficient is positive and highly significant, suggesting that (the log of)
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
231
GDP per capita may be an important determinant of the average level of happiness in countries. In column 2 we added an arbitrary number of lags to the logarithm of each nation’s GDP per capita. We experimented with different lag lengths, starting with same number (seven) as we used for the German SocioEconomic Panel. In the present case, we are not able to reject full adaptation to higher levels of GDP per capita after five years. The coefficient on current income is still positive and significant. The degree of adaptation is summarized by the sum of the coefficients on the lags (from one to five). The sum is negative ( = 1.07), and an F-test shows that it is significant at the 28 percent level. Consequently, of the initial impact of income, 83.6 percent is lost over the ensuing seven years (i.e., 1.07/1.28). Given that the estimated long-run effect of income is positive (and insignificant) we can also focus exclusively on the size of the estimated effects and ask if they are enough to explain the observed gap between happiness and income levels in Europe. In other words, can we account for the observed flat happiness levels over long-run periods of time by people’s adapting to income, using the estimated coefficients? First, we observe that the original ‘‘Easterlin Paradox’’ referred to the fact that ‘‘for the one time series studied, that for the United States since 1946 higher income was not systematically accompanied by greater happiness’’ (see p. 118 of Easterlin, 1974). A more recent calculation has been done by Blanchflower and Oswald (2004) using the U.S. General Social Surveys between 1972 & 1998. They note that there is a reasonable amount of stability in the proportion of people giving different well-being scores over this period. In terms of the Table 8.2 results, since (the logarithm of) GDP per capita increased by 0.46 between 1975 and 2002, the impact would have been to raise happiness by 0.59 units had no adaptation occurred ( = 1.28*0.46 using the coefficient on the current level of GDP per capita in column 2).16 Put another way, 9 percentage points more people would say that they were either ‘‘fairly satisfied’’ or ‘‘very satisfied with life’’ in 2002 compared to 1975. However, we would expect the increase to have been just 0.10 units after taking account of adaptation (i.e., 0.21*0.46 using the long-run effect calculated from column 2). Put another way, just 2 percentage points more people would say that they were either ‘‘fairly satisfied’’ or ‘‘very satisfied’’ with life after taking account of adaptation to income between 1975 and 2002. (The actual proportion fell 5 percentage points.) These ‘‘back of the envelope’’ calculations suggest that our estimates of adaptation may be large enough to be able to explain why no long-run trend in happiness has been observable over several decades of time. In columns 3 and 4 in Table 8.2, we divide the sample of 605,020 individuals living in 16 countries into two groups, the ‘‘poor’’ and
232
Section II: Comparisons of Income and Well-Being Through Time
Life Satisfaction
GDP per capita 11000
10000 2.8 9000
8000
2.6
7000
GDP per capita in 2000 US$
Mean life Satisfaction
3
6000
2.4 1980
1985
1990
1995
2000
2005
YEAR
Figure 8.2 Portugal’s life satisfaction and GDP per capita between 1985 and 2003. GDP per capita rose from $6,424 in 1985 to $11,008 in 2003 (in 2000 US$).
‘‘wealthy.’’ The ‘‘poor’’ and ‘‘wealthy’’ are defined as being in the bottom and top half of the income distribution, respectively, based on their real ‘‘GDP per capita’’ ranking at the start of the sample period in 1975.17 To illustrate, the life-satisfaction time series for Portugal is shown in Figure 8.2. In column 3 for the ‘‘poor’’ countries, there is significant adaptation to higher levels of income. Of the initial effect of income on happiness, 62.1% is lost over the following five years ( = 1.13/1.82). However, the remaining longrun effect is still positive and significant at the 1 percent level, suggesting that higher levels of GDP per capita may still have brought greater happiness in these nations. Column 4 shows the results for the ‘‘wealthy’’ countries. The sum of the lags is negative and significant at the 10 percent level. And the sum of the level and five lagged coefficients on GDP per capita is now negative. That is, after the initial positive (and significant) impact of higher levels of income on happiness, all of the effect is lost over the subsequent five years, leaving us unable to reject the hypothesis that there is no long-run effect for the ‘‘wealthy’’ countries. c. Cross-Country Data: The World Gallup Poll
We now focus on cross-country evidence available from the Gallup World Poll 2005. We start by looking at the data at two points in time. Figure 8.3 shows how the average responses in each country to the ‘‘ladder-of-life’’
8
DNK FIN NLD AUS BEL
Mean Life Satisfaction Ladder
NZL CRI
7
SAU
ISR ITA
ESP
MEX BRA CZE GRC JOR CHL ARG JAM PAN CYP PAK MYS GTM INDDZA THASVN COL LTU SVN POL HRV TT KOR BLR URY O SLV LBN KAZ PRT ZAF EST VNM HND IRN ROM HUN UZB EGY MRT ECU DOM SVK LAO BIH RUS IDN ZMB MDA PER UKR PRY GHA AZE NIC WBG CHN ALB NGA LVA TUR PHL BWA MOZ MAR KGZ SEN TJK NPL YEM MKD AGO BDI KEN LKA RWA BGD ARM MWI UGA TZA MDG MLI CMR SLE ETH NER BFA BGR HTI ZWE KHM GEO BEN TCD TGO
6
5
4
AUT FRA GBR ARE DEU SGP
SWE IRL
NOR CHE USA
JPN
KWT HKG
3 10000 20000 30000 40000 GDP per capita in 2005 (constant 2000 US $)
0
50000
Figure 8.3 Happiness versus GDP per capita, both measured in 2005, from the Gallup Poll.
8
DNK
CRI
Mean Life Satisfaction Ladder
7
ESP
IRL ISR
ITA MEX PRI BRASGP GRC ARG CHL PAN PAK MYS GTM IND THA COL DZA TT KOR HKGURY SLV O PRT ZAF HND HUN SVK EGY MRT ECU DOM IDN ZMB PER PRY GHA NIC CHN NGA PHL BWA MAR SEN NPL BDI KEN LKA RWA BGD MWI MDG CMR SLE BFA NER ZWE HTI BEN TCD
6
5
4
FIN NLD NZL AUS BEL AUT FRA
NOR SWE
CHE USA
GBR
JPN
TGO
3 0
5000 10000 15000 GDP per capita in 1960 (2000 US $)
20000
Figure 8.4 Happiness in 2005 versus real GDP per capita in 1960, from the Gallup Poll. 233
234
Section II: Comparisons of Income and Well-Being Through Time
satisfaction question asked in 2005 (on a 0–10 scale) vary with GDP per capita in 2005. Figure 8.4 makes one change—to the date at which GDP is being measured. It shows how the average responses in each country to the ‘‘ladder-of-life’’ question in 2005 vary with GDP per capita measured in 1960 (i.e., 45 years before). Figures 8.5 and 8.6 show the same two relationships, but instead use the logarithm of GDP per capita measured in 2005 and 1960, respectively. If anything, the relationship looks tighter between 2005 ‘‘ladder-of-life’’ satisfaction and the logarithm of 1960 GDP per capita, than with the logarithm of 2005 GDP per capita. Table 8.3 summarizes (and extends) these results by presenting a series of regression results. In column 1 the ‘‘ladder-of-life’’ question (averaged at the country level) is regressed on the logarithm of GDP per capita in 2005. The coefficient of 0.59 is significant at the 1 percent level. It suggests that a 20 percent rise in GDP per capita would move a person up 0.11 units (on the 0–10 scale). In column 2 we restrict the sample to those countries for which we also have measurements of their GDP per capita in 1960. The results are similar. In column 3 the ‘‘ladder-of-life’’ question is regressed on the logarithm of GDP per capita in 1960. The size of the coefficient rises to 0.73 (the difference with 0.62 in the previous column, is significant at the 10 percent level). What explains the results in columns 2 and 3? Why should the relationship between the ‘‘ladder-of-life’’ satisfaction in 2005 and GDP per capita
Mean Life Satisfaction Ladder
10
8
DNK FINCHE NLD NOR NZL AUS BEL SWE IRL ISRAUT USA SAU ESP CRI ITAFRA GBR MEX ARE DEU SGPJPN BRA CZE GRC JOR JAMPAN CHLARG CYP PAK KWT GTM IND THA MYS COL SVN LTU SVN DZA POL HRV HKG BLR URYTT KOR SLV KAZ ZAF LBN EST O PRT VNM HND IRN ROM HUN UZB EGY MRT ECU SVK DOM LAO BIHPER RUS IDN MDA ZMB UKR PRY GHA NGA NIC WBG AZE CHN ALB LVA PHL TUR BWA MAR TJK KGZ SEN YEM AGO NPLMOZ MKD BDI KEN RWA BGD LKA ARM MWI MDG UGA TZA MLI SLE ETH NER BFA HTI BGR ZWE CMRGEO KHM BEN TCD TGO
6
4
2
0 4
6
10 8 log (Real GDP per capita 2005)
12
Figure 8.5 Happiness and log(real GDP per capita), both measured in 2005, from the Gallup Poll.
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
235
Mean Life Satisfaction Ladder
10
8
DNK FINNLD NOR CHE NZL AUS BEL SWE USA ISR AUT ESP IRL CRI ITA FRA GBR MEX PRI BRA SGP GRCARGJPN CHL PAN PAK MYS GTMDZA IND THA COL TTURY KOR SLV HKG PRT O ZAF HND MRT EGY ECU HUN SVK DOM IDN GHA ZMB PER PRY NIC CHN NGA PHL BWA SEN MAR NPL BDI KEN RWA LKA BGD MWI MDGCMR SLE NER BFA ZWE HTI BEN TCD TGO
6
4
2
0 4
6
8
10
log (Real GDP per capita 1960)
Figure 8.6 Happiness in 2005 and log(Real GDP per Capita) in 1960, from the Gallup Poll.
strengthen when we measure GDP in 1960, instead of concurrently with the Gallup World Poll survey question in 2005? The next three columns attempt to provide the answer. We begin by estimating equation (8) in section 2 above. Under the null hypothesis that it is just the logarithm of GDP per capita in 2005 that matters for ‘‘ladder-of-life’’ satisfaction in 2005, the coefficients on the two variables, log(GDP per capita in 1960) and log(1 + growth rate of GDP per capita 1960–2005), should be equal. In column 4 the coefficient on the former is 0.67 (standard error = 0.05) and on the latter is 0.45 (standard error = 0.10). An F-test of equality of these two coefficients can be rejected at the 8 percent level. The effect of rising living standards since 1960 appear to have been of lesser importance to explaining world happiness in 2005 than the initial level of GDP per capita almost half a century before.18 In the next two columns we investigate this finding further by partitioning the Gallup sample into the ‘‘bottom half’’ and ‘‘top half’’ of the world income distribution (as defined by ranking each country’s GDP per capita in 1960). Column 5 in Table 8.3 focuses on the relatively poorer nations in our sample, and shows that the coefficients on log(GDP per capita in 1960) and log(1 + growth rate of GDP per capita 1960–2005) are both positive and significant at the 1 percent level. Since they are not significantly different
236
Section II: Comparisons of Income and Well-Being Through Time
TABLE 8.3 Cantril ‘‘Ladder of Life’’ in 2005 versus Levels and Long Run Changes in GDP, 1960–2005: The Role of Basic Needs versus Adaptation Dependent Variable: Happiness (Cantril Ladder of Life) Current level of real income log (GDP per capita in 2005) log (GDP per capita in 1960) Past levels of real income growth log (1 + growth rate of GDP per cap 1960–2005)
Wealth Proxy: 1960 GDP
(3)
(4)
Bottom half (5)
0.73 (0.05)
0.67 (0.05)
0.42 (0.13)
0.79 (0.10)
0.45 (0.10)
0.54 (0.14)
0.13 (0.17)
0.08*
0.52
All Countries (1)
(2)
0.59 (0.03)
0.62 (0.04)
Results of F tests Prob (GDP in 1960 = Growth of GDP 1960–2005) R-squared Number of observations
0.60 121
0.61 76
0.67 76
0.60 76
0.41 38
Top half (6)
0
0.60 38
Note: [1] All estimations are robust OLS. Starred tests are significant at the 10 percent level. Tests in bold face are significant at the 5 percent level. [2] Dependent variable is the country average of the individual responses to a question that asks people to imagine an eleven-rung ladder where the bottom (0) represents ‘‘the worst possible life for you’’ and the top (10) represents ‘‘the best possible life for you.’’ Respondents are then asked to report ‘‘on which step of the ladder do you feel you personally stand at the present time’’ (see the Gallup World Poll, 2005). GDP per capita is measured in constant US dollars (2000 values). [3] Wealth: Bottom half in 1960 is a country in the bottom half of the World Income Distribution in 1960. Top half in 1960 is a country in the top half of the World Income Distribution in 1960.
from one another ( = 0.42–0.54 = 0.12; standard error = 0.19) we cannot reject the null hypothesis that the happiness gains the ‘‘poor’’ nations have experienced from the past 45 years of economic growth have been the same as the happiness gains from growth prior to the 1960s. In other words, for these nations, it is still only the absolute level of the (logarithm) of income that matters for happiness. We next try a similar test in column 6 for the ‘‘rich’’ nations. Whereas the coefficient on log(GDP per capita in 1960) is
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
237
positive and significant at the 1 percent level, the coefficient on log(1 + growth rate of GDP per capita 1960–2005) is now insignificant. The two coefficients are also significantly different from each another at the 1 percent level ( = 0.79–0.13 = 0.66; standard error = 0.20). In other words, the past 45 years of economic growth (from 1960 to 2005) in the richest half of the world have not brought happiness gains above those that were already in place once the 1960s standard of living had been achieved.
4. Conclusions
We tested for whether, once ‘‘basic needs’’ are satisfied, there is adaptation to further economic growth or, more precisely, if adaptation differs across rich and poor. We used three data sets: individual panel data from Germany, a panel of countries with well-being data on over 600,000 Europeans from 1975–2002, and cross-sectional evidence from the Gallup World Poll for a large sample of countries in 2005. In the first part of our study, we found that home-owners in Germany (who are presumably wealthier than tenants) adapt fully to the effects of higher levels of the logarithm of income (after around seven years), whereas the tenants do not. We also found evidence consistent with full adaptation to the logarithm of GDP per capita in the wealthy, though not poor, countries within Europe. However, even in the wealthy European countries, full adaptation may take at least five years, so the happiness gains that they experience from higher income levels, whilst not permanent, can still be long-lasting. The final part of our study started by showing that, although there is a strong correlation between happiness in 2005 and (the logarithm of) GDP per capita in 2005 using the Gallup World Poll cross-section, the correlation between 2005 happiness and (the logarithm of) 1960 GDP per capita is significantly higher. We investigated the reason for this puzzling result by partitioning the Gallup World Poll into the rich half and poor half of nations and found that the past 45 years of economic growth (from 1960 to 2005) in the rich nations have not brought happiness gains above those that were already in place once the 1960s standard of living had been achieved. However, for the poorest half of countries, we cannot reject the null hypothesis that the happiness gains they have experienced from the past 45 years of economic growth have been the same as the gains from growth prior to the 1960s. In other words, for these nations, it is the absolute level of the logarithm of current income that matters for happiness. Overall our evidence supports the view that once basic needs have been satisfied, there is full adaptation to further economic growth, although that process may take a long period of time, in excess of five years.
238
Section II: Comparisons of Income and Well-Being Through Time
Notes 1. See, for example, the evidence in Hagerty & Veenhoven (2003) and the references cited therein, Stevenson & Wolfers (2008), as well as the (small) trends in European nations reported by Di Tella, MacCulloch, & Oswald (2003). 2. Easterlin (2005) has also concluded that ‘‘a positive happiness-income relationship typically turns up in international comparisons.’’ Note the large role played by countries with low levels of political freedom in Figure 8.1 of Inglehart & Klingemann (2000). Blanchflower & Oswald (2008) show that happiness and hypertension are correlated across countries. 3. Another interesting finding is the positive association between happiness and income in a cross-section of people within a country. 4. See Frey & Stutzer (2002), Graham & Pettinato (2002), Senik (2005) and Clark et al. (2008b) for reviews, as well as Helliwell (2002) for a broader discussion. An important precursor of the happiness literature is work on the individual welfare function of income (see, e.g., van Praag & Kapteyn, 1973). 5. This is also sometimes called the ‘‘hedonic treadmill’’ hypothesis or the ‘‘setpoint’’ model (see Costa et al., 1987; and Diener et al., 2005). Easterlin (2003) stresses that the evidence (based on small samples) is consistent only with incomplete adaptation. Using the German Panel, Clark et al. (2008a) study adaptation to labor and life events (unemployment, layoffs, marriage, and divorce). See also Clark (1999) and Lucas et al. (2003). 6. See Di Tella & MacCulloch (2008b) for a more detailed explanation. 7. Deaton (2008) finds a linear relationship between current subjective well-being and the current logarithm of GDP per capita in these data, but, conditional on current GDP per capita, a negative relationship between happiness and recent economic growth. 8. An alternative is to locate Purchasing Power Parity measures from the World Bank back to 1960 for our sample. h i DGDP2005 T T 1960 9. Since log GDP2005 ¼ log GDP1960 ð1 þ DGDP Þ we can simiGDP1960 Þð1 þ GDPT larly test for the importance of initial income conditions and GDP growth rates across several different time periods by estimating the following regression: Happiness2005 ; c ¼ log GDP1960 ; c þ log ð1 þ þ log ð1 þ
DGDPT 1960 ; c Þ GDP1960 ; c
DGDP2005 T ; c Þ þ "c : GDPT ; c
10. The number of observations is dramatically affected by the long lag structure, which requires a continuous time series that is only available for a subset of individuals (on average, we have 5.9 years of observations for each person, with a range from 1 to 15). 11. These results on adaptation effects raise the question of why individuals spend so much effort in trying to improve their economic condition. Some have argued that humans do not predict utility very well. For example, Ubel, Jepson, & Loewenstein (2001) study happiness predictions amongst people waiting for a kidney transplant. They find that those who receive one tend to report lower levels than they had predicted, whereas those who do not receive transplants report a higher quality of life that they had predicted. See also Gilbert et al. (1998) for evidence concerning predictions amongst academics concerning being denied tenure, Loewenstein &
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
12.
13.
14. 15.
16.
17.
18.
239
Schkade (1999) for a review of the evidence, and Loewenstein, O’Donoghue, & Rabin (2003) and Frey & Stutzer (2003) for detailed discussions. For example, Diener, Lucas, & Scollon (2006) argue that people adapt to different baselines, depending on their emotional dispositions. Luttmer (2004) and Stutzer & LaLive (2004) discuss the role of comparison groups. See, for example, Clark & Oswald (1994) for a study showing that the unemployed are unhappier and discussing the implications for economic theory, and Winkelmann & Winkelmann (1998) for an early study of unemployment with a panel strategy. An interesting variation on the fixed individual effects strategy has recently been explored in Kohler et al. (2005) in their study of fertility and partnership decisions. Using happiness data on identical (monozygotic) twins, the authors are able to control for unobserved endowments (ranging from preferences and abilities arising in genetic dispositions, to family history) that affect both happiness and fertility/marriage decisions. Gardner & Oswald (2001) have argued that we can use windfalls (winning the lottery and receiving an inheritance) as exogenous events. Using self-reports measured across four weeks and two years apart, life-satisfaction measures never failed to meet Campbell & Fiske’s (1959) criteria for discriminant validity from the affective components of subjective well-being. Using third-party reports on individual well-being (in which convergent validity coefficients could be expected to be lower), life satisfaction failed to meet the criterion only four times out of 32 comparisons with positive affect. The corresponding cut points between ‘‘not at all satisfied,’’ ‘‘not very satisfied,’’ ‘‘fairly satisfied’’ and ‘‘very satisfied’’ are equal to 3.0 units, 2.1units and 0.3 units, respectively. The ‘‘poor’’ individuals, based on the 1975 GDP per capita level in their country, live in Portugal, Spain, Greece, Ireland, Italy, West Germany, or Belgium. The ‘‘rich’’ individuals, based on the 1975 GDP per capita level in their country, live in France, the Netherlands, Luxembourg, Denmark, the United Kingdom, Norway, Finland, Sweden, or Austria. The sample is divided into ‘‘rich’’ and ‘‘poor’’ halves based on numbers of individuals (i.e., not numbers of countries), though the results are similar regardless of how the split is made. Note that the correlation between GDP per capita in 1960 and the growth rate of GDP per capita between 1960 and 2005 is 0.01 (and insignificant) across the 76 countries in our data set. Thus there do not appear to be sufficiently strong ‘‘catchup’’ effects (whereby poorer nations have been experiencing faster growth rates than the rich nations) to affect our estimates.
References Blanchflower, D., & Oswald, A. (2004). Well-being over time in Britain and the U.S.A. Journal of Public Economics, 88(7), 1359–1386. Blanchflower, D., & Oswald, A. (2008). Hypertension and happiness across nations. Journal of Health Economics, 27(2), 218–233. Brickman, P., Coates, D., & Janoff-Bullman, R. (1978). Lottery winners and accident victims: Is happiness relative?. Journal of Personality and Social Psychology, 36(8), 917–927. Campbell, D., & Fiske, D. (1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56, 81–105.
240
Section II: Comparisons of Income and Well-Being Through Time
Clark, A., & Oswald, A. (1994). Unhappiness and unemployment. Economic Journal, 104, 648–659. Clark, A. (1999). Are wages habit-forming? Evidence from micro data. Journal of Economic Behavior and Organization, 39, 179–200. Clark, A., Diener, E., Georgellis, Y., & Lucas, R. (2008a). Lags and leads in life satisfaction: A test of the baseline hypothesis. The Economic Journal, 529, F222–F243. Clark, A., Frijters, P., & Shields, M. (2008b). A survey of the income happiness gradient. Journal of Economic Literature, 46(1), 95–144. Costa, P., Zonderman, A., McCrae, R., Cornoni-Huntley, J., Locke, B., & Barbano, H. (1987). Longitudinal analyses of psychological well-being in a national sample: Stability of mean levels. Journal of Gerontology, 42(1), 50–55. Deaton, A. (2008). Income, health and well-being around the world: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22(2), 53–72. Diener, E., Diener, M., & Diener, C. (1995). Factors predicting the subjective wellbeing of nations. Journal of Personality and Social Psychology, 69(5), 851–864. Diener, E., & Biswas-Diener, R. (2002). Will money increase subjective well-being? A literature review and guide to needed research. Social Indicators Research, 57, 119–169. Diener, E., Lucas, R., & Napa Scollon, C. (2005). Beyond the hedonic treadmill: Revisions to the adaptation theory of well-being. American Psychologist, 61, 305–314. Diener, E., & Oishi, S. (2000). Money and happiness. In E. Diener & E. Suh (Eds.), Culture and subjective well-being. Cambridge, MA: MIT Press. Di Tella, R., Haisken-De New, J., & MacCulloch, R. (2005). Happiness adaptation to income and to status in an individual panel. NBER Working Paper no. 13159. Di Tella, R., MacCulloch, R., & Oswald, A. (2003). The macroeconomics of happiness. Review of Economics and Statistics, 95(4), 809–828. Di Tella, R., & MacCulloch, R. (2005). Partisan social happiness. Review of Economic Studies, 72(2), 367–393. Di Tella, R., & MacCulloch, R. (2008a). Happiness, contentment and other emotions for central banks. NBER Working Paper 13622. Di Tella, R., & MacCulloch, R. (2008b). Happiness adaptation to income across the rich and the poor. Harvard Business School Working Paper; Imperial College Working Paper #101. Easterlin, R. (1974). Does economic growth improve the human lot? Some empirical evidence. In P. A. David & M. W. Reder (Eds.), Nations and households in economic growth: Essays in honor of Moses Abramowitz (pp. 89–125). New York: Academic Press. Easterlin, R. (2003). Building a better theory of well-being. IZA wp 742, Bonn, Germany. Easterlin, R. (2005). Diminishing marginal utility of income? Caveat emptor. Social Indicators Research, 70(3), 243–255. Ferrer-i-Carbonell, A., & Frijters, P. (2004). How important is methodology for the estimates of the determinants of happiness? Economic Journal, 114(497), 641–659. Frederick, S., & Loewenstein, G. (1999). Hedonic adaptation. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology. New York: Russell Sage Foundation. Frey, B., & Stutzer, A. (2002). What can economists learn from happiness research? Journal of Economic Literature, 40(2), 402–435. Frey, B., & Stutzer, A. (2003). The economic consequences of mispredicting utility. Institute for Empirical Research in Economics wp 218.
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
241
Gardner, J., & Oswald, A. (2001). Does money buy happiness? A longitudinal study using data on windfalls. Warwick University mimeograph. Gilbert, D. T., Pinel, E. C., Wilson, T. D., Blumberg, S. J., & Wheatley, T. P. (1998). Immune neglect: A source of durability bias in affective forecasting. Journal of Personality and Social Psychology, 75(3), 617–638. Graham, C., & Pettinato, S. (2002). Happiness and hardship. Washington, DC: Brookings. Hagerty, M., & Veenhoven, R. (2003). Wealth and happiness revisited—growing national income does go with greater happiness. Social Indicators Research, 64, 1–26. Helliwell, J. (2002). How’s life? Combining individual and national variables to explain subjective well-being. Economic Modelling, 20, 331–360. Inglehart, R. (1997). Modernization and postmodernization: Cultural, economic, and political change in 43 societies. Princeton, N.J.: Princeton University Press. Inglehart, R., & Klingemann, H. (2000). Genes, culture, democracy and happiness. In E. Diener & E. Suh (Eds.), Culture and subjective well-being. Cambridge, MA: MIT Press. Kohler, H.P., Behrman, J., & Skytthe, A. (2005). Partner + children = happiness? An assessment of the effect of fertility and partnerships on subjective well-being. Journal of Socio-Economics, 35, 326–347. Konow, J., & Earley, J. (2008). The hedonistic paradox: Is homo-economicus happier? Journal of Public Economics, 92(1–2), 1–33. Loewenstein, G., O’Donoghue, T., & Rabin, M. (2003). Projection bias in predicting future utility. Quarterly Journal of Economics, 118(4), 1209–1248. Loewenstein, G., & Schkade, D. (1999). Wouldn’t it be nice? Predicting future feelings. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundation of hedonic psychology (pp. 85–105). New York: Russell Sage Foundation. Lucas, R., Clark, A., Georgellis, Y., & Diener, E. (2003). Re-examining adaptation and the setpoint model of happiness: Reactions to changes in marital status. Journal of Personality and Social Psychology, 84, 527–539. Lucas, R. E., Diener, E., & Suh, E. (1996). Discriminant validity of well-being measures. Journal of Personality and Social Psychology, 71(3), 616–628. Luttmer, E. F. (2004). Neighbors as negatives: Relative earnings and well-being. Quarterly Journal of Economics, 120(3), 963–1002. Pollak, R. (1970). Habit formation and dynamic demand functions. Journal of Political Economy, 78, 745–763. Rayo, L., & Becker, G. (2007). Evolutionary efficiency and happiness. Journal of Political Economy, 115, 302–337. Senik, C. (2005). Income and well-being: what can we learn from subjective data? Journal of Economic Surveys, forthcoming. Stevenson, B., & Wolfers, J. (2008). Economic growth and subjective well-being: Reassessing the Easterlin Paradox. Brookings Papers on Economic Activity, Spring. Stutzer A. (2004). The role of income aspirations in individual happiness. Journal of Economic Behavior and Organization, 54(1), 89–109. Stutzer, A., & Lalive, R. (2004). The role of social norms in job searching and subjective well-being. Journal of the European Economic Association, 2, 696–719. Ubel, P. A., Loewenstein, G., & Jepson, C. (2003). Whose quality of life? A commentary exploring discrepancies between health state evaluations of patients and the general public. Quality of Life Research, 12.
242
Section II: Comparisons of Income and Well-Being Through Time
Wathieu, L. (2004). Consumer habituation. Management Science, 50(5), 587–596. Winkelmann, L., & Winkelmann, R. (1998). Why are the unemployed so unhappy? Economica, 65(257), 1–5. van de Stadt, H., Kapteyn, A., & van de Geer, S. (1985). The relativity of utility: Evidence from panel data. Review of Economics and Statistics, 67(2), 179–187. van Praag, B., & Ferrer-i-Carbonell, A. (2004). Happiness quantified: A satisfaction calculus approach. Oxford, UK: Oxford University Press. van Praag, B., & Kapteyn, A. (1973). Further evidence on the individual welfare function of income: An empirical investigation in the Netherlands. European Economic Review, 4(1) 33–62. Veenhoven, R. (1991). Is happiness relative? Social Indicators Research, 24, 1–34.
Appendix 8.A Our Three Data Sources (with Variable Definitions)
The German Socio-Economic Panel
The GSOEP is the public use version of the Socio-Economic Panel (SOEP), a longitudinal data set begun in 1984. It was developed in a former Special Research Unit at the Universities of Frankfurt/Main and Mannheim in cooperation with the Deutsches Institut fu¨r Wirtschaftsforschung (DIW) (German Institute for Economic Research), and initially financed by the German National Research Fund (DFG). In 1990, the DIW assumed control of the panel with funding from the Joint Federal-Land Commission for Promotion of Research Activities. The SOEP began with a sample of 6,000 households living in the western states of the Federal Republic of Germany, including a disproportionate number of non-German migrant workers. In November 1990, the eastern states of Germany were reunited with the western states of the Federal Republic of Germany. In June 1990, the DIW began a survey of families in the eastern states and merged these data with the existing SOEP population to provide a representative sample of reunited Germany.
Definitions
Happiness (Life Satisfaction 0–10): The individual responses to the question: ‘‘In conclusion, we would like to ask you about your satisfaction with your life in general, please answer according to the following scale, 243
244
Section II: Comparisons of Income and Well-Being Through Time
0 means completely dissatisfied and 10 means completely satisfied: How satisfied are you with your life, all things considered? 0 1 2 3 4 completely dissatisfied
5
6
7
8 9 10 completely satisfied’’
Personal Income: Real Household Post-Government Income from the Cross-National Equivalent File 1980–2000. This variable represents the combined income after taxes and government transfers of the head, partner, and other family members. Employment state: A set of dummy variables taking the value 1 depending on the respondent’s employment state: (1) unemployed (2) retired (3) at school (4) at home (5) in the military (6) self-employed (7) public servant. The base category is ‘‘employed’’ (in the private sector). Marital state: A set of dummy variables taking the value 1 depending on the respondent got married, divorced, separated, or widowed over the course of the past year. The base category is ‘‘single.’’ Education: A generated variable determined from the following questions: ‘‘Now to a completely different topic: education and training. First, what type of school leaving certificate do you possess? Have you (successfully) completed vocational training or studies (at an institution of higher education)? Yes/No. What type of vocational or higher education degree was that? Now to the topic of further education and training. Have you participated in further education in one of the following areas within the past year?’’ Tenant, Owner: Two dummy variables that correspond to the response to the question: ‘‘Are you a tenant or an owner? 1. Tenant, or 2. Owner.’’ The Eurobarometer Survey Series
The Eurobarometer Surveys were conducted by various research firms operated within European Community nations under the direction of the European Commission. Either a nationwide multi-stage probability sample or a nationwide stratified quota sample of persons aged 15 and over was selected in each nation. The cumulative data file used contains 36 attitudinal, 21 demographic, and 10 analysis variables selected from the European Communities Studies, 1970–1973, and Eurobarometers, 3–38. Definitions
Happiness (Life Satisfaction 1–4): The individual responses to the Eurobarometer Survey question that reads: ‘‘On the whole, are you
Happiness Adaptation to Income Beyond ‘‘Basic Needs’’
245
very satisfied, fairly satisfied, not very satisfied or not at all satisfied with the life you lead?’’ Accordingly, four ordered categories were created. GDP per capita: Real GDP per capita at the price levels and exchange rates of 2000 in U.S. dollars obtained from World Development Indicators, World Bank, 2008. Employment state: A set of dummy variables taking the value 1 depending on the respondent’s employment status: unemployed, retired, housewife, in school or the military, and self-employed. The base category is ‘‘employed.’’ Male: A dummy taking the value 1 if respondent is male and 0 otherwise. Age: The respondent’s age in years. Personal Income Position: A set of 4 dummy variables that take the value 1 depending on in which income quartile the respondent’s family income lies. The base category is the lowest income quartile. Marital state: A set of dummy variables taking the value 1 depending on the respondent’s marital status: married, living as married, de facto married, divorced, separated, or widowed. The base category is ‘‘never married.’’ Education: This heading refers to a set of dummy variables which take the value 1 depending on the age at which the respondent finished full-time education: up to 15–18 years old or more than 18 years old. The base category is ‘‘education up to 14 years old.’’
The Gallup World Poll
The analysis is based on the Gallup World Poll, collected data from samples of people in each of 132 countries during 2005; with the exception of Angola, Cuba, and Myanmar, the samples are nationally representative of people aged 15 and older. To assure the Gallup World Poll survey data are representative of 95% of the world’s adult population, two primary methodological designs are employed: A RandomDigit-Dial (RDD) telephone survey design is used in countries where 80% or more of the population has landline phones. This situation is typical in the United States, Canada, Western Europe, Japan, Australia, etc. In the developing world, including much of Latin America, the former Soviet Union countries, nearly all of Asia, the Middle East, and Africa, an area frame design is used for face-to-face interviewing. The following are key aspects of the overall Gallup World Poll survey philosophy:
246
Section II: Comparisons of Income and Well-Being Through Time
• The sample represents all parts of each country, including all rural
• • •
• • •
areas. Countries are reviewed on a case-by-case basis when part of a country cannot be included in the sample design. The review determines whether the survey should be carried out. Face-to-face interviews are approximately one hour in length. Telephone interviews are considerably shorter, about 30 minutes in length. There is a standard set of questions used around the world. In the parts of the world where face-to-face surveys are conducted, the questionnaire includes questions tailored to each region. For example, the questions used in heavily indebted poor countries are tailored toward providing information about progress on the Millennium Development Goals. The questionnaire is translated into the major languages of each country. Interviewing supervisors and interviewers are trained, not only on the questionnaire, but also on the execution of field procedures. This interviewing training usually takes place in a central location. Quality control procedures are used to validate that correct samples are selected and that the correct person is randomly selected in each household. Random respondent selection uses either the latestbirthday method or the Kish grid.
Definitions
Happiness (Cantril Ladder of Life): The response to the survey question that asks: ‘‘Please imagine a ladder with steps numbered from zero at the bottom to ten at the top. Suppose we say that the top of the ladder represents the best possible life for you and the bottom of the ladder represents the worst possible life for you. If the top step is 10 and the bottom step is 0, on which step of the ladder do you feel you personally stand at the present time?’’. Accordingly a 0–10 cardinal scale was created. GDP per capita: Real GDP per capita at the price levels and exchange rates of 2000 in U.S. dollars obtained from World Development Indicators, World Bank, 2008.
Chapter 9 The Easterlin and Other Paradoxes: Why Both Sides of the Debate May Be Correct Carol Graham, Soumya Chattopadhyay, and Mario Picon1 The Brookings Institution
After decades of skepticism about using survey data—and expressed versus revealed preferences—as the basis for economic analysis, there has been a virtual explosion in the publication of papers, articles, and books based on happiness surveys in recent years.2 There is increasing consensus on the utility of survey research for answering questions that revealed preferences do not provide good answers to. Examples of these are the welfare effects of macro- and institutional arrangements that individuals are powerless to change, and the explanation of behaviors that are driven by norms, addiction, or self control, rather than by optimal or rational choices. The results of this research have challenged some standard assumptions about the determinants of individual welfare. Rather ironically, however, there is much less consensus on the first economics question that happiness research originally shed light on: the relationship between happiness and income. Richard Easterlin’s seminal work, which examined the relationship between average country happiness levels and per capita incomes over several decades, highlighted an apparent paradox: as countries grew materially wealthier—and healthier—over time, average happiness levels did not increase (Easterlin, 1974, 2003). His findings are now known as the Easterlin paradox. A number of subsequent studies confirmed the general direction of his findings—e.g., average happiness levels do not increase as average incomes increase over time. In contrast, more recent work, based on new data, questions whether the paradox exists at all. Thus there is renewed debate over Easterlin’s original question: ‘‘Will raising the incomes of all increase the happiness of all?’’ 247
248
Section II: Comparisons of Income and Well-Being Through Time
Easterlin’s paradox has been explained by rising aspirations and comparison effects. Once basic needs are met, aspirations rise as quickly as incomes, and individuals care as much about relative differences with their peers as they do about absolute gains. The importance of aspirations and comparison effects to individual well-being have been demonstrated in micro-level empirical work across a range of contexts (Graham & Felton, 2006; Kingdon & Knight, 2007; Lora & Chaparro, forthcoming; Luttmer, 2005). There is also a time dimension: most people think that they were less happy in the past and expect to be happier in the future. They judge their past living standards by their current aspirations, but fail to account for their aspirations’ adjusting over time as they predict their future happiness (Easterlin, 2001). Within countries, across cultures and levels of development, Easterlin and a host of other economists have shown that wealthier people are, on average, happier than poorer ones, although the relationship is not necessarily linear; there is diminishing marginal utility with absolute income (as in the case of the utility function in standard textbook economics). A number of studies show that the same proportional increase in income yields a lower increase in happiness at higher income levels (Easterlin, 1974, 2003; Blanchflower & Oswald, 2004; Frey & Stutzer, 2002; Graham & Pettinato, 2002a). Differences in income, meanwhile, only account for a low proportion of the differences in happiness among persons, and other economic and noneconomic factors, such as employment and health status, exert important influences on happiness. There are also differences across individual personalities. It may be that the individuals who prize material goods more highly than other things in life are less happy, and thus as their ownership of material goods increases (via higher levels of income), their happiness levels do not increase proportionately (Frey, 2008). Finally, there is also some evidence that happier people earn more (and are healthier) than unhappy people.3 As noted above, there is renewed debate over whether there is an Easterlin paradox or not. A number of scholars, such as Angus Deaton, and Betsey Stevenson and Justin Wolfers, have published papers demonstrating a clear, log linear relationship between per capita incomes and average happiness levels, with no sign that the correlation weakens, either as income levels increase or decrease over time (Deaton, 2008; Wolfers & Stevenson, 2008). Indeed, the work of both sets of authors suggests that the slope may be steeper for richer countries, most likely because wealthier people are better able to enjoy higher levels of income than are poor ones (a greed effect?).4 Both of these papers rely on the newly available Gallup World Poll, which covers over 120 countries worldwide, as well as some different data sets for
The Easterlin and Other Paradoxes
249
earlier years. Ron Inglehart, meanwhile, in a new analysis of data from the World Values Survey for 1981–2006 (presented at the Princeton conferenc) finds that subjective well-being rose in 77 percent of the 52 countries for which time series is available (Inglehart, Foa, Peterson, & Welzel, 2008). Eduardo Lora and colleagues at the Inter-American Development Bank, using Gallup data for Latin America, also find a positive relationship between per capita income levels and average happiness levels (Lora & Chaparro, forthcoming). Other studies come out somewhere in the middle. Graham and Pettinato, in an earlier and first study of happiness in a large sample of developing countries and using absolute levels of per capita GDP, find that, on average, happiness levels are higher in the developed than in the developing countries in the sample, but that within each group of countries, there is no clear income–happiness relationship (see Figure 9.1). Their work is based on the World Values survey and on the Latinobarometer poll for Latin America. Why the discrepancy? This paper posits that, for a number of reasons, the divergent conclusions may each be correct. The relationship between happiness and income is mediated by a range of factors that can alter its slope and/or functional form. These include the particular questions that are used to measure happiness; the selection of countries that is included in the 100 NETHERLANDS CANADA
90 SWITZERLAND SWEDEN
% above neutral on life satisfaction
IRELAND NEW ZEALAND
80 NIGERIA
70
PORTUGAL
CHINA
INDIA BANGLADESH
60
VENEZUELA PANAMA
ROMANIA HONDURAS
U.S.A.
JAPAN
BRAZIL MEXICO
S.AFRICA
50
FRANCE
HUNGARY
GUATEMALA
COLOMBIA URUGUAY
40
ELSALVADOR CHILE
30
ARGENTINA COSTARICA
NICARAGUA
BULGARIA PARAGUAY
RUSSIA 2
20
BOLIVIA
R = 0.14
ECUADOR PERU
10 0
2,000
4,000
6,000
8,000 10,000 12,000 14,000 16,000 18,000 20,000 GDP percapita (1998 PPP)
Figure 9.1 Happiness and Per-capita Income in the 1990s. Source: Carol Graham & Stefano Pettinato (2002). Happiness and Hardship: Opportunity and Insecurity in New Market Economies. Washington, D.C.: The Brookings Institution.
250
Section II: Comparisons of Income and Well-Being Through Time
survey sample; the specification of the income variable (log or linear); the rate of change in economic conditions in addition to absolute levels; and changing aspirations as countries go from the ranks of developing to developed economies. In addition to discussing these factors, we introduce three related macro- and micro-level phenomena: the paradox of unhappy growth, the happy peasants and frustrated achievers paradox, and the paradox of low aspirations. We provide evidence from our own analysis of Gallup and Latinobarometer data, as well as from the work of several other authors. Our objective is to contribute as much to the method as to the substance of the debate.
Question Framing Issues
Which happiness question is used makes a difference to the relationship between happiness and income. Psychologist Ed Diener and his colleagues decompose subjective well-being into an affective or emotional component and a cognitive or judgmental component (Diener, Suh, Lucas, & Smith, 1999). The first is determined and measured by how often an individual reports experiencing positive or negative affect (such as smiling), while life satisfaction is composed of an individual’s satisfaction with various life domains (such as health and work) as well as with life in general. Affect questions typically (and not surprisingly) have less of a relationship with income than do cognitive questions. More framed life satisfaction questions, such as Cantril’s Ladder of Life question, which asks respondents to compare their lives to the best possible life they can imagine on a one to ten scale, have an even closer relationship with income (Cantril, 1965). The earlier surveys that Easterlin and others used, such as the World Values Survey and the Eurobarometer, relied on open-ended happiness or life satisfaction questions, which posed very simple questions such as, ‘‘Generally speaking, how happy are you with your life?’’ or ‘‘How satisfied are you with your life?’’ with possible answers on various scales, ranging from one to four to one to ten. Answers to general happiness and life satisfaction questions are highly correlated (Blanchflower & Oswald, 2004; Graham & Pettinato, 2002a).5 In contrast, the ‘‘life satisfaction’’ variable that is used in the Gallup World Poll—which is the basis for the Deaton and the Stevenson and Wolfers papers— is Cantril’s ‘‘best possible life’’ question. The ‘‘best possible life’’ question provides much more of a reference frame than does an open-ended life satisfaction question. Surely, when asked to compare their lives to the best possible life, respondents in very poor countries are aware that life is likely to be better
The Easterlin and Other Paradoxes
251
in wealthier ones, not least because of widespread access to the media and the internet. In his paper, Deaton makes the point that the simplest interpretation of the Gallup World Poll findings is that when asked to imagine the best and worst possible life for themselves on a ten-point scale, people use a global standard, and the Danes understand how bad life in Togo is; and the Togolese, through television and other media, understand how good life is in high-income countries (Deaton, 2008). John Helliwell, meanwhile, compares results based on the Cantril Ladder in the 2006 Gallup poll with those on life satisfaction as a whole in the World Values Survey, and finds that the correlation between income and life satisfaction is stronger with the more framed Ladder of Life question. At the same time, there is striking consistency in the other factors that contribute to life satisfaction across the two surveys (Helliwell, 2008). As a simple test of the extent to which question framing matters, we compared the income and happiness relationship across several different life satisfaction questions for the Latin America sub-sample of the Gallup World Poll.6 Latin America is a good testing ground, as the region includes countries with a wide range of income levels, with some wealthier countries such as Chile and Brazil near Organization for Economic Cooperation and Development (OECD) levels, and others, such as Guatemala and Honduras, on the other extreme of the development spectrum. The questions included: the ‘‘best possible life’’ question described above; an economic satisfaction question (Are you satisfied with your standard of living, all the things that you can buy and do?); a poor–rich scale question (In a scale from zero to ten, with zero the poorest people and ten the richest people, in which cell would you place yourself?); an affect question (Did you smile or laugh a lot yesterday?); a life purpose question (Do you feel your life has purpose or meaning?)7; and a freedom/opportunity question (Are you satisfied with your freedom to choose what to do with your life?). (For exact question phrasing and the distribution of responses across questions, see Appendix 9.1.) Given that accurately measuring incomes in a context such as Latin America is difficult, we used both income and wealth variables. A large percentage of respondents work in the informal sector or in unsteady jobs and have difficulty accurately recalling earnings, for example. And the most recent income levels that are reported may misrepresent more permanent income flows, due to seasonality, economic shocks, and/or job instability. Wealth indices, on the other hand, while adjusting better for temporary fluctuations, are less effective at capturing variability across households, particularly at the high end of the income scale, nor do they account for the quality of assets. Access to water may be irregular, or televisions and refrigerators that are owned may be functioning poorly, if at all.
252
Section II: Comparisons of Income and Well-Being Through Time
For our income variable, we used log of per capita household income measured in 2005 Purchasing Power Parity (PPP) U.S. dollars. In each country, Gallup includes a categorical question on total household income. Each respondent is asked to identify the household’s monthly income within a bracket expressed in local currency. The number of brackets is different in each country, and in Latin America it ranges from four brackets in Colombia to twenty in Bolivia. We relied on an adjustment to the Gallup income variable introduced by colleagues at the Inter-American Development Bank. Each household was assigned a normalized random value within the bracket that they self-reported. Income was transformed to U.S. PPP dollars and then divided by household size, resulting in a monthly per capita household income variable, which is normally distributed across the sample.8 (For the distribution of the income variable, see Appendix 9.1.) We use the log of per capita household income. This specification helps control for outliers and conforms to standard economic assumptions that an extra unit of income is more significant for those at the bottom of the distribution with less available resources than for those at the top. For our country-level analysis, we used the average log of per capita household income (as opposed to the log of the average per capita household income). Some of the earlier studies of income and happiness used absolute rather than log income levels, and showed a curvilinear relationship between income and happiness, suggesting the satiation point that is part of the explanation of the Easterlin paradox. Easterlin himself notes that the specification of the income variable has important effects, and that ‘‘the supposed attenuation of the income–happiness relation does not occur when happiness is regressed on log income’’ (Easterlin, 2005). Later studies, such as Wolfers & Stevenson (2008), and Deaton (2008), use either log of average GDP per capita or the average of log income per capita. We constructed our wealth index, based on the list of assets in the Gallup poll, using both the simple, unweighted Rasch scale of asset ownership, and then a principal components analysis (PCA)–based index in which the assets that are more unequally distributed across households are weighted more. (For details on the assets and the construction of the indices, see Appendix 9.1.) Our results are essentially identical using alternative methods; the results on wealth reported in the tables are those based on the PCA approach. To test the cross-country relationship between income and happiness across questions, we use the following model: average happiness (as measured by each separate question) in country i = f (average log of per
The Easterlin and Other Paradoxes
253
capita income or wealth in country i + characteristics of the average individual in country i). We find that the log-linear income and happiness relationship holds across countries for the ‘‘best possible life’’ and the poor to rich economic ladder questions, but not for affect (smiling a lot the previous day), life purpose and ‘‘freedom to choose in life’’ questions. (See Table 9.1.) Indeed, the relationship between income and the smiling and life purpose questions is negative and significant when we use our income variable, and insignificant when we use our wealth variable. The relationship between freedom to choose and income (and wealth), meanwhile, is positive, but insignificant. At the individual level, our basic model was: individual happiness = f (household log income or wealth + personal controls + country dummies). We ran the model sequentially, first looking at just happiness and income or wealth, then adding the country dummies, and finally adding the personal controls. At this level, we find that income and wealth were positively correlated with most measures of happiness, except for the life purpose and ‘‘freedom to choose’’ questions, which were insignificant with some specifications. (See Table 9.1.) This confirms the work of many other studies, which consistently finds a cross-sectional relationship between happiness and income within countries, regardless of whether or not the Easterlin paradox holds across countries or through time. Across the questions, we find that the highest effect size—based on the size of the coefficients—is between satisfaction with one’s standard of living and income (and/or wealth: .43 on log income/.39 on wealth). This is followed by the poor-to-rich scale economic ladder question (.32/.35), and the ‘‘best possible life’’ question (.27/.29). Income and wealth do a good job of explaining the distribution of responses on the ladder of life question, including when other controls are used, but they do not explain answers on smiling, life purpose, and ‘‘freedom to choose’’ questions. The first three questions provide more of an economic frame for respondents, while the latter are vaguer and more open-ended. Our results comparing across the questions, both across countries and within them, support our intuition that question framing can have important effects on the measured relationship between income and happiness. Questions that provide more tangible economic or status frames seem to have a closer relationship with income than do more open-ended questions that capture either affect and/or life chances. The work of several other scholars supports the basic direction of our findings. Psychologists Ryan and Colleen Howell, using a meta-analysis of 111 samples from 54 developing countries, find evidence of an Easterlin effect: the relationship between income and subjective well-being is stronger for the poor countries than it is for the rich countries in
TABLE 9.1 Question phrasing, income and happiness across and within Latin American countries Summary of Results Happiness questions vs. Income per capita (Yes means significant at the 10% level) Individual cross-sectional analysis
Cross-country analysis
Subjective variable
Model 1
Model 2
Model 3
Model 4
Model 5
Best possible life Living Standard Satisfaction Place in socio-economic scale Life Purpose Smiled Satisfaction with personal freedom
Yes (þ) Yes (þ)
Yes (þ) Yes (þ)
Yes (þ) Yes (þ)
Yes (þ) No (þ)
Yes (þ) No (þ)
Yes (þ)
Yes (þ)
Yes (þ)
Yes (þ)
Yes (þ)
No (þ) Yes (þ) Yes (þ)
Yes (þ) Yes (þ) Yes (þ)
Yes (þ) Yes (þ) No (þ)
Yes () Yes () No (þ)
Yes () No () No ()
Summary of Results Happiness questions vs. Wealth Index (Yes means significant at the 10% level) Individual cross-sectional analysis
Cross-country analysis
Subjective variable
Model 1
Model 2
Model 3
Model 4
Model 5
Best possible life Living Standard Satisfaction Place in socio-economic scale Life Purpose Smiled Satisfaction with personal freedom
Yes (þ) Yes (þ)
Yes (þ) Yes (þ)
Yes (þ) Yes (þ)
Yes (þ) Yes (þ)
Yes (þ) Yes (þ)
Yes (þ)
Yes (þ)
Yes (þ)
Yes (þ)
Yes (þ)
Yes (þ) Yes (þ) Yes (þ)
Yes (þ) Yes (þ) Yes (þ)
Yes (þ) Yes (þ) Yes (þ)
No () No () No (þ)
Yes () Yes () No ()
Model 1: Probit/ Oprobit: Subjective variable ¼ f(logpcincome) Model 2: Probit/ Oprobit: Subjective variable ¼ fflogpcincome, country dummies) Model 3: Probit/Oprobit: Subjective variable ¼ fflogpcincome, country dummies, demographics) Model 4: Oprobit Country Average Subjective variable ¼ ffCountry average hh per capita income) Model 5: Oprobit: Country Average Subjective variable ¼ ffCountry average hh per capita income, demographics) Source: Gallup World Poll, 2007 and authors’ calculations.
254
The Easterlin and Other Paradoxes
255
their sample. Yet they also find that the income-subjective well-being relationship is stronger when subjective well-being is measured as life satisfaction than when it is measured as happiness. They describe the former as a more cognitive assessment and the latter as a more emotional assessment (Howell & Howell, 2008). Jim Harter, also using the Gallup World Poll, finds that different kinds of questions display different results depending on the income levels of the countries. Poor countries slope higher on the relationship between the best possible life question and pain and sadness questions, while rich countries slope higher for enjoyment questions (such as ‘‘How often did you smile yesterday?’’). This probably reflects different concerns at different levels of development. There are different concerns for those for whom basic needs and health are precarious, survival-level issues, such as avoiding pain and death, and paramount: while respondents in wealthier countries, who for the most part take basic needs for granted, have higher expectations about more expensive components of life, such as enjoying leisure time (Figure 9.2).
WB Income Groups
3.00000
Zscore: Life today
2.00000
1.00000
0.00000
–1.00000
–2.00000
Low Income Lower Middle Income Upper Middle Income High Income
Denmark
Switzerland Netherlands Canada Swden Australia Belgium Israel Norway New Zealand Spain Ireland Italy Saudi Arabia France Costa RicaAustria United Arab Emirates Puerto Rico Mexico United Kingdom Czech Republic Brazil Germany Japan Singapore Jordan Pakistan Cyprus Argentina Greece Lithuania Panama Trinidad & Tobago Croatia Slovenia Kuwait Colombia South Korea Thailand Poland Uruguay Hong Kong Lebanon Portugal El Salvador Honduras Belarus Estonia India Iran Bolivia Myanmar Mauritania Cuba Slovakia Moldova Romania Laos Uzbekistan Montenegro Indonesia South Africa Hungary Russia Bosnia Palestine Azerbaijan Herzegovina Peru Botswana Philippiness Nigeria China Ukraine Latvia Serbia Paraguay Turkey Kyrgyzstan Morocco Armenia Mozambique Tajikistan Nicaragua Sri Lanka Macedonia Bangaldesh Burundi Kenya Mali Cameroon Malawi Haiti Madagascar Zimbabwe Georgia Uganda Burkina Faso Niger Ethiopia Chad Tanzania Sierra Leone Cambodia Togo Benin Finland
–3.00000
–2.00000
–1.00000
0.00000
1.00000
2.00000
Zscore: Net affect
Figure 9.2 Evaluated Well-being and Net Affect. Source: James K. Harter, ‘‘Gallup World Poll: Methodology and Validation’’. Gallup Institute of Global Well-Being.
256
Section II: Comparisons of Income and Well-Being Through Time
These findings echo Deaton’s steeper slope on the income–happiness relationship for rich countries than for poor ones. Both sets of findings reflect that respondents in wealthier countries are better able to enjoy income because they are better positioned to use it for a broader range of things other than basic needs (and may have more time to enjoy their incomes as well). As a result, questions about survival issues seem to be more relevant to respondents in poor countries, while those concerning enjoyment or quality of life issues resonate more among rich country respondents. Other possible factors muddying the waters are measurement error and the role of non-market income. Measurement error is a particular problem for developing countries with large informal sectors, as it is difficult for workers with unstable salaries to estimate their earnings accurately, and at the same time, weak tax systems provide strong incentives to underreport. Non-market income is a factor in both contexts. In rich countries, better public goods and assets that facilitate leisure, such as inherited assets, may contribute to higher happiness levels but are not directly correlated with per capita income levels. In poor countries, meanwhile, non-market income in the form of subsistence agriculture, home-grown food, and family networks can all contribute to happiness—or mitigating unhappiness—although it is not reported as income. These factors vary significantly across countries. Work on other domains also highlights the role of question framing. Mauricio Cardenas and colleagues, using the Gallup poll for Latin America, look at the determinants of satisfaction with education. They find, for example, that satisfaction with education does not vary much by income levels, but that it is negatively correlated with education levels. This surely reflects higher aspirations among the more educated respondents, as well as less available information (such as test scores) upon which to base assessments among poorer ones. Another explanation for the mixed findings on education is the extent to which education is endogenous to many other variables that are also correlated with happiness, such as income, social trust, and health. When they compare satisfaction with life across different education levels, they find that education levels are positively correlated with life satisfaction when a Ladder of Life question is used, but they are negatively correlated when a ‘‘satisfaction with standard of living’’ question is used. When they include controls for differences in the innate optimism of individuals, the positive correlation on the ladder question becomes insignificant, while the negative correlation between education satisfaction and satisfaction with standard of living holds.9 Their findings reflect both higher expectations among the more educated, as well as question framing.
The Easterlin and Other Paradoxes
257
Country selection issues
Even the opposite extremes of the debate recognize that there is some relationship between income and happiness across countries, with wealthier countries generally showing higher levels of happiness than poorer ones. Yet different surveys sample different selections of countries, and that, too, seems to affect the strength of the relationship. Graham’s above-mentioned work with Pettinato, based on a large sample of Latin American and OECD countries and an open-ended life satisfaction question, finds a relationship between income levels and happiness, although within each of the poor and rich country sets there is no clear pattern (Figure 9.1). Howell and Howell’s meta-analysis finds that the income-subjective well-being relationship is weakest among the wealthiest developing countries in their 54-country sample, and strongest among the poorest and least educated countries. They also find that the effect is strongest when economic status is defined as a stock variable (wealth or assets) rather than as a flow variable (income) (Howell & Howell, 2008). Education seems to mediate the income–happiness relationship. The strongest R-squared, not surprisingly, is for the least educated respondents in the poorest countries. In contrast, there is very little difference in the strength of the relationship between the most educated respondents in the poor or the rich countries (Howell & Howell, 2008).10 The Gallup World Poll clearly contains the largest and most regionally and developmentally diverse set of countries of any of the surveys that have been used to study happiness to date. Yet that presents methodological challenges as well as analytical diversity. A large number of the new countries in the Gallup poll are small, poor countries in sub-Saharan Africa and/or the transitional economies, which have seen the dismantling of existing social welfare systems and dramatic falls in happiness. Thus the steeper sloped income and happiness relationship that appears in research based on these data may not be driven by rising incomes and happiness in the rich countries, but by falling incomes and happiness in a large number of small countries at the bottom of the distribution. A similar point has been made by Easterlin about the sample of countries in the World Values Survey. It is not clear, however, how to resolve the problem of different country selection, and dropping large parts of the sample—for example the transitional economies—in order to retain comparability eliminates some of the most important trends in the global economy in recent decades. The transitional economies do have a distinct experience, however. Easterlin examined happiness in Eastern Europe from 1989 to 1998 and found that life satisfaction followed the V-shaped pattern of GDP for those same years, but failed to recover commensurately. Across domains,
258
Section II: Comparisons of Income and Well-Being Through Time
increased satisfaction with material standards of living occurred at the same time that satisfaction with work, health, and family life decreased. Disparities across cohorts increased, with the unhappiest respondents being the least educated and those over age 30, not surprisingly those cohorts that were least able to protect themselves from economic dislocation and take up new opportunities offered by the transition (Easterlin, 2008). Since Easterlin did his work, happiness levels have recovered in at least some of the former Soviet Union (FSU) countries, such as Russia (Eggers, Graham, & Gaddy, 2006). Whether or not they will recover fully in all of them is an open question. But surely their inclusion in cross-country analysis during a time period when happiness levels were unusually low, as well as that of a large number of small African economies that are likely to perform poorly for the foreseeable future, will affect the slope of the cross-country income and happiness relationship.
The Paradox of Unhappy Growth
The relationship between happiness and income may be affected as much by the pace and nature of income change as it is by absolute levels. Based on the Gallup World Poll in 122 countries around the world, Eduardo Lora and his collaborators find that countries with higher levels of per capita GDP have, on average, higher levels of happiness. Yet, controlling for levels, they find that individuals in countries with positive growth rates have lower happiness levels. We have chosen to call this negative correlation between economic growth and happiness the ‘‘paradox of unhappy growth’’ (see Lora & Chaparro, forthcoming; Deaton, 2008; Wolfers & Stevenson, 2008).11 A simple scatter plot shows that the relationship between per capita incomes and life satisfaction (as measured by the ‘‘best possible life’’ question) is linear (when incomes are logged), while that between growth rates and life satisfaction is negative (Figures 9.3a and b). Econometric analysis confirms the visual relationship in the scatter plot. In an ordinary least squares (OLS) regression with average life satisfaction in each of the 122 countries in the Gallup World Poll as the dependent variable, they find that the coefficient on GDP per capita is positive, while that on economic growth—defined as the average rate of growth over the past five years—is negative (and significant in both cases) (See Table 9.2). Deaton, and Stevenson and Wolfers, also find evidence of an unhappy-growth effect based on the full sample of the Gallup World Poll. Stevenson and Wolfers find insignificant effects of growth in general, but strong negative effects for the first stages of growth in ‘‘miracle’’ growth economies, such as Ireland
National average life satisfaction without economic growth effect
9.0 8.0 7.0 6.0 5.0 4.0 3.0 2.0 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0 Natural Log, GDP percapita, US$ PPP
National average life satisfaction without GDP per capita effect
Figure 9.3a The relationship between satisfaction and GDP per capita. Source: Eduardo Lora. ‘‘Beyond Facts: Understanding Quality of Life in Latin America and the Caribbean’’. IADB.
0.5 0.0 –0.5 –1.0 –1.5 –2.0 –2.5 –3.0 -5.0
-2.5
0.0
2.5
5.0
7.5
10.0 12.5 15.0
Economic growth, GDP percapita average real growth rate 2000-2005
Figure 9.3b The relationship between satisfaction and economic growth. Source: Eduardo Lora. ‘‘Beyond Facts: Understanding Quality of Life in Latin America and the Caribbean’’. IADB. 259
260
Section II: Comparisons of Income and Well-Being Through Time
TABLE 9.2 The paradox of unhappy growth The relationship among satisfaction, income per capita and economic growth 122 countries
Life satisfaction Standard of living Health satisfaction Job satisfaction Housing satisfaction
GDP per capita
Economic growth
0.788 0.108 0.017 0.077 0.084
0.082 0.018 0.017 0.006 0.006
*** *** * *** ***
*** *** ***
Source: IDB - RES using Gallup World Poll 2006–2007 data. Eduardo Lora. ‘‘Beyond Facts: Understanding Quality of Life in Latin America and the Caribbean’’. IADB Notes:
1. OLS regression; dependent variable is average life satisfaction per country, growth rates are averaged over the past five years. N ¼ 122 2. The coefficients on GDP per capita are marginal effects; how much does the satisfaction of two countries differ when one has two times the incomes ofanother. The coefficients on growth imply how much an additional percentage point of growth affects life satisfaction. 3. The life satisfaction variable is on a 0 to 10 scale; all others are the percentage of respondents that are satisfied.
and South Korea during their takeoff stages. The negative effect becomes insignificant in later stages (Deaton, 2008; Wolfers & Stevenson, 2008). Deaton finds that the inclusion of region dummies make a major difference to the results, with the significance being taken up by Africa and Russia, regions that are both fast-growing and very unhappy. In Lora’s sample, economic growth is also negatively correlated with perceived standard of living and with health satisfaction. When the sample is split into rich and poor countries (above and below median income), the effect holds for the rich but not for the poor countries. The only variable that is significant and negative on growth for poor countries is health satisfaction. One can imagine the factors associated with structural change and rapid growth rates in poor countries—such as long working hours and new industries without provision for worker safety or environmental externalities— that could have negative effects on health. Meanwhile growth is negatively correlated with economic, health, job, and housing satisfaction, in addition to life satisfaction for the rich countries.12 When they split the sample into above and below median growth rates, meanwhile, the unhappy growth effect only holds for those that are growing at rates above the median. Graham and Chattopadhyay, using Latinobarometer data, also find hints of an unhappy growth effect, or at least an irrelevance of growth effect. In
The Easterlin and Other Paradoxes
261
contrast to Lora et al., they use individual happiness on the left-hand side, with the usual socio-demographic and economic controls and clustering the standard errors at the country level. When they include the current GDP growth rate in the equation, as well as the lagged growth rate from the previous year (controlling for levels), they find that the effects of growth rates—and lagged growth rates—are, for the most part, negative but insignificant (Graham & Chattopadhyay, 2008) (Table 9.3). It may well be that this unhappy growth is driven by its nature in rapidly changing economies, where growth is often accompanied by changes in rewards to different skill sets and increased job insecurity for some groups, TABLE 9.3
Happiness immune to country level economic growth?
Dependent variable: happy age age2 gender married edu edu2 socecon subinc ceconcur unemp poum domlang vcrime els growth_gdp gini
0.0240 (4.40)** 0.0000 (3.53)** 0.0330 1.5500 0.0790 1.7800 0.0410 1.5300 0.0010 0.8800 0.2110 (5.22)** 0.2900 (8.78)** 0.2340 (9.04)** 0.1810 (2.05)* 0.1800 (4.48)** 0.5380 (2.73)** 0.1160 (2.30)* 0.0900 (5.48)** 0.0170 0.5300 0.0170 0.7000
0.0230 (4.34)** 0.0000 (3.88)** 0.0070 0.4800 0.0910 (2.40)* 0.0260 1.1800 0.0010 0.7000 0.2160 (5.76)** 0.2900 (8.02)** 0.2260 (9.50)** 0.1760 (3.45)** 0.1890 (5.42)** 0.4810 (2.48)* 0.1060 (2.98)**
0.0230 (4.23)** 0.0000 (3.72)** 0.0070 0.5200 0.0940 (2.56)* 0.0280 1.2900 0.0010 0.7900 0.2150 (5.77)** 0.2940 (8.36)** 0.2360 (7.66)** 0.1900 (3.59)** 0.1830 (5.56)** 0.4840 (2.48)* 0.1060 (2.89)**
0.0220 (4.29)** 0.0000 (3.76)** 0.0070 0.4800 0.0930 (2.60)** 0.0260 1.2800 0.0010 0.7600 0.2170 (5.78)** 0.2920 (8.41)** 0.2370 (8.92)** 0.1880 (3.69)** 0.1840 (5.59)** 0.4810 (2.48)* 0.1080 (3.08)**
0.0090 1.1100 0.0270 1.2400
0.0040 0.6000 0.0240 1.1200
0.0060 0.7700 0.0240 1.1900 (continued )
TABLE 9.3
(Continued)
0.0190 1.4000
gdpgrhl1 gdpvol2 Observations
34808
67308
67308
0.0180 0.9900 0.0030 0.1400 67308
Absolute value of z statistics in parentheses * significant at 5% ** significant at 1% Regressions dustered at a country level
• Ordered logit regressions, clustered at the level of the country. • els (economic ladder scale) question asked only in select iterations of the survey, and hence reduced observations when included in the regression.
• The results were robust and consistent with different durations for lagged growth rates (1–3 years) and growth rate volatility (2–5 years).
• Description of the variables is appended below. Variable Name happy age age2 gender married edu edu2 socecon subinc ceconcur unemp poum els vcrime domlang growth_gdp gini gdpgrll gdpgrE gdpgrl3 gdpvol2 gdpvol3 gdpvol5
Variable Description Happiness/Life Satisfaction: 1 ¼ Lowest 4 ¼ Highest Age Age squared Gender: 1 ¼ Male 0 ¼ Female DV Married: 1 ¼ Yes 0 ¼ No (Single/Separated) Years of education Years of education squared Socioeconomic Level: 1 ¼ Verybad 5 ¼ Verygood Subjective Income: Does family salary fulfil family needs: 1 ¼ No, great difficulty 3 ¼ Yes, without much difficulty Country Economic Situation, Current: 1 ¼ Very bad 5 ¼ Very good DV: Unemployed 1 ¼ Yes 2 ¼ No Prospect of upward mobility for self: 1 ¼ Worse 2 ¼ Same 3 ¼ Better Economic ladder scale: 1 ¼ Poorest 10 ¼ Richest Victim of crime (indiv or in the family) in the last 12 months: 1 ¼ Yes 2 ¼ No DV: Dominant language based on mothertongue: 1 ¼ Spanish/ Portuguese 0 ¼ Other/minority GDP growth rate (WDI) Gini coefficient (WDI) GDP growth rate, lag 1 GDP growth rate, lag 2 GDP growth rate, lag 3 GDP growth rate volatility, last 2 years GDP growth rate volatility, last 3 years GDP growth rate volatility, last 5 years
Source: Graham and Chattopadhyay based on Latinobarometro data, The Brookings Institution, 2008.
262
The Easterlin and Other Paradoxes
263
and by related increases in vertical or horizontal inequality. Latin America in recent decades certainly fits this pattern, which may help explain unexpected pockets of frustration in relatively prosperous countries like Chile. Rapid growth in newly reforming economies, meanwhile, as in the case of Korea and Ireland and in the case of many more recent examples in the emerging market economies, is typically even more uneven in terms of rewards. Given that cross-country analysis of the income–happiness relationship usually captures some sample of countries in this particular stage, some of the seeming outliers in the analysis may at least in part be explained by the ‘‘unhappy growth’’ paradox.
The Happy Peasants and Frustrated Achievers Paradox
There is an overall happiness and income relationship within countries, and wealthier people are, on average, happier than poor people. Yet the withincountry story is more complicated than the averages suggest, as in the case of the cross-country income and happiness relationship. It is typically not the poorest people who are most frustrated or unhappy with their conditions or the services that they have access to, for example. Graham and Pettinato, based on research in Peru and Russia, identified a phenomenon that is now termed the ‘‘happy peasant and frustrated achiever’’ problem (Graham, & Pettinato, 2002b). This is an apparent paradox, where very poor and destitute respondents report high or relatively high levels of well-being, while much wealthier ones with more mobility and opportunities report much lower ones. This may be because the poor respondents either have a higher natural level of cheerfulness or have adapted their expectations downwards, while the wealthier ones have constantly rising expectations (or are naturally more curmudgeonly). Regardless of the balance between objective conditions and individual character traits driving the paradox, it mirrors some of the puzzles in the cross-country relationship. These include the apparent greed effect that Deaton finds at the highest income levels, where the slope in the income– happiness relationship is steepest, and the relatively large number of unusually happy countries at the bottom of the income distribution (Deaton, 2008). A closer look at Graham and Pettinato’s frustrated achievers shows that they were more likely to have had upward mobility than the average respondent, and they were of average incomes and education for their relative samples, of similar gender, and more likely to be living in urban than rural areas. Yet they reported that their current economic situation was worse or much worse than the past. And when compared with upwardly mobile respondents who did not report frustration, they had lower levels of general
264
Section II: Comparisons of Income and Well-Being Through Time
life satisfaction, they had higher fear of unemployment, and they were more concerned about relative income differences (as assessed by their scores on the poor-to-rich ladder question). Some of this may be driven by less optimistic character traits or lower innate set points, which probably have a similar distribution across most population samples; e.g., the percentage of natural curmudgeons is likely to be similar, regardless of what the dependent variable is. Yet we also posit that at least some of it is contextually driven, not least because of the nature of the contexts under study. The behavioral economics literature highlights the extent to which individuals value losses disproportionately to gains. It is not a stretch of the imagination to assume that upwardly mobile respondents who managed to escape poverty or near-poverty in the volatile macroeconomic context of both Peru and Russia in the late 1990s would be loss-averse, not least because of the absence of any safety net or social insurance system. Their income mobility, while having an overall positive trajectory, may have been punctuated with spells of unemployment or unstable income flows. If they were recent migrants, meanwhile, they also probably left strong family or other support networks behind, which are not readily available in crowded urban or peri-urban contexts. The poor, some of whom rely on subsistence agriculture rather than earnings, have much less income to lose and have probably adapted to constant insecurity. Some of the literature on job insecurity, for example, shows that reported insecurity is actually higher among more formal-sector workers with more stable jobs than it is among informal-sector workers. The latter have either adapted to higher levels of income and employment insecurity (and/or have selected into jobs with less stability but more freedom) (Lora & Chaparro, forthcoming). John Knight and Ramani Gunatilaka, and Martin Whyte and Chunping Hun, each find an analogous urban effect in China, where urban migrants who are materially much better off than they were in their pre-migration stage neverthless report higher levels of frustration with their material situation. Their reference norm, meanwhile, quickly becomes other urban residents rather than their previous peers in rural areas (Knight & Gunatilaka, 2007; Whyte & Hun, 2006). In addition to comparison effects, there may also be new costs related to urban living that erode the positive effects of income gains. The same literature highlights the extent to which individuals adapt much more to income gains than to status gains. Based on the German Socio-Economic Panel, Rafael di Tella shows that most individuals adapt to a significant income gain or salary increase within a year (in econometric terms, the effect of income gains becomes insignificant after a one-year lag), while status gains (such as a promotion) have a positive effect that
The Easterlin and Other Paradoxes
265
lasts up to five years (di Tella & MacCulloch, 2006). In the context of the frustrated achievers in very volatile emerging markets contexts, where currencies are often shifting in value and where the rewards to particular skill and education sets are in flux, as are social welfare systems, income gains may seem particularly ephemeral. More generally, the paradox highlights the extent to which slightly raised expectations in the context of rapid economic change may result in more frustration and risk aversion than do static poverty levels. To the extent that there are macro-level implications to this micro-level phenomenon, it supports a scenario where growth rates and economic development are associated with less rather than more happiness, at least in the short term, until higher income levels are stabilized. Over the long term, however, there does seem to be a generalized levels effect, with countries with higher levels of GNP on average happier than poorer ones, albeit with a great deal of variance within the subsets of rich and poor countries.
The Aspirations Paradox
A third, related paradox is the ‘‘aspirations paradox.’’ This paradox is most evident in the health satisfaction arena, although there are also some hints of it in the findings on education, job, and financial satisfaction. Across countries, there is higher tolerance for poor health in the poorer countries, and less satisfaction with better health in the rich ones. Within countries, while rich people are slightly more satisfied with their health than poor ones, and more ‘‘objective’’ measures of health, such as the EQ-5D description system, also track with socioeconomic status, the gaps in the assessments of satisfaction are much smaller than gaps in objective conditions (quality, access, outcomes) would predict.13 The same often holds across education, job, and economic satisfaction domains, depending on the sample.14 Lora and his collaborators and Graham and Chattopadhyay (using different data sets for Latin America) find that respondents in poor countries are more likely to be satisfied with their health systems than are respondents in wealthier ones, while respondents in some very poor countries, such as Guatemala, have much higher levels of health satisfaction than do those in much wealthier ones with better health systems, such as Chile. Deaton finds the same pattern—or lack of one—with satisfaction with health systems in the worldwide Gallup poll. While there are surely outliers, objective health conditions—as measured by indicators such as morbidity and life expectancy—are materially better in the wealthier countries (Lora & Chaparro, forthcoming; Graham & Chattopadhyay, 2008; Deaton, 2008). Cross-country comparisons of average levels of
266
Section II: Comparisons of Income and Well-Being Through Time
personal health satisfaction demonstrate a similar, although not as notable, pattern. Within countries, wealthier respondents are more likely to be both happier and more satisfied with their health than are poor ones. Despite the aggregate pattern, though, there is clearly an ‘‘optimism bias’’ in the responses of the poorest respondents across many domains—health as well as many others, at least in Latin America. For example, those in the highest quintile in the region hold 57 percent of the income (on average), while those in the poorest quintile hold 4 percent. But the differences in their perceptions are much smaller. Seventy-nine percent of individuals in the highest quintile declare themselves satisfied with their material or economic quality of life, while 57 percent of those in the lowest quintile say they are satisfied. There is a similar ‘‘optimism bias’’ in the responses of the poor as they assess their living conditions and public policies in their countries. The gaps in the assessments of satisfaction with health, education, and jobs between the rich and the poor are much smaller than the gaps in objective conditions. This paradox is probably due to lower expectations and available information among those living in poorer contexts, as well as to lower expectations. For wealthier individuals and respondents in wealthier countries, aspirations and awareness of those in wealthier countries may go up as much if not more rapidly than improvements in service provision (and/or economic growth). At the same time, there is also inconsistent usage of available information—such as test scores—among slightly wealthier respondents. A surprisingly small amount of school choice, for example, is informed by test score results (Lora & Chaparro, forthcoming). This, among other things, may contribute to increased public frustration in the face of improvements in service quality. The gaps caused by aspirations bias are greater for more subjective things like personal health status than they are for education. In the case of education, there is usually more objective information available to make assessments, and parents are most likely evaluating the education that their children are receiving rather than their own. In the case of financial satisfaction, income levels provide a benchmark for making such assessments, in contrast to personal health status and satisfaction, which lacks an analogous general indicator, other than mortality rates, which are an expost measure, at least for the average individual. The gaps between perceptions and objective measures seem to be greater at the individual level than at the average country level (perhaps not a surprise as there is more variance); for richer rather than poorer countries (as relative deprivation effects seem to increase as average wealth increases); and for poorer rather than richer individuals (perhaps because they have less
The Easterlin and Other Paradoxes
267
good information to make assessments, as well as lower expectations) (Lora & Chaparro, forthcoming; Graham & Felton, 2006). These gaps between subjective and objective measures across various domains may be part of the explanation for the divergent conclusions over the income–happiness relationship and, like that relationship, also vary according to which questions are used and the nature of the objective indicators.
Relative incomes and inequality—part of the paradox?
Happiness may not be all relative, but relative differences do seem to matter. While wealthier people are happier than less wealthy ones on average, people of similar income levels are less happy when the incomes of those in a relevant reference group, ranging from neighbors to professional cohorts, to towns and cities, are higher (Graham & Felton, 2006; Luttmer, 2005; Kingdon & Knight, 2007).15 These concerns for relative income differences, which in theory are greater as average income levels rise, are often cited as part of the explanation for the Easterlin paradox. The intuition is that until basic needs are met, people are not concerned about relative differences, and the relationship between happiness and income resembles a linear one. At higher levels, however, it curves off and resembles a logarithmic function. At the same time, micro-level empirical work suggests that concerns for relative income differences arise at surprisingly low levels of income. This is apparent, for example, in the lack of correlation between average per capita incomes and happiness among the less developed country sample in Graham and Pettinato’s work, as well as in within country work on Latin America by Graham and Felton and by Lora and colleagues, as well as on South Africa by Kingdon and Knight.16 Ravallion and Lokshin test for relative deprivation effects in a much poorer context—Malawi—where basic needs are an issue for the average respondent, and find that they do not matter for most respondents in their sample, but do matter for those who are comparatively better off (Ravallion & Lokshin, 2005). For the Gallup poll for Latin America, Eduardo Lora and his colleagues find that reference group income—defined as similar age, income, and education cohorts—is positively correlated with life satisfaction (the Cantril ‘‘best possible life’’ question) but negatively correlated with satisfaction with one’s standard of living, job, and housing (Lora & Chaparro, forthcoming) (Table 9.4). It is likely that both question-framing and variance across domains mediates the extent to which comparison effects matter. Indeed, one hypothesis—which could be tested—is that the frames, which provide more visible or tangible reference points across
268
Section II: Comparisons of Income and Well-Being Through Time
TABLE 9.4 Relative income effects for Latin America The relationship among satisfaction, own income and income of others Household per capita monthly income, US$ PPP, natural log Life satisfaction Standard of living Health satisfaction Job satisfaction Housing satisfaction
0.410 0.370 0.196 0.379 0.261
*** *** *** *** ***
Reference group average monthly per capita income, US$ PPP, natural log 0.254 0.217 0.003 0.429 0.236
* * ** **
Source: fDB - RES using Gallup World Poll 2006–2007 data. Notes: Each individual belongs to one reference group. A reference group is composed by every individual inside a country of the same gender, within the same age range and with a similar educational level.
The relationship between satisfaction, own income, income of others: the effect of gender, residency and relative income Monthly average per capita income, US$ PPP, natural logarithm
People with income above the regional median
People with income below the regional median
People in the urban areas
People in the rural areas
Life satisfaction Peronal economic situation Health satisfaction Job satisfaction Housing satisfaction
0.129 0.933 ***
0.549 0.578 ***
0.149 0.328 *
0.500 0.044
0.306 1.810 *** 0.970 *
0.921 ** 0.142 0.697 **
0.014 0.847 ** 0.251
0.007 0.609 0.092
Source: RES- IDB based on Gallup (2007).
jobs, housing, and education levels, matter more for comparison effects than they do for absolute ones, while inherent character traits/optimism are more important to open-ended life satisfaction questions. (Alternatively, of course, naturally less happy people might be more likely to be concerned about comparison effects.) John Helliwell and colleagues, working with the Gallup poll (the world sample) and also in this volume, find that average per capita income levels are negatively correlated with life satisfaction (again the Ladder of Life question), controlling for individual levels. The significance goes away when additional questions about basic needs, corruption, and freedom to
The Easterlin and Other Paradoxes
269
choose are added to the model specification. When the sample is split according to region, the coefficient on average per capita GDP only remains negative (and significant) for Eastern Europe and the FSU and for Africa, and negative (but not significant) for Latin America. Helliwell posits that, for the sample as a whole, relative income effects are likely to be mediated by taxes and by the public goods that accompany particular countries’ distribution. Concerns for relative differences do not seem to be mediated by public goods in the three regions where the ability of the state to provide public goods has either been dramatically shaken or was very weak to begin with. In earlier work based on the Latinobarometer, Graham and Felton find that average country-level incomes do not matter to individual happiness, but relative income differences—measured as the distance from the mean for the average income in one’s country—do matter. Thus, even though a poor peasant in Honduras is half as wealthy as a poor Chilean, the former is happier because his/her distance from mean income is smaller (Figure 9.4). Their findings depart from those of Alesina, di Tella, and MacCulloch for the Untied States and Europe, where the effects of inequality (albeit measured very differently) on individual happiness are very modest. The starkest contrast is the United States, where the only group that is made less happy by inequality is left-leaning rich people! In the United States, inequality remains for many respondents a sign of future opportunities and mobility, even though the data on mobility rates no longer support that perception (Alesina, di Tella, & MacCulloch, 2004; Benabou & Ok, 1998; Graham & Young, 2003). Graham and Felton posit that, in Latin America in contrast, inequality is still a sign of persistent advantage for the rich and disadvantage for the poor, even though the data show more mobility than public perceptions suggest (see also Graham, forthcoming). Country-level aggregations may not be the most relevant ones for studying concerns for relative income differences; the average person may be more concerned with reference groups such as neighbors or the workplace, where comparisons are more visible. Graham and Felton find that the relative income effect holds and is indeed more notable across cities of different sizes. It is stronger for large cities, where there is more income variance, and smaller for small cities, where average income levels are positively and significantly correlated with happiness (as opposed to insignificant, as for the rest of the sample) Relative incomes are still negatively correlated with happiness in the small cities, meanwhile.17 The relevant reference group probably also varies across cohorts and countries and may then still change. Graham and Pettinato find that their frustrated achievers assess their living standards favorably in comparison
270
Section II: Comparisons of Income and Well-Being Through Time
Figure 9.4 A tale of two countries – Comparing a rich and a poor person in Honduras to those in Chile.
to others in their community, but much more negatively when the reference group is expanded beyond the community to others in their country, a reference group that became more relevant as information and technology such as the internet became more widely available. The Kingdon and Knight work on China shows that recent migrants quickly change their reference group to their new urban counterparts rather than the living standards in their towns of origin. When the Gallup Latin American sample is split into ‘‘above’’ and ‘‘below the regional median income’’ groups, the comparison effect holds for both groups, with the difference being that health satisfaction is significant and negative for the below median group, but insignificant for the
The Easterlin and Other Paradoxes
271
rich, while job satisfaction is significant and negative for the rich but not for the poor, implying that the poor and the rich value different domains as they make comparisons. When the sample is split into urban and rural groups, the effects largely hold for the urban cohorts but not for the rural ones (analogous to Graham and Felton’s city size findings) (Lora & Chaparro, forthcoming). The evidence suggests that concerns for relative income differences matter and can erode the positive effects of higher absolute income levels on happiness, thus helping to explain the Easterlin paradox. It also suggests that they hold at surprisingly low levels of income, as is suggested by the lack of a clear income–happiness relationship within some less-developed countries (LDC) samples. Yet it is difficult to be conclusive about how relative differences mediate the Easterlin paradox. One reason for this is that different reference groups matter to different cohorts or cultures, and country-level incomes may not be the most relevant comparator group in many instances. In addition, concerns for relative income differences are mediated by perceptions about what inequality signals as well as the availability (or not) of public goods.
Mediating Factors in the Income–Happiness Relationship:Social Capital, Freedom, and Corruption
There are a number of non-income variables that correlate strongly with happiness. The Gallup poll has data on variables such as the importance of friendships. Friendships matter to the well-being of the average Latin American respondent more than health, employment, or personal assets, and only slightly less than food security (of course it could be that happier people are more likely to have and value friendships). This varies according to income levels, with the rich valuing work and health more, and the poor valuing friendships (Figure 9.5). These friendships most likely provide important coping mechanisms for the poor in the absence of publicly provided safety nets. Whether they serve as strong or weak ties in the Granovetter sense, with weak ties being more important for upward mobility, is an open question (Granovetter, 2009). The life domains that are most relevant to happiness in Latin America are economic satisfaction, the importance of friends, and work, health, and housing satisfaction (in that order of importance). Reporting religion to be important and having access to a telephone, meanwhile, are also positively correlated with happiness. A number of studies show that those who have religious faith are, on average, happier than others, although it is not clear whether happier people are likely to
272
Section II: Comparisons of Income and Well-Being Through Time
Monetary valuation of some life satisfaction determinants (If someone suffers a change in his life conditions, what is the new income requiered to compensate for the related effects?) Monthly per capita household income, US$ PPP 0
200
400
600
800 1,000 1,200 1,400 1,600 1,800
163
Original income
1,675
Food insecurity Losing your friends
1,261 581
Losing your good health
534
Losing your faith
491
Losing most of your assets
462
Housing insecurity 342
Losing your telephone connection Losing your job Gaining a college degree
264 80
Fuente: IDB calculation using Gallup (2007). The person for this example is a single 30 year old woman, with no children, a high school degree, employed, with friends and religious believes.
Figure 9.5 Valuation of life satisfaction components.
have faith, or whether having religion makes them happy, or if there is a more generalized effect that comes from the social networks that often accompany religiosity. Meanwhile, there has been a proliferation of cell phones in Latin America in recent years, as a result of the privatization of telecommunications. Cell phones are both a status good, and provide important linkages with job and other networks that large numbers of poor in urban shantytowns previously lacked (Labonne & Chase, 2008). While a full discussion of why these variables matter is beyond the scope of this paper, it is likely that their relative importance varies significantly across countries and cultures, as well as socio-economic levels. Because they mediate the income–happiness relationship, they are likely to play some role in explaining cross-country discrepancies and outliers. John Helliwell has done extensive research into whether living in contexts with greater freedom and social capital plays a role in individual wellbeing. The basic answer is a resounding positive on both fronts. In a recent paper based on the Gallup poll, Helliwell and colleagues compare the various determinants of well-being across 120 countries in the five regions covered by the poll.18 They find that all measures of social connections are significantly correlated with life satisfaction, across the countries and regions in the sample. And respondents seem to value both the support that they get from others and the support that they give to others.
The Easterlin and Other Paradoxes
273
They also test for interregional differences in the effects of income, social connections—as measured by the importance of friendships and memberships in associations, among others—and corruption. They find that the income coefficient is weakest in Africa—most likely due to the likelihood of mismeasurement of the income variable and the importance of subsistence agriculture. The effects of social connections are lower in Asia and Africa and higher in Region 1 (the United States, Western Europe, Australia, and New Zealand) than in any other region. The negative effects of corruption are weakest in Asia and Africa and strongest in Region 1, as are the positive effects of personal freedom. It seems that the well-being effects of living in countries where the perceived level of corruption is high are lower where corruption is a long-established feature of the status quo, and the well-being value attached to a sense of personal freedom is slightly higher in societies classified as individualistic rather than collectivist. A recent paper by Ronald Inglehart and his colleagues also finds that the well-being effects of freedom are greater in countries that have more of it and are more accustomed to it (Inglehart et al., 2008). In the same vein, Graham and Chattopadhyay, using the Latinobarometer data, find that the well-being effects of being a crime victim are lower—as are reporting rates—in countries in Latin America where crime rates are higher. As crime rates go up, respondents seem to adapt, which is evidenced in lower reporting rates (reporting of petty crimes is less likely to result in corrective action as overall rates go up) and less stigma attached to being a victim. Nick Powdthavee’s work on crime in South Africa suggests similar dynamics (Graham & Chattopadhyay, forthcoming; Powdthavee, forthcoming). In the same way that individuals adapt to the benefits (and also to the negative externalities) of overall rising income trends, they also adapt to the costs of rising crime and corruption trends. In the same way that income increases across time may not result in commensurate increases in wellbeing, increasing crime and corruption may not result in commensurate decreases in well-being as societies adapt to these phenomena.19 There are surely tipping points in both instances, as levels of crime and corruption become unsustainable, for example, and/or as rising income levels result in positive externalities that increase happiness (and/or greed?).
Health Gains and Happiness: Adaptation, Again?20
Analogous to the Easterlin paradox, the Preston curve shows that income matters much more to health and longevity at lower levels of income than at higher ones (Figure 9.6). Income gains in poor countries are associated with
274
Section II: Comparisons of Income and Well-Being Through Time
90
Life expectancy at birth, years in 2000
Japan Mexico
China
Korea
Spain
Italy
Argentina Switz.
Norway USA
Germany
75
France UK Bahamas
Brazil Russian Fed. Indonesia
60
Saudi Arabia
India Pakistan Bangladesh Namibia South Africa
45 Botswana
Zambia
Equitorial Guinea
30 0
5000
10000
15000 20000 25000 GDP percapita in 2000 (current PPP$)
30000
35000
40000
Figure 9.6 Life expectancy and GDP GDP per capita (current PPP$) in 2000 versus average life expectancy at birth of total population in 2000. The trend line is the estimated line using the logarithmic form: life expectancy ¼ f[ln(GDP per capita)] (R2 ¼ 0.6467). Source: World Development Indicators, 2007, replicating A. Deaton (2003). ‘‘Health, Inequality, and Economic Development.’’ Journal of Economic Literature. March.
rapid improvements in basic health and in defeating preventable diseases and lowering infant mortality rates.21 The availability of clean water and electricity can make a huge difference in the diarrheal diseases that claim so many infant deaths in poor countries (Adrianzen & Graham, 1974). At higher levels of per capita income, technology and scientific innovation play more of a role than income in generating cures for the types of diseases that are more typical of developed economies, such as cancer. Gains in longevity at higher levels of life expectancy are much harder to achieve. At the same time, due to technological advances, poor countries today are able to enjoy much higher levels of life expectancy at lower income levels than were their predecessors in the development process (Deaton, 2004; Preston, 1975). The health and happiness relationship may well reflect these trends, if not exactly mirroring the paradox. People no doubt adapt to better health conditions, and in turn expect them. Angus Deaton finds that satisfaction with health (which is highly correlated with happiness) and per capita income are surprisingly uncorrelated across countries. A higher percentage of Kenyans (82%) are satisfied with their personal health than Americans (81%), and the United States ranks 81st out of 115 countries in public confidence in the health system; lower than countries such as India,
The Easterlin and Other Paradoxes
275
Malawi, and Sierra Leone (Deaton, 2008). Lora and colleagues and Graham and Chattopadhyay’s findings on health satisfaction across Latin American countries, discussed above, echo these results. Arie Kapteyn and colleagues find that even though objective health conditions are better in the Netherlands than they are in the United States, self reports of work disability are significantly higher in the former context, even when differences in response scales and differences in the generosity of government programs (which are much more generous in the Netherlands) are controlled for. Their research, based on novel vignette methodology, finds that American respondents have a ‘‘tougher’’ standard than do Dutch ones when assigning a work disability status, particularly in the more subjective health domains of pain and emotion (Kapteyn, Smith, & van Soest, 2007).22 These differences reflect norms and expectations about health standards more than they do objective conditions. Along a similar line, Graham and Felton find that the well-being costs related to obesity in the United States vary significantly across socioeconomic, racial, and professional cohorts, and are much higher in cohorts where obesity is not the norm. They posit that these ‘‘unhappiness’’ effects, which are above and beyond the measured health effects of obesity, probably stem from stigma, which they show is higher for individuals who depart from the weight norm for their cohort, and argue that the incentives to reverse the condition in cohorts with higher weight norms are lower. They also find that being obese is associated with a lower probability of upward income mobility. This may stem from lower expectations or from more discrimination in the job market, or both. Regardless, the finding is suggestive of how norms and expectations affect health standards and related outcomes (see Felton & Graham, 2005; Graham, 2008). Once a certain level of health standards and longevity is achieved, there is no consistent cross-country relationship between health and happiness. What that level is remains an open question (as it does for income in the Easterlin paradox). Within countries, however, healthier people are happier—similar to difference in the across and within country relationship between income and happiness. A recent study, based on a sub-sample of wealthy European countries, finds that happiness and longevity are negatively correlated (Blanchflower & Oswald, 2007). Health expenditures and happiness are also negatively correlated for this sample. All of the countries in the sample have widely available care and relatively high life expectancies. At these socioeconomic levels, where people have come to expect good health, factors other than longevity, such as norms about health standards, may mediate the happiness and health relationship. In addition, longevity is only one measure of health, and slightly shorter but healthier life years may matter more to happiness
276
Section II: Comparisons of Income and Well-Being Through Time
than extending already long life expectancies. Similar to income, after a certain point more health may not buy more happiness, and other factors related to quality of life matter more.23 Meanwhile, it is also possible that, given an overall high standard and widely available health care, less healthy (and less happy) people demand more health expenditures. At the bottom end of the income scale, some countries with extremely poor health standards, such as Nigeria, Pakistan, Honduras, and Guatemala, have relatively high average happiness scores. Yet within each set of these same countries, healthier people are happier, again echoing the within and cross-country differences in the Easterlin paradox. While the discussion in this section is far from conclusive, and perhaps risks muddying the waters, it suggests the complexity of the relationship between happiness and income and a range of other mediating factors such as health increase as countries go up the development ladder. As levels go up, rising aspirations and increasing awareness interact with pre-existing cultural and normative differences, as well as both the extent and quality of public goods, which are in part endogenous to these cultural and normative differences. At the same time, because global information and access to a range of technologies are now available to countries at much lower levels of per capita incomes, benefits associated with higher incomes, such as better health care, are often available in contexts with much lower levels of per capita incomes than was previously possible. Not surprisingly, all of these factors—and how they are or are not captured in the survey data that are used—come into play in the debate over how much happiness levels increase as countries grow wealthier.
Conclusions
Our aim in this chapter was to help disentangle the debate on the Easterlin paradox and, more generally, the income–happiness relationship, both within and across countries. Our review, based on some of our work and that of many others, finds that while in general rich countries are happier than poor ones, there is a great deal of variance among the countries within the rich and poor clusters, as well as in the slope of the relationship. The results are quite sensitive to the method selected, the choice of micro- or macro-data, and the way that happiness questions are framed, thus supporting divergent conclusions about the importance of the paradox. We find, for example, that question-framing makes a major difference to the relationship, both in terms of direction and slope. Analysis based on questions that are framed in economic or status terms, for example, seems much more likely to yield a positive and linear relationship between
The Easterlin and Other Paradoxes
277
income and happiness, across and within countries, than are open-ended happiness or affect questions. What countries are in the sample also matters. Respondents in poorer countries, who are still struggling to meet basic needs, display a stronger income–well-being link than do those in wealthy countries, where that relationship is mediated by factors such as relative differences and rising aspirations. Education levels may also mediate that difference, with the least educated respondents in poor countries demonstrating the steepest slope, but more educated respondents in both rich and poor countries having a similar one. There is some evidence, based on the Ladder of Life question in the Gallup poll, which suggests that the slope of the income–happiness relationship is steepest at the top of the country wealth distribution, where respondents are either better positioned to enjoy wealth and/or are more aware of how their lives compare to those of others in poor countries. It is not clear that the same steep slope would hold with a more open-ended life satisfaction or affect question. An additional question is the extent to which the linear income–happiness relationship in more recent work based on the Gallup Poll is driven by lower happiness levels in a large number of small, poor countries in subSaharan Africa, which have been experiencing falling GDP per capita, and in the turbulent transitional economies, versus by rising happiness levels in the wealthier countries. Both factors may be at play. A number of paradoxes in the data support our basic propositions. The paradox of unhappy growth, for example, suggests that the rate of change matters as much to happiness as do per capita income levels, and that rapid growth with the accompanying dislocation may undermine the positive effects of higher income levels, at least in the short term. The number of countries experiencing these kinds of changes at the time a survey is conducted could surely affect results. A mirror image of this paradox at the micro level—the happy peasant and frustrated achiever phenomenon— again suggests that the nature and pattern of economic growth, and in particular instability and inequality issues—can counterbalance the positive effects of higher income levels for a significant number of respondents. Finally, low aspirations among the poorest respondents in the poorest countries can bias their responses upwards on a number of questions, particularly those that are more personal and open-ended, such as health satisfaction and open-ended happiness questions. The income–happiness relationship is also mediated by factors such as inequality levels and institutional arrangements, particularly as countries get beyond the basic needs level. Norms and aspirations are also at play, as citizens of particular countries not only adapt to the benefits (and possibly some of the negative externalities) associated with rising incomes, but also
278
Section II: Comparisons of Income and Well-Being Through Time
to the costs and complexities of things like rising crime and corruption. There is also significant evidence of adaptation to better and more widely available health care, and of an Easterlin paradox of sorts in the relationship between happiness and health. The complexity of the relationship between happiness and income—and the range of other mediating factors—seems to increase as countries go up the development ladder. Rising aspirations and increasing knowledge and awareness interact with pre-existing cultural and normative differences, as well as the extent and quality of public goods, which are in turn endogenous to the cultural and normative differences. At the same time, because global information and access to a range of technologies is now available to countries at much lower levels of per capita income than was previously the case, they have access to the benefits associated with higher income levels, such as better health care, quite early on in the development process. These complexities, coupled with different conceptualizations of happiness, which are captured differently by the various questions that are used to measure happiness, as well as important differences in the sampling of countries that are studied, are alone sufficient to explain divergent conclusions about the Easterlin paradox. Notes 1. The authors would like to thank Peyton Young, Andrew Felton, and Charles Kenny, as well as the participants and reviewers from the Princeton conference for very helpful comments. 2. Economists typically use the terms happiness, life satisfaction, and reported wellbeing synonymously, while psychologists typically make more specific distinctions between them, and see them as separate components of the broader concept of subjective well-being. 3. This finding holds for people who are, on average, happier, but not necessarily for those who are the happiest in every sample. See Diener & Biswas-Diener (2008) and Graham, Eggers, & Sukhtankar, (2004). 4. Deaton gets a positive and significant coefficient on a squared specification of the income variable. Stevenson and Wolfers split their sample into those countries above and below $15,000 per capita (in year 2000 U.S. dollars); they get a slightly steeper slope for the rich countries than for the poor ones. 5. Blanchflower and Oswald find a correlation coefficient of .50 for the two questions in Europe and the United States; Graham and Pettinato find one of .55 for Latin America, where the questions were used interchangeably in various years of the Latinobarometer poll. 6. The initial results and details on the methodology for this analysis are in Mario Picon (2008), Explaining subjective well-being by objective well-being measures: An application for Latin America; mimeograph; The Brookings Institution. 7. This question is slightly problematic, at least in econometric terms, as 96 percent answer yes to the yes–no question, providing very little variance for analytical purposes.
The Easterlin and Other Paradoxes
279
8. The most common adaptation for scale is to divide total household income by the square root of the number of people in the household, under the assumption that there are some economies of scale and that children, for example, consume less than equivalent adults. The IADB adaptation divided income by the number of household members; we ran the same regressions with the income variable adjusted for the square root of household size rather than on a per capita basis, and got identical results. 9. They control for individual optimism levels/person fixed effects, to the extent possible in a cross section, via a principal components analysis of domain satisfactions. See Mauricio Cardenas, Carolina Mejia, and Vincenzo Di Maro, Education and life satisfaction: Perception or reality? in Carol Graham and Eduardo Lora, Paradox and perception: Measuring quality of life in Latin America (The Brookings Institution Press, 2009). 10. They also find that the effect is stronger for men! 11. Deaton and Wolfers and Stevenson also find evidence of an unhappy-growth effect in the Gallup World Poll. It is also possible that initially happier countries grew faster than initially unhappy countries with the same income (because they had happier, more productive workers?) and thus the coefficient on growth in a regression which compares the two with final income and final happiness is negative. I thank Charles Kenny for raising this point. 12. This finding based on reported well-being departs from Ben Friedman’s more general proposition that some growth is necessary for overall welfare, even as levels increase, in order to keep economies and societies from stagnating and to generate productivity increases and technological advance. See Benjamin Friedman, The moral consequences of growth (NewYork: Alfred Knopf, 2005). 13. The EQ5D is a five-part question developed for the British general population, and now widely used in other contexts. The descriptive dimensions are: mobility, self care, usual activities, pain/discomfort, and anxiety/depression, with the possible answers for each being: no health problems, moderate health problems, and extreme health problems. See James W. Shaw, Jeffrey A. Johnson, and Stephen Joel Coons, ‘‘U.S. valuation of the EQ-5D health states: Development and testing of the D1 valuation model, in Medical Care, Vol. 43, No.3, March 2005. 14. Of course, this could also be considered a pessimism bias of the rich. 15. Luttmer’s work is based on U.S. PUMA’s, geographic units that are established in census data, which proxy for neighborhoods. 16. For example, Graham & Felton (2006) and Paradox and perception: Measuring quality of life in Latin America highlight the extent to which this holds for very poor countries in Latin America; Kingdon & Knight (2007) show that it holds for poor communities in South Africa. 17. Because there is not a good income variable in the Latinobarometer, the authors use an index of assets to proxy for wealth/income. See Graham & Felton (2006). 18. They drop roughly eight countries that do not have specifications for income. See John Helliwell, Haifang Huang, and Anthony Harris, International differences in the determinants of life satisfaction, mimeograph (University of British Columbia, 2008). For more detail on social capital and trust, and how it varies across cohorts and ethnic boundaries, see Stuart N. Soroka, John F. Helliwell, and Richard Johnson, Measuring and modelling interpersonal trust, in Fiona Kay and Richard Johnson, eds., Diversity, Social Capital, and the Welfare State (Vancouver: UBC Press, 2007).
280
Section II: Comparisons of Income and Well-Being Through Time
19. For a discussion of how people adapt and how these strategies may vary across socioeconomic cohorts, see Rafael Di Tella, Sebastian Galiani, and Ernesto Shargrodsky, Crime distribution and victim behavior during a crime wave, mimeograph (Harvard University, November 2007). 20. For a more detailed discussion of the relationship between happiness and health, see Carol Graham, Happiness and health: Lessons—and questions—for public policy, Health Affairs, Vol. 27, No. 1 (January–February 2008). 21. For a detailed discussion on the role of technology and institutions in making health improvements, see Charles Kenny’s work, summarized in http://charleskenny. blog.com/weblog/2008/02/theres-more-to.html (June 2008). 22. For differences in the level of benefits and the incentives that they have for overreporting, see Arie Kapteyn and Klaas de Vos, Social security and labor participation in the Netherlands, The American Economic Review, Vol. 88, No.2, Papers and Proceedings (May 1998). 23. On QALYs, see John Broome, Ethics and economics (Cambridge, UK: Cambridge University Press, 1999); and Daniel M. Hausman, What’s wrong with health inequalities? The Journal of Political Philosophy, Vol.15, No.1 (2007).
REFERENCES Adrianzen, B., & Graham, G. (1974). The high costs of being poor. Archives of Environmental Health, 28, 312–315. Alesina, A., di Tella, R., & MacCulloch, R. (2004). Inequality and happiness: Are Europeans and Americans different? Journal of Public Economics, 88. pp. 2009–2034. Benabou, R., & Ok, E. (1998). Social mobility and the demand for redistribution: The POUM hypothesis. NBER Working Paper 6795. Cambridge, MA. Blanchflower, D., & Oswald, A. (2004). Well-being in Britain and the U.S.A. Journal of Public Economics, 88. Blanchflower, D., & Oswald, A. (2007). Hypertension and happiness across nations. Mimeograph, NBER. Cambridge, MA, February. Broome, J. (1999). Ethics and economics. Cambridge, UK: Cambridge University Press. Cantril, H. (1965). The pattern of human concerns. New Brunswick, NJ: Rutgers University Press. Cardenas, M., Mejia, C., & Di Maro, V. (2009). Education and life satisfaction: Perception or reality? In Carol Graham & Eduardo Lora, Paradox and perception: Assessing quality of life in Latin America. Washington, DC: The Brookings Institution Press. Deaton, A. (2004). Globalization and health. In Susan M. Collins & Carol Graham (Eds.), Brookings trade forum: Globalization, poverty, and inequality. Washington, DC: The Brookings Institution Press. Deaton, A. (2008). Income, health, and well-being around the world: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22(2). Diener, E., & Biswas-Diener, R. (2008). Happiness: Unlocking the mysteries of psychological wealth. Boston: Blackwell-Wiley. Diener, E, Suh, E. M., Lucas, R. E., & Smith, H. L. (1999). Subjective well-being: Three decades of progress. Psychological Bulletin, 125. Di Tella, R., Galiani, S., & Shargrodsky, E. (2007). Crime distribution and victim behavior during a crime wave. Mimeograph. Cambridge, MA: Harvard University, November.
The Easterlin and Other Paradoxes
281
di Tella, R., & MacCulloch, R. (2006). Happiness and adaptation to income and status: Evidence from an individual panel. Mimeograph. Cambridge, MA: Harvard University. Easterlin, R. (1974). Does economic growth improve the human lot? Some empirical evidence. In Paul David & Melvin Reder, Nations and households in economic growth: Essays in honor of Moses Abramowitz. New York: Academic Press. Easterlin, R. (2001). Income and happiness: Towards a unified theory. Economic Journal, 111. Easterlin, R. (2003). Explaining happiness. Proceedings of the National Academy of Sciences, Vol. 100, No. 19. Easterlin, R. (2005). Feeding the illusion of growth and happiness: A reply to Hagerty and Veenhoven. Social Indicators Research, 74(3). Easterlin, R. (2008). Lost in transition: Life satisfaction on the road to capitalism. Institute for the Study of Labor (IZA) Discussion Paper Series, No.3409. Bonn, Germany; March. Eggers, A., Graham, C., & Gaddy, C. (2006). Unemployment and happiness in Russia: Can society’s suffering provide individual solace? Journal of Socioeconomics, 35. Felton, A., & Graham, C. (2005). Variance in obesity across cohorts and countries: Some evidence from happiness surveys in the United States and Russia. CSED Working Paper #44. The Brookings Institution, Washington, DC. Frey, B. (2008). Happiness: A revolution in economics. Cambridge, MA: MIT Press. Frey, B., & Stutzer, A. (2002). Happiness and economics: How the economy and institutions affect well-being. Princeton, NJ: Princeton University Press. Friedman, B. (2005). The moral consequences of growth. New York: Alfred Knopf. Graham, C. (2008). Happiness and health: Lessons—and questions—for public policy. Health Affairs, 27(1). Graham, C. (forthcoming). Comments on Alejandro Gaviria: Social mobility and preferences for redistribution in Latin America. Economica. Graham, C., & Chattopadhyay, S. (2008). Public opinion trends in Latin America (and the U.S.): How strong is support for markets, democracy, and regional integration? Paper prepared for the Brookings Partnership for the Americas Commission, Washington, D.C. Graham, C., & Chattopadhyay, S. (forthcoming). How much can citizens adapt to rising crime? Some evidence from happiness surveys from Latin America. Mimeograph. Washington, DC: The Brookings Institution; September. Graham, C., Eggers, A., & Sukhtankar, S. (2004). Does happiness pay? Some evidence from panel data for Russia. Journal of Economic Behavior and Organization, 55. Graham, C., & Felton, A. (2006). Inequality and happiness: Some insights from Latin America. Journal of Economic Inequality, 4. Graham, C., & Lora, E. (2009). Paradox and perception: Measuring quality of life in Latin America. Washington, DC: The Brookings Institution Press. Graham, C., & Pettinato, S. (2002a). Happiness and hardship: Opportunity and insecurity in new market economies. Washington, DC: The Brookings Institution Press. Graham, C., & Pettinato, S. (2002b). Happy peasants and frustrated achievers: Mobility, opportunity, and perceptions in the development process. Journal of Development Studies, 38(4). Graham, C., & Young, P. (2003). Ignorance fills the income gulf. The Boston Globe, June 23. Granovetter, M. (2009). The strength of weak ties. American Journal of Sociology, 78, 1360–1379.
282
Section II: Comparisons of Income and Well-Being Through Time
Hausman, D. M. (2007). What’s wrong with health inequalities? The Journal of Political Philosophy, 15(1). Helliwell, J. F. (2008). Life satisfaction and quality of development. Mimeograph. Vancouver: University of British Columbia. Helliwell, J., Huang, H., & Harris, A. (2008). International differences in the determinants of life satisfaction. Mimeograph. Vancouver: University of British Columbia. Howell, R. T., & Howell, C. J. (2008). The relation of economic status to subjective wellbeing in developing countries: A meta-analysis. Psychological Bulletin, 134(4). Inglehart, R., Foa, R., Peterson, C., & Welzel, C. (2008). Development, freedom, and rising happiness: A global perspective (1981–2007). Perspectives on Psychological Science, 3(4). Kapteyn, A., & de Vos, K. (1998). Social security and labor participation in the Netherlands. The American Economic Review, 88(2). Papers and Proceedings, May. Kapteyn, A., Smith, J. P., & van Soest, A. (2007). Vignettes and self-reports of work disability in the United States and the Netherlands. The American Economic Review. Kingdon, G., & Knight, J. (2007). Communities, comparisons, and subjective wellbeing in a divided society. Journal of Economic Behavior and Organization, 64(1). Knight, J., & Gunatilaka, R. (2007). Great expectations? The subjective well-being of rural-urban migrants in China. Discussion Paper Series No. 322, Department of Economics, University of Oxford; April. Labonne, J., & Chase, R. (2008). So you want to quit smoking: Have you tried a mobile phone? Policy Research Working Paper Series, No. 4657. Washington, DC: The World Bank; June. Lora, E., & Chaparro, J. C. (2009). The conflictive relationship between satisfaction and income. In Understanding quality of life in Latin America. Washington, DC: InterAmerican Development Bank. Luttmer, E. (2005). Neighbors as negatives: Relative earnings and well-being. Quarterly Journal of Economics, 120(3). Picon, M. (2008). Explaining subjective well-being by objective well-being measures: An application for Latin America. Mimeograph. Washington, DC: The Brookings Institution. Powdthavee, N. (forthcoming). Happiness and standard of living. Economica. Preston, S. (1975). The changing relation between mortality and level of development. Population Studies, 29(2), 239–248. Ravallion, M., & Lokshin, M. (2005). Who cares about relative deprivation effects. Policy Research Working Paper Series, No. 3782. Washington, DC: The World Bank. Dec. (http//:ideas.repec.org/p/wbk/wbrwps/3782.html) Shaw, J. W., Johnson, J. A., & Coons, S. J. (2005). U.S. valuation of the EQ-5D health states: Development and testing of the D1 Valuation Model. Medical Care, 43(3). Soroka, S. N., Helliwell, J. F., & Johnson, R. (2007). Measuring and modelling interpersonal trust. In Fiona Kay & Richard Johnson (Eds.), Diversity, social capital, and the welfare state. Vancouver: UBC Press. Stevenson, B. & Wolfers, J. (2008). Economic growth and subjective well-being: Reassessing the Easterlin Paradox. Brookings Panel on Economic Activity. Whyte, M., & Hun, C. (2006). Subjective well-being and mobility attitudes in China. Mimeograph. Cambridge, MA: Harvard University.
Appendix 1 Question wording differences in the survey questionnaires Type of Subjective Wellbeing measured
Question on World Poll Did you smile or laugh a lot yesterday?
Affective.
Do you feel your life has an important purpose or meaning?
Affective.
Are you satisfied or dissatisfied with your income, all the things you can buy and do?
Life-domain satisfaction.
Are you satisfied or dissatisfied with your job?
Life-domain satisfaction.
Are you satisfied or not with your standard of living, all the things you can buy and do?
Life as a whole satisfaction.
Are you satisfied with your freedom to choose what to do with your life?
Life as a whole satisfaction.
Please imagine a ladder with steps from zero to ten, if the higher the step, the best possible life, on which step of the ladder you personally feel you stand?
Life as a whole satisfaction, in relation to others.
In a scale from zero to ten, with zero the poorest people and ten the richest people, in which cell you put yourself?
Life as a whole satisfaction, in relation to others.
Measure reflecting changes in best possible life ladder question comparing present to five years ago.
Life as a whole satisfaction compared to the past.
Measure reflecting changes in best possible life ladder question comparing what is expected in five years to present.
Life as a whole satisfaction expected in the future.
Source: Gallup World Poll, 2007.
Distribution of responses across questions Variable Smiled Life Purpose Best possible life Satisfaction living standards Poor to Rich Scale Freedom
Obs
Average
Std. Dev.
18,816 18,786 18,952 18,804 17,982 18,519
0.82 0.97 5.83 0.69 4.24 0.73
0.385657 0.1724025 2.374283 0.4611318 1.847543 0.4430965
Source: Gallup World Poll, 2007.
283
0
1000
Frequency 2000 3000
4000
5000
Best possible life question
0
2
4
6
8
10
wp16 please imagine a ladder/mountain with steps numbered from zero at the botto
Source: Gallup World Poll, 2007.
Frequency 5000 1.0e + 04
1.5e + 04
Satisfaction with living standards
0
Freq.
–.5 0 .5 1 satisfied or not with standard of living things you can do
Dissatisfied Satisfied Total
Percent
5767 13037 18804
30.67 69.33 100
Cum. 30.67 100
Source: Gallup World Poll, 2007.
0
1000
Frequency 2000 3000
4000
5000
Perception of location on a socioeconomic scale
0
2
4
6
8
10
self-assessment of personal situation poorest richest 10 point scale
Source: Gallup World Poll, 2007.
284
Poorest 1 2 3 4 5 6 7
Freq. 543 934 1573 2809 3435 4786 2113 1133
Percent 3.02 5.19 8.75 15.62 19.1 26.62 11.75 6.3
Cum. 3.02 8.21 16.96 32.58 51.69 78.3 90.05 96.35
8 9 Richest Total
497 80 79 17982
2.76 0.44 0.44 100
99.12 99.56 100
1.0e+ 04 5000
Frequency
1.5e+ 04
2.0e+ 04
Life purpose
Life purpose Frequency 0
No Yes Total
–.5
0
.5
1
lifepurp
576 18,210 18,786
% 3.07 96.93 100
Source: Gallup World Poll, 2007.
1.0e+04 5000
Frequency
1.5e+04
Frequency of smiling the previous day
Freq. 0
No Yes Total
–.5
0 .5 did you smile a lot yesterday?
1
Percent
3420 15396 18816
Cum.
18.18 81.82 100
18.18 100
Percent
Cum.
Source: Gallup World Poll, 2007.
1.0e + 04 5000
Frequency
1.5e + 04
Satisfaction with personal freedom
0
Freq. –.5 0 .5 1 Satisfied with your freedom to choose what to do with your life
Source: Gallup World Poll, 2007.
285
Dissatisfied Satisfied Total
4969 13550 18519
26.83 73.17 100
26.83 100
Descriptive statistics of survey variables: sorted by country Mean HH income per capita (US$)
Mean value Wealth PCA (from 0 to 0.3)
Smiled (Yes/No)
Life purpose (Yes/No)
Best possible life (0–10)
Satisfaction with living standards (Yes/No)
Poorest richest scale (0 to 10)
Freedom (Yes/No)
286
Venezuela Brasil Mexico Costa Rica Argentina Belize Bolivia Chile Colombia Dominican Republic Ecuador EI Salvador Guatemala Guyana Honduras Nicaragua Panama Paraguay peru Uruguay
191.51 299.88 221.90 416.89 408.02 618.17 173.32 340.85 394.28 288.33
0.14 0.14 0.12 0.15 0.14 0.16 0.08 0.15 0.14 0.10
0.88 0.86 0.81 0.87 0.82 0.75 0.77 0.76 0.82 0.78
1.00 0.96 0.95 0.98 0.95 0.96 0.95 0.94 0.99 0.96
6.47 6.17 6.56 7.41 6.01 6.41 5.37 5.77 6.23 4.93
0.83 0.71 0.76 0.85 0.68 0.72 0.67 0.62 0.72 0.62
4.18 5.03 5.32 4.75 5.12 4.35 4.42 3.97 3.22
0.80 0.78 0.67 0.92 0.66 0.70 0.79 0.71 0.79 0.88
170.39 166.74 177.13 369.92 213.52 227.04 233.50 221.16 156.74 391.14
0.10 0.09 0.12 0.13 0.09 0.07 0.11 0.07 0.09 0.14
0.83 0.88 0.81 0.77 0.79 0.84 0.85 0.86 0.79 0.77
0.98 0.98 0.99 0.98 0.97 0.98 0.99 0.99 0.98 0.92
4.93 5.28 6.31 6.02 5.11 4.93 6.84 5.22 5.33 5.67
0.72 0.63 0.84 0.65 0.70 0.64 0.73 0.58 0.54 0.62
4.04 3.81 4.32 5.16 3.89 2.82 4.52 4.07 4.19 4.34
0.66 0.64 0.63 0.69 0.68 0.83 0.64 0.70 0.64 0.79
Total Sample
263.12
0.11
0.82
0.97
5.85
0.69
4.24
0.73
Source: Gallup World Poll, 2007.
Descriptive statistics of assets used to generate wealth indices Variable Electricity Running water TV set Landline phone Cable TV Savings Acc. Automobile Computer Credit Card Internet
Obs
Average
Std. Dev.
19087 19086 19081 19074 19041 18670 19014 19057 18595 19023
0.961021 0.894949 0.915308 0.500839 0.323355 0.260793 0.25497 0.236186 0.1274 0.114072
0.193551 0.306627 0.27843 0.500012 0.467769 0.439079 0.435856 0.424749 0.333429 0.317908
Source: Gallup World Poll, 2007.
2000 0
1000
Frequency
3000
4000
Descriptive statistics of survey variables: sorted by country
–5
0
5
10
linpc
Variable HH Monthly Income
Obs 14,329
Mean 895.74
Std. Dev. 1054.962
Min 0.06
Max 23,164.16
Per Capita HH Monthly Income Log of Per Capita HH Income
14,329
263.12
341.591
0.01
8,541.94
14,329
5.05
1.078811
-4.44
9.05
Source: Gallup World Poll, 2007.
287
Details of control variables used in individual level analyses
Source: Gallup World Poll, 2007.
Details of control variables used in cross-country analyses
Source: Gallup World Poll, 2007.
288
Section III
International Differences in the Social Context of Well-Being
This page intentionally left blank
Chapter 10 International Evidence on the Social Context of Well-Being John F. Helliwell, Chris Barrington-Leigh, Anthony Harris, and Haifang Huang1
Together, the chapters of this book argue that there are many useful and interrelated ways of measuring well-being. The chapter by Diener, Kahneman, Tov, and Arora (2010) argues that there appear to be systematic differences between types of measure, with subjective evaluations of life as a whole being more systematically related to life circumstances than are measures of positive or negative affect. They focus on the relative strength of linkages with income, while in this chapter we extend the analysis to estimate and emphasize the importance of the social context as a determinant of well-being both within and among nations. We find that life evaluations accommodate a robust explanatory role for both income and social variables, and thereby permit precise estimates of their relative importance as factors explaining international differences in well-being. We consider two different ways of evaluating life as a whole. One is the Cantril Ladder question (also known as the Cantril self-anchoring striving scale) used in the Gallup World Poll and the other is the more widely used assessment of satisfaction with life (SWL), as used in several decades of the World Values Survey, and now also included in the recent waves of the Gallup World Poll. In our previous work comparing the Cantril Ladder with measures of life satisfaction (Helliwell, 2008; and Helliwell, Huang, & Harris, 2009), we argued that life evaluations provide a useful way to assess the quality of development within and across communities and nations. We made the case that previous doubts about the usefulness of comparing measures of subjective well-being across cultures and over time are being resolved in favor of subjective measures2. 291
292
Section III: Differences in the Social Context of Well-Being
In Helliwell (2008), we compared results from the first wave of the Gallup World Poll and the World Values Survey, focusing on modeling differences among nations in average scores from different measures of the quality of life. That paper also estimated two-level regressions based on individual data, and argued that most of the cross-country variance in survey measures of life satisfaction can be explained by measurable differences in life circumstances in those countries, under the assumption that people all over the world have similar basic preferences and answer life satisfaction questions in roughly comparable ways. In Helliwell, Huang, & Harris (2009) we dug further into the data to see to what extent the assumption of common preferences is justified. More particularly, we used Gallup World Poll wave 1 data on the quality of life (as measured by the Cantril Ladder) to estimate cross-sectional life evaluation equations in each of 105 countries. We found that the results on a country-by-country basis were broadly consistent with the use of two-level analysis in which coefficients are assumed to be the same for residents in all countries. The coefficients were strikingly consistent among countries, cultures and regions, although we found confirmation of several experimental results on cross-cultural differences in reporting styles. While these differences were interesting, they were in general not large enough to obscure the broad commonality of results. Thus it would appear that the large international differences in life evaluations are not due to differences in underlying preferences but rather to identifiable differences in life circumstances. In the current chapter, we extend our earlier analysis in several key ways. We now have three waves of data from the Gallup World Poll, thus increasing substantially the number of countries represented in our analysis and adding to the sample size for those countries included in two or three waves. From our perspective, one of the most valuable features of the second and third waves of the Gallup World Poll (conducted in 2007 and 2008) is that for 103 countries the survey included not just the Cantril Ladder question on Life Today, but also the standard life satisfaction question used in the World Values Survey and many other national and international surveys. The recent availability of both Ladder and life satisfaction responses for large global samples is of great value, as it permits important issues to be more systematically evaluated. In our earlier comparisons of the wave 1 Gallup World Poll Ladder data with the World Values Survey data on life satisfaction, we found that the Gallup World Poll Ladder data had a tighter relation to income than did the World Values Survey satisfaction with life (SWL) data. We hypothesized that this difference might be due to the ladder framing of the Cantril Ladder question. In particular, we suspected that the vertical nature of the Ladder, with the best possible life being that
International Evidence on the Social Context of Well-Being
293
on the top rung, might encourage respondents to think in more materialistic terms, and in a more relative way. We were puzzled that the framing effects would be this large, so we strongly supported the introduction of the SWL question into the Gallup World Poll, so as to be able to analyze the differences between the questions for the same individuals within the context of the same survey. Our initial hypothesis was encouraged by our first results comparing SWL and Ladder responses from the same respondents, drawn from 53 countries in wave 2 of the Gallup World Poll. We found, and reported in the Princeton conference version of this chapter, that SWL responses were systematically higher than those for the Cantril Ladder, and that the amount by which the SWL responses exceeded the Ladder responses was smaller in the richer countries, even within separate groupings of industrial and developing economies. This was just as we would expect if the Ladder responses were more inclined to be relative in nature, and more incomeoriented, especially if the relative comparisons were global in nature. Diener et al. (2010) also used the same data, found similar results, and added another possible difference between the Ladder and the SWL responses: that satisfaction is an emotion, and is therefore likely to share some of the characteristics of the answers to the more explicitly emotional questions, such as those relating to happiness and measures of positive and negative affect. Given this apparent difference between the results of SWL and Ladder questions, some of the conference discussions then turned to which, if either, was to be preferred. Kahneman argued that the Ladder and measures of affect were at opposite ends of a continuum between emotional and cognitive measures, with SWL and happiness being interesting but less informative intermediate measures. Helliwell put on an Aristotelian hat and argued that SWL and the Ladder had equivalent claims as life evaluations, but if a choice between them had to be made it should favor life satisfaction, to avoid the possible framing effects of the Ladder and to provide more overlap with other surveys. Recognizing the value of larger samples, we deferred our revised analysis for this chapter to await the arrival of data from the 2008 wave 3 of the Gallup World Poll. This has permitted us to almost double the number of countries with data for both SWL and the Cantril Ladder. The expanded coverage permits more precise and hopefully more robust conclusions about the nature of the similarities and differences between SWL and the Cantril Ladder. We find that with the larger global samples the SWL and Ladder responses show very similar correlations with presumed structural factors related to well-being. In addition, the differences in framing and possible emotional content of the two questions seem to be such that a simple
294
Section III: Differences in the Social Context of Well-Being
average of the individual-level answers to the two questions is more tightly correlated with the presumed structural determinants of wellbeing than is either measure on its own. Thus we are now inclined to suggest that an average of the individual-level answers to the life satisfaction and Ladder questions provides a measure that is more robust and informative than either on its own. The most encouraging feature of our new results is that while the differences in the framing of the two questions may be enough to make it useful to average the two measures, the separate and average measures all provide consistent evaluations of the relative importance of the circumstances of life, both within and among nations. In the following section we make use of this new-found capacity to compare the similarities and differences in these alternative measures of well-being. We shall make principal use of the satisfaction with life and Ladder data from the Gallup World Poll, but shall also link these results with the life satisfaction data and analysis based on the World Values Survey. In all cases we shall make special efforts to assess the importance of variables measuring the quality and nature of the social context facing individuals and nations. First, a word of clarification about the key dependent variables we are using here to evaluate well-being. For all waves of the Gallup World Poll, we use the Cantril Ladder question asking respondents to evaluate their lives at present using a ladder with steps numbered from zero at the bottom to 10 at the top, with zero representing the worst possible life and 10 the best possible life. For 93 countries in the second and third waves of the Gallup World Poll, and for 92 countries in the fourth and fifth waves of the World Values Survey3, we use the answers to the standard life satisfaction question, measured for Gallup on a scale of zero to 10, and for the World Values Survey on a scale of one to 10. We show in Figure 10.1 the distribution of individual responses for each of the three alternative measures (Gallup Ladder responses are pooled data from all three waves, while the Gallup SWL responses, while also pooled, are available only in some of the wave 2 and wave 3 country surveys). The WVS satisfaction with life responses are drawn from the largest sample possible from the surveys conducted around the late 1990s and the early 2000s. The total sample size, sample means, and the number of countries represented are shown in each case. In Figure 10.2, for greater comparability, we show the distributions only for respondents in countries with data for all three measures. Note that for the Gallup Ladder and standard life satisfaction responses shown in Figure 10.2, the questions are being asked of precisely the same respondents, so that the only differences should relate to question framing and placement.
Ladder, Gallup All Waves
Fraction 0 .05 .1 .15 .2
Fraction 0 .05 .1 .15 .2 .25
SWL, Gallup Wave-2&3
0
2
4
6
8
10
0
Satisfied with life, GWP Sample size: 115402; Mean: 5.95; # of Countries: 104
2
4
6
8
10
Ladder, GWP Sample size: 342329; Mean: 5.43; # of Countries: 145
Fraction 0 .05 .1 .15 .2
SWL, WVS Wave-4
0
2
4 6 8 10 Satisfied with life, WVS4 Sample size: 117264; Mean: 6.53; # of Countries: 81
Figure 10.1 Distribution of responses to alternative measures.
Ladder, Gallup Wave-2&3
Fraction 0 .05 .1 .15 .2
Fraction 0 .05 .1 .15 .2
SWL, Gallup Wave-2&3
0
2
4
6
8
10
0
Satisfied with life, GWP Sample size: 68805; Mean: 6.49; # of Countries: 59
2
4
6
0
Fraction .05 .1 .15 .2
SWL, WVS Wave-4
0
2 4 6 8 10 Satisfied with life, WVS4 Sample size: 91789; Mean: 6.52; # of Countries: 59
Figure 10.2 Figure 10.1 reproduced for countries with all three measures.
295
8
10
Ladder, GWP Sample size: 169600; Mean: 5.87; # of Countries: 59
296
Section III: Differences in the Social Context of Well-Being
The distribution shapes of the SWL responses from WVS and Gallup are more similar than either is to the Gallup Ladder. The distribution of the Ladder responses has a larger central tendency, with a strongly defined mode at the midpoint of the distribution, much as one would expect to find if the Ladder framing led respondents to think of their lives in relative terms. The two SWL distributions, by contrast, have modes higher up in the distribution (at least when the country samples are matched), and much fuller right-hand shoulders. Most of the difference between the means of Gallup distributions for the Ladder and for SWL is due to differences in the number and distribution of middle and top-half answers, rather than in the number and distribution of those rating their life satisfaction below the midpoint of the scale. We turn now to our estimation of global, regional, and national equations for both Ladder and SWL measures of well-being. The basic observations are at the individual level, and we are interested in estimating the extent to which individual life satisfaction depends on circumstances and events at the individual, household, community, and national levels. We have developed three interrelated ways of unravelling the data. The most general is to use two-level estimation to account for individual-level and national-level effects simultaneously. The second is to use the individual-level data in equations that are separate for each country, and then to look for international differences in the resulting coefficients. The third is to use country average data to explain international differences in well-being. We shall do all three, starting with the latest two-level results based on the newly expanded Gallup sample.
Two-Level Global and Regional Equations Based on Individual Responses
The basic estimation form for the two-level analysis of the ordered life satisfaction responses is: LSij ¼ þ ln ðyij Þ þ Xij þ Zj þ Eij
ð1Þ
where LSij is some measure of life satisfaction, for respondent i in country j, yij is the level of household income of the respondent, the Xij are other individual or household-level variables, and the Zj are national-level variables, with the same value being used for all individual observations in country j. We use the log form for both household and national average income, to reflect standard economic assumptions and many empirical results suggesting that less affluent agents derive greater utility from
International Evidence on the Social Context of Well-Being
297
extra income. In general, we employ national averages of variables for which we also have household-level observations, in which case the g coefficients represent contextual effects, or, in other terms, the extent of positive or negative externalities4. In all equations robust standard errors are estimated assuming errors to be clustered by country, and a dummy variable is included to permit differences in life satisfaction from one wave to the next. When we calculate compensating differentials for non-financial determinants of life satisfaction, we take into account the functional form of equation (1). Thus in our theoretically and empirically preferred case where income is in log form and X is in linear form, b = m/ d will be the log change in income that has for the average respondent the same life satisfaction effect as a change in the non-financial life characteristic X. Table 10.1 shows subjective well-being equations based on global samples ranging from about 50,000 to over 140,000 respondents in 125 countries, with the smaller sample sizes resulting from missing countries and observations for some variables5. All equations include gender, age (in quadratic form), marital status, the logarithm of household income, a measure of unmet food needs, a measure of social connectedness (having someone to count on), a measure of the individual’s sense of freedom to choose, the individual’s perception of the prevalence of corruption in business and government, several measures of pro-social behavior (donations of time, donations of money, and providing help to a stranger), and two measures of religious attachment—one a measure of the importance of religion in the respondent’s life and the other the amount of time devoted to the practice of religion. The exact wording of each question is shown in the appendix to this chapter, and the basic results for the same equation fitted separately to each of seven world regions are shown in Table 10.2. The first equation in Table 10.1 shows the two-level results for the largest possible three-wave sample of responses to the Ladder question, a global sample of more than 140,000 respondents from 116 countries. The next three equations show the results of fitting the same equation to a smaller sample of 52,600 respondents from 80 countries, all of whom provided answers on the same 0–10 scale for the Ladder and for satisfaction with life. Equation 2 uses the Ladder responses as the dependent variable, equation 3 uses SWL, and equation 4 uses an equally weighted average of the Ladder and SWL responses. The first important finding is that the key structural coefficients are very similar in all four equations. The second key result is that the equally weighted average of the Ladder and life satisfaction responses produces tighter coefficients, and explains a significantly greater share of the total variance than does either measure on its own. To explain more than 44 percent of the individual-level variance of life evaluations on a global basis,
TABLE 10.1
Global equation with full set of social context variables Significance:
1%*
5%
10%*
region fixed effects
log (household income) male
298
age (age/10)2 (as) married separated, divorced, or widowed not enough money: food(net) friends to count on freedom to choose
ladder (all)
ladder (matched)
SWL (matched)
(1)
(2)
(3)
.42* (.025) .11* (.019) .030* (.004) .024* (.004) .008 (.036) .13* (.045) .63* (.025) .46* (.027) .31* (.023)
.37* (.034) .12* (.025) .028* (.005) .022* (.005) .076 (.058) .11 (.071) .68* (.043) .49* (.038) .36* (.033)
.44* (.055) .095* (.025) .031* (.005) .029* (.006) .078 (.064) .10 (.086) .73* (.036) .45* (.045) .47* (.052)
1 2 (ladder + SWL) (4)
ladder (all)
.40* (.042) .11* (.022) .029* (.004) .025* (.005) .077 (.054) .10 (.071) .71* (.032) .47* (.034) .42* (.039)
.43* (.025) .11* (.019) .030* (.004) .023* (.004) .014 (.034) .14* (.043) .63* (.025) .46* (.027) .31* (.022)
(5)
1 2 (ladder + SWL) (6)
.41* (.042) .12* (.022) .029* (.004) .024* (.004) .073 (.046) .12* (.063) .70* (.032) .47* (.032) .41* (.037)
country fixed effects
(7)
1 2 (ladder + SWL) (8)
.44* (.025) .12* (.018) .031* (.004) .025* (.004) .009 (.022) .14* (.031) .62* (.024) .46* (.027) .32* (.022)
.41* (.046) .12* (.021) .026* (.004) .022* (.004) .045 (.030) .16* (.047) .71* (.032) .46* (.031) .41* (.035)
ladder (all)
perception of corruption donated time donatedMoney helped a stranger
299
importance of religion attended church/etc nation: log(GDP per cap) nation: not enough money: food(net) nation: friends to count on
.26* (.025) .077* (.023) .16* (.020) .083* (.022) .010 (.023) .047 (.019) .051 (.064) .17 (.36) 1.60* (.48)
.22* (.041) .15* (.031) .19* (.034) .070* (.023) .026 (.038) .058 (.028) .19 (.082) .21 (.46) l.12* (.66)
.25* (.047) .22* (.040) .32* (.046) .054* (.031) .13* (.040) .040 (.035) .13 (.13) .25 (.64) 1.19 (.01)
.24* (.038) .18* (.030) .25* (.035) .062* (.024) .078 (.035) .049* (.027) .16* (.095) .23 (.48) 1.16 (.73)
.26* (.025) .078* (.023) .16* (.020) .081* (.021) .013 (.023) .046 (.019) .006 (.060) .69* (.37) 1.05 (.46)
.24* (.038) .18* (.030) .25* (.035) .057 (.024) .083 (.034) .046* (.027) .11 (.080) .83* (.47) .56 (.71)
.26* (.023) .082* (.023) .17* (.020) .085* (.020) .005 (.022) .042 (.018)
.22* (.037) .18* (.029) .24* (.032) .052 (.023) .076 (.032) .035 (.027)
TABLE 10.1 (Continued)
Global equation with full set of social context variables Significance
ladder (all) (1)
300
nation: freedom to choose nation: perception of corruption nation: donated time nation: donated Money nation: helped a stranger nation: importance of religion nation: attended church/etc healthy life expectancy (net) wave 2 wave 3
.55* (.30) 1.53* (.31) .64 (.50) .12 (.27) .032 (.35) .57 (.42) 1.33* (.48) .020* (.007) .27* (.073) .24*
ladder (matched) (2) .76* (.45) 1.34* (.39) .27 (.64) .25 (.36) .10 (.45) .65 (.48) 1.38* (.49) .020 (.010)
.013
SWL (matched) (3) 1.24 (.59) .97* (.56) .92 (.86) .85* (.48) .77 (.55) 1.04 (.76) 1.39 (.62) .051* (.012)
.79*
1%*
1 2 (ladder + SWL) (4)
1.00 (.42) 1.15* (.37) .33 (.67) .55 (.38) .33 (.38) .84 (.54) 1.39* (.52) .035* (.009)
.40*
5% 10%*
ladder (all) (5) .79 (.34) .1.01 (.40) .73 (.46) .20 (.29) .18 (.32) .55 (.39) 1.12* (.42) .013 (.009) .27* (.071) .25*
1 2 (ladder + SWL) (6)
ladder (all) (7)
1 2 (ladder + SWL) (8)
1.14 (.47) .73 (.36) .21 (.54) .13 (.37) .049 (.38) .74* (.43) 1.29* (.39) .019* (.010)
.081
.18* (.070) .20*
.078 (.063)
constant
(.058) 5.96* (.56)
(.14) 6.39* (.85)
1(.18) 6.17* (1.28)
(.15) 6.28* (1.01)
dRegion1 dRegion2 dRegion3 dRegion4
301
dRegion5 dRegion6 region fixed effects country fixed effects obs. R2(adj) Nclusters
140267 .296 116
52657 .338 80
52657 .387 80
52657 .441 80
(.055) 5.76* (.63) .051 (.47) .38 (.44) .54 (.44) .33 (.43) .12 (.45) .075 (.41)
(.15) 5.51* (.82) .60 (.39) .84 (.39) .79 (.39) 1.18* (.42) .48 (.43) .37 (.31)
[
[
140267 .299 116
52657 .452 80
(.063) 6.53* (.13)
6.98* (.15)
[ 146217 .321 124
[ 53174 .481 81
TABLE 10.2 Sample separated by world region (ladder). Regions: 1: Former Soviet Union Countries and Eastern European Countries; 2: European Countries; 3: United States,Canada, Australia, New Zealand; 4: Latin America and Caribbean; 5: Asia; 6: Africa; 7: Persian and Mid-east, including Isreal. Countries in regions 3 and 6 were not asked the SWL question. OLS standard errors are calculated using clustering at the country level. Significance: 1%*
log( household income) male
302
age (age/10)2 (as) married separated, divorced, or widowed not enough money: food(net) friends to count on
5% 10%*
ladder (1)
ladder (2)
ladder (3)
ladder (4)
ladder (5)
ladder (6)
ladder (7)
.56* (.045) .017 (.029) .049* (.010) .037* (.010) .051 (.082) .007 (.094) .76* (.057) .42* (.051)
.37* (.051) .19* (.039) .041* (.009) .036 (.008) .21* (.060) .29* (.081) .86* (.094) .74* (.056)
.42* (.023) .32* (.055) .002* (.011) .067* (.009) .24* (.14) .28* (.10) .90* (.15) .73* (.12)
.54* (.043) .17* (.036) .051* (.005) .040* (.006) .099* (.029) .12* (.064) .66* (.044) .60* (.055)
.53* (.039) .23* (.039) .025* (.008) .025* (.008) .088 (.056) .29* (.080) .61* (.052) .32* (.051)
.30* (.033) .022 (.024) .001 (.006) .006 (.006) .071 (.054) .073 (.064) .48* (.034) .42* (.043)
.65* (.14) .24* (.078) .056* (.017) .048 (.019) .27* (.083) .17 (.23) .64* (.11) .54* (.093)
freedom to choose perception of corruption donated time donatedMoney helped a Stranger importance of religion
303
attended church/etc log(GDP per cap.,PPP,2003/5) nation: not enough money: food(net) nation: friends to count on nation: freedom to choose nation: perception of corruption
.42* (.042) .39* (.061) .057 (.063) .22* (.061) .12* (.033) .005 (.058) .027 (.034) .23 (.18) 1.52 (.94) 2.92 (1.19) .72 (1.01) .041 (.60)
.50* (.068) .34* (.060) .16* (.035) .20* (.043) .041 (.020) .004 (.034) .007 (.045) .18 (.50) 3.13* (1.87) 4.47 (3.40) .28 (2.28) 1.10 (.48)
.62* (.17) .17 (.083) .19* (.11) .21* (.047) .028 (.042) .013 (.048) .047 (.085) 1.15* (.17)
3.06* (.45)
.27* (.050) .24* (.045) .13* (.065) .15* (.032) .11* (.038) .054 (.041) .096* (.031) .25 (.17) 2.84* (.85) 3.56 (2.30) .072 (1.46) .55 (.91)
.23* (.038) .22* (.047) .007 (.043) .11* (.025) .096* (.037) .079 (.059) .072 (.034) .13 (.13) 1.14 (.49) .92 (1.86) 1.13 (.47) 1.10 (.83)
.27* (.040) .24* (.042) .048 (.035) .17* (.045) .088* (.050) .12 (.050) .010 (.043) .007 (.083) .043 (.58) 1.23 (.55) .43 (.44) 1.07* (.63)
.36* (.041) .083 (.23) .15 (.062) .10 (.20) .13 (.17) .078 (.092) .011 (.10) 1.02* (.28)
2.80* (.68)
TABLE 10.2 (Continued) Sample separated by world region (ladder). Regions: 1: Former Soviet Union Countries and Eastern European Countries; 2: European Countries; 3: United States, Canada, Australia, New Zealand; 4: Latin America and Caribbean; 5: Asia; 6: Africa; 7: Persian and Mid-east, including Isreal. Countries in regions 3 and 6 were not asked the SWL question. OLS standard errors are calculated using clustering at the country level. Significance:
nation: donated time nation: donatedMoney
304
nation: helped a stranger nation: importance of religion nation: attended church/etc wave 2 wave 3 constant region obs. R2(adj) Nclusters
ladder (1)
ladder (2)
.49 (1.19) 1.87 (.85) .99 (.95) 1.17 (.55) .61 (.82) .049 (.14) .17* (.097) 3.00* (1.06)
.76 (.91) .26 (.72) .43 ( .80) .33 (1.17) .51 (1.12) .23 (.095) .33* (.075) 3.01 (2.58)
1 18024 .199 22
2 15860 .323 19
1%*
5%
10%*
ladder (4)
ladder (5)
ladder (6)
.18* (.033) .083 (.034) 8.55* (.39)
1.25 (.97) .46 (.45) 1.61 (1.12) 1.71 (.84) 2.47* (.60) .15 (.14) .21 (.14) 3.92 (1 .54)
1.41* (.79) .10 (.55) .50 (.42) .48 (.59) .060 (.92) .37 (.18) .22 (.17) 5.61* (1.15)
1.01 (1.24) 1.12 (1.06) .45 (.58) .81 (1.02) .63 (.71) .31 (.12) .24 (.11) 5.25* (1.30)
3 4688 .166 4
4 32125 .178 23
5 26211 .190 18
6 44855 .146 34
ladder (3)
.93* (.081)
ladder (7)
2.40* (.31) .094 (.14) 12.3* (.64) 7 4454 .309 4
International Evidence on the Social Context of Well-Being
305
using globally uniform coefficients and making no allowance for country or region effects, is quite remarkable. It leads us to suggest that this average might, in due course when full global samples permit, be a preferred measure for life evaluations. For most of this chapter, we shall take advantage of the close similarities of coefficients, and concentrate our analysis on equation 1, which has a much larger sample size and corresponds better with our subsequent analysis based on national equations. Age effects are estimated by a quadratic form in age. With the exception of Africa, there is a significant U-shape in age in all of the regional equations of Table 10.2, in all of the regional averages of the age coefficients from the national equations, and in most of the individual national equations. The first age variable is age in years and the second is the square of age/100. Thus the age at which life satisfaction is lowest is younger than 50 if the coefficient on the squared age variable exceeds that on age, and vice versa. As was found by Blanchflower & Oswald (2008) using data from the WVS and other surveys, life evaluations tend to have their lowest point within a few years on either side of age fifty. Marital status is divided into three categories: married or equivalent; single; and a combination of divorced, separated, and widowed; with single being treated as the base case in estimation. The coefficient for married or equivalent is insignificant in the global equation, but significantly positive in regions 2 and 3, comprising Western Europe, North America, and Australasia, in comparison with the never-married base case. The coefficient on the combination of divorced, separated, and widowed is significantly negative in the largest global sample and in Western Europe, North America, Australasia, and Asia. The log of household income is a very strong correlate of individual life satisfaction in all equations. To ensure that as much as possible of the direct and indirect effects of higher income are captured by the income variables, we have defined the food inadequacy variable to be the residual from an equation explaining the raw variable by the log of household income. As shown in Table 10.2, which divides the sample by regions, the income coefficient is if anything higher in the richer countries (as previously noted by Deaton, 2008) and shows no obvious tendency to drop as individual income rises, beyond the substantial non-linearity implied by the logarithmic form for income. As already noted, the food inadequacy variable is defined net of its very significant correlation with household income. This has the intended effect of raising the estimated coefficient on household income above what it would have been if the food variable had not been redefined to exclude the variance of income.6 Looking across countries, the different variance of the income and food adequacy variables can be interpreted as a measure of
306
Section III: Differences in the Social Context of Well-Being
inequality of income distribution, and it is strongly correlated7 with international differences in the inequality of income distribution, as measured by the Gini coefficient. Our primary measure of social connections is provided by the answers to a question asking whether respondents have relatives or friends they can count on to help them whenever help is needed. In all parts of the world, most respondents in the Gallup sample combining waves 1 to 3 report that they have family or friends they can count on, ranging from 69% in Africa, to 78% in Asia, to over 90% in most of the industrial countries in the Organization for Economic Cooperation and Development (OECD). In all regions this social support is tightly linked to life satisfaction, with a global coefficient that exceeds that on log income. As will be shown later, this implies income-equivalent life satisfaction values for social connections that are very high indeed. It would appear from the regional differences in the data and estimated coefficients that respondents in Western Europe, North America, Australia, and New Zealand are richer in social as well as economic terms than those living elsewhere, and attach even higher absolute and relative values to such social support. On the basis of the largest global sample, obtained by combining all three waves of the Gallup World Poll, the coefficient on having someone to count on is 0.68 in Western Europe, and 0.69 for United States + Canada + Australia + New Zealand, compared to .30 in Asia and 0.40 in Africa (see Table 10.2). Other variables indicative of personal or community-level social capital show the high values attached to mutually supportive social connections. Respondents appear to value, not only the support they get from others, but also their own support for others. For instance, those who in the last month had donated money or time to an organization or aided a stranger needing help were systematically more satisfied with their lives, especially for donations (.15), followed by donating time (.074) and helping a stranger (.072), as shown by the first equation in Table 10.1. Trust questions were included in only a sub-sample of the wave 1 Gallup countries. As suggested by earlier work (e.g., Helliwell & Putnam, 2004) the well-being effects of living in an environment where other people can be trusted8 are very substantial. For example, as shown in Helliwell (2008), the Gallup respondents who think that their lost wallets would be returned by a neighbor or the police evaluated the quality of their lives more highly (by .15 and .22 points), as did those who expressed confidence in the police (.22). Returning to the largest global samples represented by the first equation of Table 10.1, an individual who thinks that corruption is widespread in business and government has life satisfaction that is lower by 0.26 points, more than half the size of the coefficients on income and having family or friends to rely on. Table 10.2 shows that these coefficients are fairly similar in all regions.
International Evidence on the Social Context of Well-Being
307
The prevalence of perceived corruption, and hence the average life evaluation effect, does differ considerably among regions. The perceived prevalence of corruption is highest in the transitional countries (0.88) and lowest in the United States + Canada + Australia + New Zealand grouping (0.45) and Western Europe (0.60). Regional averages for Asia, Latin America, and Africa range from 0.80 to 0.83. There is a large variation among countries in the level of perceived corruption, both within and across regions, with Bulgaria perceived as most corrupt (0.98), and Finland the least (0.15). How well are the large international differences in life evaluations explained by international differences in life circumstances, whether related to income, social networks, or corruption? We shall later use national aggregate equations to address that question, but there are other ways. The middle equation panel of Table 10.1 adds regional dummies to the standard two-level model, while the equations in the right-hand panel include dummy variables for all countries, thus forcing out all of the country-level variables. If the global model with uniform coefficients were seriously mistaken, then the explained variance would increase substantially with the addition of regional or national dummy variables, while the other coefficients might well be unstable. Even though the equation with dummy variables for 100 countries has a larger total explained variance than the one with only a few national variables, this increase is small, and the main structural parameters are unaffected. This provides a first line of evidence that international differences in life evaluations are due to differing life circumstances rather than different structural relations between circumstances and life evaluations. Table 10.5 (in the electronic appendix; http://wellbeing.econ.ubc.ca/ helliwell/papers/OUP-Chapter10-2010-Appendix.pdf) provides a more precise way of testing for interregional differences in coefficients. The equation is estimated using region 1 (the former Soviet Union [FSU] and Eastern Europe) as the base case, and tests for individual-level coefficients in other regions that are different from those in region 1. The equations are estimated for the largest pooled sample for the Ladder, as well as the matched samples for the Ladder and SWL. Looking only at the differences that are pervasive and significant, we find that the importance of having someone to rely on is greater in region 2 (Europe), region 3 (United States + Canada + Australia + New Zealand) and in region 4 (Latin America and the Caribbean), while being lower in Asia, in each case relative to the base region. A sense of freedom matters more in region 3 and less in Asia and Africa. The effects of nationallevel corruption are great everywhere, with Table 10.5 showing no significant differences among regions. The effects of a sense of personal freedom are also lower in Asia, Latin America, and Africa than in region 2 and Western Europe, while once again being significantly positive in all regions, as shown by the regional equations in Table 10.2.
308
Section III: Differences in the Social Context of Well-Being
We turn now to consider contextual effects, as measured by the national averages of variables also included at the individual level. One of the more striking results that we found in our earlier studies using the first wave of the Gallup World Poll, with only the Ladder variable available as a measure of life satisfaction, was that average per capita income had little effect. Earlier research using more local data has tended to find significant relative income effects9, and this was matched by the earlier WVS results. In Tables 10.1 and 10.2, household incomes are measured as log levels, converted into common units by the use of purchasing power parities used in the preparation of the Penn World Tables estimates of average GDP per capita10. Thus if there are any significant relative income effects at the national level, we would expect to find the contextual national GDP per capita entering with a negative sign. The results in Table 10.1 suggest that any relative income effects at the national level are being substantially offset by the effects of other excluded variables that support life satisfaction in the richer countries11. In particular, the national average should reflect all the tax-funded public-good consumption and income supports that are largely missing from measured variables. The estimation of contextual effects at the national level is most reliably done using Ladder responses pooled from all three waves of the Gallup World Poll, since that gives the largest number of individual and country observations. We have discussed above the contextual effect of income, which is uneven in sign and generally insignificant. In contrast, the contextual effects of the other variables generally take the same sign as the individual-level effects, showing some significant evidence of positive contextual effects. The contextual effects are large and significant in nations where corruption is low, where more people have others to rely on, and where religious participation is higher. The last result confirms the recent European findings of Clark & Lelkes (2009). The greater importance of participation than of religious beliefs echoes the recent U.S. results of Lim & Putnam (2009). Thus the growing body of international evidence tends to support the earlier findings of Helliwell & Putnam (2004) that the externalities of social context variables are more likely to be positive than are those of income.
Country-by-Country Modeling of Life Satisfaction
The basic estimation form for analysis of individual life satisfaction within each country is: LSij ¼ j þ j ln ðyij Þ þ j Xij þ Eij
ð2Þ
International Evidence on the Social Context of Well-Being
309
where LSij is individual life satisfaction measured on a scale of 0 to 10, yij is household income, and the Xij are other individual-level variables. The estimates aj, dj, and mj are specific to country j. The entire explanatory power of equation (2) comes from explaining cross-sectional individuallevel variance within a specific country, with differences between countries showing up as differences in constant terms and the estimated coefficients. The raw national samples are in the first instance approximately 1000 for each survey wave, but are rendered smaller by lack of data on key variables, especially household income. Figure 10.3 shows histograms of the coefficients from all 125 country equations. Table 10.6 (in the electronic appendix; http://wellbeing.econ.ubc.ca/helliwell/papers/OUP-Chapter102010-Appendix.pdf) displays the estimated coefficients and standard errors for the 125 countries. The model in these estimations is identical12 to that in the first equation of Table 10.1, except that all country-wide contextual variables drop out because there is no within-country variance. Figure 10.3 also includes the mean coefficients and 99% confidence range for the corresponding parameters estimated in the two-level global equation. These coefficients are very close to the means of the distributions of national coefficients. The quadratic pattern of age effects is nearly universal, with almost all countries having coefficients that are negative on age and positive on age squared. The pairs of coefficients are significant for all regional groupings
Figure 10.3 Distribution of coefficient values from within-country regressions.
310
Section III: Differences in the Social Context of Well-Being
Figure 10.3 (Continued).
of countries, although not for many individual countries. The gender effect for males is negative in 102 of the 125 countries, although significantly so only for 35. The other demographic variables are also fairly weakly defined in the national samples, reflecting the small sample sizes and the variety of individual experiences. The log of household income receives a positive coefficient in all but one of the 125 country regressions, and the coefficient is more than twice its standard error for 114 countries. To ensure that all of the effects of income flow through the coefficient on household income, we have redefined the other income-dependent variable, lack of enough money for food, to remove its correlation with income. The variable thus measures the extent to which lack of money for food is greater than would be expected for an average household with the same level of income. The lack of sufficient money for food takes the expected negative coefficient in 123 of the 125 countries, and it is more than double its standard error at least three-quarters of the country equations. The effects of a sense of freedom, and the absence of perceived corruption are also pervasively important, as shown by the coefficient patterns in Figure 10.3. For all variables the means of the country coefficients are very close to the values estimated in Table 10.1, as would be expected if the national samples were drawn from a global population with broadly similar responses to these variables.
International Evidence on the Social Context of Well-Being
311
Finally, it is necessary to address more directly the experimental (e.g., Heine & Norenzayen, 2006) and other evidence (e.g., Kahneman, 1999; Diener & Suh, 2000; Kahneman & Riis, 2005) that cross-national comparisons of retrospective assessments of subjective well-being are rendered difficult or possibly uninformative by cultural differences in the ways in which questions are interpreted, scales are used, values are determined and answers are framed (Heine et al., 2002; Schmidt & Bullinger, 2007; Oishi, 2009). What is meant by culture in this context? Matsumoto (2000) defines culture as ‘‘a dynamic system of rules—explicit and implicit—established by groups in order to ensure their survival, including attitudes, values, beliefs, norms, and behaviors . . . communicated across generations, relatively stable but with the potential to change across time.’’ This bears striking similarities to the OECD (2001) definition of social capital (Putnam, 1993, 2000; Halpern, 2005) as ‘‘networks together with shared norms, values and understandings that facilitate co-operation within and among groups.’’ In international research into the well-being consequences of differences in the quality of social capital (Helliwell & Putnam, 2004), it is presumed that key aspects of social norms (e.g., trust) can be meaningfully measured and compared across cultures and over time. The use of pooled international samples also assumes that the underlying structural relations are reasonably uniform across nations and regions. Our research and results suggest that some of the key intercultural differences in norms and values emphasized in the literature are supported in the subjective well-being data of the Gallup World Poll. For example, the well-being costs of living in a society with high perceived levels of corruption in business and government appear to be slightly less in countries where corruption is an established feature of the status quo (–.27 vs –.33, if we divide the global sample into countries with above-average and below-average levels of perceived corruption). Similarly, the well-being value attached to a sense of personal freedom is slightly higher in societies classed as individualistic rather than collectivist. But while these differences qualitatively confirm some key experimental cross-cultural findings, what appears to us remarkable is that application of the same well-being equation to 125 different national societies shows the same factors coming into play in much the same way and to much the same degree. This is illustrated by Figure 10.4 (in the electronic appendix; http://wellbeing.econ.ubc.ca/helliwell/papers/OUP-Chapter10-2010-Appendix.pdf), which shows actual and predicted values of life satisfaction obtained by applying the same model, with coefficients restricted to be the same for all countries13. One interesting exception is the significant positive boost to life satisfaction in South and Central America14. Is this perhaps related to some as yet unmeasured features of the Latin American family or broader social
312
Section III: Differences in the Social Context of Well-Being
context? With this interesting exception, the international differences in predicted values are entirely due to differences in their underlying circumstances, even without making any provision for international differences in equation structure and coefficients. The international similarity of the estimated structure of life evaluations means that the large international differences in average life evaluations are not due to different approaches to the meaning of a good life, but to differences in the social, institutional, and economic circumstances of life in different nations. This is exactly in line with the chapter by Kahneman et al. (2009) comparing the lives of French and American women. They find, consistent with our results, that international differences in life satisfaction are due to differences in the content of life rather than in the structure of the process of evaluation.
How Much Does the Social Context Matter for Life Evaluations?
It is time to review and summarize our findings on the social determinants of well-being. There are several ways of doing this: from two-level models pooling individual-level data from all countries, from the averages of national-level coefficients, and from aggregate equations making use of national average data. We can also choose among different ways of evaluating the quality of life: satisfaction with life, the Cantril Ladder, and our preferred index averaging the two measures. For any sample of dependent variable and model type, the importance of social context variables can be assessed either individually or in groups, and measured either in terms of their direct life satisfaction effects or in relation to the effects of income. Some of our main findings seem to apply broadly across alternative models and data, and are hence worth spelling out in summary form. First, we now find that the two alternative life evaluation measures we have assessed in large samples produce very similar estimates of the relative importance of the economic and non-economic correlates of well-being, especially in models based on individual data. Indeed, the remaining differences between the two measures15 can be seen as an advantage, because a simple average of the two measures appears to increase the signal-to-noise ratio significantly, resulting in models with tighter structural form and lower standard errors of estimation. Second, we find that the log of household income is a robust explanatory variable in almost all countries and regions, with a global coefficient in the region of .4 to .5 in equations where food adequacy is redefined to exclude the effect of income, thus forcing income effects to flow though the income
International Evidence on the Social Context of Well-Being
313
variable. This is about ten times its standard error, making it possible to evaluate other factors reasonably precisely in income-equivalent terms. Third, since we now find that SWL and the Ladder produce quite similar estimates, and since we do not yet have large enough samples for SWL to employ our preferred measure that averages SWL and the Ladder, we can without too much likely bias use global, national, and country average equations for the Ladder to estimate the importance of the social context. In doing this evaluation, we shall use three methods: the first based on the global two-level equation of Table 10.1, the second based on average coefficients from the national-level equations of Table 10.6, and the third based on an equation using only national average data, as shown in Table 10.3. All three methods show that there are several features of the social context that have large and significant effects on well-being in all regions of the world. For example, the individual-level ‘‘friends to count on’’ coefficient in the global two-level equation is as large as that on the log of household income, implying that having someone to count on is more than twice as important as a 50 percent higher income. But even that ignores the social context effect—that people make a more positive evaluation of their lives if they live in a society where others, and not just themselves, have people to rely on. This is shown by the large and significant positive coefficient on the national average for the ‘‘count on friends’’ variable. In order to assess the total effects of differences in the social context using two-level modeling, we need to calculate the income-equivalent changes in well-being through both channels. We do this by assuming a social context change that moves some fraction, say one-tenth, of the respondents from a 0 to a 1, thus raising the value of a typical individual’s social context variable by 0.1, which is also the amount by which the national average increases. The standard deviations of the national averages of the social context variables for the 140-odd Gallup World Poll countries range from .1 to .2, so that a change of .1 is of moderate size in relation to the current range of differences across countries. Our estimate of the direct plus contextual effects from method 1 thus involves simply summing the two coefficient ratios from the left-hand column of Table 10.4. Our second method for estimating compensating differentials is to base them entirely on the national-level equations. This requires the use of the averages of coefficients from equations applied separately to each national sample. The averages are shown in Figure 10.3, as vertical lines in the distributions of estimated coefficients16. The ratios of these coefficients to those on the log of household income, and estimates of the standard errors of those ratios, are shown in the middle column of Table 10.4. Our third method relies on equations based entirely on national average data for the countries for which there are values for average per capita
TABLE 10.3
Macro-level estimates Significance:
nation: log(GDP per cap) nation: not enough money: food(net)
314
nation: friends to count on nation: freedom to choose nation: perception of corruption nation: donated time nation: donatedMoney
1%*
5%
10%*
ladder
SWL
1 2 (ladder + SWL)
ladder
SWL
(1)
(2)
(3)
(4)
(5)
.61* (.066) .55 (.42) 2.42* (.65) 1.47* (.45) .72* (.28) .31 (.53) .22 (.35)
.54* (.058) 1.74* (.38) .98 (.61) 1.24* (.39) .70 (.33) .21 (.48) .079 (.39)
.58* (.057) 1.03* (.39) 1.57 (.61) .94* (.33) 1.14* (.33) .055 (.51) .13 (.33)
.62* (.090) .20 (.63) 3.05* (.84) 1.94* (.62) .51 (.32) .26 (.61) .28 (.46)
.52* (.072) 1.35* (.44) 1.58 (.66) 1.75* (.52) .60* (.33) .25 (.53) .50 (.50)
1 2 (ladder +SWL)
(6)
.52* (.053) 1.32* (.36) 1.50 (.61) 1.46* (.44) .67 (.29) .38 (.50) .32 (.42)
nation: helped a stranger nation: importance of religion nation: attended church/etc healthy life expectancy (net)
315
constant region fixed effects obs. R2(adj)
45 (.37) .31 (.38) 1.27* (.43) .024* (.007) 4.88* (.74) 128 .849
.74 (.48) .024 (.50) 1.29 (.53) .070* (.010) 2.63 (1.03) 99 .863
.64* (.38) .32 (.39) 1.38* (.44) .047* (.007) 3.71* (.81)
.15 (.35) .42 (.33) .90 (.38) .017 (.008) 5.28* (.72)
.43 (.43) .14 (.37) .86* (.45) .029* (.009) 4.42* (.77)
.35 (.38) .33 (.32) 1.13* (.35) .022* (.008) 4.80* (.70)
99 .897
[ 128 .869
[ 99 .921
[ 99 .922
TABLE 10.4
Compensating differentials Significance:
perception of corruption nation: perception of corruption freedom to choose nation: freedom to choose
316
friends to count on nation: friends to count on not enough money: food(net) nation: not enough money: food(net) donatedMoney nation: donatedMoney donatedTime
1%*
5%
Method-1, From two-level regressions
Method-2, From within-nation regressions
0.62* [0.07] 3.62* [0.77] 0.73* [0.07] 1.30 [0.71] 1.08* [0.09] 3.77* [1.17] 1.50* [0.08] 0.19 [0.83] 0.38* [0.05] 0.37 [0.63] 0.18* [0.05]
0.43* [0.03]
10% Method-3, From regressions on national averages
1.92* [0.53] 0.45* [0.03] 1.57 [0.66] 0.61* [0.03] 2.64* [0.99] 1.04* [0.03] 1.74* [0.62] 0.22* [0.03] 0.21 [0.61] 0.13* [0.03]
nation: donated time helped a stranger nation: helped a stranger importance of religion nation: importance of religion attended church/etc
317
nation: attended church/etc nation: healthy life expectancy (net)
1.59 [1.17] 0.20* [0.05] 0.04 [0.82] 0.02 [0.05] 1.37 [0.98] 0.11* [0.04] 3.14* [1.12] 0.05* [0.02]
0.09 [0.9] 0.13* [0.02] 0.76 [0.66] 0.01 [0.03] 0.52 [0.58] 0.10* [0.03] 2.13* [0.68] 0.04* [0.01]
Note:
1. Compensating differentials are defined as the ratio of the coefficient of the variable to the left of the table, over the coefficient of the log of household income. In the case of method 2, the compensating differential is the weighted average of the country-specific compensating differentials derived from within-country regressions. The weight is the inverse of the square of standard errors. 2.The standard error for a compensating differential, which is defined as a ratio of coefficients, is calculated from the variance-covariance matrix of estimated coefficients using the Delta method. In the case of method 2, the standard error for the weighted average is calculated from the standard errors of country-specific compendsating differentials, while assuming that each nation-sample represents a random sample from the same distribution.
318
Section III: Differences in the Social Context of Well-Being
national income, food inadequacy, health-adjusted life expectancy, and eight social context variables. One advantage of using estimates based on national average data, is that they are less open to the possibility that individual-level personality differences are responsible for individuals’ reporting more frequent and higher-quality social interactions as well as higher life satisfaction. A similar argument has been applied for income; people with more optimistic and outgoing personalities may be more likely to find and report higher-paying jobs and higher levels of life satisfaction. It is not clear what the net effect would be on the relative importance of income and social context variables. In any event, averaging across all 1000 respondents in each nation-wave should reduce substantially or eliminate that source of personality-driven bias. Another advantage of using the national level data is that the average individual-level and contextual effects are automatically (if indistinguishably) combined to provide estimates of total effects from social and economic variables. The corresponding disadvantage, in contrast to our two-level modeling, is that it is not possible to separate the individual-level and context effects of the key variables. Another disadvantage of using the national average data is the limited number of observations, coupled with a large number of correlated candidate variables. The resulting coefficients are thus sensitive to changes in the number and nature of countries and variables. Table 10.3 shows the results of re-estimating the first equation of Table 10.1 using national average data for the countries with available data. Since we wish our estimates of the value of the income-equivalent value of the social context to err if anything on the conservative side, the two variables that have high correlations with income: inadequate money for food, and health-adjusted life expectancy, have been redefined to exclude their correlation with income, thus giving GDP per capita the whole credit for circumstances that might be responsible for both higher incomes and greater life expectancy. The social context variables by themselves explain 73.4% of the cross-country variance of the Gallup Ladder, adjusted for degrees of freedom. This rises slightly to 73.5% when food adequacy (net of income effects) is added, no further with the addition of net healthy life expectancy and then to 84.9% when average per capita income is added. Starting from the other side, per capita income explains 71.2% of the cross-country variance of the Ladder. This rises to 74.2% with the addition of food adequacy, no further with the addition of healthy life expectancy, and finally once again to 84.9% with the addition of the social context variables. Thus, even with the national level Ladder data, the social context matters as much as income in explaining international differences in well-being. Finally, we can use the coefficients from the aggregate equation, just as we have used the two-level global model and the results from the country
International Evidence on the Social Context of Well-Being
319
regressions, to calculate compensating differentials, and these have been entered in the third column of Table 10.4. Table 10.4 brings together estimates of compensating differentials of various social context variables obtained from three different sources: the two-level modeling of Table 10.1, from the weighted averages of coefficients estimated in national equations, and finally, from the Table 10.3 equation using national average data. As we have already noted, there are good reasons why these estimates should differ. Methods one and three both take externalities into account, and may for that reason be higher if externalities are more positive for the social context variables than for income. In method two, based on coefficient averages from separate national equations, the national-level contextual effects are implicitly contained in the constant terms of each national equation. However, in methods one and two, there might be an upward bias to the individual-level estimates of compensating differentials if individual-level personality differences are more likely to generate positive covariance between life evaluations and social reports than between life evaluations and income. Method three avoids this risk by using only national-level data for social context variables in estimating the coefficients used to calculate the compensating differentials, but at the cost of small samples and inability to distinguish individual-level and contextual effects. In summary, if methods one and three give higher estimates of compensating differentials than does method two, then we would infer that contextual effects (externalities) are more positive for the social variables than for income. Method one, which includes both individual-level and contextual effects, should probably provide the largest estimate of their combined size, but only if the externalities are positive, so that the individual-level and national-level receive coefficients of the same sign. Table 10.4 shows the ratios of the social context coefficients to the coefficient on the relevant log income variable, with the estimated standard error of that ratio shown below. These ratios are estimates of the change in the (log of) household income that would have a life satisfaction effect equivalent to a unit change in the corresponding measure of social structure. For example, consider a change where 10 percent more of a country’s population thought that they had someone to count on. The individual effect for those having the improved social connections would be a log change of income of 1.08, roughly equivalent to a trebling of family income. On top of this, and received by all people, including those whose own social support had not improved, would be the contextual effect, with a log equivalent value of .1*3.77 ¼ .377, or more than a one-third increase of family income. The typical individual in this more-connected country would have life satisfaction increase by .1*(3.77 + 1.08) = .485, equivalent to an increase of family income of more than 60%.
320
Section III: Differences in the Social Context of Well-Being
These very large compensating differentials must be viewed in the context of actual individual and international differences in per capita incomes and social circumstances. In a typical country, 80 percent of the respondents have someone to rely on, and the standard deviation of this measure across countries is just over one-sixth as large, while the national average per capita income, measured in purchasing power parity as a fraction of that in the United States, is .30, with a standard deviation across countries of .34, reflecting the large number of very poor countries. Thus it is not too surprising to find that even very large differences in life satisfaction across countries are dwarfed by the corresponding differences in material income. One of the things that supports life satisfaction in the poorer countries is that while they are also more socially deprived, on average, according to almost all of the Gallup measures, than are those living in the richer countries, their relative social deprivation is less than their relative economic deprivation17. The very large size of the estimated compensating differentials for social variables in all regions suggests the importance of focussing on ways of supporting and improving the social context in developing countries. In the richer countries, the case for giving increased importance to the social context is even stronger, given the diminishing life satisfaction effect of higher incomes implied by the empirically preferred log-linear form.
Conclusion
We find strong evidence for the importance of both income and social context variables in explaining differences in well-being. For most specifications tested, the combined effects of a few measures of the social and institutional context exceed that of income in equations explaining international differences in life satisfaction. Calculation of compensating differentials also reveals large income-equivalent values for improvements in the social context, with much of this value flowing via positive national spillover effects for key social variables. In a preliminary version of this chapter, based on data for a smaller number of countries, we found evidence suggesting that average answers to the standard life satisfaction question were higher than for the Cantril Ladder, with this difference being larger for the poorer countries. We hypothesized that this was because the Ladder framing encouraged respondents to think in more relative terms, and that these comparisons were global in nature. However, as the sample of countries with both SWL and Ladder data has grown, we have found that both measures show similar correlations with the key structural variables. The SWL answers remain
International Evidence on the Social Context of Well-Being
321
higher on average than those for the Ladder, but by a smaller amount than we found using data from a smaller number of countries, and the difference between the two measures is no longer related significantly to per capita incomes. Reflecting the facts that the Ladder and the SWL questions are related in very similar ways to the underlying structural determinants, that they are both evaluations of life as a whole, that they are framed differently, and asked at different parts of the interview, we hypothesized that an average of the two measures might have a higher signalto-noise ratio than either on its own18. Our tests to date appear to confirm this hypothesis, as the average is significantly better predicted than is either measure alone, and has a smaller standard deviation and a smaller standard error of the estimate. The global cross-sectional equation for the average of life satisfaction and the Ladder explains 44 percent of the variance for the 52,000 individuals for whom both assessments are currently available. This is substantially the highest explained variance we have ever seen for a cross-sectional explanation for subjective well-being. This is especially remarkable since the simple model used constrains parameters to be equal in all countries. This suggests that both measures are tapping into life in similarly evaluative ways, with their slightly different framing, and their different positions in the body of the survey, giving some random differences that become smaller when the two measures are averaged. It also suggests that despite cross-cultural differences in the ways in which social and economic institutions are designed, and in how people assess and are influenced by these characteristics, a relatively small number of key structural variables appears to explain a large fraction of differences in subjective life evaluations around the world.
Notes 1. Helliwell and Barrington-Leigh are with the Canadian Institute for Advanced Research and the Department of Economics, University of British Columbia. Huang is in the Department of Economics at the University of Alberta, and Harris is at Nuffield College, Oxford. This paper is part of the ‘‘Social Interactions, Identity and Well-Being’’ research program of the Canadian Institute for Advanced Research, and is also supported by grants from the Social Sciences and Humanities Research Council of Canada. We are grateful to conference participants for comments and advice. In revising the paper we have been especially aided by the post-conference review by Richard Layard. 2. First, earlier claims that each person has a psychological set-point for subjective well-being to which he or she invariably returns (Brickman & Campbell, 1971; Brickman et al., 1978; Lucas et al., 2003) have been replaced by research showing
322
3. 4.
5. 6.
7.
8. 9. 10.
11.
12.
Section III: Differences in the Social Context of Well-Being that adaptation to most changes in life circumstances is partial in nature (Lucas, Diener, & Scollon, 2006; Lucas, 2007). Second, experimental evidence that retrospective assessments of well-being differ from Bentham-like (Kahneman, Wakker, & Sarin, 1997) integrals of momentary assessments (Kahneman, 1999; Frederickson & Kahneman, 2003; Kahneman & Riis, 2005) was held not to threaten the usefulness of retrospective evaluations of satisfaction, especially as the latter are what govern future decisions (Wirtz et al., 2003). Third, in response to suggestions that freedom and capabilities, which were held to be of fundamental value to well-being (Sen,1990, 1999), would be left out of account by measures of life satisfaction, it was shown that measures of life satisfaction appear to differ from assessments of positive and negative affect in just the ways that make life satisfaction an appropriate measure. Indeed, in this paper we show that a sense of personal freedom is highly significant as a support for higher measures of life satisfaction. These two sub-samples contain 63 countries in common. We do not include national level values for gender, the age variables, and marital status. Although there are some differences among countries and regions in population age structure and marital status, experiments adding the national averages to equation 1 do not reveal significant effects or materially alter the sizes of other coefficients. The lack of responses to the question of household income is responsible for most of the reduction in sample size. The Gallup variable (WP40) for not having enough money for adequate food was regressed on the log of household income, and the residuals are used as the net food variable. If this transformation is not done, the income coefficient in the first equation in Table 10.1 would be lower by 0.07. For the 60 countries for which Gini coefficients are available, the correlation between the Gini and the food inadequacy variable (WP40) is +0.70, and +.35 between the Gini and the food variable adjusted to remove its relation to GDP per capita. This is especially so for trust in the workplace, as discussed in Helliwell and Huang (2008). See Luttmer (2005) and Barrington-Leigh & Helliwell (2008). See also Easterlin (1974). More precisely, the individual household incomes in the Gallup data are divided by their country means to get relative incomes within each country. These figures are then converted into common level form by adding the resulting relative income to the average GDP per capita in 2005 measured at Purchasing Power Parity (from the World Bank ICP). The contextual variable is the same World Bank series. Thus, if there are significant relative income effects at the national level, the contextual variable should attract a negative coefficient. Alternatively, since the biggest reduction in the coefficient on national income happens when the basic needs variables are added, the reduction in the relative income effect may be due to the large positive cross-country correlations between national income and the attainment of basic needs. It is one more reason for the issue to remain open. One small exception is the foodnet variable, for which the definition changes slightly with the change in level of estimation. For the global equations, foodnet
International Evidence on the Social Context of Well-Being
13. 14. 15. 16.
17. 18.
323
is defined net of globally estimated effects of income on food adequacy, while for the national equations it is the national relation that is used to take the variance of income out of the food variable to create foodnet. The equation is that shown in the first column of Table 10.1. As shown by the significant positive coefficient on the dummy variable for region 4 in the (SWL + Ladder)/2 equation in the center panel of Table 10.1. Measured across all individuals asked both questions on the same survey, the correlation between SWL and the Ladder is 0.61. The global averages of the national-level coefficients are weighted by the inverse square of the estimated standard error of each of each coefficient. The exact data used are shown at the bottom of Table 10.6, in the electronic appendix. In the macro equation of Table 10.3, the standardized beta on income is .66 for the log of GDP per capita, compared to .43 for the non-income variables. This tends to confirm the point, made at several places in this volume, that multiple measures of well-bring should be collected where feasible, and used to form combined measures and/or guide the selection among alternative measures,
References Barrington-Leigh, C. P., & Helliwell, J. F. (2008). Empathy and emulation: Subjective well-being and the geography of comparison groups. NBER Working Paper No. 14593. Blanchflower, D. G., & Oswald, A. J. (2008). Is well-being U-shaped over the life cycle? Social Science and Medicine, 66, 1733–1749. Brickman, P., & Campbell, D. T. (1971). Hedonic relativism and planning the good society. In M. H. Appley (Ed.), Adaptation level theory: A symposium (pp. 287–302). New York: Academic Press. Brickman, P., Coates, D., & Janoff-Bullman, R. (1978). Lottery winners and accident victims: Is happiness relative? Journal of Personality and Social Psychology, 36, 917–927. Clark, A. E., & Lelkes, O. (2009). Let us pray: Religious interactions in life satisfaction. Working Paper: Paris School of Economics. Deaton, A. (2008). Income, health and well-being around the world: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22(2), 53–72. Diener, E., & Suh, E. M. (Eds.) (2000). Culture and subjective well-being. Cambridge, MA: MIT Press. Diener, E., Lucas, R. E., & Scollon, C. N. (2006). Beyond the hedonic treadmill: Revising the adaptation theory of well-being. American Psychologist, 61(4), 305–314. Diener, E., Kahneman, D., Tov, W., & Arora, R. (2010). Income’s differential impact on judgments of life versus affective well-being. In E. Diener, J. F. Helliwell, & D. Kahneman (Eds.), International differences in well-being. Oxford, UK: Oxford University Press. Easterlin, R. (1974). Does economic growth improve the human lot? Some empirical evidence. In P. A. David & M. Reder (Eds.), Nations and households in economic growth (pp. 89–125). New York: Academic Press. Frederickson, B. L., & Kahneman, D. (2003). Duration neglect in retrospective evaluations of affective episodes. Journal of Personality and Social Psychology, 65, 45–55.
324
Section III: Differences in the Social Context of Well-Being
Fowler, J. H., & Christakis, N. A. (2008). The dynamic spread of happiness in a large social network. British Medical Journal 337,a2338 (4 December, 2008) Gallup Organization. (2007). The state of global well-being. New York: Gallup Press. Halpern, D. (2005). Social capital. Cambridge, UK: Polity Press. Heine, S., & Norenzayan, A. (2006). Towards a psychological science for a cultural species. Perspectives on Psychological Science, 1(3), 251–269. Heine, S., et al. (2002). What’s wrong with cross-cultural comparisons of subjective Likert Scales?: The reference group effect. Journal of Personality and Social Pyschology, 82(6), 903–918. Helliwell, J. F., & Putnam, R. D. (2004). The social context of well-being. Philosophical Transactions of the Royal Society of London B 359: 1435–46. Reprinted in F. A. Huppert, N. Baylis, & B. Keverne (Eds.), The science of wellbeing (pp. 435–59). Oxford: Oxford University Press, 2005. Helliwell, J. F., & Huang, H. (2008). Well-being and trust in the workplace. NBER Working Paper No. 14589. Helliwell, J. F. (2008). Life satisfaction and quality of development. NBER Working Paper No. 14507. Helliwell, J. F., Huang, H., & Harris, A. (2009). International differences in the determinants of life satisfaction. In Tridip Ray, E. Somanathan, & Bhaskar Dutta (Eds.), New and enduring themes in development economics (pp. 3–40). Singapore: World Scientific. Kahneman, D. (1999). Objective happiness. In E. Diener, D. Kahneman & N. Schwarz, (Eds.), Well-being: The foundations of hedonic psychology (pp. 3–25). New York: Russell Sage. Kahneman, D., & Riis, J. (2005). Living, and thinking about it: Two perspectives on life. In F. A. Huppert, N. Baylis, & B. Keverne (Eds.), The science of well-being (pp. 285–304). Oxford, UK: Oxford University Press. Kahneman, D., Wakker, P. P., & Sarin, R. (1997). Back to Bentham? Exploration of experienced utility. Quarterly Journal of Economics, 112, 375–405. Kahneman, D., Schkade, D. A., Fischler, C., Krueger, A. B., & Krilla, A. (2009). The structure of well-being in two cities. In E. Diener, J. F. Helliwell & D. Kahneman, (Eds.), International differences in well-being. Oxford, UK: Oxford University Press. Lim, C., & Putnam, R.D. (2009). Praying alone is not fun: Religion, social networks and subjective well-being (working paper). Lucas, R. E. (2007). Long-term disability is associated with lasting changes in subjective well-being: Evidence from two nationally representative longitudinal studies. Journal of Personality and Social Psychology, 92(4), 717–730. Lucas, R. E., Clark, A. E., Georgellis, Y., & Diener, E. (2003). Re-examining adaptation and the set point model of happiness: Reactions to changes in marital status. Journal of Personality and Social Psychology, 84(3), 527–539. Luttmer, E. F. P. (2005). Neighbours as negatives: Relative earnings and well-being. Quarterly Journal of Economics (January), 963–1002. Matsumoto, D. (2000). Culture and psychology. 2nd ed. Pacific Grove, CA: Brooks Cole. Oishi, S. (2009). Culture and well-being: Conceptual and methodological issues. In E. Diener, J. F. Helliwell & D. Kahneman, (Eds.), International differences in wellbeing. Oxford, UK: Oxford University Press. Organization for Economic Co-operation and Development. (2001). The well-being of nations: The role of human and social capital. Paris: OECD.
International Evidence on the Social Context of Well-Being
325
Putnam, R. D. (1993). Making democracy work. Princeton, NJ: Princeton University Press. Putnam, R. D. (2000). Bowling alone: The collapse and revival of American community. New York: Simon and Schuster. Sen, A. (1990). Development as capability expansion. In K. Griffin & J. Knight (Eds.), Human development and the international development strategy for the 1990s. London: Macmillan. Sen, A. (1999). Development as freedom. New York: Knopf Press. Schmidt, S., & Bullinger, M. (2007). Cross-cultural quality of life assessment approaches and experiences from the health care field. In I. Gough & J.A. McGregor (Eds.), Well-being in developing countries: From theory to research (pp. 219–241). Cambridge, UK: Cambridge University Press. Stevenson, B., & Wolfers, J. (2008). Economic growth and subjective well-being: Reassessing the Easterlin Paradox. Brookings papers on economic activity, spring 2008, 1–87. Wirtz, D., Kruger, J., Scollon, C. N., & Diener, E. (2003). What to do on spring break? The role of predicted, on-line and remembered experience in future choice. Psychological Science, 14(5), 520–524.
Appendix 10.A Wording of Key Questions From Gallup World Poll
Life Today as Ladder (Cantril’s self-anchoring striving scale) Please imagine a ladder with steps numbered from zero at the bottom to ten at the top. Suppose we say that the top of the ladder represents the best possible life for you and the bottom of the ladder represents the worst possible. If the top step is 10 and the bottom step is 0, on which step of the ladder do you feel you personally stand at the present time? SWL: All things considered, how satisfied are you with your life as a whole these days? Use a 0 to 10 scale, where 0 is dissatisfied and 10 is satisfied. Count On Friends If you were in trouble, do you have relatives or friends you can count on to help you whenever you need them, or not? Cannot Afford Food Have there been times in the past twelve months when you did not have enough money to buy food that you or your family needed? Corrupt: average of the following two responses 1. Is corruption widespread within businesses located in (county of interview), or not? 2. Is corruption widespread throughout the government in (county of interview), or not? 326
International Evidence on the Social Context of Well-Being
327
Freedom In (county of interview), are you satisfied or dissatisfied with your freedom to choose what you do with your life? Donate Money Have you done any of the following in the past month? How about Donated money to a charity? Donate Time Have you done any of the following in the past month? How about Volunteered your time to an organization? Help Stranger Have you done any of the following in the past month? How about Helped a stranger or someone you didn’t know who needed help? Religion Importance Is religion an important part of your daily life? Attended church/etc Have you attended a place of worship or religious service within the last seven days? From World Values Survey SWL: All things considered, how satisfied are you with your life as a whole these days? Please use this card to help with your answer. Dissatisfied (1) 2 3 4 . . . Satisfied (10)
Chapter 11 How Universal Is Happiness? Ruut Veenhoven Erasmus University, Rotterdam, The Netherlands
Introduction
The recent rise of interest in happiness has revived classic discussions about the nature of happiness. One of these discussions centers on whether happiness is similar for all humans or rather something that varies across cultures. In the universalist view, happiness is comparable to ‘‘pain.’’ All humans know what pain is, will experience pain when touching a hot stove, and tend to avoid pain. In the relativistic view, happiness is more comparable to ‘‘beauty,’’ the idea of which varies across time and culture. Picasso’s paintings are not appreciated by everybody, nor does everybody seek only beauty1. This discussion links up with wider issues, among which is the longstanding debate about the merits of utilitarian moral philosophy. Its ‘‘greatest happiness principle’’ assumes that happiness is something universal. If different in different cultures, happiness cannot serve for the evaluation of cultures. If culturally variable, the definition of happiness can also change over time, and happiness is therefore not a strong criterion for public choice within cultures. These arguments have been presented repeatedly, with few conclusions arising from the discussions, due to a lack of empirical proof for either position. In this chapter, I inspect what our new knowledge about happiness can tell us about this old controversy. Has a decade of empirical research made us any wiser on this matter, or are we still as much in the dark as the nineteenth-century armchair philosophers who criticized utilitarianism on this ground? 328
How Universal Is Happiness?
329
Concept of happiness
A preliminary step is to define happiness, since some of the things denoted using this word can be less universal than others things called by the same name. I use the word ‘‘happiness’’ for a subjective state of mind, which I define as the overall appreciation of one’s life as–a–whole. I have elaborated this definition in earlier publications (Veenhoven, 1984 chapter 2; Veenhoven, 2000). This definition fits Jeremy Bentham’s classic notion of happiness as ‘‘the sum of pleasures and pains.’’ Happiness in this sense is synonymous with ‘‘life satisfaction’’ and ‘‘subjective well-being’’2. Additionally, I distinguish two ‘‘components’’ of happiness: an affective component and a cognitive component. The affective component is how well one typically feels. I call this the hedonic level of affect. The cognitive component is the perceived difference between what one has and what one wants in life, which I call contentment. I assume that these components serve as subtotals in the overall evaluation of life. Sub-questions
The question ‘‘How universal is happiness?’’ is too broad to answer, since there are different facets of happiness, which may be more or less universal. Hence I will break down the main question into the following sub-questions: 1) 2) 3) 4) 5) 6) 7)
Do we all appraise how much we like life? Do we appraise life on the same grounds? Are the conditions for happiness similar for all humans? Are the consequences of happiness similar around the globe? Do we all seek happiness? Do we seek happiness in similar ways? Are we about equally happy in all cultures?
Since the focus of this chapter is on cultural variations in the nature of happiness, I do not deal with the cross-cultural measurement of happiness. Cultural measurement bias may distort the data on which this chapter builds in several ways, but the literature suggests that the degree of distortion is not alarming (see, e.g., Diener & Oishi, 2004; Veenhoven, 2008c). Some issues in cross-cultural measurement of happiness are discussed in the chapter by Oishi (2010) in this volume. Data source
Most of the empirical data used in this chapter are taken from the World Database of Happiness (Veenhoven, 2008), which is a collection of
330
Section III: Differences in the Social Context of Well-Being
research findings on happiness as defined above. References are made to sections of this database. In each of these sections one can find standardized descriptions of research findings and links to the original publication. Citing these all separately would be unwieldy.
Question 1: Do We All Appraise How Much We Like Life?
Above, I distinguished between overall happiness and its components and assumed that the components serve as subtotals in the overall evaluation of life. Do all humans appraise their life in these ways?
Hedonic level of affect
Like other higher animals, humans experience positive and negative affects. This is not just something we know from our own experience, it is also something we can recognize in the facial expressions of other people all over the world (Ekman, 1970). Using brain imaging we can now also observe part of the neural processes that make us feel so (e.g., Davidson, 2004) and these neurological structures do not differ across cultures either3. The balance of positive and negative affects is reflected in the hedonic tone of ‘‘mood.’’ Though mood is something we are aware of, it is mostly not in the foreground of our consciousness. Still, it is assessable, and we can estimate how well we feel most of the time. Babies are not yet able to engage in such reflection, but they still experience happy or unhappy moods. Although they cannot report how they typically feel, their mood level can be assessed using behavioral indications. This case of babies illustrates that one can be happy without having a concept of happiness in mind. Adult humans know typically how well they feel most of the time and this appears in the practice of measurement. When asked how well they usually feel, people answer instantly. The non-response rate tends to be small. Self-ratings of average hedonic level do not differ much from the balance scores scientists compute from responses to multiple questions about specific affects4 and do not differ substantially from ratings based on experience sampling5 or from ratings by intimates6.
Contentment
Unlike their fellow animals, humans can develop ideas of what they want from life and then compare these aspirations with the realities of their life.
How Universal Is Happiness?
331
This faculty is not present from birth on, but develops on the road to adulthood. There is no doubt that all adults have wants, even ascetics who want to denounce all wants still have the desire to denounce wants. There is also no doubt that most adults have an idea of how well their wants are being met, at least about important wants. Wants are often not very specific, and few people have clear priorities in mind; nevertheless, most people have no problem in estimating of how successful they are in getting what they want from life. Several survey studies have involved questions about what one wants from life and the degree to which one sees these wants being met. A common question is: ‘‘So far, I have gotten the important things I want in life’’ (item in Diener’s ‘‘Satisfaction with Life Scale,’’ Diener et al., 1985). The responses tend to be prompt, and the percentage of respondents who use the ‘‘Don’t know’’ option is very low. So, apparently, this question links up with something people have in mind. Even if people have no overall judgment of success already in mind, they appear able and willing to make one when asked. This appears in the practice of focused interviews, in life-review interviews in particular. Like in the case of hedonic level it is not required that people have made up their mind: an external observer can estimate someone’s overall contentment based on that person’s reported success in meeting specific wants.
Overall happiness
Given the above, it is no surprise that people have no problem in reporting how much they like their life-as-a-whole. Responses to questions on overall happiness are typically prompt. If not, happiness would not be such a common item in survey research. The non-response level to questions on happiness is typically low. Fewer than 1 percent use the ‘‘Don’t know’’ option7, and few people skip the question8. (See Table 11.1.) Non-response TABLE 11.1 ‘Don’t Know’ response to survey questions about happiness USA Zimbabwe China France Russia India Average in 78 nations
0.19% 0.26% 0.49% 0.65% 1.42% 2.69% 0.75%
Source: World Values Surveys, life-satisfaction item, average waves 1 to 4
332
Section III: Differences in the Social Context of Well-Being
is much higher on questions about other issues such as income and political preference. Questions on life-satisfaction are also easily answered in nonmodern societies, such as the Inughuit, the Amish, and the Maasai (BiswasDiener et al., 2005).
Question 2: Do We Appraise Life on the Same Grounds?
As mentioned above, I assume that we appraise our life in two ways: affectively, by assessing how well we feel; and cognitively, by comparing life-as-itis with how we want-life-to-be. This theory is summarized in Table 11.2. Hedonic level depends on gratification of universal ‘‘needs’’
Why can we experience pleasure and pain? The biological function is evidently to signal that things are good or bad for us. Evolution has programmed us this way. What, then, is the function of mood? Clearly not to signal specific benefit or danger: unlike pleasure and pain, moods are typically not related to specific stimuli and certainly not average mood level over longer periods of time. Mood level seems to function as a meta-signal and indicates how well we are doing on the whole. Feeling good means that all lights are on green and that we can go ahead, while feeling bad means that there is something wrong and that we should check what that is. This affective signal mechanism seems to exist in all higher animals, and its neural basis is found in the evolutionarily eldest parts of the human brain. TABLE 11.2 Concepts of happiness: Overall happiness and components.
Source: Veenhoven, 1984.
How Universal Is Happiness?
333
What, then, is ‘‘doing well’’? I assume, but cannot prove, that this is meeting innate ‘‘needs.’’ Needs are requirements for functioning that are so essential that evolution has safeguarded their gratification by linking these functions to affective signals. This is pretty evident in the case of ‘‘deficiency needs’’ such as hunger, but it seems also to apply to ‘‘growth needs’’ such as curiosity9. In this view, happiness is rooted in the gratification of basic needs that are part of human nature. In that respect happiness draws on universal grounds. I have discussed this theory in more detail elsewhere (Veenhoven, 1991, 2009).
Contentment depends on meeting culturally variable ‘‘wants’’
Why do we have wants? Mainly to gratify universal needs. In lower animals, needs are met by means of instinctive behaviors. The human strategy is more flexible than that and allows need gratification though planned behavior. ‘‘Wants’’ are a part of that planning. What do we want? Part of the answer is that we tend to adopt current standards of the good life; e.g., the standard of what material level of living is desirable and possible. These standards vary across time and culture; today we want more material comfort than our great-grandparents could dream of, and standards are higher in American business circles than in Tibetan monasteries. In this view, happiness is rooted in social standards and in this respect is culturally relative. For a recent statement of this view, see Chambers (1999).
Affective experience dominates in the overall evaluation of life
In this line of thought, the question of how universal ‘‘happiness’’ is boils down to the question which of these two ways of appraising life is the most important. I have considered this question in earlier publications (Veenhoven, 1991, 2009) and concluded that affective experience dominates the overall evaluation of life. Below I will summarize the main arguments and present some more evidence. Theoretical plausibility
From an evolutionary point of view it is not plausible that cognitive contentment dominates our overall appraisal of life. Cognition developed much later and serves as an addition to affective appraisal rather than a substitute. Reason helps explain why we feel good or bad and allows detection of false
334
Section III: Differences in the Social Context of Well-Being
affective signals, although it is difficult to ignore these, as depressives can tell you. Affective appraisal tends to precede cognitive decision (Zajonc, 1984), and without affective appraisal we cannot come to a decision, as cases of brain damage demonstrate (Damasio, 1994). From this perspective it is also unlikely that humans orient by variable cultural standards in the first place, rather than by needs that are rooted in biological evolution10. The limited role of cognitive comparison is also illustrated by the fact that it does not exist in little children, who as yet have no idea of what they want from life. Still, it is clear that children can be happy or unhappy, and there is typically no great change in happiness when they develop wants. Empirical indications
Since we cannot (yet) look into people’s heads, there is no direct empirical evidence of the relative strength of both ways of appraising life. Still, there are several indirect indications. Overall happiness more correlated to affect than contentment. If affective experience dominates the overall appraisal of life, this must appear in sizable correlations with overall happiness and more sizable correlations than with contentment. Unfortunately, there are no reports of studies involving measures of all three of these variants of happiness, so we must make do with studies that correlated either happiness with affect or overall happiness with contentment. The findings of such studies are stored in the World Database of Happiness, which distinguishes measures of overall happiness (coded ‘‘O’’), measures of affect level (coded ‘‘A’’) and measures of contentment (coded ‘‘C’’). Eight studies link self-ratings of overall happiness and average affect and find an average correlation of +.7011. Another 13 studies relate responses to global questions on overall happiness and contentment and find an average correlation of +.4612. Not surprisingly, the correlation between hedonic level and contentment is weaker. The average in three studies is +.4013. An even lower correlation was observed in the recent Gallup World Poll, the correlation between Best–Worst possible life and Yesterday’s Affect being around +.20 (Harter & Arora, 2009). Happy with unfulfilled aspirations. If happiness depends on seeing one’s wants met, people must be unhappy when they have unfulfilled aspirations and increasingly unhappy the more unfulfilled aspirations they have. Yet people with unfulfilled aspirations appear to be happier than people without, and more so the more unfulfilled aspirations they have (Wessman, 1965, p. 210)14. This finding fits better with the theory that we have an innate need to use our potentials, since unfulfilled aspirations provide an aim to achieve.
How Universal Is Happiness?
335
Happy in spite of value–reality gap in nation. If contentment drives happiness in the first place, we can expect that people are happier in nations where the values endorsed are perceived to be met than in nations where a gap between value and reality is perceived to exist. This is not always the case; for instance, not with ‘‘gender equality’’ and ‘‘human orientation’’ as measured in the Globe study in 62 societies (House et al., 2004). Average happiness is higher in nations where the widest gaps between ideal and reality are perceived to exist on these issues, probably because this marks respect for humanistic values. Happiness drives contentment rather than the reverse. The right arrow in Table 11.2 denotes a ‘‘bottom-up’’ effect of contentment on overall happiness. Above, I have interpreted the observed correlations in this way. Yet causality can also be ‘‘top-down,’’ overall happiness affecting the perception of the gap between what one wants and what one has. Analysis of a panel study has shown that causality typically works this way. In this study, discrepancies (gaps) were assessed between how respondents rated their present life on a 20-step scale and ratings of what they wanted from life (expectations, aspirations, entitlements) on the same ladder scale.
TABLE 11.3
Theories of happiness: significance of two causal paths
Source: Veenhoven, 2009.
336
Section III: Differences in the Social Context of Well-Being
Comparison over time showed a significant top-down effect but no bottomup effect (Headey & Veenhoven, 1989, p. 117). So it seems that contentment is largely driven by happiness. If we feel good, we infer that we have most of the things we want, and if we feel bad we start looking for what we might miss. Though affect seems to dominate the overall appraisal of life, it does not dominate equally everywhere. Correlations between overall happiness and affect balance tend to be stronger in individualistic nations than in collectivist ones (Suh et al., 1998). Likewise, the relative weight of positive and negative affect differs somewhat across cultures. Negative affect is more strongly correlated to overall happiness in individualistic nations than in collectivist ones, while positive affect correlates more with overall happiness in nations where self-expression values are endorsed than in nations where the focus is more on survival (Kuppens et al., 2008).
Question 3: Are Conditions for Happiness Similar Across Cultures?
Do we need the same conditions to be happy? Or can some people be happy in conditions that render other people unhappy? Below, I will consider this question on two levels: the macro level of nations and at the micro level of individuals within nations.
Much uniformity in societal requirements for happiness
Average happiness differs markedly across nations: the highest average on a 0 to 10 scale is currently observed in Denmark (8.4) and the lowest in Zimbabwe (3.2)15. There is a clear system in these differences. People live more happily in the most modern nations, in particular in nations characterized by economic development, freedom, rule of law and good governance. The societal characteristics set out in Table 11.4 explain no fewer than 75 percent of the differences in average happiness in nations16. Societal progress in these matters is likely to have fostered the recent rise of happiness in modern nations (e.g., Inglehart et al., 2008). Interestingly, the societal conditions that make people happy are not always the conditions they value. For instance, average happiness is markedly lower in nations where women are discriminated against (ChinHonFoei, 2007), but this practice is widely approved in most of
How Universal Is Happiness?
TABLE 11.4
337
Average happiness by societal characteristics in 136 nations 2006 Zero order
Wealth Income p/c Freedom economic freedom political freedom Peace Justice corruption rule of law Equality income equality gender equality Education school enrolment intelligence
wealth controlled
+.79
–
+.62 +.50 +.39
+.11 +.07 +.15
.77 +.70
.14 +.06
+.27 +.67
.33 +.19
+.57 +.63
+.12 +.21
Source: World Database of Happiness States of Nations 40
these countries. Likewise, corruption brings down happiness even in societies where favoritism is seen as a moral obligation.
Much uniformity in required living conditions within nations
There are also differences in individual happiness within nations. In a happy country like Denmark, 5 percent of the people still rate 5 or lower on the 0–10 scale, and in an unhappy country like Zimbabwe, some 13 percent score 8 or higher. Are the reasons for high and low scores similar across nations? Below, I consider some living conditions for which crossnational data are available. Freedom
Not only is average happiness higher in free countries, but within countries individuals are also happier the more control they have over their life. This appears, among other things, in strong correlations between personal happiness and perceived freedom and control all over the world17. Social rank
People are typically happier on the upper steps of the social ladder than at the bottom. This appears in findings on relative income position18,
338
Section III: Differences in the Social Context of Well-Being
occupational prestige19, subjective class identification20 and indexes of socio-economic status21. The differences tend to be bigger at the lower end of the hierarchy. Though the correlations with happiness differ in size, they are positive all over the world. This finding fits the view that we have an innate need for social respect. Like other group animals, we are hardwired to avoid a bottom position. Marriage
Adults are typically happier when living with a spouse than when single. The difference is around half a point on scale from 0–10 and is largely independent of income, gender, and age. Again the size of the difference varies somewhat across time and nations, but the pattern is clearly universal22. This finding fits the view that we are social animals, hardwired to form pairs. Personality
Cross-national research on the relationship between happiness and personality is limited as yet, but the available data suggest that extroverted people tend to be happier23 across a variety of nations (Lucas et al., 2000) and that neurotics tend to be less happy in all cultures. Once more, there is difference in the size of the effects. For instance, the effect of self-esteem appears to be stronger in individualistic cultures than in collectivist cultures (Oishi et al., 1999). Still the direction is the same everywhere. This is not to say that all conditions for happiness are universal. One notable exception is ‘‘education.’’ Although there is a correlation between average happiness and level of education in countries, the most highly educated individuals are not always happier. Correlation between happiness and education vary between –.08 and +.2724.
Question 4: Are the Consequences of Happiness Similar?
Research into happiness has focused on its determinants in the first place; however, there is also a strand of investigation into the consequences of enjoying life or not (Veenhoven, 1989a; Lyubomirsky et al., 2005). Fredrickson (2004) has summarized much of the findings in the ‘‘broaden and build theory’’ of positive affect. Although most of this research has been done in Western nations, the observed effects are also likely to exist in other parts of the world.
How Universal Is Happiness?
339
Happiness fosters functioning
Happiness appears to encourage engagement, while unhappiness tends to instigate withdrawal. This appears as greater engagement in activity at work25 and in leisure26. The energizing effect of happiness manifests also in social behavior: happiness predicts the formation of friendships27, entering marriage28and participation in voluntary organizations29. There is also experimental evidence of happy moods’ broadening perception and enhancing creativity30. All this is compatible with the above-mentioned theory that happiness works as a ‘‘go signal’’, and that this effect seems to exists also in other higher animals. If so, the effect is likely to be universal.
Happiness lengthens life
Another illustrative finding is that happiness fosters physical health31 and that happiness therefore lengthens life considerably32. One of the mechanisms seems to be that happiness encourages the full functioning of mind and body and thus keeps us in shape. Another mechanism is probably that unhappiness triggers the fight or flight response, since it signals that there is something wrong. It is well known that this automatic reaction makes an organism economize on other functions, among them the immune response. In this line, Cohen (1995) has demonstrated experimentally that unhappiness makes people more susceptible to catching a common cold. The above are essentially biological reactions that are unlikely to differ much across cultures. Possibly there are effects of happiness that do differ across cultures, but for the time being, it is the universality strikes the eye.
Question 5: Do We All Seek Happiness?
It is rather evident that most humans prefer a happy life to an unhappy one. Still, this does not mean that happiness is the main driver in human motivation, nor that happiness is valued universally.
Happiness is a universal human striving, though not innate
In the first lines of his famous Principles of Morals and Legislation, Jeremy Bentham (1789) stated that human behavior is governed by the pursuit of pleasure and the avoidance of pain. There is much truth in this theory of motivation, yet happiness is not the only driver of human behavior, at least not happiness in the sense of overall life-satisfaction.
340
Section III: Differences in the Social Context of Well-Being
Like other mammals, we are driven by different needs, such as hunger, sex, love, and curiosity, which have developed subsequently through evolution. All these needs are linked to hedonic signals, but their gratification is not only sought for the sake of pleasure. For instance, we want sex for the sex, and we do not settle for the esthetic pleasure of reading poetry instead. Still, pleasure is a main driver of human behavior. Yet pleasure is not quite the same as life-satisfaction. Short-term pleasures can be at the cost of long-term happiness, and in such cases humans do not always look to the long term. Our fellow animals are driven by primary motives, but in humans consciousness also gives rise to secondary motives, such as figuring out who we are and seeking an answer to questions about the meaning of life. Wentholt (1980) calls this ‘‘universal strivings’’, which he distinguishes from ‘‘organic needs’’. The pursuit of long-term happiness is one of these universal strivings. Though not ‘‘innate’’ as such, it is an inclination that develops in most humans as a result of their consciousness. While this inclination manifests in all cultures, it does not necessarily appear in all individuals. Happiness is typically not an issue for people who are trying to survive in the first place, and some opt to forsake happiness for ideological reasons.
Happiness is valued in most societies, though possibly not in all
Happiness seems to be positively valued in all nations of our time. This is at least suggested by a study among university students in 47 nations in the 1990s (Diener, 2004). These students were asked to rate the importance of several values, such as wealth, health, and love. Happiness ranked highest in the importance rating, with an average of 8 on a scale of 1 to 9. Ratings ranged between 8.7 and 7.3, and there is thus no country in this study where happiness is deemed unimportant. This is not to say that happiness has always been prized in all human cultures. Though all humans have a natural inclination to pursue happiness, cultures can go against that inclination just as some cultures go against the natural drive for sex. What cultures denounce happiness? Unfortunately cultural anthropology cannot tell us, since this discipline has a blind eye for happiness (Thin, 2006). Still, there are indications that in the past, miserable societies tended to glorify suffering rather than happiness33, and that collectivist cultures emphasize the well-being of groups rather than the well-being of individuals. One can think of reasons why cultures come to depreciate happiness. When life is miserable, it may be comforting to believe that happiness is no
How Universal Is Happiness?
341
good after all, and renouncing happiness may be functional for engaging people in common causes such as war. Next to such macro-societal functions, internal factors can be involved, such as cultural distinctions between groups in a society. This seems to have been one of the reasons for the sexual abstention of religious people during the Middle Ages. The campaign ‘‘against happiness’’ of some present-day philosophers34 could be placed in a similar vein as intellectual ‘‘distinction’’.
Question 6: Do We Seek Happiness in a Similar Way?
To the extent that they seek happiness, do humans do this in similar ways? This question is easier put than answered, since behavior is typically guided by multiple motives that cannot be observed as such. Still we can get a clue by looking at beliefs about conditions for happiness. Do universal themes dominate these beliefs or do these lay theories of happiness tend to be culture-specific? This question can be answered in principle, but a shortage of data sets limits.
Variation in philosophical views gives no answer
Can we answer this question on the basis of what prominent thinkers have said on this subject in different times and cultures? We can draw on a large philosophical literature about ways to lead a happy life35, and reviews of this literature show wide differences. Yet even though philosophers differ so much, public opinion need not be equally diverse. Philosophers often seek the difference from common opinion. Moreover, philosophers often use the term happiness in a broader sense than defined here and typically write about how we should seek happiness rather than how we actually do. So, to answer this question we need to look at survey research into beliefs about happiness.
Survey research shows much similarity
As yet there is more comparative research on degrees of happiness around the world than on beliefs about conditions for happiness. The available research on presumed conditions for happiness is limited to present-day nation states. These data suggest that there is quite some similarity across nations. Most of the available research findings are about perceived sources of one’s own happiness36. The bulk concerns modern Western people, who
342
Section III: Differences in the Social Context of Well-Being
tend to believe that happiness depends on health and good relations in the first place, and less so on material affluence and social prestige. Only a few studies have compared cultures. Chiasson & Dube´ (1997) found striking similarities in North America and Latin American countries. Likewise, Lee et al. (1999) found that students in Canada and Korea employed essentially the same ordered set of perceived sources for happiness, in spite of considerable difference in average level of happiness. A recent Gallup World Survey found that a happy family life and good health are ranked high all over the world (see Table 11.5). Belief about conditions for happiness was addressed indirectly in Cantril’s (1965) landmark study, ‘‘The Pattern of Human Concerns’’ in 16 nations in the early 1960s. This study involved open-ended questions about what constitutes the ‘‘best possible life’’ and the ‘‘worst possible life.’’ Analysis of the responses showed much similarity; the same themes popped up in all countries, though not equally frequent everywhere (pp. 162–167). The observed differences appeared to correspond more with the country’s stage of societal development than with its ideology (p. 302). Cantril explained these observations in terms of need theory (Chapter 16).
TABLE 11.5
Perceived sources of happiness
Source: Gallup Millennium World Survey (Spoga´rd 2005)
How Universal Is Happiness?
343
Studies by Tsai et al. (2007) address differences in both ideal and actual affect across cultures and suggest that culture influences ideal affect more than actual affect. Still, a look at their findings shows much similarity in ideal affect, and in particular, similarly high ratings for the desirability of happiness.
Question 7: Are We About Equally Happy in All Cultures?
The last question is about degree of happiness. Are humans about equally happy in all cultures, or do they live happily in some cultures and unhappily in others? Some variants of the comparison theory of happiness imply that humans are about equally happy in all cultures. If happiness depends on social comparison with one’s compatriots, the average level will be about neutral in all societies. If happiness depends on comparison with earlier experiences over the lifetime, the average will tend to be neutral as well. Yet other variants of comparison theory imply that happiness can differ across cultures. If happiness depends on meeting local standards of the good life, happiness can be high in cultures where these standards can be easily met and low where the meeting of these standards is out of reach for most people. Need-theory also implies that the level of happiness can differ across cultures. If happiness depends on the degree to which human needs are met, average happiness will be higher in societies that fit human nature well than in societies that do not. From a functional point of view, it is unlikely that we are equally happy irrespective of conditions. At best, evolution has resulted in a tendency to feel happy once conditions are tolerable. The data are quite clear on this matter. There are wide differences in average happiness across nations (see Table 11.6). Average happiness37 is 8.2 in Denmark and only 3.2 in Zimbabwe. Average happiness is above neutral in the present-day world38. As we have seen in Table 11.4, most of these differences can be explained by national characteristics such as wealth, freedom, and security, which are part of ‘‘modernity’’. Average happiness not only differs among contemporary cultures, but also varies over time. The level of happiness has risen in most nations over the last 30 years, but declined in some (Veenhoven & Hagerty, 2006; Inglehart, 2008). Average happiness fell dramatically in the Russia and China following the great social changes that have taken place since the 1980s (Brockmann et al., 2008). Though people live more happily in modern society, the change to modernity tends to reduce happiness temporarily.
344
Section III: Differences in the Social Context of Well-Being
TABLE 11.6 Happiness in nations around 2000, averages on scale 0–10 Denmark Sweden USA Germany France Philippines Japan Turkey Russia Zimbabwe Average in 86 nations
8.2 7.7 7.4 7.2 6.5 6.4 6.2 5.2 4.4 3.3 6.5
Source: World Database of Happiness, Happiness in Nations (Veenhoven 2008)
There are also indications that average happiness has varied considerably over human history. Our forefathers seem to have been fairly happy when living as hunter-gatherers, but less so in the agrarian phase of societal evolution. The industrial revolution brought not only more material comfort, but also an unprecedented rise in happy life years (Veenhoven, 2005). Discussion Limitations
In his book The Pattern of Human Concerns, Headley Cantril (1965, p. 315) notes that ‘‘differences between individuals and groups are often easier to detect than the similarities they obscure.’’ An illustrative case is eating; cultural differences in eating behavior catch the eye: for example, when you find snake on your plate at a business dinner in China. Yet all humans want to eat, do eat, and need about the same nutrients. It is difficult to express this universality in numbers, since it depends rather on an interpretation of what is most essential. As such, my argument may not convince everybody. A more tangible limitation is in the data used in this chapter. Since anthropology has failed to inform us about happiness in pre-modern societies (Thin, 2007), we must largely make do with data gathered in more or less modern societies, in particular in rich Western societies. The few studies of pre-modern societies I have mentioned (Biswas Diener, 2005; Kilpatric & Cantril, 1960) do not cover all the questions addressed here. Furthermore, the data are not free from cultural measurement bias. For instance: Latin Americans value positive affect more and may for that
How Universal Is Happiness?
345
reason report more positive affect than Asians do (Diener & Oishi, 2004). Lastly, the question about variation across cultures is largely answered using data on happiness in nations. Why is the idea of cultural relativity so popular?
In spite of these limitations, it is pretty clear that happiness is not only in the minds of Western people, and that there is a striking similarity in conditions for happiness across cultures. This elicits the question of why so many social scientists believe that happiness is culture-specific. One answer to that question is that theory plays them false; most social scientists have been raised with the idea that human experience is socially constructed and are trained to see human behavior as guided by malleable preferences. Another answer is a moral aversion to utilitarianism that gives rise to discounting the significance of happiness. I have discussed this question in more detail in Veenhoven (2006).
Conclusion
The available data suggest that all humans tend to appraise how much they like the life they live. In appraising life we draw on how well we feel in the first place, which in its turn draws on how well our universal human needs are gratified. The overall appraisal of one’s life draws less on cognitive comparison with cultural standards of the good life. Consequently, conditions for happiness appear to be quite similar across the world. The consequences of enjoying life are also largely universal. There is more cultural variation in the valuation of happiness and in beliefs about conditions for happiness. The greatest variation is found in how happy people are.
Notes 1. This is not to say that appraisals of beauty are entirely culturally specific, since there is good evidence of universal tendencies on this matter. 2. In other chapters of this book the term subjective well-being is used for this concept, while the terms happiness and life-satisfaction denote survey questions using these words. I cannot follow that terminology since this chapter draws on the World Database of Happiness, which is organized on the basis of a different idiom. 3. There is some cultural variation in recognition of ‘‘fear,’’ ‘‘anger’’ and ‘‘disgust’’ in facial expressions (see, e.g., Russell, 1994). Still, the pattern is largely universal, and recognition of ‘‘happy’’ emotion stands out as the most universal. Expression of emotion may be more universal than the recognition of it in other people.
346
Section III: Differences in the Social Context of Well-Being
4. World Database of Happiness, Correlational Findings, subject section ‘‘Current Happiness’’ (H6.2.2.2), one study that observed a correlation of +.56. 5. World Database of Happiness, Correlational Findings, subject section ‘‘Current Happiness’’ (H6.2.2.3), one study. 6. World Database of Happiness, Correlational Findings, subject section ‘‘Reputation of Happiness’’ (H8.2), four studies that observed correlations between +.43 and +.64. 7. World Database of Happiness, States of Nations, Variable Happiness_DKLS_ 1980_2000. Average percent ‘‘don’t know’’ in 76 nations: 0.75; range, 0.4 to 4.5. 8. Analysis of non-response to the questions on happiness and life-satisfaction in the World Values Survey shows some variation across cultures. The percentage of ‘‘don’t know’’ responses is slightly higher in nations where unhappiness prevails (r ¼ +.10) and also higher in collectivist cultures (r ¼ +.21). Still, these are variations on an otherwise universal pattern. 9. The distinction between deficiency needs and growth needs is part of Maslow’s (1970) theory of human motivation. 10. Rayo & Becker (2007) have a different view and argue that we are hardwired to compare and prefer the best, since this is evolutionarily advantageous. Their argument is appealing. Still, more is not always better and can even be detrimental. The tendency to compare can also be explained on the basis of innate needs that are not exclusively human, such as the need for social status and the need to use and develop one’s potentials. I do buy that we tend to see things in a comparative perspective, but see that rather as a consequence of cognition than as a ‘‘need’’ in itself. The distinction between consequences of human consciousness and innate ‘‘needs’’ is further discussed in the response to Question 5. 11. This analysis involved eight studies, the results of which are summarized in the World Database of Happiness, Collection of Correlational Findings, Section H6.1.2, ‘‘Current Happiness: Overall Happiness by Hedonic Level of Affect.’’ The analysis is limited to studies among general population samples using comparable single, direct questions on overall happiness (type O-HL, O-SL, O-DT, OQOL) and affect (type A-AOL). Correlation with affect balance (type A-AB) is lower (+.50 in 70 studies), but in this case the correlation is depressed by the timeframe of the questions, which is typically ‘the last few weeks’. 12. This analysis involved seven studies, the results of which are summarized in the World Database of Happiness, Collection of Correlational Findings, Section H6.1.3, ‘‘Current Happiness: Overall Happiness by Contentment.’’ The analysis is limited to studies among general population samples using comparable single, direct questions on overall happiness (type O-HL, O-SL, O-DT) and contentment (type C-BW). 13. This analysis involved three studies, the results of which are summarized in the World Database of Happiness, Collection of Correlational Findings, Section H6.23 ‘‘Current Happiness: Hedonic Level of Affect by Contentment.’’ The analysis is limited to studies among general population samples. Hedonic level was measured using affect balance scales (type A-AB) and contentment using the Cantril Ladder (type C-BW) and questions about perceived realization of wants (type C-RW). 14. Wessman wrongly interpreted Table 44 as showing that unfulfilled aspirations go with unhappiness. 15. World Database of Happiness, Happiness in Nations, Rank Report 2009-1.
How Universal Is Happiness?
347
16. Similar results are obtained when average happiness is measured using an affect balance scale instead of a single direct question on satisfaction with life-as-a-whole (Variable in data file ‘‘states of Nations’’: HappinessYesterdaysAffect3_2006). 17. World Database of Happiness, Correlational Findings, subject section ‘‘Personality, Inner Locus of Control (P4.58). For a recent cross-national study, see Verne (2008). 18. World Database of Happiness, Correlational Findings, subject section ‘‘Current Income’’ (1.2). For a recent cross-national comparison, see Ball and Chernova (2008). 19. World Database of Happiness, Correlational Findings, subject section ‘‘Current Occupational Level’’ (O1.3.1). 20. World Database of Happiness, Correlational Findings, subject section ‘‘Subjective Social Rank’’ (S9.2.2). 21. World Database of Happiness, Correlational Findings, subject section ‘‘Objective Social Status’’ (S9.2.1). 22. World Database of Happiness, Correlational Findings, subject ‘‘Marital Status’’ (M2.1). See also Diener et al. (2000). 23. Schimmack et al. (2002) found that the link between extroversion and hedonic level of affect is more universal than the link between extroversion and overall happiness and suggest that the influence of personality on the emotional component of happiness is pan-cultural, whereas the influence of personality on the cognitive component of happiness is more moderated by culture. 24. World Database of Happiness, States of Nations, variable r_LS_Education_1990. 25. World Database of Happiness, Correlational Findings, subject ‘‘Later Work Performance’’(code W6.1.4). 26. World Database of Happiness, Correlational Findings, subject ‘‘Later Leisure’’(code L3.1.4). 27. World Database of Happiness, Correlational Findings, subject ‘‘Later Friendships’’ (code F6.1.4). 28. World Database of Happiness, Correlational Findings, subject ‘‘Later Marriage’’ (code M1.4). See also Veenhoven (1989b). 29. World Database of Happiness, Correlational Findings, subject ‘‘Later Organizational Participation’’(code S7.1.4). 30. See, e.g., Isen (1998). 31. World Database of Happiness, Correlational Findings, subject ‘‘Later Physical Health’’ (code P6.1.4). 32. World Database of happiness, Correlational Findings, subject ‘‘Longevity’’ (code P6.1.4.1). See also Veenhoven (2008b). 33. This appears, for instance, in the history of philosophical thought. Happiness was a common theme in the prosperous Antique city societies, but disappeared in the dark Middle Ages and popped up again in the seventeenth century, together with a rise in quality of life of the new middle class. 34. See, e.g., Bruckner (2000), and Wilson (2008). 35. World Database of Happiness, Bibliography of Happiness, subject ‘‘Perceived Sources of Happiness’’ (code 15a2). For a recent review, see MacMahon (2008). 36. World Database of Happiness, Bibliography of Happiness, subject ‘‘Perceived Sources of One’s Own Happiness (code Rco1). 37. Overall happiness measured with a survey question on life-satisfaction. Average contentment, as measured with Cantril’s Best-Worst Possible Life question, is closer to neutral, which fits the prediction of comparison theory.
348
Section III: Differences in the Social Context of Well-Being
38. World Database of Happiness, Happiness in Nations, Rank Report 2009-1. See also Diener, 1996. 40. The data file ‘‘States of Nations’’ is available on request. The variables used here are: HappinessBWLS11_2006, RGDP_2000_2004, FreeEconIndex2_2002, DemocracyIndex2_2004, PeaceIndex_2007, Corruption3_2006, RuleLaw_2006, IncomeInequality1_2005, GenderEqualIndex2_2005, EduEnrolGross_2000_04, IQ_2006. These variables are described on the website: http://www.worlddatabaseofhappiness.eur.nl/statnat/statnat_fp.htm
REFERENCES Ball, R., & Chernova, K. (2008). Absolute income, relative income and happiness. Social Indicators Research, 8, 497–529. Bentham, J. (1789). Introduction to the Principles of Morals and Legislation. Original edition, London: Payne. Reprinted by Althone Press, London, 1970. Biswas-Diener, R., Diener, E., & Vitterso, J. (2005). Most people are pretty happy, but there is cultural variation: The Inughuit, the Amish and the Maasai. Journal of Happiness Studies, 6, 205–226. Brockmann, H., Delhey, J., Welzel, C., & Yuan, H. (2008). The China puzzle: falling happiness in a rising economy. Journal of Happiness Studies, 10, 387–405. Bruckner, P. (2000). L’Euphorie perpetuelle: essai sur le devoir de bonheur. Grasset, France. Cantril, H. (1965). The pattern of human concern. New Jersey: Rutgers University Press. Chambers, J. C. (1999). Situating psychological well-being: Exploring cultural roots of its theory and research. Journal of Counseling and Development, 77, 141–152. Chiasson, N., & Dube´, L. (1997). La Conception du bonheur: similitudes et differences interculturelles. Revue que´be´coise de psychologie,18, 121–144. ChinHonFoei, S. S. (2007). Gender equality and happiness in nations. Paper presented at Dag van de Sociologie, Tilburg, Netherlands, June 8th 2006. Cohen, S., Doyle, W. J., Skomer, D. P., Fireman, P., Gwaltney, J. M., & Newson, J. P. (1995). State and trait negative affect as predictors of subjective and objective symptoms of respiratory viral infections. Journal of Personality and Social Psychology, 68, 159–169. Damasio, A. (1994). Descartes’ error. New York: Putman. Davidson, R. J. (2004). Well-being and affective style: Neural substrates and biobehavioural correlates philosophical transactions. Biological Sciences, 359, 1395–1341. Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with life scale. Journal of Personality Assessment, 49, 71–75. Diener, E. (1996). Most people are happy. Psychological science, 7, 181–185. Diener, E., Gohm, C. L., Oishi, S., & Suh, E. M. (2000). Similarity of the relation between marital status and subjective well-being across cultures. Journal of CrossCultural Psychology, 31, 419–436. Diener, E., & Oishi, S. (2004). Are Scandinavians happier than Asians? Issues in comparing nations on subjective well-being. Asian Economic and Political Issues, 10, 1–25. Ekman, P. (1970). Universal facial expression of emotion. Californian Mental Health Research Digest, 8, 151–158. Fredrickson, B. L. (2004). The Broaden-and-Build Theory of Positive Emotions. Philosophical Transactions, Biological Sciences, 359, 1367–1377.
How Universal Is Happiness?
349
Harter J. K., & Arora, R. (2009). The impact of time spent working and job-fit on wellbeing around the world. In E. Diener, D. Kahneman, and J. F. Helliwell (Eds.), International differences in well-being. Oxford, UK: Oxford University Press. Headey, B., & Veenhoven, R. (1989). Does happiness induce a rosy outlook? In R. Veenhoven (Ed.), How harmful is happiness? Consequences of enjoying life or not (pp. 106–127). Rotterdam: University Press. Heady, B., Veenhoven, R., & Wearing, A. (1991). Top-down versus bottom-up: Theories of subjective well-being. Social Indicators Research, 24, 81–100. House, R. J., Hanges, P. J. Javadin, M., Dorfman, P. W., & Gupta, V. (2004). Culture, leadership and organizations. The GLOBE study of 62 societies. London: Sage. Inglehart, R., Foa, R., Peterson, C., & Wetzel, C. (2008). Development, freedom and rising happiness: A global perspective: 1981–2007. Perspectives on Psychological Science, 3, 268–285. Isen, A. M. Doubman, K. A., & Nowiki, G. P. (1998). Positive affect facilitates creative problem solving. In: J. M. Jenkins, K. Oatley, & N. L. Stein (Eds.). Human emotions: A reader (pp. 288–297). Malden, MA: Blackwell Publishers. Kilpatrick, F. P., Cantril, H. (1960). Self-anchoring scaling: A measure of individuals’ unique reality worlds. Journal of Individual Psychology, 16, 158–173. Kuppens, P., Realo, A., & Diener, E. (2008). The role of positive and negative emotions in life-satisfaction judgment across nations. Journal of Personality and Social Psychology, 95, 66–75. Lee, D. Y., Park, S. H. Uhleman, M. R., & Patsula, P. (1999). What makes you happy? A comparison of self-reported criteria between two cultures. Social Indicators Research, 50, 351–362. Lucas, R. E., Diener, E., Grob, A., Suh, E. M., & Shao, L. (2000). Cross-cultural evidence for the fundamental features of extraversion. Journal of Personality and Social Psychology, 79, 452–468. Lyubomirsky, S., Diener, E., & King, L. A. (2005). The benefits of frequent positive affect: Does happiness lead to success? Psychological Bulletin, 131, 803–855. Maslow, A. H. (1970). Motivation and personality. New York: Harper & Row. Michalos, A. C. (1991). Global report on student well-being. Vol. 1: Life satisfaction and happiness. New York: Springer. McMahon, D. M. (2008). The pursuit of happiness in history. In: M. Eid & R. J. Larsen (Eds.), The science of subjective well-being (pp. 80–93). New York: The Guilford Press. Oishi, S. (2010). Culture and well-being: Conceptual and methodological issues. In E. Diener, D. Kahneman, & J. F. Helliwell (Eds.), International differences in wellbeing. Oxford, UK: Oxford University Press. Oishi, S., Diener, E., Lucas, R. E., & Suh, E. (1999). Cross-cultural variation in predictors of life-satisfaction: Perspectives from needs and values. Personality and Social Psychology Bulletin, 25, 980–990. Rayo, L., & Becker, G. S. (2007). Evolutionary efficiency and happiness. Journal of Political Economy, 115, 302–337 Russell, J. A. (1994). Is there universal recognition of emotion from facial expression? A review of cross-cultural studies. Psychological Bulletin, 115, 102–141. Spoga´rd, R. (2005). World opinion: A happy family and good health is what matters most in life. Retrieved on June 6th 2009 from http://www.gallup-international.com/ ContentFiles/millenium16.asp.
350
Section III: Differences in the Social Context of Well-Being
Schimmack, U., Radharkrishan, P., Oishi, S. Dzokoto, V., & Ahadi, S. (2002). Culture, personality and subjective well-being: Integrating process model of life satisfaction. Journal of Personality and Social Psychology, 82, 582–593. Suh, M. E., Diener, E., Oishi, S., & Triandis, H. C. (1998). The shifting basis of life satisfaction judgments across cultures: Emotion versus norms. Journal of Personality and Social Psychology, 74, 482–493. Thin, N. (2005). Happiness and the sad topics of anthropology. WeD Working Paper 10 ESCR. Research Group on Well-Being in Developing Countries. Thin, N. (2008). ‘‘Realising the substance of their happiness’’: How anthropology forgot about Homo gauisus. In A. Corsin Jimenez (Ed.), Culture and well-being (pp. 134–155). London: Pluto Press. Tsai, J., Knutson., & Fung, H. H. (2006). Cultural variation in affect valuation. Journal of Personality and Social Psychology, 90, 288–307. Veenhoven, R. (1984). Conditions of happiness. Dordrecht, Netherlands: Reidel. Veenhoven, R. (1989a). How harmful is happiness? Consequences of enjoying life or not. Rotterdam, Netherlands: University Press. Veenhoven, R. (1989b). Does happiness bind? Marriage chances of the unhappy. In R. Veenhoven (Ed.), How harmful is happiness? Consequences of enjoying life or not. Rotterdam, Netherlands: University Press. Veenhoven, R. (1991). Is happiness relative? Social Indicators Research, 24, 1–34. Veenhoven, R. (2000). The four qualities of life: Ordering concepts and measures of the good life. Journal of Happiness Studies, 1, 1–39. Veenhoven, R. (2005). Is life getting better? How long and happily do people live in modern society? European Psychologist (special section, ‘‘Human development and well-being’’), 10, 330–343. Veenhoven, R. (2006). Sociology’s blind eye for happiness. Paper presented at the 16th World Congress of Sociology. Durban, South Africa. Available at: http://www2.eur.nl/ fsw/research/veenhoven/Pub2000s/2006g-full.pdf. Accessed June 6th 2009. Veenhoven, R. (2008). World Database of Happiness: Continuous register of scientific research on subjective enjoyment of life. Retrieved on June 6th 2009 from http:// worlddatabaseofhappiness.eur.nl. Veenhoven, R. (2008b). Healthy happiness: Effects of happiness on physical health and the consequences for preventive health care. Journal of Happiness Studies, 9, 449–464. Veenhoven, R. (2008c). Comparability of happiness across nations. School of Sociology and Social Work Journal, 104, 211–234. Veenhoven, R. (2009). How do we appraise how happy we are? In A. Dutt & B. Radcliff (Eds.), Happiness, economics and politics. Cheltenham, UK: Edward Elgar Publishers (in press). Veenhoven, R., & Hagerty, M. (2006). Rising happiness in nations, 1946–2004. A reply to Easterlin. Social Indicators Research, 79, 421–436. Verne, P. (2008). Happiness, freedom and control. Journal of Economic Behaviour and Organization, 71: 146–161 Wilson, E. G. (2008). Against happiness: In praise of melancholy. New York: Farrar, Straus & Giroux. Wentholt, R. (1980). Map motivatieleer. Erasmus Universiteit Rotterdam. Wessman, A. E. (1956). A psychological inquiry into satisfaction and happiness. Unpublished doctoral dissertation. Princeton University, New Jersey. Zajonc, R. B. (1984), On the primacy of affect. American Psychologist, 39, 117–123.
Chapter 12 Faith and Freedom: Traditional and Modern Ways to Happiness Ronald F. Inglehart Department of Political Science, University of Michigan, Ann Arbor
The author gratefully acknowledges support from the National Science Foundation, and the foreign ministries of Sweden and The Netherlands, each of which supported fieldwork in several countries in the 2005–2007 wave of the World Values Survey.
Two main routes to happiness exist, one linked with modernization and another with belief systems. As Sen (2001) has argued, the underlying theme of development is that it increases people’s freedom of choice. Insofar as modernization brings greater income, and political and personal freedom, it increases freedom of choice—which is universally linked with subjective well-being (Inglehart & Welzel, 2005). In recent decades, modernization has actually made people happier. Although it was widely believed that that neither individuals nor societies can lastingly increase their happiness levels, during the past 25 years, economic development, democratization, and rising social tolerance have led to rising happiness in most countries around the world (Inglehart, Foa, Peterson, & Welzel, 2008). Economic development is conducive to subjective well-being, but it is only one of many causal factors, with social tolerance and political freedom playing even more important roles. Moreover, the impact of rising gross-national-product follows a curve of diminishing returns. Consequently, economic development by itself does not necessarily bring rising subjective well-being. The relationship between development and subjective well-being is further complicated by the fact that belief systems also shape subjective 351
352
Section III: Differences in the Social Context of Well-Being
well-being. Even in the absence of prosperity and freedom, religion seems to enhance subjective well-being. Thus, in most countries, religious people are happier than non-religious people, despite the fact that they tend to have lower incomes. Similarly, although the people of poor countries tend to be (1) less happy and (2) more religious than the people of rich countries, religion can help compensate for poverty, so that the people of low-income countries that are strongly religious are substantially happier than the people of low-income countries that are less religious. The negative correlation between economic development and religiosity tends to conceal the impact of religion on happiness—which only becomes fully apparent when one controls for a society’s level of economic development. But religions are not the only type of belief system that helps shape subjective well-being. Most ex-Communist countries show weak or negative correlations between religion and subjective well-being. This seems to reflect a recent influx of unhappy people who have turned to religion, in the wake of the collapse of faith in Communist ideology, which once provided a sense of meaning and certainty for many people. A strong belief system, whether religious or secular, is linked with relatively high levels of subjective well-being. This chapter will discuss each of these points in greater detail, starting with the relationship between modernization and happiness.
Freedom and Happiness
Until recently, it was widely believed that neither rising prosperity nor severe misfortune permanently affect happiness. After a period of adjustment, individuals return to their baseline levels of well-being, leaving humanity on a ‘‘hedonic treadmill’’ (Brickman & Campbell, 1981; Diener, Suh, Lucas, & Smith, 1999; Kahneman, Krueger, Schkade, Schwartz, & Stone, 2004). Similarly, as entire countries become richer, relative gains and losses neutralize each other, bringing no overall increase in overall happiness (Easterlin, 1974; Kenny, 2004). Thus, Inglehart (1990) demonstrated that life satisfaction levels were very stable from 1973 to 1988 in most West European countries. Diener & Oishi (2000), Inglehart & Klingemann (2000), Easterlin (2005) and Kahneman & Krueger (2006) presented similar findings. And despite considerable economic growth, U.S. subjective well-being levels showed a flat trend from 1946 to the present. Because the happiness levels of given societies do not seem to change over time, the idea that economic development brings rising happiness was widely rejected.
Faith and Freedom: Traditional and Modern Ways to Happiness
353
Recent research questions the view that happiness levels remain constant. Diener, Lucas, & Scollon (2006); Fujita & Diener (2005); and Lucas, Clark, Georgellis, & Diener (2005) found that on average, groups regressed toward their baseline levels of satisfaction, but that significant numbers of individuals remained above their original level, while others remained below it. Individuals are not necessarily trapped on a hedonic treadmill. But what about nations? Social comparison theory holds that the relative gains and losses of different individuals in a given nation cancel each other out, resulting in no significant shifts, upward or downward, for a society as a whole. This hypothesis seems to contradict the fact that cross-sectional comparisons of nations show strong linkages between economic development and subjective well-being. Analyzing data from 24 countries, Inglehart (1990) found a .67 correlation between per capita GNP and life satisfaction. He interpreted this as implying that economic development is conducive to rising happiness. Easterlin (1974), on the other hand, held it to be an enduring paradox that although rich countries show relatively high levels of happiness, economic growth is not conducive to rising happiness. Questioning the Easterlin paradox, Hagerty & Veenhoven (2003) demonstrated that income was positively correlated with happiness in 14 of the 21 nations for which data were available from 1972 to 1994. But Easterlin (2005) argued that their findings were not robust and relied on different measures of happiness administered to different types of samples. Subsequently, Hagerty &Veenhoven (2006) demonstrated statistically significant increases in subjective well-being in four of eight high-income countries, and three of four low-income countries for which a long time series was available, but the evidence did not seem decisive. Recent results from the World Values Surveys provide more conclusive evidence that happiness has increased in most countries. These surveys asked the same questions, in the same format, of representative national samples of respondents in scores of countries from 1981 to 2007, providing longitudinal data that are not undermined by problems of methodological incomparability. How might economic development bring rising happiness? Inglehart (1997) hypothesized that economic development brings a societal-level shift from maximizing economic growth to maximizing subjective wellbeing. This shift is linked with individual-level value changes, from giving top priority to economic and physical security toward giving top priority to self-expression values, which emphasize participation, freedom of expression, and quality of life. Under conditions of scarcity, people focus on survival needs, giving top priority to economic and physical security. Economic development increases people’s sense of existential security, leading them to shift their emphasis from survival values toward self-expression values and free choice, which is a more direct way to maximize happiness and life
354
Section III: Differences in the Social Context of Well-Being
satisfaction. People shift from the pursuit of happiness through economic means, toward the pursuit of happiness by maximizing free choice in all aspects of life. Happiness is linked with one’s sense of freedom across all major cultural zones (Inglehart & Welzel, 2005, p. 140). In many societies, people value free choice as much as they value economic security (Sen, 2001), and emphasis on freedom increases with rising economic security. The fact that people change the way they pursue happiness does not necessarily mean that they will attain it. But, since 1981, these shifts in individual-level values have contributed to societal changes that are conducive to human happiness. Self-expression values have become increasingly widespread around the world, contributing to democratization, growing support for gender equality, and growing acceptance of outgroups such as gays and lesbians (Inglehart & Welzel, 2005). Happiness is strongly related to democracy (Inglehart, 1990; Barro, 1999; Frey & Stutzer, 2000; Inglehart & Klingemann, 2000). Democracy increases people’s free choice, which is conducive to subjective well-being (Haller & Hadler, 2004; Inglehart & Welzel, 2005; Ott, 2001; Veenhoven, 2000; Welsch, 2003). Like democratization, social tolerance broadens the range of choices available to people, also enhancing happiness. Accordingly, Inglehart & Welzel (2005) found that support for gender equality and tolerance of outgroups were strongly linked with happiness—not just because tolerant people are happier, but because living in a tolerant society enhances everyone’s freedom of choice. Intolerant social norms can narrowly restrict people’s life choices, reducing subjective well-being. During the late 1980s and early 1990s, dozens of societies experienced transitions to democracy that enhanced freedom of expression, freedom to travel, and free choice in politics. Moreover, from 1981 to 2007, support for both gender equality and tolerance of outgroups increased substantially in most of the countries monitored by the Values Surveys (Inglehart & Welzel, 2005). Furthermore, during the past two decades, low-income countries containing fully half of the world’s population have experienced the highest rates of economic growth in history, allowing them to emerge from subsistence-level poverty. These changes have increased the prosperity of people in less-prosperous societies and the political and social freedom of people in middle-income and high-income societies, enhancing the extent to which people in both types of societies have free choice in how to live their lives, bringing rising levels of happiness in most societies. Inglehart et al. (2008) examined the happiness levels of nations from 1981 to 2007, using data from the World Values Survey and European Values Study, which have carried out five waves of surveys in scores of
Faith and Freedom: Traditional and Modern Ways to Happiness
355
countries containing almost 90 percent of the world’s population1. All five waves of surveys included two widely used indicators of subjective wellbeing—(a) happiness and (b) overall life satisfaction—administered in the same format in equivalent translations. The longitudinal analyses presented here are based on data from the 52 nations for which substantial time series data are available. Appendix 12.A shows the year of fieldwork for each of these 194 surveys, and the levels of happiness, life satisfaction, and the subjective well-being index score from that survey. Life satisfaction was assessed by asking respondents to indicate how satisfied they were with their life as a whole, using a scale that ranged from one [not at all satisfied] to ten [very satisfied]. Happiness was assessed by asking respondents to indicate how happy they were, using four categories: very happy; rather happy; not very happy; and not at all happy. These items are sensitive indicators of a broad, subjective well-being dimension (Andrews & Withey, 1976), capturing most of the common variance in scores of domain-specific indicators. To test whether happiness levels have risen, Inglehart et al. (2008) constructed a subjective well-being (SWB) index using these two indicators.2 The index provides a broader-based indicator of the subjective wellbeing levels of given societies than either of its two components: because it reflects both the affective and cognitive components of subjective wellbeing, it provides a relatively reliable measure of well-being that is less likely to be influenced by idiosyncratic factors than a more narrowly focused measure. They examined trends of this indicator and its two components in 52 societies, using ordinary least squares (OLS) panel regression analysis to analyze societal-level effects, and Hierarchical Linear Modeling (HLM) regression analysis to test the interaction of individual-level and societal-level effects, using the SWB index and its two components as dependent variables, and measures of the feeling that one has free choice, and related attitudes, as independent variables— controlling for democratization and growth in GDP per capita. They found that economic development is linked with rising levels of subjective well-being—but also that economic factors are only part of the story. Pooled time series regression analysis suggests that religion, tolerance of outgroups, and a society’s level of democracy are strong predictors of subsequent levels of subjective well-being, controlling for economic development and a society’s initial level of subjective well-being. Analysis of changes over time suggest that all of the foregoing factors influence subjective well-being mainly insofar as they give people a wider range of free choice. The years from 1981 to 2007 were a period of global economic growth, widespread democratization, and rising social tolerance, which implies that subjective well-being should have increased. The strong
356
Section III: Differences in the Social Context of Well-Being
version of the hedonic treadmill model would dismiss these facts as irrelevant: this model attributes any difference between the subjective well-being levels of rich and poor countries to fixed cultural differences in the meaning of happiness, rather than to differences in prosperity. The low-ranking countries have always been low and will remain so. But they have not. Data from the Values Surveys show that during the past two decades, the SWB index rose in fully 77 percent of the countries for which a substantial time series is available.3 These surveys now provide time series data covering at least a decade for 52 countries (with an average of 17 years between the earliest and latest surveys). Contrary to the belief that happiness remains constant, subjective well-being rose in 40 countries and fell in only 12, with a median increase of .35 on this index. The average percentage of people in these countries saying they were ‘‘very happy’’ increased by almost seven points. Figure 12.1 shows which countries rose and fell in this overall rising trend. Ukraine Moldova Slovenia Nigeria Argentina Turkey Russia Czech R. Mexico Brazil Bulgaria Spain Italy Poland Japan Uruguay Malta France N. Ireland W.Germany Slovakia South Africa New Ireland Austria Belgium Portugal Denmark U.S.A. Estonia Finland E.Germany Britain Serbia Netherlands Canada Sweden Norway Iceland Colombia South Belarus Lithuania Switzerland Latvia Chile Romania Taiwan China India Australia Hungary
–3.0
–2.5
–2.0
–1.5
–1.0
–0.5
0.0
0.5
1.0
1.5
2.0
2.5
3.0
Figure 12.1 Change on subjective well-being index* from earliest to latest survey for all countries with a substantial time series covering 17 years on average. * This index is based on responses to questions measuring life satisfaction and happiness, equally weighted. Life satisfaction is measured on a ten point scale, and happiness on a four-point scale, and the two questions have opposite polarity, so the index is constructed as follows: subjective well-being = (life satisfaction – 2.5* happiness). Source: Inglehart et al. 2008 : p. 266.
Faith and Freedom: Traditional and Modern Ways to Happiness
357
The chapter by Diener et al. in this volume presents converging findings. They find that long-term changes in income are correlated with rising subjective well-being, as measured by four different indicators. This finding flatly contradicts the claim that rising income does not bring increased subjective well-being. Moreover, short-term changes in income have a negative impact on subjective well-being while long-term changes have a positive impact, which contradicts set-point theory (which holds that the exact opposite should be true). The chapter by Graham et al. in this volume helps explain why this is true: as they point out, in the short term, economic growth can actually produce declining happiness, because it leads to rising inequality and aspirations that outpace objective improvement. Their elegant presentation of the paradoxes of unhappy growth, the happy peasants and frustrated achievers, and the paradox of low aspirations all demonstrate the complexity of the relationship between economic development and subjective well-being. As Graham et al. point out, many other factors besides income shape subjective well-being, and they can cloud or conceal the linkage. Large bodies of time series evidence from both the Gallup World Poll and the Values Surveys indicate that subjective well-being has been rising in recent decades, and that—contrary to the Easterlin paradox—the increase was partly due to rising income levels. If this is true, it implies that the converse proposition should also hold true: if the current economic recession proves to be deep and lasting like the Great Depression of the 1930s, it should bring falling levels of subjective well-being. Future waves of the Gallup World Poll and the World Values Survey will enable us to test this prediction. In any event, it seems clear that subjective well-being rose from 1981 to 2007. But happiness rose more consistently than did life satisfaction, for they tap different aspects of subjective well-being. A society’s level of life satisfaction seems to be more strongly influenced by economic conditions than is happiness, which is more sensitive to religion and democratization. We agree with Diener et al. that life satisfaction taps a cognitive evaluation of one’s circumstances that is more closely linked with one’s income, while happiness taps a more affective response—so the two aspects of subjective well-being can move on somewhat different trajectories. During the period analyzed here, many ex-Communist countries experienced democratization that was accompanied by economic collapse—with the result that personal happiness rose, while life satisfaction fell. Russia is a striking example. In the years since 1981, Russia experienced both political liberalization and economic trauma. Real income fell to less than half its pre-transition level, and life expectancy declined by several years. Consequently, happiness rose, but life satisfaction fell sharply until 2000. Subsequently, Russia experienced a strong economic recovery and life satisfaction rebounded.
358
Section III: Differences in the Social Context of Well-Being
Figure 12.2 shows the trajectories of happiness and life satisfaction in four types of societies. They include high-income societies, low-income and middle-income societies, and two types of ex-Communist societies: the Soviet successor states plus China; and nine central European ex-Communist societies.1 Figure 12.2a shows how happiness and life satisfaction evolved in 20 high-income countries. The changes observed in high-income countries from 1981 to 2007 are relatively small. We find a modest rise in happiness, High-income countries (Australia, Belgium, Britain, Canada, Denmark, Finland, France, Iceland, Ireland, Italy, Japan, Malta, N. Ireland, Netherlands, Norway, Spain, Sweden, Switzerland, U.S. and W. Germany)
8.5
Happiness 8
Life Satisfaction 7.5
7
6.5 1980
1985
1990
1995
2000
2005
2010
Low and Middle-income countries (Argentina, Brazil, Chile, India, Mexico, Nigeria, Peru, South Africa, South Korea, Turkey)
8.5
8
Happiness
7.5
7
Life Satisfaction
6.5 1985
1990
1995
2000
2005
2010
Figure 12.2 Trajectories of Happiness and Life Satisfaction, 1981–2007. (Life Satisfaction reflects the mean score on 10-point scale; Happiness reflects the mean score on 4-point happiness scale, multiplied by 2.5 for comparability with life satisfaction, and polarity reversed)
Faith and Freedom: Traditional and Modern Ways to Happiness
359
Ex-USSR and China 7
(Belarus, China, Estonia, Latvia, Lithuania, Moldova, Russia, Ukraine)
Happiness 6.5
6
5.5
Life Satisfaction 5
4.5 1985
1990
1995
2000
2005
2010
2005
2010
East-European ex-communist countries (Bulgaria, Czech Rep, E. Germany, Hungary, Poland, Romania, Serbia, Slovakia, Slovenia)
7.5
Happiness 7
6.5
Life Satisfaction 6
5.5 1985
1990
1995
2000
Figure 12.2 (Continued).
while life satisfaction is almost flat. The flatness of the life satisfaction trajectory shown here fits the implications of the diminishing returns curve linked with economic development: as the cross-sectional curve implies, high-income countries show relatively modest gains over time. Until recently, virtually all of the time series data on subjective well-being came from high-income countries, and the flatness of the trajectory found among them helps explain why it was widely believed that subjective wellbeing remains constant, both among individuals and among nations. This
360
Section III: Differences in the Social Context of Well-Being
idea was so widely accepted that several theories emerged to explain why it was impossible for subjective well-being to rise. Although subjective well-being actually did rise in most high-income countries, the change would be small enough to be disregarded in the absence of other evidence. The next section of Figure 12.2 shows the changes in happiness and life satisfaction that were observed in ten low-income and middle-income countries. As the principle of diminishing returns implies, we find a larger increase in subjective well-being in these countries than we do among the high-income countries. But happiness and life satisfaction responded differently: while happiness rose steadily and substantially (moving from 7.43 to 8.03), life satisfaction showed a modest initial decline and then a sharp rise, producing a net gain from 6.91 to 7.15. The last two sections of Figure 12.2 show the trajectories of happiness and life satisfaction in two types of ex-Communist societies. The events that occurred from 1990 to 2007 had a dramatic impact on the people of ex-Communist countries, with sharply different effects on happiness and life satisfaction. As suggested above, the combination of democratization plus severe socio-economic decline that accompanied the collapse of Communism in the Soviet successor states led to a steady rise in happiness, but a sharp initial decline in life satisfaction, which recovered with the return of economic growth after 2000. Economic recovery occurred earlier in the East European ex-Communist countries (nearly all of which are now members of the European Union) than in the Soviet successor states, but in both types of societies, by 2007, life satisfaction had recovered, reaching higher levels than those of 1990. In the world as a whole, during the years covered by these surveys, subjective well-being rose in the great majority of countries. The highincome countries started out with much higher levels of well-being than those found elsewhere, and they maintained these high levels, even showing modest gains in most cases. But the gains were much larger in low- and middle-income non-Communist countries, and in ex-Communist countries, in many of which we find substantial gains. The Easterlin paradox holds that, as countries grow wealthier, average happiness levels do not increase. This claim does not hold up in the light of the empirical evidence now available. Average happiness levels can and do increase—and in the years since 1981, subjective well-being rose in most countries. But there are good reasons why, until recently, this paradox seemed to hold true. For there is nothing like a one-to-one relationship between economic growth and subjective well-being. Material prosperity is only one of many relevant factors, and not necessarily the most important one: living in a free and tolerant society seems to have an even greater impact on happiness than does wealth. Moreover, economic growth seems to have a
Faith and Freedom: Traditional and Modern Ways to Happiness
361
diminishing effect on happiness, with its impact being greatest in lowincome societies—but until recently, almost all of the available time-series data came from a handful of high-income countries. Still other effects complicate the relationship between economic development and happiness. The chapter by Graham et al. in this book discusses the ‘‘paradox of unhappy growth’’: on average, people in countries with relatively high growth rates show lower levels of happiness (cf. Deaton, 2008; and Wolfers & Stevenson, 2008). The initial stages of growth produce negative effects, perhaps linked with rapidly rising inequality. Similarly, they report the paradox of ‘‘happy peasants and frustrated achievers,’’ in which destitute peasants report higher levels of well-being than much wealthier respondents who have achieved upward mobility (cf. Graham & Pettinato, 2002). This phenomenon seems to reflect differences in the belief systems and aspiration levels of the respective groups—factors that will be given more attention later in this chapter. Clearly, the relationship between economic development and subjective well-being is complex. But in the long haul, across large numbers of societies, development does seem to contribute to happiness.
Faith and Happiness
The first half of this chapter has examined the impact of modernization on subjective well-being—finding that, since 1981, unprecedented rates of economic development, democratization, and rising social tolerance contributed to an environment in which subjective well-being rose in most countries around the world. But modernization is not the only route to happiness. Belief systems, particularly religion, also seem to play an important role in shaping happiness. Religion does not account for the recent rise in happiness: high or low levels of religiosity tend to be relatively constant features of given societies, and cannot account for the rapid recent rise of happiness. But religious beliefs help explain enduring cross-national differences in levels of subjective well-being. Previous studies have found that subjective well-being varies a great deal across societies. Recent evidence, drawn from a wider range of countries, illustrates this point even more strongly. To minimize the impact of shortterm changes, the combined data from 1995–2007 Values Surveys are used here and in Figure 12.3. As Table 12.1 indicates, the subjective well-being (SWB) index shows a huge range of variation, from a high in Denmark to a global low in Zimbabwe. In Denmark, 76 percent of the public place themselves at the high end of a life satisfaction scale (at points 8 to 10); in Zimbabwe, only 14 percent do so. Conversely, in Denmark only 5 percent of the population describe themselves as unhappy, but in
TABLE 12.1 Subjective well-being in 97 societies based on reported happiness and life satisfaction, equally weighted (Ranked from happiest to least happy. Negative scores indicate that a majority of the population is unhappy/dissatisfied with life). Country
Mean
Denmark Puerto Rico Colombia Iceland N Ireland Ireland Switzerland Netherlands Canada Austria El Salvador Malta Luxemburg Sweden New Zealand U.S.A. Guatemala Mexico Norway Belgium Britain Australia Venezuela Trinidad Finland Saudi Arabia Thailand Cyprus Nigeria Brazil Singapore Argentina Andorra Malaysia W. Germany Vietnam France Philippines Uruguay Indonesia Chile
4.24 4.21 4.18 4.15 4.13 4.12 3.96 3.77 3.76 3.68 3.67 3.61 3.61 3.58 3.57 3.55 3.53 3.52 3.50 3.40 3.39 3.26 3.25 3.25 3.24 3.17 3.02 2.96 2.82 2.81 2.72 2.69 2.64 2.61 2.60 2.52 2.50 2.47 2.43 2.37 2.34 (continued )
362
TABLE 12.1
(Continued)
Dominican Rep Japan Spain Israel Italy Portugal Taiwan E. Germany Slovenia Ghana Poland Czech Rep China Mali Kyrgyzstan Jordan Greece S Africa Turkey Peru S Korea Hong Kong Iran Bangladesh Bosnia Croatia Morocco India Uganda Zambia Algeria Burkina Faso Egypt Slovakia Hungary Montenegro Tanzania Azerbaijan Macedonia Rwanda Pakistan Ethiopia Estonia Serbian Bosnia Lithuania Latvia Romania Russia
2.29 2.24 2.16 2.08 2.06 2.01 1.83 1.78 1.77 1.73 1.66 1.66 1.64 1.62 1.59 1.46 1.45 1.39 1.27 1.24 1.23 1.16 1.12 1.00 0.94 0.87 0.87 0.85 0.69 0.68 0.60 0.60 0.52 0.41 0.36 0.19 0.13 0.13 0.06 0.15 0.30 0.30 0.36 0.45 0.70 0.75 0.88 1.01 (continued )
363
364
Section III: Differences in the Social Context of Well-Being
TABLE 12.1
(Continued)
Country
Mean
Georgia Bulgaria Iraq Albania Ukraine Belarus Moldova Armenia Zimbabwe Mean:
1.01 1.09 1.36 1.44 1.69 1.74 1.74 1.80 1.92 1.57
Source: Combined data from 1995–2007 Values Surveys. To maximize reliability, data from the last three waves are combined.Latin American countries are shown in italics, and ex-communist countries are shown in bold face.
Zimbabwe, 44 percent do so. A considerable share of the cross-national variation is linked with economic differences: the people of high-income societies show much higher levels of subjective well-being than do the people of low-income societies—partly because high-income societies also tend to have higher levels of democracy and social tolerance. Across the 97 societies for which we have data, the correlation between subjective wellbeing and gross domestic product (GDP) per capita is a sturdy r ¼ 0.62. Accordingly, virtually all of the high-income societies shown in Table 12.1 rank above the median on subjective well-being. But these differences in development, democracy, and social tolerance are not the whole story. Even controlling for their effects, the people of certain types of societies show substantially higher levels of subjective well-being than do the people of others. Long-standing differences in the belief systems of these societies help explain these differences. Thus, although they are not high-income countries as classified by the World Bank, two of the three countries with the highest levels of subjective well-being are Latin American countries, and eleven of the twelve Latin American societies for which we have data rank above the median level for the world as a whole. Putting it another way, all twenty-five societies with the highest levels of subjective well-being are either (1) high-income countries or (2) Latin American countries. Being Latin American seems to provide a bonus in subjective well-being that raises their scores toward the level of high-income countries. As Figure 12.3 demonstrates, subjective well-being increases with rising GNP per capita, but the relationship is curvilinear, as the regression line on this figure indicates. All twelve Latin American societies fall above the
Puerto Ric o
4.25 Colombia
3.75
El Salv ador LATIN Guatemal a Venezuel a
3.25 2.75
365
S W B Index
2.25
Domin. Brazil Rep. Argentina Vietnam Uruguay Trinidad Ghana Chile Indonesia Philippines
Peru
1.25 0.75 0.25
Zambia Uganda Pakist an
Croa ti a
Serb ia
Egypt
Jorda n RwandaTanzania Macedonia Eth
Latv ia
–0.75 Georgia
–1.25
Belarus
–1.75
Iraq
Belgium West Germany France
Spain Por tugal East Germany Slovenia Greece S. Ko rea
Taiw an
Israel
Japa n
Ital y
Hungary
Slov akia
EXCOMMUNIST
Azerba ij an
–0.25
Aust ralia Finland
Cyprus Singapore
Poland Turkey South Iran Af rica
Bangla desh India Al geria Bur kina
New Zealand
Saudi Arabia
Czech R ep.
China
Mali
Kyrgy zst an
Netherlands U.S.A. Luxembourg Canada Britain Au stria Norw ay
Sweden
AMERICAN
Nigeria
1.75
Denma rk Switzerland
Iceland
Ireland
Mexi co
Estonia Lithuania
Russia Romania Albania Ukra ine
Bulga ria
Moldova
Ar menia
Zimbabwe
–2.25 0
5
10
15
20
25
30
GDP per capita, five years before survey in $ thousands, purchasing power parity Well-being index is based on reported Life Satisfaction and Happiness,using mean results from all available surveys conducted 1995 - 2007 Cubic curve plotted (r = .62)
Figure 12.3 Subjective well-being, per capita GDP and different types of society.
366
Section III: Differences in the Social Context of Well-Being
regression line, showing higher levels of subjective well-being than their economic level would predict. Puerto Rico, Mexico, and Colombia show some of the highest levels of subjective well-being in the world, though their economic levels are much lower than those of the United States, Canada, Western Europe, and Japan. Conversely, the ex-Communist societies in general, and the Soviet successor states in particular, show lower scores than their economic level would predict. In countries that never experienced Communist rule, 33 percent of the public describe themselves as ‘‘very happy,’’ in contrast to only 7 percent in the ex-Soviet societies and 16 percent in the other former Communist societies. For the former Communist societies, this finding is not surprising. Communism collapsed throughout Central and Eastern Europe around 1990. In the Soviet successor states, this brought drastic decreases in standards of living, stagnant or falling life expectancies, and the collapse of the social, political, and belief systems their people knew. The prevailing sense of existential security and individual control over one’s life fell sharply, and the collapse of faith in the Communist ideology seems to have left a spiritual vacuum. Although these societies are far from the poorest countries in the world, their publics experienced a sharp downward shift from their previous levels of income and security, and their sense of meaning. Thus, although Vietnam, China, East Germany, Slovenia, Poland, and the Czech Republic have middle-range levels of subjective well-being, all of the other ex-Communist societies, especially the ten Soviet successor states, have low scores on subjective well-being. Ex-Communist countries comprise thirteen of the fifteen lowest-ranking countries on subjective well-being. In contrast, all of the Latin American societies except Peru rank above the global mean on subjective well-being, and even Peru ranks higher than its economic level would predict. The people of the Nordic countries (Denmark, Sweden, Norway, Iceland, and Finland) also show high levels of subjective well-being, but this is easily understandable: they live in high-income countries with advanced welfare states, high levels of social tolerance, and stable and effective democratic governments. The Nordic countries exemplify the effects of successful economic, social, and political modernization. The Latin American countries show levels of subjective well-being that are comparable to those of the Nordic countries although they have not experienced comparable prosperity or good governance, indicating that other factors must be at work. One reason why the Latin American societies rank so high on subjective well-being is linked with the fact that they rank high on religiosity, a traditional pathway to relatively high levels of subjective well-being. Table 12.2 shows responses to a sensitive indicator of how strongly people emphasize religion: the question, ‘‘How important is God in your life?’’ As
TABLE 12.2 Percentage saying ‘‘God is very important in my life’’ (choosing 10 on a ten-point scale) China Japan E. Germany Hong Kong Denmark Sweden Estonia France Czech Rep Taiwan Netherlands Norway W. Germany Bulgaria S Korea Vietnam Iceland Finland Luxemburg Slovenia Latvia Spain Russia Thailand Britain Belgium Andorra Belarus Serbia Hungary Switzerland Montenegro Serb Bosnia Australia Ukraine New Zealand Lithuania Albania Austria N Ireland Croatia Slovakia Greece Armenia Italy Moldova
5 6 6 6 7 8 9 10 10 10 11 11 12 12 14 14 15 15 15 16 16 18 18 18 19 19 19 20 20 21 21 22 23 24 24 24 25 25 26 28 28 30 31 31 33 35 (continued )
367
TABLE 12.2
(Continued)
Uruguay Portugal India Georgia Ireland Canada Macedonia Malaysia Bosnia Kyrgyzstan Poland Israel Singapore Romania USA Argentina Cyprus Chile Azerbaijan Mexico Peru Ethiopia Malta Mali S Africa Burkina Faso Zambia Uganda Turkey Rwanda Venezuela Domin. Rep Iran Colombia Nigeria Brazil Ghana Philippines Guatemala Zimbabwe El Salvador Bangladesh Egypt Trinidad Tanzania Indonesia Puerto Rico Iraq
36 37 38 38 40 40 40 40 42 48 54 55 55 57 58 58 58 61 61 64 67 67 68 68 70 70 71 73 78 79 80 80 82 83 86 87 87 87 87 89 89 89 89 89 90 90 91 92 (continued )
368
Faith and Freedom: Traditional and Modern Ways to Happiness
TABLE 12.2
369
(Continued)
Saudi Arabia Morocco Algeria Pakistan Jordan Mean:
94 95 95 97 98 48
Source: 1995–2007 Values Surveys
this table demonstrates, societies vary immensely in their religiosity. In the world as a whole, about 36 percent of the public says that God is very important in their lives. Among the twelve Latin American societies, the figure ranges from a low of 37 percent in Uruguay to a high of 91 percent in Puerto Rico. The Islamic societies rank even higher in religiosity than Latin America, but with much lower levels of democracy and social tolerance, they do not show comparable levels of subjective well-being. Communist rule inculcated secular values, and the publics of the Soviet successor states are less than half as likely to say that ‘‘God is very important in my life’’ as are the publics of non-Communist societies. But high levels of economic development in Western Europe had a secularizing impact fully as strong as that of seven decades of Communist rule. Thus, the publics of Denmark, Norway, Sweden, France, the Netherlands, and Japan are even less likely to say that God plays an important role in their lives than are the publics of most exCommunist countries. Previous empirical studies have found that emphasis on religion is linked with subjective well-being (Ferriss 2002; Kahneman & Krueger, 2006; Limm & Putnam, 2009), but it is not entirely clear why: do happy people become religious, or are religious beliefs conducive to happiness? As we will see, the evidence examined here suggests that initially, unhappy people turn to religion (as has happened recently in many ex-Communist countries)—but that in the long run, religious beliefs are linked with relatively high happiness levels. Why would religious beliefs be conducive to subjective well-being? There are several possible reasons: 1. Religion dampens aspirations. If a person’s level of subjective wellbeing reflects a balance between aspirations and attainment, one can increase happiness either by raising attainments, or by lowering one’s aspirations. Many religions tend to do the latter, encouraging people to be content with their station in life.
370
Section III: Differences in the Social Context of Well-Being
2. Religion provides a sense of solidarity: it encourages sharing and mutual support (a form of social insurance where the welfare state is absent). 3. Religion provides a sense of certainty and stability in an unpredictable and dangerous world. This has been found to be a strong motive for mass adherence to religion, especially among low-income countries (Norris & Inglehart, 2004). 4. Religion provides a sense of meaning and purpose in life. This is a universal human need, which tends to become more salient when survival seems secure (Norris & Inglehart, 2004).
Thus, religion serves important psychological functions for its believers. It can provide security, social solidarity, and a sense of meaning in life. Consequently, religiosity tends to be linked with subjective well-being. But this relationship is complicated and partly concealed by the fact that, within most countries, low-income people are more likely to be religious than wealthier people; and that the people of low-income countries tend to be much more religious than the people of high-income countries. Since, as we have seen, successful modernization is conducive to high levels of subjective well-being, the traditional and modern routes to happiness have a tendency to cancel each other out—though this is not inevitable. Table 12.3 shows the results of a multiple regression analysis that examines the TABLE 12.3
Religion, economic development and subjective well-being Dependent variable 1. Happiness
Independent variables
2. Life Satisfaction
Model 1.1
Model 1.2
Model 1.3
Model 2.1 Model 2.2 Model 2.3
GDP/capita PPP in 1995
___
.019 (.025)***
.027 (.002)***
___
.088 (.007)***
.106 (.009)***
Importance of religion
.106 (.036)***
___
.225 (.028)***
.086 (.137)
___
.362 (.108) ***
Constant
1.77 (.081)
2.16 (.025)
1.78 (.061)
6.35 (.313)
5.74 (.088)
6.29 (.230)
Adjusted R2 N
.043 89
.221 89
.487 89
.003 89
.432 89
.473 89
Notes: Cell entries are unstandardized regression coefficients (standard errors in parentheses). Significance levels:*** p < .001;** p < .01;* p < .05. Source: attitudinal variables from 1981–2007 Values Surveys; economic data from the World Bank, World Development Indicators.
Faith and Freedom: Traditional and Modern Ways to Happiness
371
national-level relationship between religion, economic development, and two components of subjective well-being—happiness and life satisfaction respectively. Models 1.1 through 1.3 deal with happiness, while Models 2.1 through 2.3 deal with life satisfaction. Our indicator of each nation’s level of religiosity is the mean score in response to the question, ‘‘How important is religion in your life? Would you say it is very important, rather important, not very important, or not at all important?’’ Our indicator of a country’s level of economic development is its per capita gross domestic product in 1995, using the World Bank’s purchasing power parity estimates.4 As Model 1.1 indicates, religiosity by itself has a relatively weak zeroorder relationship with happiness, explaining only 4.3 percent of the cross-national variance in happiness scores. Economic development has a considerably stronger zero-order relationship, explaining 22.1 percent of the variance (Model 1.2). But when we include both religiosity and economic development in the equation, the two variables explain fully 48.7 percent of the cross-national variance in happiness (see Model 1.3). The negative correlation between religiosity and economic development tends to conceal their full impact. When we add the impact of religiosity to that of economic development, we more than double the explained variance, which rises from 22 percent to almost 49 percent. To some extent, religion and economic development can substitute for each other in producing happiness. But this holds true much more strongly of happiness than of life satisfaction. As Model 2.1 indicates, religiosity by itself has almost no impact on life satisfaction. Economic development, on the other hand, has a strong impact on life satisfaction—by itself, it accounts for 43.2 percent of the cross-national variation (almost twice the impact that it has on happiness), as Model 2.2 indicates. When we add religiosity to the equation, it shows a significant though modest impact, raising the explained variance to 47.3 percent. Together, the two independent variables explain fully as much of the variance in happiness as they do in life satisfaction, but their relative impact is very different: while religiosity seems to have as much impact on happiness as does economic development, the latter has far more impact on life satisfaction. Religion has a strong impact on people’s affective evaluations of their lives, but their cognitive appraisals are mainly shaped by economic factors. Religiosity tends to be linked with high levels of happiness. But not in all societies. Table 12.4 shows the strength of the correlation between subjective well-being and emphasis on religion (as measured by the proportion saying that God is important in their lives). Across 98 countries, 78 percent show positive correlations between religiosity and subjective
TABLE 12.4 countries
Religiosity and Subjective Well-being among the populations of 98
Country
Corr. with SWB index
Mexico Ireland S Korea Thailand Malaysia Singapore Chile Malta N Ireland Venezuela USA Argentina Azerbaijan Burkina Faso El Salvador India Indonesia Italy Armenia Britain Canada Croatia Iceland Japan Turkey Uganda Uruguay Austria Brazil Colombia E. Germany Finland Hong Kong Kyrgyzstan Taiwan Greece Guatemala Latvia Ukraine Dominican Rep Ethiopia Puerto Rico Saudi Arabia Iran
0.28 0.21 0.21 0.18 0.17 0.17 0.16 0.15 0.15 0.14 0.12 0.11 0.11 0.11 0.11 0.11 0.11 0.11 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.10 0.09 0.09 0.09 0.09 0.09 0.09 0.09 0.09 0.08 0.08 0.08 0.08 0.07 0.07 0.07 0.07 0.06 (continued )
372
TABLE 12.4
(Continued)
Luxemburg Peru Switzerland W. Germany Australia New Zealand S Africa Tanzania Morocco Nigeria Philippines Poland Slovakia Sweden Zimbabwe Czech Rep Denmark Netherlands Romania Zambia Jordan Lithuania Mali Russia Serbia Spain Andorra Bangladesh Bosnia France Georgia Ghana Algeria Belgium Bulgaria Estonia Hungary Israel Pakistan Vietnam Belarus China Cyprus Egypt Norway Trinidad Moldova
0.06 0.06 0.06 0.06 0.05 0.05 0.05 0.05 0.04 0.04 0.04 0.04 0.04 0.04 0.04 0.03 0.03 0.03 0.03 0.03 0.02 0.02 0.02 0.02 0.02 0.02 0.01 0.01 0.01 0.01 0.01 0.01 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.01 0.01 0.01 0.01 0.01 0.01 0.02 (continued )
373
374
Section III: Differences in the Social Context of Well-Being
TABLE 12.4
(Continued)
Country Slovenia Rwanda Macedonia Portugal Montenegro Iraq Albania
Corr. with SWB index 0.02 0.03 0.04 0.04 0.06 0.07 0.08
Cell entry is correlation between mean response to ‘‘How important is God in your life?’’ rated on a scale from 1 to 10, and the subjective well-being index based on reported happiness and life satisfaction, equally weighted. Relatively strong negative correlations are shown in bold face. Source: Values Surveys, 1995–2007.
well-being, 8 percent show no correlation, and 14 percent show negative correlations. Most countries show positive correlations—but almost all of the ex-Communist countries show weak or negative ones; five of the eight countries showing substantial negative correlations (from –.02 to –.08) are ex-Communist countries. This may be linked with the fact that until recently, religion was severely repressed in ex-Communist countries. Marxist ideology once filled the function of a religion, providing psychological security, predictability, and a sense of meaning and purpose in life for many people. It is impossible to understand the rise to power of Communist movements in these countries without recognizing the motivating power that the Communist worldview once had. Many thousands of true believers sacrificed their lives for the Communist cause during the Russian revolution and civil war, during the Long March in China, and during the Vietnam war. But during the 1970s and 1980s this ideology began losing credibility; fewer and fewer people believed that Communist regimes were building an ideal society that represented the wave of the future. By 1990 Communism was generally discredited, and Communist regimes collapsed throughout the Soviet Union and Eastern Europe. In China and Vietnam, hard-line Communist regimes were replaced by nominally Communist regimes that have become increasingly pragmatic and market-oriented. The collapse of Communist regimes was accompanied by severe economic and social decline in the former Soviet Union and Eastern Europe and left an ideological vacuum everywhere. Communist regimes systematically repressed religion—but since their collapse, religion has been making a comeback. It has not recruited equally from all strata: it has tended to attract the least happy people—those who
Faith and Freedom: Traditional and Modern Ways to Happiness
375
feel the greatest need for a sense of meaning, reassurance, predictability, and social support. We hypothesize that in the ex-Communist countries, a disproportionate share of those who emphasize religion are new recruits who have been drawn to religion because they are unhappy and disoriented. If this is true, it would create a negative correlation between religion and happiness in these countries. In Eastern Europe and the Soviet successor states, the free practice of religion has emerged only since 1989–1991, and is only emerging now in China and Vietnam. As we will see, in recent years, religion has been growing rapidly in many ex-Communist countries. These facts are consistent with the interpretation that the weak or negative correlations between religiosity and subjective well-being found in most ex-Communist countries, reflect a recent influx of relatively unhappy newcomers, seeking to fill the spiritual vacuum left by the collapse of Communist ideology. Rwanda and Iraq also show relatively strong negative correlations between religiosity and subjective well-being. Both countries have had traumatic recent experiences, linked (respectively) with genocidal civil war, and with invasion, ethnic conflict, and almost daily suicide bombings. The negative correlations found in Rwanda and Iraq may also reflect an influx of unhappy people who have turned to religion.5 Figure 12.4 shows the relationship between subjective well-being and religion in countries at different levels of economic development.. The vertical axis shows the zero point where there is no correlation between religiosity and the SWB index. Countries to the right of this line show positive correlations between religion and subjective well-being, while countries to the left of this line show negative correlations. In an overwhelming majority of countries, we find a positive correlation: religious people tend to be happier than those who are not. Virtually all of the highincome countries show positive correlations between religion and subjective well-being.6 But among the low-income and middle-income countries, we find two groups of countries with roughly equal levels of economic development but contrasting relationships between religion and happiness: an ex-Communist cluster and a Latin American cluster. All of the Latin American countries show positive correlations between religion and subjective well-being; but almost all of the ex-Communist countries show weak or negative correlations. While the free practice of religion has emerged only recently in the ex-Communist societies, most Latin American countries have been strongly religious for centuries and remain so today. This, too, is consistent with the interpretation that the weak or negative correlations found in ex-Communist countries reflect a recent influx of unhappy people who have been turning to religion to fill the ideological void left by the collapse of Communism.
376
Section III: Differences in the Social Context of Well-Being
30
Luxemburg
United States
Singapore
GNP/capita 1995 (PPP)
Switzerland
Germany Norway Austria Canada France Denmark Japan Belgium Iceland 20 Netherlands Italy Great Britain Sweden Australia Finland New Zealand Israel
RELIGIOUS PEOPLE ARE LESS HAPPY
Spain Portugal
N. Ireland Ireland
Greece
RELIGIOUS PEOPLE ARE Puerto HAPPIER
S. Korea
Rico
10 Czech
Taiwan
Malaysia Chile
Slovenia
ArgentinaVenezuela Uruguay Mexico LATIN Colombia Brazil Estonia Romania Thailand Croatia BulgariaRussia S.AfricaTurkey AMERICA Belarus Domin.Rep. Lith.Slovakia Iran Indonesia Morocco Ukraine Macedonia Egypt Vietnam China Armenia Monte- Moldova Georgia Jordan El Salvador negro Nigeria Peru India Iraq Albania Pakistan ZimbaweUganda Bangladesh Tanzania Rwanda 0
EXHungary Poland COMMUNIST
-0.14
Correlation between religiosity and Subjective Well-Being Index
Figure 12.4 Linkage between Religiosity and Subjective Well-being Index among the Populations of Rich and Poor Countries.
This explanation depends on the assumption that religion has been growing recently in former Communist countries. Figure 12.5 shows the extent to which the publics of given countries have come to emphasize religion more strongly (or less strongly) since 1981. The graph shows the difference between each country’s mean score on the ‘‘Importance of God’’ scale in the earliest and latest available survey for each country from which we have at least two surveys spanning a substantial period of time (the average number of surveys per country is 3.7, and the mean time span is seventeen years).7 For example, the mean score of the Bulgarian public (at the top of the graph) increased from 3.56 in the 1990 survey to 5.70 in the 2006 survey—a gain of more than two points on the ten-point scale.
Faith and Freedom: Traditional and Modern Ways to Happiness
377
Bulgaria Russia China Belarus Serbia Romania Argentina Ukraine Moldova Italy Portugal Slovenia Uruguay Slovakia Taiwan S.Africa Turkey Mexico Finland Hungary Japan Czech R. Lithuania Brazil Austria S.Korea France Britain Canada Colombia U.S.A. W.German Chile Iceland India Poland Australia Nigeria Denmark Sweden Ireland N. Zealand E. Germany Switzerland Belgium Netherland N. Ireland Spain Norway
–2
–1
0
1
2
Change in mean score on Importance of God scale, earliest to latest survey
Figure 12.5 Changes in Emphasis on Religion, 1981–2007.
Russia rose from 4.00 in 1990 to 6.02 in 2006. And China started just above the bottom of the scale (point 1.0) in 1990, with a score of 1.62, but showed a large proportional gain, rising to 3.58 in 2007. We do not find a global resurgence of religion, as some observers have claimed. Most West European countries show declining emphasis on religion, as do several other countries. But many countries show increases, and all six of the countries showing the greatest gains are ex-Communist: Bulgaria, Russia, China, Belarus, Serbia, and Romania. Overall, the publics of 13 of the 15 ex-Communist countries for which we have a substantial time series increased their emphasis on religion.8 This is consistent with the hypothesis that religiosity is linked with unhappiness in these countries because those who are religious consist disproportionately of new, relatively unhappy recruits. This hypothesis gains further support from the fact that emphasis on religion increased most in countries with relatively low levels of happiness (r = 0.5). Unhappy people tend to turn to religion. Conversely, in most countries with a long religious tradition and high levels of subjective well-being, religiosity is linked with happiness.
378
Section III: Differences in the Social Context of Well-Being
How much impact does religion have?
In almost four-fifths of the countries for which we have data, religion is positively correlated with subjective well-being. Several ex-Communist countries show negative correlations, but even in the ex-Communist world as a whole, the relationship is positive. Sorting out the impact of religion on happiness could involve many variables, but the two most important factors to be controlled are: (1) the society’s economic level, and (2) whether it has a history of Communist rule. High-income societies show substantially higher levels of subjective well-being than low-income societies, not only because they are richer, but also because they tend to be more democratic and have higher levels of social tolerance. And the relationship between religion and subjective well-being is complicated by the fact that richer countries tend to be happier and less religious than poorer ones— which tends to conceal the fact that religion is linked with happiness. Moreover, religion itself is correlated with a number of other traditional values such as high levels of national pride, which may also have an impact on subjective well-being. The impact of Communist rule also is clearly important: (1) ex-Communist societies show markedly lower levels of subjective well-being than other societies; and (2) ex-Communist societies show substantially lower levels of religiosity than other societies. The following analysis gives a first approximation to an answer by measuring to what extent religious people are happier than less religious ones when we control for their society’s level of per capita GNP, and whether or not it has experienced Communist rule. By restricting ourselves to these four variables, the relationships can be shown on a pair of twodimensional graphs, providing an intuitively clear sense of what matters, and where. 9 Figure 12.6 shows the two graphs. Each graph compares the life satisfaction levels of those for whom religion is important (those choosing points 8 to 10 on the ‘‘importance of God in my life’’ scale) with those for whom religion is less important (those choosing points 1–7). This dichotomy separates the world’s population into two roughly equal groups. The top graph on Figure 12.6 examines the relationship between religion and life satisfaction in non-Communist countries. As it demonstrates, religious people show higher levels of life satisfaction in societies at all levels of economic development, but religion makes even more difference in high-income and upper-middle-income countries than it does in low-income and lower-middle-income societies. Averaged across the four types of societies, religious people are likelier to rank ‘‘high’’ on life satisfaction (placing themselves at points 8 to 10 on the ten-point scale) than are nonreligious people, by a margin of 9 percentage points.
Faith and Freedom: Traditional and Modern Ways to Happiness
379
A. Non-Communist countries:
% high (8–10) on life satisfaction
70 60
God is very important (8–10)
50 40 God is less important (1–7) 30 20 10
0
1 Low
2 3 Lower-middle Upper Middle World Bank income group (2000) Non-Communist countries
4 High
5
B. Ex-Communist countries:
% high (8–10) on life satisfaction
70 60 50
God is very important (8–10)
40 God is less important (1–7)
30 20 10 0
1 Low
2 3 Lower-middle Upper Middle World Bank income group (2000) Ex-Communist countries
4 High
5
Figure 12.6 Life Satisfaction by Importance of God.
The lower graph on Figure 12.6 examines the impact of religion on life satisfaction in ex-Communist countries. The absolute levels are lower than they are in non-Communist countries, and the difference between religious and non-religious people is somewhat smaller, but the basic pattern is
380
Section III: Differences in the Social Context of Well-Being
similar: again, religious people show higher levels of life satisfaction in societies at all levels of economic development, and the differences are greater among wealthier countries. The average difference between religious and non-religious people is about 6 percentage points. Averaging out these differences for the world as a whole, those who emphasize religion are about 8 percentage points higher on life satisfaction than those who do not. This is a modest but significant part of the total cross-national difference. Moreover, religion does not have the same amount of impact among all groups: being religious makes considerably more difference for the intensely religious than it does for those who are lukewarm, as we will see below. Finally, there is another rather surprising finding: under certain conditions, those at the atheist end of the scale are the happiest of all.
Happy Atheists
We suggested that religious beliefs might be conducive to subjective wellbeing insofar as they do the following: (1) they dampen aspirations; (2) they encourage a sense of sharing and solidarity; (3) they provide a sense of certainty in an insecure world; and (4) they provide a sense of meaning and purpose in life. But the last three factors are not unique to religion: they can be provided by any strong belief system—including the militant atheistic world-view espoused by Communism. This implies that we should find a somewhat U-shaped relationship between the strength of religion and subjective well-being: instead of rising steadily from the atheistic end of the scale to its highest level at the religious end of the scale, insofar as an alternative belief system such as Communism is available, we should find relatively high levels of subjective well-being at both ends of the Importance of God scale, with lower levels in between, since strong conviction is registered at both ends of the scale, with relative uncertainty in between. The data support this expectation. Even among low-income countries (as classified by the World Bank in 2001), there is a small group (4 percent of the sample) for whom God is completely unimportant (rated ‘‘1’’ on a ten-point scale)—and this group shows a higher level of life satisfaction than those near the middle, ranking only below those at the very top of the scale. Even in low-income countries, religion is not the only route to relatively high subjective well-being. Who are these happy atheists, among the low-income countries? Mostly Vietnamese. Fully 55 percent of the relatively satisfied atheists are from Vietnam; India and Ukraine come next with respectively 15 percent and 10 percent of the group—and among the 14 other low-income countries, satisfied atheists are virtually
Faith and Freedom: Traditional and Modern Ways to Happiness
381
non-existent. Similarly, in the 23 lower-middle-income countries for which we have data, the majority of satisfied atheists consist of Chinese (42 percent) and Russians (19 percent). And in the 21 upper-middle-income countries, a majority of the satisfied atheists are Czechs (40 percent), Hungarians (14 percent) and Slovaks (8 percent). Among low-income and middle-income countries, satisfied atheists are mainly found in exCommunist countries, where a Communist ideology once provided a powerful challenge to religion as a source of meaning, and a sense of security and solidarity among people. This curvilinear pattern suggests that any strong belief system is more conducive to well-being than an uncertain outlook. We suspect, that several decades ago, a strong belief in Communism was linked with relatively high levels of subjective wellbeing. Today, the people of ex-Communist countries show significantly lower levels of happiness and life satisfaction than people in non-Communist countries with similar income levels. But the phenomenon of satisfied atheists persists to some extent—and in low- and middle-income countries, it is mainly linked with Communism. Figure 12.7 examines the relationship between religiosity and life satisfaction among the people of high-income countries, upper-middle-income countries, lower-middle-income countries, and low-income countries.
A. Non-Communist countries:
% high (8–10) on life satisfaction
70 60
high-income countries
50 upper-middle income countries
40
lower-middle income countries
30 20 low-income countries
10 0 0
1
2
3
4
5
6
7
8
9
10
Importance of God in my life (non-communist countries)
Figure 12.7 Life Satisfaction by Importance of God, in Rich and Poor Countries.
11
382
Section III: Differences in the Social Context of Well-Being B. Ex-Communist countries:
% high (8–10) on life satisfaction
60 50 high-income countries upper-middle income countries
40 30
lower-middle income countries
20 low -inco me countries
10 0 0
1
2
3
4
5
6
7
8
9
10
11
Importance of God in my life (ex-communist countries)
Figure 12.7 (Continued).
The upper graph focuses on non-Communist countries, while the lower graph shows the relationship in ex-Communist countries. As the upper graph indicates, among high-income countries there is only a very faint curvilinear tendency: those for whom God is completely unimportant in their lives have slightly higher levels of life satisfaction than those at points 2 through 6, but life satisfaction rises steadily from that point on, reaching a peak at point 10. The pattern is more strongly curvilinear among the publics of upper-middle-income, lower-middle-income, and low-income countries, with those at the atheist end of the scale showing substantially higher levels of life satisfaction than those closer to the middle of the scale. In both upper- and lower-middle-income countries, the strongly religious people rank much higher on life satisfaction than those at the low end of the scale, but there is no clear trend among the publics of low-income countries. The lower graph of Figure 12.7 focuses on the people of ex-Communist countries, and here the curvilinear pattern is much more pronounced than in other countries. Both in high-income and upper-middle income ex-Communist countries, those at the atheist end of the scale show considerably higher levels of life satisfaction than those at the middle of the scale—though the strongly religious rank highest of all. In richer exCommunist countries, religion today is more conducive to happiness than
Faith and Freedom: Traditional and Modern Ways to Happiness
383
is atheism. But in lower-middle-income and low-income ex-Communist countries, we not only find a clear U-shaped pattern, but those at the atheist end of the spectrum rank highest of all. When we recall that the happy atheists in the lower-middle-income countries consist largely of Chinese, and in the low-income countries they consist mainly of Vietnamese, the findings are less surprising. Both China and Vietnam are still governed by Communist regimes and their economies were thriving when these surveys were carried out. Communist ideology has lost its credibility for most people and the state-run economies that were once considered the wave of the future are being abandoned. But pockets of true believers still remain, and for them, atheism is a core element of their beliefs. For most of the world, strong religious beliefs are linked with happiness, but among Communism’s remaining true believers, Communism still brings a sense of meaning and certainty, though the next highest level is found among the strongly religious. In high-income countries, an alternative route to subjective well-being has emerged that is not linked with either religion or Communism. In these countries, people for whom God is totally unimportant are relatively numerous—and many of them show high levels of subjective well-being. Satisfied atheists are concentrated in some countries. Among the 34 highincome countries for which we have data, those with the largest numbers of satisfied atheists are (in this order) Sweden, East Germany, Norway, Netherlands, Denmark, Belgium, France, Britain, West Germany, Spain, Slovenia, and Canada. They are relatively rare in the U.S. and Italy, and almost nonexistent in Ireland, Northern Ireland, Portugal, Cyprus, Greece, and Israel. The quest for a sense of purpose in life seems to reflect a universal human need, and it has not diminished with economic development—on the contrary, during the past 25 years, a growing share of the population in most countries reports that they spend time thinking about the meaning and purpose of life (Norris & Inglehart, 2004). The happy atheists in highincome countries presumably find a sense of meaning in such causes as environmental protection, which often takes on quasi-religious tones, to a fascination with the majesty and beauty of science that seems to motivate militant atheists such as Richard Dawkins (1998, 2006).
Conclusion
There seem to be two main routes to happiness, one linked with modernization and another with traditional belief systems.
384
Section III: Differences in the Social Context of Well-Being
Throughout human history there have been two strategies for reducing unhappiness: the first is to lower one’s expectations, and accept the inevitability and dignity of suffering, which is achieved through one’s religion and spiritual life. The second is to expand people’s range of material, political and social opportunities, a strategy often called modernization, which occurs through a profound set of transformations in a society’s economic, political, and value systems. Economic development is conducive to rising subjective well-being, especially insofar as it also tends to bring rising social tolerance and democratic political institutions. Until recently it was widely believed that neither individuals nor societies can lastingly increase their happiness levels, but recent research indicates that, during the past 25 years, economic development, democratization, and rising social tolerance have led to rising happiness around the world. Since 1981, overall subjective well-being increased in 40 of the 52 countries for which a substantial time series is available. The link between economic development and subjective wellbeing is complex, and its impact can be concealed by the fact that it is only one of many causal factors, with social tolerance and political freedom playing even more important roles. Moreover, belief systems also play a crucial role in shaping people’s levels of subjective well-being. Evidence from scores of societies containing almost 90 percent of the world’s population indicates that, in an overwhelming majority of countries, religious people are happier than non-religious people, even though they tend to have lower incomes. In the United States, for example, 46 percent of those for whom God is very important describe themselves as very happy, as compared with 34 percent of those who are less religious. Some ex-Communist countries show negative correlations between religion and subjective well-being. This seems to reflect a recent influx of unhappy people who have turned to religion after the collapse of faith in Communist ideology—which once provided a sense of meaning and certainty for many people, and still does for a small group of true believers. People have an enduring need for a sense of meaning in life, and a strong belief system, whether religious or secular, tends to be linked with relatively high levels of subjective well-being. At this point in history, happy atheists are heavily outnumbered by those who find a sense of meaning in religion. The evidence indicates that people can attain happiness by optimizing external conditions or by developing a belief system that inculcates a positive response to existing conditions: in other words, by getting what one likes, or by liking what one gets. One approach is linked with modernization and the other with traditional society. At this point in history, the Nordic countries constitute the leading example of successful modernization, maximizing prosperity, social solidarity, and political and personal
Faith and Freedom: Traditional and Modern Ways to Happiness
385
freedom. Both the Values Surveys and the Gallup World Poll identify Denmark as the country with the world’s highest level of subjective wellbeing, with Iceland, Sweden, Norway, and Finland also ranking high. But—surprising as it may seem to people with a modern world-view—a group of Latin American countries ranks almost equally high, despite having substantially lower levels of prosperity and good governance. Interestingly, the Nordic countries (and most of the other leading cases of successful modernization) are located in Northern Europe, while all of the highest-ranking Latin American countries are located on or near the Caribbean. And while the Nordic countries are among the world’s most secular societies, all of the high-ranking Latin American countries have strongly religious publics. When asked to comment on these findings, Latin American colleagues have emphasized the role of strong networks of family and friends that supply social support and financial support that help offset the absence of effective welfare states—factors that the literature on Gemeinschaft and Gesellschaft has emphasized since the era when early modernization brought a breakdown of traditional society that led to feelings of anomie and meaninglessness. Successful modernization can bring high levels of prosperity, social solidarity, tolerance, and democracy, producing high levels of subjective well-being. But long before modernization became possible, traditional societies evolved ways of coping with the stresses of human existence and the need for a sense of meaning. Physiological measurements suggest that Buddhist monks attain some of the highest levels of subjective well-being ever recorded.10 Both faith and freedom can lead to happiness. To some extent they tend to substitute for each other—but there is no reason why a society could not attain both high levels of autonomy and a belief system conducive to happiness. This chapter has discussed some of the factors linked with subjective wellbeing. The findings suggest that we might learn more by in-depth study of the reasons why certain types of societies—such as the Nordic countries, on one hand, and the high-ranking Latin American societies, on the other hand— have attained relatively high levels of subjective well-being.
Notes 1. Over the past few decades, the Values Surveys have interviewed representative national samples of scores of countries, with an average sample size of 1,400 respondents. More than three thousand publications have used these data. Extensive information about publications, findings, fieldwork and the organization of these surveys can be found at http://worldvaluessurvey.org. Fieldwork information, questionnaires in the original languages, and reports of findings can be downloaded from this site.
386
Section III: Differences in the Social Context of Well-Being
2. Because life satisfaction is measured on a ten-point scale, and happiness on a fourpoint scale, and because the two questions have opposite polarity, the SWB composite was constructed as follows: SWB = life satisfaction – 2.5 * happiness. If 100 percent of its people were very happy and extremely satisfied, a country would get the maximum score of 7.5. If happiness and life satisfaction were evenly balanced, the country would get a score of zero. If more people were dissatisfied/ unhappy than are satisfied/happy, the country would get a negative score. 3. Operationally, we define ‘‘substantial’’ as including surveys from at least two waves scheduled to be held at least ten years apart. Because actual fieldwork sometimes took place earlier or later than scheduled, in a few cases the actual time span between surveys is less than ten years, but in other cases the span covers as much as 26 years. For the average country, this figure reflects changes that took place over a period of 17 years. 4. We use a society’s economic level in 1995, at the start of the period during which religiosity and subjective well-being were measured, on the assumption that it is an independent variable that should be measured prior to the dependent variable, subjective well-being. 5. We do not have the time series data from Rwanda and Iraq that would be needed to test this hypothesis, but we do have a substantial body of time series data from a number of ex-Communist countries that support this hypothesis. 6. Norway is a high-income country with a faintly negative correlation between religiosity and SWB. We have no explanation for this anomaly. 7. Appendix B shows the religiosity scores for each country in each year from which data are available, together with the year of the survey. 8. Poland is a prominent exception: in reaction to Soviet repression, adherence to the Roman Catholic faith became a symbol of Polish identity, leading to uniquely high levels of religiosity among Communist countries. Since regaining independence, Polish religiosity has declined somewhat, but remains high. 9. More complex multivariate analyses indicate that these results are in the right ballpark. 10. Using imaging devices to study brain activity, Davidson (2001) finds that a pattern of left prefrontal activation is associated with positive affect. Testing the brain activity of a Buddhist monk, he found a pattern of left prefrontal activation that was more intense than that of any of the other 175 individuals he had ever tested.
REFERENCES Andrews, F. M., & Withey, S. B. (1976). Social indicators of well-being. New York: Plenum. Barro, R. J. Determinants of democracy. Journal of Political Economy, 107, 158–183. Brickman, P., & Campbell, D. T. (1981). Hedonic relativism and planning the good society. In M. Appley (Ed.), Adaptation-level theory (pp. 287–305). New York: Academic Press. Davidson, R. J. (2001). Toward a biology of positive affect and compassion. In R. J. Davidson and A. Harrington (Eds.), Visions of compassion: Western scientists and Tibetan Buddhists examine human nature. New York and Oxford: Oxford University Press. Dawkins, R. (1998). Unweaving the rainbow: Science, delusion and the appetite for wonder. Boston: Houghton, Mifflin. Dawkins, R. (2006). The god delusion. New York: Bantam.
Faith and Freedom: Traditional and Modern Ways to Happiness
387
Deaton, A. (2008). Income, health, and well-being around the world: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22(2). Diener, E., & Oishi, S. (2000). Money and happiness: Income and subjective well-being across nations. In E. Diener & E. Suh (Eds.), Subjective well-being across cultures, Cambridge, MA: MIT Press. Diener, E., Suh, E., Lucas, E., & Smith, H. (1999). Subjective well-being: Three decades of progress. Psychological Bulletin, 125, 276–302. Diener, E., Lucas, R. E., & Scollon, C. N. (2006). Beyond the hedonic treadmill: Revising the adaptation theory of well-being. American Psychologist, 61, 305–314. Easterlin, R. A. (2005). Feeding the illusion of growth and happiness: A reply to Hagerty and Veenhoven. Social Indicators Research, 75, 429–443. Easterlin, R. A. (1974). Does economic growth improve the human lot? In P. A. David & M. Reder (Eds.), Nations, households and economic growth (pp. 98–125). New York: Academic Press. Ferriss, A. L. (2002). Religion and the quality of life. Journal of Happiness Studies, 3(3), 19–25. Frey, B., & Stutzer, A. (2000). Happiness prospers in democracy. Journal of Happiness Studies, 1, 79–102. Fujita, F., & Diener, E. (2005). Life satisfaction set point: Stability and change. Journal of Personality and Social Psychology, 85, 155–164. Graham, C., & Pettinato, S. (2002). Happiness and hardship: Opportunity and insecurity in new market economies. Washington, DC: The Brookings Institution Press. Graham, C., Chattopadhyay, S., & Picon, M. (2008). The Easterlin and other paradoxes: Why both sides of the debate may be correct. In E. Diener, D. Kahneman, and J. F. Helliwell (Eds.), International differences in well-being. Oxford, UK: Oxford University Press. Hagerty, M., & R. Veenhoven (2006). Rising happiness in nations, 1946–2004. A reply to Easterlin. Social Indicators Research, 79, 421–436. Hagerty, M. R., & Veenhoven, R. (2003). Wealth and happiness revisited: Growing national income does go with greater happiness. Social Indicators Research, 64, 1–27. Haller, M., & Hadler, M. (2004). Happiness as an expression of freedom and selfdetermination: a comparative multilevel analysis. In W. Glatzer, S. von Below, & M. Stoffregen (Eds.), Challenges for quality of life in the contemporary world (pp 207–229). Dordrecht, Netherlands: Kluwer Academic Publishers. Inglehart, R. (1990). Culture shift in advanced industrial society. Princeton: Princeton University Press. Inglehart, R. (1997). Modernization and postmodernization: Cultural, economic and political change in 43 societies. Princeton, NJ: Princeton University Press. Inglehart, R., & Welzel, C. (2005). Modernization, cultural change and democracy: The human development sequence. New York: Cambridge University Press. Inglehart, R., Foa, R., Peterson, C., & Welzel, C. (2008). Social change, freedom and rising happiness: A global perspective, 1981–2007. Perspectives on Psychological Science, 3, 264–285. Inglehat, R., & Klingemann, H.D. (2000). Genes, culture, democracy and happiness. In E. Diener & E. Suh, (Eds.), Subjective well-being across cultures (pp. 165–183). Cambridge, MA: MIT Press. Kahneman, D., & Krueger, A. B. (2006). Developments in the measurement of subjective well-being. Journal of Economic Perspectives, 20, 3–24.
388
Section III: Differences in the Social Context of Well-Being
Kahneman, D., Krueger, A. B., Schkade, D.A., Schwartz, N., & Stone, A. A. (2004). A survey method for characterizing daily life experience: The Day Reconstruction Method. Science, 3, 1776–1780. Kenny, C. (2004). Does development make you happy? Subjective well-being and economic growth in developing countries. Social Indicators Research, 65, 1–22. Limm, C. & Putnam, R. . (2009) ‘‘Praying Alone is not Fun: Religion, Social Networks and Subjective Well-Being.’’ (Harvard University, working paper). Lucas, R. E., Clark, A. E., Georgellis, Y., & Diener, E. (2003). Reexamining adaptation and the set point model of happiness: Reactions to changes in marital status. Journal of Personality and Social Psychology, 84, 527–539. Norris, P., & Inglehart, R. (2004). Sacred and secular: Religion and politics worldwide. New York: Cambridge University Press. Ott, J. (2001). Did the market depress happiness in the U.S.? Journal of Happiness Studies, 2, 433–443 Sen, A. (2001). Development as freedom. New York: Alfred Knopf. Veenhoven, R. (2000). Freedom and happiness: A comparative study in forty-four nations in the early 1990s. In E. Diener & E. Suh (Eds.), Subjective well-being across cultures (pp. 257–288). Cambridge: MIT Press: Welsch, H. (2003). Freedom and rationality as predictors of cross-national happiness patterns: The role of income as a mediating variable. Journal of Happiness Studies, 4, 295–321. Wolfers, J., & Stevenson, B. (2008). Economic growth and subjective well-being: Reassessing the Easterlin Paradox. Brookings Panel on Economic Activity.
Appendix A Subjective Well-being in 52 countries, 1981–2007
Country & Year Argentina (1984) Argentina (1991) Argentina (1995) Argentina (1999) Argentina (2006) Australia (1981) Australia (1995) Australia (2005) Austria (1990) Austria (1999) Belarus (1990) Belarus (1996) Belarus (2000) Belgium (1981) Belgium (1990) Belgium (1999) Brazil (1991) Brazil (1997) Brazil (2006) Bulgaria (1990) Bulgaria (1997) Bulgaria (1999) Bulgaria (2006) Canada (1982) Canada (1990) Canada (2000) Canada (2006) Chile (1990) Chile (1996) Chile (2000)
SWB index 1.69 2.43 2.16 2.65 3.30 3.65 3.47 2.95 3.32 3.68 0.82 2.11 0.92 3.05 3.44 3.40 2.24 2.23 3.25 1.66 1.37 1.10 0.77 3.63 3.01 3.78 3.74 2.65 2.07 2.53
Happiness
Life Satisfaction
2.05 1.93 1.91 1.87 1.80 1.70 1.63 1.73 1.81 1.75 2.54 2.58 2.31 1.75 1.69 1.67 2.06 1.97 1.76 2.67 2.42 2.59 2.40 1.69 1.96 1.61 1.59 1.97 1.94 1.84
6.79 7.25 6.92 7.33 7.79 7.88 7.55 7.28 7.80 8.02 5.52 4.35 4.81 7.37 7.65 7.56 7.39 7.15 7.65 5.03 4.66 5.34 5.22 7.84 7.88 7.80 7.72 7.55 6.91 7.12 (continued )
389
(Continued)
Country & Year Chile (2005) China (1990) China (1995) China (2001) China (2007) Colombia (1997) Colombia (1998) Colombia (2005) Czech Rep. (1990) Czech Rep. (1991) Czech Rep. (1998) Czech Rep. (1999) Denmark (1981) Denmark (1990) Denmark (1999)
Country & Year Estonia (1990) Estonia (1996) Estonia (1999) Finland 1981 Finland (1990) Finland (1996) Finland (2000) Finland (2005) France (1981) France (1990) France (1999) France (2006) E. Germany (1990) E. Germany (1997) E. Germany (1999) E. Germany (2006) W. Germany (1981) W. Germany (1990) W. Germany (1997) W. Germany (1999) W. Germany (2006) Great Britain (1981) Great Britain (1990) Great Britain (1998) Great Britain (1999) Great Britain (2006)
SWB index
Happiness
Life Satisfaction
2.37 2.16 1.97 1.20 1.61
1.92 2.05 1.95 2.13 2.06
4.18 4.19 0.76 1.45 1.18 1.95 3.90 4.07 4.24
1.70 1.65 2.24 2.16 2.10 2.04 1.74 1.64 1.61
7.16 7.29 6.83 6.53 6.76 8.19 8.42 8.31 6.36 6.83 6.39 7.06 8.21 8.16 8.24
SWB index
Happiness
Life Satisfaction
0.02 0.87 0.18 3.15 2.95 3.17 3.20 3.35 1.89 2.21 2.48 2.54 1.62 1.41 2.11 1.81 2.21 2.40 2.30 2.82 2.68 3.51 3.17 3.12
2.42 2.36 2.30 1.90 1.91 1.85 1.87 1.79 1.91 1.84 1.78 1.75 2.05 2.10 2.04 2.04 2.04 1.96 1.97 1.95 1.91 1.67 1.72 1.79
3.68
1.57
6.00 5.00 5.90 7.91 7.68 7.78 7.87 7.84 6.66 6.78 6.93 6.91 6.67 6.64 7.19 6.88 7.25 7.22 7.22 7.70 7.39 7.66 7.47 7.59 7.40 7.60 (continued )
390
(Continued)
Hungary (1982) Hungary (1991) Hungary (1998) Hungary (1999) Iceland (1984) Iceland (1990) Iceland (1999) India (1990) India (1995) India (2001) India 2006 Ireland (1981) Ireland (1990) Ireland (1999) Italy (1981) Italy (1990) Italy (1999) Italy (2005)
1.58 0.34 0.57 0.23 4.11 3.99 4.15 1.52 1.68 0.03 0.85 3.74 3.78 4.12 1.24 2.29 2.06 2.07
2.15 2.28 2.12 2.19 1.60 1.62 1.56 2.07 1.96 2.05 1.98 1.64 1.64 1.62 2.16 2.01 2.05 1.93
6.93 6.03 5.86 5.69 8.09 8.02 8.05 6.70 6.53 5.14 5.79 7.82 7.88 8.17 6.62 7.31 7.17 6.89
Country & Year
SWB index
Happiness
Life Satisfaction
2.02 2.00 1.75 1.83 1.82 2.48 2.27 2.39 2.47 2.44 2.21 2.05 1.74 1.84 1.86 2.05 2.08 1.52 1.51 2.60 2.47 2.52 1.70 1.62 1.59
6.59 6.53 6.72 6.48 6.99 5.70 4.90 5.27 6.01 4.99 5.09 7.95 8.28 8.21 7.97 7.41 7.54 8.13 8.23 3.73 4.57 5.45 7.70 7.76 7.88
Japan (1981) Japan (1990) Japan (1995) Japan (2000) Japan (2005) Latvia (1990) Latvia (1996) Latvia (1999) Lithuania (1990) Lithuania (1997) Lithuania (1999) Malta (1983) Malta (1991) Malta (1999) Mexico (1981) Mexico (1990) Mexico (1996) Mexico (2000) Mexico (2005) Moldova (1996) Moldova (2002) Moldova (2006) Netherlands (1981) Netherlands (1990) Netherlands (1999)
1.62 1.58 2.37 1.96 2.46 0.44 0.78 0.70 0.10 1.10 0.23 2.90 3.95 3.61 3.33 2.32 2.36 4.34 4.48 2.78 1.61 0.85 3.48 3.71 3.89
(continued )
391
(Continued)
Country & Year Netherlands (2006) New Zealnd (1998) New Zealnd (2004) Nigeria (1990) Nigeria (1995) Nigeria (2000) N. Ireland (1981) N. Ireland (1990) N. Ireland (1999) Norway (1982) Norway (1990) Norway (1996) Norway (2007) Poland (1989) Poland (1990) Poland (1997) Poland (1999) Poland (2005) Portugal (1990) Portugal (1999) Romania (1993) Romania (1998) Romania (1999) Romania (2005) Russia (1990) Russia (1995) Russia (1999) Russia (2006) Serbia (1996) Serbia (2001) Serbia (2006) Slovakia (1990) Slovakia (1991) Slovakia (1998) Slovakia (1999) Slovenia (1992) Slovenia (1995) Slovenia (1999) Slovenia (2005) South Africa (1982) South Africa (1990) South Africa (1996) South Africa (2001) South Africa (2007) South Korea (1982) South Korea (1990)
SWB index
Happiness
Life Satisfaction
3.65 3.39 3.80 1.52 2.31 3.32 3.53 3.60 4.13 3.41 3.25 3.25 3.78 1.56 0.67 1.49 1.21 2.38 1.67 2.01 0.05 1.26 1.30 0.33 0.72 1.79 1.59 0.53 0.09 0.21 0.26 0.08 0.89 0.45 0.39 0.38 1.10 2.02 2.18 1.95 0.75 0.65 1.11 2.40
1.64 1.72 1.64 2.02 1.72 1.42 1.66 1.71 1.58 1.80 1.78 1.76 1.67 2.03 2.38 1.98 2.07 1.88 2.17 2.00 2.37 2.45 2.61 2.44 2.46 2.50 2.54 2.24 2.20 2.17 2.31 2.49 2.38 2.25 2.26 2.38 2.15 2.09 2.03 1.95 2.18 1.98 1.88 1.85
1.39
2.14
7.76 7.70 7.89 6.59 6.60 6.87 7.69 7.88 8.07 7.89 7.68 7.66 7.96 6.64 6.53 6.42 6.37 7.02 7.08 6.98 5.88 4.86 5.23 5.75 5.37 4.45 4.74 6.09 5.56 5.62 6.01 6.15 6.81 6.07 6.03 6.29 6.46 7.23 7.24 6.79 6.20 5.59 5.81 7.03 5.33 6.69 (continued )
392
(Continued)
South Korea (1996) South Korea (2001) South Korea (2005) Spain (1981) Spain (1990) Spain (1995) Spain (1999) Spain (2000) Spain (2007) Sweden (1982) Sweden (1990) Sweden (1996) Sweden (1999) Sweden (2006) Switzerland (1989) Switzerland (1996) Switzerland (2007) Taiwan (1994) Taiwan (2006) Turkey (1990) Turkey (1996) Turkey (2001) Turkey (2007) Ukraine (1996) Ukraine (1999) Ukraine (2006) U. S. (1982) U. S. (1990) U. S. (1995) U. S. (1999) U.S. (2006) Uruguay (1996) Uruguay (2006)
2.00 2.04 2.01 2.02 1.96 1.95 1.94 1.94 1.95 1.76 1.64 1.66 1.71 1.61 1.70 1.66 1.64 1.81 1.96 1.92 1.63 2.08 1.81 2.55 2.56 2.17 1.78 1.71 1.60 1.68 1.62 2.02 1.85
1.12 1.34 1.57 2.26 1.73 2.29 2.16 2.45 3.62 3.89 3.64 3.37 3.72 4.14 4.00 3.91 2.06 1.68 1.60 2.14 0.42 2.94 2.41 1.78 0.30 3.24 3.48 3.68 3.47 3.52 2.03 2.83
393
6.21 6.35 6.60 7.14 6.61 7.09 6.99 7.32 8.01 7.97 7.77 7.65 7.74 8.39 8.14 8.01 6.56 6.58 6.41 6.20 5.62 7.46 3.95 4.56 5.67 7.67 7.76 7.67 7.65 7.57 7.06 7.46
Appendix B Changes in mean scores on ‘‘Importance of God’’ scale, 1981–2007
Argentina (1984) Argentina (1991) Argentina (1995) Argentina (1999) Argentina (2006) Australia (1981) Australia (1995) Australia (2005) Austria (1990) Austria (1999) Belgium (1981) Belgium (1990) Belgium (1999) Brazil (1991) Brazil (1997) Brazil (2006) Bulgaria (1990) Bulgaria (1997) Bulgaria (1999) Bulgaria (2006) Belarus (1990) Belarus (1996) Belarus (2000) Canada (1982) Canada (1990) Canada (2000) Canada (2006) Chile (1990) Chile (1996)
7.00 7.91 8.37 8.49 8.32 6.17 5.65 6.09 6.36 6.65 5.84 5.31 5.39 9.43 9.61 9.63 3.56 4.88 5.15 5.70 4.15 5.83 5.97 7.29 6.89 7.45 7.42 8.62 8.65 (continued )
394
(Continued)
Chile (2000) Chile (2005) China (1990) China (2007) Taiwan (1994) Taiwan (2006) Colombia (1998) Colombia (2005) Czech Rep (1991) Czech Rep (1998) Czech Rep (1999) Denmark (1981) Denmark (1990) Denmark (1999) Finland 1981 Finland (1990) Finland (1996) Finland (2000) Finland (2005) France (1981) France (1990) France (1999) France (2006) Germany (1990) Germany (1997) Germany (1999) Germany (2006) Hungary (1982) Hungary (1991) Hungary (1998) Hungary (1999) Iceland (1984) Iceland (1990) Iceland (1999) India (1990) India (1995) India (2001) India 2006 Ireland (1981) Ireland (1990) Ireland (1999) Italy (1981) Italy (1990) Italy (1999) Italy (2005) Japan (1981) Japan (1990) Japan (1995)
8.75 8.66 1.62 3.58 5.52 6.13 9.56 9.67 3.54 4.01 3.76 4.14 3.92 4.02 5.60 5.60 5.86 5.76 6.01 4.53 4.44 4.37 4.68 4.68 4.36 4.39 4.48 4.84 5.41 5.40 5.24 6.23 6.11 6.24 7.53 8.30 8.53 7.52 7.98 7.93 7.74 6.77 7.05 7.43 7.84 4.79 4.79 4.88 (continued )
395
(Continued)
Japan (2000) Japan (2005) Mexico (1981) Mexico (1990) Mexico (1996) Mexico (2000) Mexico (2005) Moldova (1996) Moldova (2002) Moldova (2006) Netherlands (1981) Netherlands (1990) Netherlands (1999) Netherlands (2006) N. Zealand (1998) N. Zealand (2004) Nigeria (1990) Nigeria (1995) Nigeria (2000) Norway (1982) Norway (1990) Norway (1996) Norway (2007) Poland (1989) Poland (1990) Poland (1999) Poland (2005) Portugal (1990) Portugal (1999) Romania (1993) Romania (1998) Romania (1999) Romania (2005) Russia (1990) Russia (1995) Russia (1999) Russia (2006) Slovakia (1991) Slovakia (1998) Slovakia (1999) Slovenia (1992) Slovenia (1995) Slovenia (1999) Slovenia (2005) South Africa (1982) South Africa (1990) South Africa (1996) South Africa (2001)
5.02 5.01 8.99 8.13 7.89 9.39 9.43 6.99 7.51 8.13 5.17 4.89 4.92 4.69 5.67 5.35 9.73 9.63 9.62 5.19 4.55 4.65 4.21 8.78 8.43 8.37 8.70 7.10 7.83 7.45 8.03 8.63 9.17 4.00 5.39 5.34 6.02 5.97 6.60 6.63 4.75 5.17 5.02 5.42 8.54 8.99 9.12 9.11 (continued )
396
(Continued)
South Africa (2007) Spain (1981) Spain (1990) Spain (1995) Spain (1999) Spain (2000) Spain (2007) Sweden (1982) Sweden (1990) Sweden (1996) Sweden (1999) Sweden (2006) Switzerland (1989) Switzerland (1996) Switzerland (2007) Turkey (1990) Turkey (1996) Turkey (2001) Turkey (2007) Ukraine (1996) Ukraine (1999) Ukraine (2006) Great Britain (1981) Great Britain (1990) Great Britain (1998) Great Britain (1999) Great Britain (2006) U.S. (1982) U.S. (1990) U.S. (1995) U.S. (1999) U.S. (2006) Uruguay (1996) Uruguay (2006) N Ireland (1981) N Ireland (1990) N Ireland (1999)
9.16 6.32 6.07 6.88 5.97 5.83 5.34 4.08 3.75 3.94 4.09 3.93 6.68 6.10 6.35 8.84 9.18 9.16 9.36 5.96 6.29 7.19 5.43 5.37 5.28 4.91 5.58 8.34 8.06 8.22 8.53 8.40 6.66 7.32 7.44 7.82 6.94
397
Chapter 13 The Impact of Time Spent Working and Job Fit on Well-Being Around the World James K. Harter and Raksha Arora Gallup, Inc.
Is work a necessary evil, mostly comprising drudgery, grim duties, and toil? If so, more work should mean lower well-being, all other variables held constant. Here in this chapter we present analyses examining the relationship between time spent working and well-being. In a subset of countries spread across seven regions in the first wave of the Gallup World Poll, we have asked representative samples of respondents how much time they typically work in a week, and how much time they worked yesterday. Coupled with these data on work time are subjective assessments of life evaluation and daily experiences (positive and negative affect and daily experiences such as being treated with respect, choice, pride, feeling rested, etc). Additionally, we include subjective assessments of the respondent’s job fit (satisfaction with their job or work, and having the opportunity to do what they do best every day). From these data, we are better able to understand the relationships between the experience of work, work-time, and well-being. In addition, we were able to study the interactive nature of these relationships, examining whether the relationship between time spent working and well-being depends on the congruence between the worker and the work. Work is imbedded in a complex mix of domains that impact our wellbeing. Given its dominance of awake time for most adults, work is a central variable in our lives. Thousands of research studies have documented variables that impact quality work experience and performance. But more research is needed to understand how our use of work time interacts with quality experience to give us the best chance for a good life. 398
Work and Well-Being Around the World
399
In addition to the obvious link between work and our pecuniary needs, research has uncovered other relationships between work and well-being. For instance, there is evidence that job satisfaction and mood at work impacts or spills over to mood after work (Judge & Ilies, 2004) and that justice at work impacts the ability of individuals to manage both their work and family lives, thereby impacting levels of stress. Further, longitudinal research has revealed substantial relationships between attitudes at work and health outcomes such as risk of coronary heart disease (Kivimaki et al., 2005). Since day reconstruction research is showing work time to be among the least pleasurable times of the day (where the peak emotion is frequently a negative emotion; Krueger & Kahneman, 2006), the interaction of quantity and quality of work is particularly important. Time at Work
Two lines of research attempt to describe how quantity and quality of work might impact workers ill-being or well-being. The first line originated with the study of stress and health and is best represented by the theory of person–environment fit (see French, Caplan, & Van Harrison, 1982). Here it is argued that worker performance and quality of life are hindered by strain or boredom. When demands exceed or fall below the resources, individuals experience undesirable states (e.g., strain or boredom) that detract from the quality and quantity of performance as well as their well-being. From this perspective, a healthy work force means the absence of strain or boredom (see also Edwards, Caplan, & Van Harrison, 1998). As Andrew Clark reports in Chapter 14, not having a job when you want one is a major source of low well-being. But having too much work can present a different set of challenges. As defined by the International Labor Organization (ILO), ‘‘hours of work relates to any period of time spent on activities which contribute to the production of goods and services.’’ While the hours worked have declined since the dawn of the industrial revolution, and the 40-hour work week is the norm for most countries, there is still much cross-cultural variation in hours of work. While the countries of Western Europe have shorter work weeks on average on account of policy and institutional frameworks, legal mandates on number of hours worked, tax regimes, and a public preference for leisure time, this has not been the case in other industrialized nations—the United States, Japan, and Korea being notable exceptions. The implications of hours worked for economic productivity, leisure time, social life, family life, health, and well-being of workers are wide-ranging, and there is a considerable literature on these subjects. A large number of these studies focus on how long hours at work negatively impact the marriage and family life of workers and result in
400
Section III: Differences in the Social Context of Well-Being
greater stress, disease burden, and occupational injury. In particular overwork has been linked to greater risk of hypertension, heart disease, stress, fatigue, depression, and poor lifestyle habits, including smoking, poor diet, and insufficient exercise (Spurgeon, A., Harrington, J. M., Cooper, C. L., 1997; Sparks, K., Cooper, C., Fried, Y., & Shirom, A., 1997; Crouter, A. C., Bumpus, M. F., Head, M. R., & McHale, S. M., 2001; Dembe, A. E., Erickson, J. B., Delbos, R. G., & Banks, S. M., 2005). However some research also shows that the terms and conditions of one’s work mediate the relationship between one’s work and such negative outcomes. Factors such as job-satisfaction, autonomy, flexibility and whether the nature of overtime is voluntary or mandatory have been shown to mediate the relationship between overwork and well-being (Gray, M., Qu, L., Stanton, D., Weston, R., 2004: Golden, L. Wiens-Tuers, B. 2005). A second line of research on worker quality of life and performance originated with the behavioral, cognitive, and health benefits of positive feelings and positive perceptions (Isen, 1987; Warr, 1999). This perspective says the presence of positive emotional states or experiences accentuates the worker’s well-being and performance. When workers have interesting, meaningful, and challenging tasks, they are likely to have what Maslow (1943) described as ‘‘self actualization’’ and what Csikszentmihalyi (1997) has described as optimal states or ‘‘flow.’’ When job demands and the individual are aligned, people experience positive emotional states (e.g., pleasure, joy, energy) and they perceive themselves as engaged, productive, and progressing. Surveys of generations of employees show many workers desire greater meaning and personal development from their work and suggest many workers see their work as a calling—enjoyable, socially useful, and fulfilling (Avolio & Sosik, 1999; Wrzesniewski, McCauley, Rozin, & Schwartz, 1997). This implies that workers can find work that is meaningful to them personally. A growing body of research now suggests workers and their work can become aligned both through the matching of their talents to the tasks they perform, and through the environment that exists around them.
Job Fit
Organizational psychologists have defined the concept of ‘‘job fit’’ as the fit between the abilities or desires of a person and the job attributes or demands (Edwards, 1991). Abilities can be thought of as personality, mental ability, and a variety of skills. Desires include goals, psychological needs, interests, and values.
Work and Well-Being Around the World
401
Research on the psychological fit of the person to the job has both genetic and environmental implications. Studies of identical twins reared apart show a high genetic component in such attributes as interests, personality, and cognitive characteristics (Bouchard, Lykken, McGue, Segal, & Tellegen, 1990; Bouchard, 1997). Since jobs vary substantially in what they require the person to do (from self-subsistence farming, to educating children, to conducting research studies, to running a company), it is readily apparent that potential efficiency within any given job is partially dependent on matching talents to tasks. That is, leveraging the individual differences by aligning traits with requirements for the job. In economically developed countries, systematic selection systems have proven to affect the fit of the person to the job as measured through prediction of job performance, turnover, engagement, customer ratings, safety, and financial outcomes (Schmidt & Hunter, 1998; Schmidt & Rader, 1999; Harter, Hayes, & Schmidt, 2004). These findings, documenting metaanalytic predictive validity effect sizes that often range from .20–.50, reveal substantial practical economic benefits for organizations that use systematic selection processes. Interest in work, research suggests, is influenced by both the psychological fit of the person to the job and the surrounding environment. Largescale meta-analyses have shown that work attitudes correlate with performance at the individual level (Judge, Thoresen, Bono, & Patton, 2001) and at the business-unit level (Harter, Schmidt, & Hayes, 2002; Harter, Schmidt, & Keyes, 2002; Harter, Schmidt, Killham, & Asplund, 2006). These correlations are in the .30–.45 range (somewhat higher at the business or work unit–level than the individual level). The fact that job satisfaction and work engagement help explain performance means what is good for people is often good for organizational economics. Over the past two decades, these meta-analyses have found linkages between work attitudes and various outcomes, including profitability, productivity, customer perceptions of service, employee turnover, absenteeism, theft, and safety (including risk adjusted mortality rates in health care). Various work environment elements have been shown to exhibit consistent importance across supervisor-employee work situations in developed countries, such as making expectations clear, recognizing good work, encouraging workers’ development, listening to opinions, and getting workers the right materials. But as one considers the context of work across the globe, not all work situations resemble the typical organizational setting that has been studied in the thousands of industrial-organizational psychology studies over the decades. In addition to general satisfaction with work, one element that transcends the work situation across the globe, from self-subsistence work to organized labor, is ‘‘having the opportunity
402
Section III: Differences in the Social Context of Well-Being
to do what you do best every day.’’ This implies the matching of talents to tasks, or the fit of the person to the job. At its core, this element has its roots in strengths theory, first researched by Donald Clifton (Clifton, Hollingsworth, & Hall, 1952). Strengths theory can be seen as ‘‘identification of positive personal and interpersonal traits (talents) in order to position and develop individuals to increase the frequency of positive subjective experience’’ (Clifton & Harter, 2003, p.114). The approach suggests people can develop most efficiently through their natural talents, by integrating learning and skills to inherent talents, as opposed to identifying and attempting to fix the weak or missing traits. The theory does not imply ignorance of weaknesses, but is a shift in focus. Follow-up research suggests employees whose managers focus on their strengths have more than double the probability of being engaged relative to those whose managers either ignore them or focus on their weaknesses. Organizations that have used strengths-based interventions have seen significantly higher employee engagement growth in comparison to a control group of organizations (Clifton & Harter, 2003). And based on a study of over 65,000 respondents, those who received a strengths intervention experienced 14.9 percent less turnover than a control group, even after controlling for job-type and tenure (Asplund, Lopez, Hodges, & Harter, 2007). Whether viewed through the lens of selection or development, the notion of job fit is one that applies to a variety of work situations around the globe. Given the capability and breadth of recent survey work around the globe, we now have the opportunity to study the impact of job fit on wellbeing across a wide variety of regions and work settings, from economically developed to self-subsistence situations. Going back to the organizational psychology definition of ‘‘person–job fit,’’ our definition here is intentionally broad. For purposes of this world study, examining a wide variety of work settings in a population survey format, our definition is closest to ‘‘perceived fit,’’ as opposed to ‘‘actual fit.’’ Estimating ‘‘actual fit’’ would involve more extensive job analysis for each job studied (Edwards, 1991). As outlined above, the concepts included in our present definition of ‘‘perceived job-fit’’ have meta-analytic evidence of their importance in a variety of work settings.
Well-Being and the World Poll
This study focuses on the hedonic well-being of workers. Hedonic wellbeing has been defined as ‘‘the study of what makes experiences and life pleasant or unpleasant. It is concerned with feelings of pleasure and pain, of interest and boredom, of joy and sorrow, and of satisfaction and dissatisfaction. It is also concerned with the whole range of circumstances, from
Work and Well-Being Around the World
403
the biological to the societal, that occasion suffering and enjoyment’’ (D. Kahneman, E. Diener, & N. Schwarz, 1999). The two elements of hedonic well-being represented in the Gallup World Poll include evaluative well-being and experienced well-being. Evaluative well-being is measured by the Cantril striving scale, which asks respondents to rate their lives on the Ladder of Life (‘‘If the top step is 10 and the bottom step is 0, on which step of the ladder do you feel you personally stand at the present time?’’). Experienced well-being, on the other hand, is measured by a series of questions relating to feelings and emotions experienced in the time-frame of the last twenty-four hours.
Research Questions
The primary research questions examined in this chapter pertain to the relationships between hours worked, job fit, and both conceptualizations of well-being discussed above. The primary research questions are as follows:
• To what extent does hours worked (weekly and daily) impact wellbeing (life evaluation and daily experience)?
• To what extent does job fit (satisfaction with job/work and opportunity to do what you do best) impact well-being (life evaluation and daily experience)? • Is there an interaction, such that optimum number of hours worked depends on the degree of the worker’s job fit? • Do the answers to the above questions vary for respondents in different regions of the world? Methods
The Gallup World Poll uses two primary methodological designs: A random-digit-dial (RDD) telephone survey design is used in countries where 80 percent or more of the population has landline phones. This situation is typical in the United States, Canada, Western Europe, Japan, Australia, etc. In the developing world, including much of Latin America, the former Soviet Union countries, nearly all of Asia, the Middle East, and Africa, an area frame design is used for face-to-face interviewing. In countries where face-to-face surveys are conducted, census listings of Primary Sampling Units (PSU), consisting of clusters of households, are used as the basis for random sampling. In the RDD surveys, at least five call attempts are made to reach a person, aged 15 and older, in each household. Typically the design is not stratified, but otherwise, the other processes and
404
Section III: Differences in the Social Context of Well-Being
procedures follow those used in the face-to-face design. Once data are collected, aggregate data are scientifically weighted according to national population parameters. The typical World Poll survey in a country consists of 1,000 completed questionnaires. However, in some countries, oversamples are collected in major cities. Prior research suggests respondents to aural modes are significantly more likely than are respondents to visual modes to give extreme responses (Dillman et al., 2001). To reduce possible mode effects, both in-person and telephone interviews utilized verbal/aural (rather than visual) response choices and scales. The analyses presented in this chapter are conducted at the individual level. To alleviate concerns about cross-cultural differences across countries, we standardized all variables within each country, prior to conducting analyses across countries and regions of the world. To increase sample size and statistical power, we combined data across countries into regions. We include data from seven regions of the world, but given limitations on income data, did not include data from the Middle East.
Region Definitions
Table 13.1 shows the countries comprising each of the seven regions included in this analysis. While these represent the total number of countries in a region where polling was conducted in wave 1 of the Gallup World Poll, the countries eventually included in the regression analysis results are somewhat fewer on account of the listwise deletion of cases. The African countries include all the countries of sub-Saharan Africa where Gallup polls were conducted in the 2006 wave. The East Asian region has been defined broadly to contain Hong Kong, China, Singapore, Japan, South Korea, and Taiwan. The Eastern Europe and Former Soviet Union region includes countries from the former Union of Soviet Socialist Republics, as well as the Baltic states, post-Communist Balkan states and the Central and Eastern European countries of the former Eastern bloc. The Western countries have been combined into one regional group that includes North America, Western Europe, and Australia and New Zealand. The South Asian region includes the south and southeastern countries of Bangladesh, India, Sri Lanka, Vietnam, Thailand, Cambodia, Laos, Myanmar, Malaysia, Nepal, and Indonesia. The Southern Europe region includes Italy, Greece, Israel, Cyprus, Portugal, and Spain, while the Latin American region consists of twenty countries in Central and South America where data were collected in wave 1.
TABLE 13.1
Region Definitions
405
Africa
East Asia
• Nigeria • Kenya • Tanzania • Ghana • Uganda • Benin • Madagascar • Malawi • South Africa • Angola • Botswana • Ethiopia • Mali
• Hong Kong • Singapore • Japan • China • South Korea • Taiwan
Eastern Europe, FSU
• Hungary • Czech Republic • Romania • Belarus • Georgia • Kazakhstan • Kyrgyzstan • Moldova • Russia • Ukraine • Albania • Armenia • Azerbaijan
Latin America
• Brazil • Mexico • Costa Rica • Argentina • Bolivia • Chile • Colombia • Cuba • Dominican Republic
• Ecuador • El Salvador • Guatemala
Western Europe, North America, Australia
• United Kingdom • France • Germany • Netherlands • Belgium • Sweden • Denmark • Canada • Australia • New Zealand • Austria • Finland
South Asia
• Bangladesh • India • Sri Lanka • Vietnam • Thailand • Cambodia • Laos • Myanmar • Malaysia • Nepal • Indonesia
Southern Europe
• Italy • Greece • Israel • Cyprus • Portugal • Spain
(continued )
TABLE 13.1
(Continued)
Africa
406
• Mauritania • Mozambique • Niger • Rwanda • Senegal • Zambia • Burkina Faso • Cameroon • Sierra Leone • Zimbabwe • Burundi • Chad • Togo
East Asia
Eastern Europe, FSU
• Bosnia • Herzegovina • Bulgaria • Croatia • Estonia • Latvia • Lithuania • Macedonia • Montenegro • Serbia • Slovakia • Slovenia • Tajikistan • Uzbekistan • Kosovo • Poland
Latin America
• Honduras • Nicaragua • Panama • Paraguay • Peru • Puerto Rico • Uruguay • Venezuela
Western Europe, North America, Australia
• Ireland • Norway • Switzerland • United States
South Asia
Southern Europe
Work and Well-Being Around the World
407
Study Variables
The demographic variables used in the regression analysis include age, gender (coded male = 0, female = 1), education (coded elementary = 1, some high school = 2, high school or above = 3), income (log of household income), and marital status (dummy coded, single = 1, other = 1, divorced/ separated/widowed = 1, other = 0). In a separate analysis, in order to control for the possible nonlinear relationship between age and the well-beingdependent variables, we also added a quadratic term for age in each of the overall regression models—however, this did not change any of the main results. After entering the demographic variables, we also included a block of subjective variables independent of the primary study variables, including items relating to health problems (‘‘Do you have health problems that prevent you from doing any of things people your age normally can do?’’ yes = 1, no = 0), safety (‘‘Do you feel safe walking alone at night in the city or area where you live?’’ yes = 1, no = 0), social networks (‘‘If you were in trouble, do you have relatives or friends you can count on to help you whenever you need them, or not?’’ yes = 1, no = 0) and community satisfaction (‘‘Are you satisfied or dissatisfied with the city or area where you live?’’ Yes = 1, no = 0). Among the primary study variables, the three work-related independent variables used are hours worked yesterday, hours worked in a typical week, and job fit. We defined job fit as the sum of responses on two dichotomous items—job satisfaction (‘‘Are you satisfied or dissatisfied with your job or the work you do?’’ yes = 1, no = 0) and doing what one does best at work (‘‘In your work, do you have an opportunity to do what you do best every day or not?’’ yes = 1, no = 0). The well-being-dependent variables used are the Ladder of Life item (‘‘Please imagine a ladder with steps numbered from zero at the bottom to ten at the top. Suppose we say that the top of the ladder represents the best possible life for you and the bottom of the ladder represents the worst possible life for you. Which step comes closest to the way you feel?’’ scored 0–10) which represents the evaluative well-being (life evaluation). The positive and negative experience/affect variables summarize results from a series of items about yesterday, which relate to individual positive and negative emotions and experiences reported by the respondent over the course of previous day. We have defined positive experience as the average of items that measure whether the respondent had feelings of enjoyment, smiling and laughter, respect, being well-rested, pride, learning, the ability to choose how one’s time was spent and would like more days like yesterday (8 items, each scored yes = 1, no = 0). The negative experience variable has been defined as the mean of the emotion items measuring the experience of feelings of boredom, depression, anger, stress, sadness, and worry yesterday (six items, each scored yes = 1, no = 0). The two sets of daily
408
Section III: Differences in the Social Context of Well-Being
experience items (positive and negative) were confirmed as distinct through factor analysis. Cronbach’s alpha reliabilities (at the individual level) are .77 (positive experience) and .75 (negative experience). Analyses
Table 13.2 provides unstandardized descriptive statistics for each region for the focal study variables (time spent working, job fit, and life TABLE 13.2
Descriptives by Region Hours Worked Per Week
Africa East Asia Eastern Europe, FSU Latin America Western Europe, North America, Australia South Asia Southern Europe
N
Mean
Standard Deviation
9160 2232 8418 7514 5099 6064 1570
45.63 46.25 45.42 51.07 39.55 50.66 45.49
19.09 17.32 16.26 20.07 15.09 20.48 15.71
Hours Worked Yesterday
Africa East Asia Eastern Europe, FSU Latin America Western Europe, North America, Australia South Asia Southern Europe
N
Mean
Standard Deviation
9395 2839 9323 8048 9849 6548 3051
6.50 6.40 7.04 6.58 5.25 7.19 5.40
3.40 4.23 4.39 4.32 4.53 3.45 4.50
Job Fit
Africa East Asia Eastern Europe, FSU Latin America Western Europe, North America, Australia South Asia Southern Europe
N
Mean
Standard Deviation
9287 2797 12550 7988 9664 6829 2992
1.30 1.54 1.47 1.65 1.72 1.45 1.69
0.78 0.68 0.75 0.63 0.55 0.72 0.60 (continued )
Work and Well-Being Around the World
TABLE 13.2
409
(Continued)
Life Evaluation N Africa East Asia Eastern Europe, FSU Latin America Western Europe, North America, Australia South Asia Southern Europe
Mean
9464 2850 13246 8111 9865 7078 3034
Standard Deviation
4.17 6.06 5.38 5.95 7.41 4.96 6.61
1.80 1.80 1.92 2.31 1.46 1.82 1.95
Negative Experiences N Africa East Asia Eastern Europe, FSU Latin America Western Europe, North America, Australia South Asia Southern Europe
9504 2866 13276 8167 9877 7142 3074
Mean
Standard Deviation
0.22 0.26 0.22 0.25 0.20 0.21 0.25
0.27 0.27 0.26 0.28 0.23 0.28 0.26
Positive Experiences N Africa East Asia Eastern Europe, FSU Latin America Western Europe, North America, Australia South Asia Southern Europe
9520 2866 13350 8177 9881 7158 3079
Mean 0.68 0.65 0.62 0.77 0.74 0.64 0.68
Standard Deviation 0.28 0.25 0.30 0.25 0.22 0.29 0.25
evaluation). Table 13.A provides unstandardized descriptive statistics and intercorrelations for all study variables (correlations were calculated across respondents after standardization within countries). Average hours worked for each country included in the study are included in Table 13.B. In addition to examining bivariate relationships, we conducted hierarchical regression analysis (see Tables 13.4A–C) for each well-being-dependent variable (life evaluation, positive and negative daily experiences). We wanted to understand the contribution of work hours and job fit to well-being after accounting for demographics and other important subjective domains. We first entered a block of demographic variables (age, gender, household
410
Section III: Differences in the Social Context of Well-Being
income, marital status, and education). Next, we entered a block of subjective domains (perceptions of health, safety, relationships, and community satisfaction). We next entered hours worked, followed by job fit, and the hours worked–job fit interaction term, each as separate blocks. When examining life evaluation as the dependent variable, we used ‘‘hours worked per week’’ as the ‘‘hours worked’’ variable. When examining daily experiences (both positive and negative), we used ‘‘hours worked yesterday’’ as the ‘‘hours worked’’ variable. Examining the bivariate correlations in the appendix, ‘‘hours worked yesterday’’ correlates more highly with positive and negative affect/daily experience than with life evaluation. The hierarchical regression analysis was completed for each of the three dependent variables for the overall sample, and then again for each region of the world. To look more closely at the impact of hours worked and job fit on specific daily experiences, we also conducted hierarchical regression analysis, treating each daily experience item as a dependent variable.
Results
Average hours worked per week exceeds 50 hours for individuals sampled in Latin America and South Asia, and just below 40 hours per week for workers in Western Europe and Australia (hours worked per week was not available for North America and many Western European countries, as the variable was added partially through the data collection period of World Poll wave 1). Hours worked yesterday exceeded 7 hours for respondents in South Asia and Eastern Europe/Former Soviet Union (Latin Americans worked 6.6 hours), and below six hours for respondents in Western Europe/ North America/Australia, and Southern Europe. Job fit is much more predominant in the more developed parts of the world, such as Western Europe, North America, Australia, Latin America, and Southern Europe; and less frequent in sub-Saharan Africa, South Asia, and Eastern Europe/FSU. It is most likely the case that in less economically developed parts of the world, residents are less able to be selective about what they do to earn money, relying either on self-subsistence, difficult physical work, and perhaps less advanced business or supervisory practices for those in organizations. As others have reported (Deaton, 2007) there are substantial differences in life evaluation across regions of the world, with average ratings below 5 on the ladder of life in sub-Saharan Africa and South Asia, and above 7 in Western Europe, North America, and Australia. However, daily experiences differ much less across regions with slightly more negative daily experiences in East Asia, Latin America, and Southern Europe, and fewer
Work and Well-Being Around the World
411
negative daily experiences in Western Europe, North America, and Australia. Positive daily experiences are expressed more frequently in Latin America and Western Europe/North America/Australia, and less frequently in Eastern Europe/FSU. Still, there are wide ranges in daily experiences across individuals within all regions of the world. Table 13.3 presents the correlations between hours worked, job fit, and well-being for each region. As anticipated, more hours worked are associated with lower well-being (lower life evaluation, less positive daily experience/ affect, and more negative daily experience/affect). These findings are relatively, but not entirely, robust across regions. The most striking difference is in Africa, where the pattern is the opposite of other regions on life evaluation (the Ladder of Life). More hours worked per week are associated with higher life evaluation. This may be reflective of more work relating to higher status when the well-being norm is low for a region. But, interestingly, putting in more work hours on a given day does detract from positive experience, but does not increase negative experience in Africa. In Western Europe and Australia, the number of hours worked yesterday and per week is unrelated to life evaluation but is related to positive and negative daily experiences. Hours worked per week, and on any given day, do not seem to impact ratings on the Ladder of Life, but more hours worked coincides with less positive and more negative daily experiences. In South Asia, the correlations are also close to zero between hours worked and life evaluation, and are only slightly related to daily experiences/affect. Across regions, the relationship between job fit and well-being is remarkably consistent. Higher subjective job fit is associated with higher life evaluation, more positive daily experiences, and less negative daily experiences, in every region. In the lower part of Table 13.3, we also report the correlation between job fit and hours worked. In East Asia, individuals with more hours worked perceive lower job fit. In the West (North America, Western Europe, and Australia), we find just the opposite, where those with more hours worked have higher perceived job fit. But, regardless of the bivariate relationships, it could be argued that perceptions of job fit are also determined by other objective factors, such as one’s income, age, marital status, and education, and subjective factors in other domains such as feelings of safety, relationships, surrounding community, or health. So the next step was to study the independent contribution of job fit to well-being after controlling for these other variables. Second, it is important to understand whether job fit and hours worked interact in contributing to well-being. After entering demographic and subjective domain blocks into the regression analysis, number of hours worked explained significant variance in life evaluation, as did job fit. But the interaction term was nonsignificant.
TABLE 13.3
Region Level Correlations
Correlations Between Hours Worked Yesterday and Well-Being Variables Region Africa East Asia Eastern Europe, FSU Latin America Western Europe, North America, Australia South Asia Southern Europe
Life Evaluation
Positive Experiences
Negative Experiences
.06** .07** .05** .05** .01
.06** .20** .20** .10** .17**
.01 .16** .10** .05** .15**
.00 .06**
.03* .20**
.01 .12**
Correlations Between Hours Worked Per Week and Well-Being Variables Region Africa East Asia Eastern Europe, FSU Latin America Western Europe, North America, Australia South Asia Southern Europe
Positive Experiences
Negative Experiences
.04** .08** .03** .02 .01
.01 .11** .07** .02 .03*
.01 .04 .05** .04** .04**
.06** .12**
.06** .05*
.03* .04
Life Evaluation
Positive Experiences
Negative Experiences
.15** .18** .19** .18** .15**
.25** .35** .28** .28** .18**
.16** .16** .21** .21** .17**
.22** .18**
.29** .24**
.20** .23**
Life Evaluation
Correlations Between Job Fit and Well-Being Variables Region Africa East Asia Eastern Europe, FSU Latin America Western Europe, North America, Australia South Asia Southern Europe **p<.01 *p<.05
Correlations Between Job Fit and Hours Worked
Region
Hours Worked Yesterday
Hours Worked Per Week
Africa 0.01 .03(*) East Asia .06(**) .10(**) E. Europe, FSU .00 0.01 Latin America 0.02 0.01 W. Europe, North America, .03(**) .08(**) Australia South Asia .03(*) 0.01 Southern Europe 0.01 0.01 Correlations calculated after first standardizing variables within each country. Sample Sizes range from 2783 to 12522
412
Work and Well-Being Around the World
Low Job Fit 10.00 9.00 8.00 7.00 6.00 5.00 4.00 3.00 2.00 1.00 0.00
413
High Job Fit
5.96
5.94
5.84
6.03
5.90
5.97
5.78
5.78
5.66
5.60
4.41
4.39
4.56
4.58
4.47
4.63
4.41
4.33
4.06
4.13
<20 20-29 30-34 35-39 40 41-44 45-49 50-54 55-60 >60 Hours Hours Hours Hours Hours Hours Hours Hours Hours Hours
Figure 13.1 Life Evaluation for Workers with High and Low Job Fit by Hours Worked in a Typical Week Note: Figure presents unstandardized marginal means on life evaluation (ladder of life) adjusted based on age, gender, log income, marital status, and subjective domains (health, community, relationships, and safety).
As such, respondents with fewer hours worked and better job fit had higher life evaluation. Higher job fit was associated with better life evaluation regardless of hours worked. Marginal means on the Ladder of Life (after controlling for demographics and other subjective domains) are illustrated in Figure 13.1, both for those with high job fit (satisfied with their work, and have the opportunity to do what they do best) and low job fit (dissatisfied and do not have the opportunity to do what they do best). These results show that life evaluation peaks for respondents with high job fit and between 35 and 44 hours of work per week. Hours worked explains differences of less than one half step on the ladder of life, if one compares respondents with typical work hours (35–44 hours) with those with extremely high work burden (55 or more hours). Job fit explains approximately 1.5 steps on the ladder of life, if one compares the differences between low and high job fit across the hours-worked continuum in Figure 13.1. The prediction of hours worked and life evaluation was not consistent for every region. In Africa, Latin America, and Southern Europe, for instance, job fit predicted life evaluation, whereas hours worked did not. In East Asia, Eastern Europe, Western Europe/Australia, and South Asia, both hours worked and job fit predicted life evaluation (hours worked negatively and job fit positively). Hours worked and job fit interact significantly in East Asia and Western Europe/Australia. As hours worked increased in these regions, job fit was a more important variable in defining life evaluation on the Ladder of Life. Respondents in these regions with poorer job fit have increasingly lower levels of well-being as the number of hours worked increased.
TABLE 13.4A
Hierarchical Regression Analysis: Dependent Variable – Life Evaluation (Ladder of Life)
1. Overall Sample Beta (Sig.)
414
Age Female Log Household Income Single Divorced/Separated Education Health Problems Safety Count on Others Community Satisfaction Hours Worked Per Week Job Fit Hours Worked Per Week, Job Fit Interaction Term R R Square R Square Change F Change Sig. F Change
0.03 (.000) 0.03 (.000) 0.22 (.000) 0.02 (.006) 0.02 (.000) 0.10 (.000)
0.28 0.08 0.08 353.66 0.00
Beta (Sig.)
Beta (Sig.)
0.02 (0.01) 0.04 (.00) 0.21 (.00) 0.02 (.00) 0.02 (.00) 0.10 (.00) 0.05 (.00) 0.04 (.00) 0.09 (.00) 0.09 (.00)
0.02 (0.01) 0.04 (.00) 0.21 (.00) 0.02 (.00) 0.02 (.00) 0.10 (.00) 0.05 (.00) 0.04 (.00) 0.08 (.00) 0.09 (.00) 0.03 (.00)
0.32 0.10 0.02 153.11 0.00
0.32 0.10 0.00 23.49 0.00
Beta (Sig.)
Beta (Sig.)
0.03 (.00) 0.04 (.00) 0.20 (.00) 0.02 (.00) 0.02 (.00) 0.09 (.00) 0.05 (.00) 0.03 (.00) 0.07 (.00) 0.07 (.00) 0.03 (.00) 0.11 (.00)
0.03 (.00) 0.04 (.00) 0.20 (.00) 0.02 (.00) 0.02 (.00) 0.09 (.00) 0.05 (.00) 0.03 (.00) 0.07 (.00) 0.07 (.00) 0.03 (.00) 0.11 (.00) 0.01 (.09)
0.33 0.11 0.01 323.44 0.00
0.33 0.11 0.00 2.86 0.09
2. Significant Effects by Region (Beta, Sig.) Africa
East Asia
Eastern Europe, FSU
415
0.04 (.01)
Hours Worked Per Week
0.01 (.29)
0.08 (.00)
Job Fit
0.09 (.00)
0.16 (.00)
0.12 (.00)
Hours Worked Per Week, Job Fit Interaction
0.00 (.71)
0.01 (.76)
0.04 (.01)
Latin America
Western Europe, North America, Australia
South Asia
Southern Europe
0.02 (.18) 0.13 (.00) 0.02 (.21)
0.06 (.00)
0.05 (.00) 0.12 (.00) 0.01 (.44)
0.05 (.15) 0.10 (.00) 0.05 (.10)
0.09 (.00) 0.04 (.03)
416
Section III: Differences in the Social Context of Well-Being
Low Job Fit 100 90 80 70 60 50 40 30 20 10 0
75
56
0 Hours
75
57
1-4 Hours
High Job Fit
74
73
71
70
68
67
52
48
47
44
43
41
8 Hours
9 Hours
10 Hours
11-12 Hours
13+ Hours
5-7 Hours
Figure 13.2 Positive Daily Experience/Affect for Workers with High and Low Job Fit by Hours Worked Yesterday Note: Figure presents unstandardized marginal means on percentage of positive daily experience adjusted based on age, gender, log income, marital status, and subjective domains (health, community, relationships, and safety).
Both hours worked yesterday and job fit predicted positive daily experiences, and the interaction was significant. The direction of the main effects was consistent across regions. Fewer hours worked yesterday and better job fit were associated with more positive daily experiences. And as hours worked increased, the impact of job fit on positive daily experiences increased. For residents expressing lower job fit, as hours of work increased, positive daily experiences decreased. This interaction is significant in all regions. Figure 13.2 shows the percentage of respondents with positive daily experiences (after controlling for demographics and other subjective domains) for high and low job fit. The difference between respondents with high and low job fit is substantial, regardless of number of hours worked, but the difference in positive experience/affect increases as hours of work increase. There is a 34 percent difference in positive experiences/ affect for those with high job-fit for respondents who worked no hours on the day reported on. This difference increases to 57 percent for those working eight hours, and to 63 percent for those working thirteen or more hours. Similarly, both hours worked yesterday and job fit predicted negative daily experiences, and the interaction was again significant, although less strong and less consistent across regions. Figure 13.3 illustrates the percentage of respondents with negative daily experiences (after controlling for demographic and other subjective domains) for different amounts of hours worked yesterday, for both those with high and those with low job fit. The interaction is only slight; the difference between those with high and low
TABLE 13.4B
Hierarchical Regression Analysis: Dependent Variable – Positive Daily Experiences/Affect
1. Overall Sample
417
Age Female Log Household Income Single Divorced/ Separated Education Health Problems Safety Count on Others Community Satisfaction Hours Worked Yesterday Job Fit Hours Worked Yesterday, Job Fit Interaction Term
Beta (Sig.)
Beta (Sig.)
0.01 (.48) 0.01 (.04) 0.09 (.00)
0.01 (.10) 0.01 (.40) 0.08 (.00)
0.00 (.81) 0.02 (.00) 0.03 (.00)
Beta (Sig.)
Beta (Sig.)
Beta (Sig.)
0.01 (.14) 0.01 (0.07) 0.08 (.00)
0.00 (.88) 0.02 (.01) 0.06 (.00)
0.00 (.92) 0.02 (.01) 0.05 (.00)
0.00 (.88) 0.02 (.00)
0.01 (.43) 0.02 (.00)
0.00 (.92) 0.02 (.01)
0.00 (.81) 0.02 (.01)
0.03 (.00) 0.04 (.00) 0.05 (.00) 0.09 (.00) 0.14 (.00)
0.03 (.00) 0.04 (.00) 0.05 (.00) 0.08 (.00) 0.14 (.00)
0.02 (.00) 0.04 (.00) 0.04 (.00) 0.06 (.00) 0.10 (.00)
0.03 (.00) 0.04 (.00) 0.04 (.00) 0.06 (.00) 0.09 (.00)
0.11 (.00)
0.11 (.00)
0.11 (.00)
0.22 (.00)
0.20 (.00) 0.07 (.00)
(continued )
TABLE 13.4B
(Continued)
1. Overall Sample Beta (Sig.) R R Square R Square Change F Change Sig. F Change
Beta (Sig.)
Beta (Sig.)
Beta (Sig.)
Beta (Sig.)
418
0.11 0.01 0.01
0.22 0.05 0.03
0.24 0.06 0.01
0.32 0.10 0.05
0.33 0.11 .00
53.14 0.00
235.52 0.00
344.14 0.00
1345.57 .00
117.14 0.00
Western Europe, North America, Australia
South Asia
2. Significant Effects by Region (Beta, Sig.) Standardized Coefficients
Hours Worked Yesterday Job Fit Hours Worked Yesterday, Job Fit Interaction
Africa
East Asia
Eastern Europe, FSU
Latin America
0.08 (.00)
0.20 (.00)
0.18 (.00)
0.11 (.00)
0.15 (.00)
0.04 (0.00)
0.17 (.00) 0.07 (.00)
0.31 (.00) 0.10 (.00)
0.24 (.00) 0.08 (.00)
0.21 (.00) 0.08 (.00)
0.15 (.00) 0.06 (.00)
0.19 (.00) 0.04 (0.01)
Southern Europe
0.17 (.00) 0.20 (.00) 0.10 (0.00)
Work and Well-Being Around the World
Low Job Fit 100 90 80 70 60 50 40 30 20 10 0
419
High Job Fit
32
31
32
33
34
37
40
31
17
17
19
18
21
20
22
23
0 Hours
1-4 Hours
5-7 Hours
8 Hours
9 Hours
10 Hours
11-12 Hours
13+ Hours
Figure 13.3 Negative Daily Experiences/Affect for Workers with High and Low Job Fit by Hours Worked Yesterday Note: Figure presents unstandardized marginal means on percentage of negative daily experience adjusted based on age, gender, log income, marital status, and subjective domains (health, community, relationships, and safety).
job fit is greatest (17 percentage points) when respondents had heavy work days (13 hours or more). Job fit explained about 14 percentage points of negative daily experience, whereas hours worked explained only slight differences. The impact of daily hours worked on negative daily experiences is significant in East Asia, Eastern Europe/FSU, Western Europe/North America/Australia, and Southern Europe, but not in Africa, Latin America, and South Asia. Job fit was a significant predictor of negative daily experience (higher job fit, lower negative experience) across all regions. The interaction was significant in Africa, Eastern Europe/FSU, Western Europe/North America/Australia, and South Asia. For residents in these regions, those with lower job fit had increasing levels of negative daily experiences as hours worked increased. For the full sample, we conducted the same hierarchical regression analysis, using each of the 14 daily experience items as an independent variable, to better understand the specific experiences that are most impacted by number of hours worked and job fit. A summary of the main effects and interactions is presented in Table 13.5. Interestingly, ‘‘feeling well-rested,’’ ‘‘choice,’’ and ‘‘stress’’ were most impacted by hours worked. Residents with more hours worked on a given day feel less well-rested and less able to choose what they do with their time, and more stress. But those who work more hours also feel more pride. Those with better job fit have consistently better days across the positive and negative
TABLE 13.4C
Hierarchical Regression Analysis: Dependent Variable – Negative Daily Experiences/Affect
1. Overall Sample
420
Beta (Sig.)
Beta (Sig.)
Beta (Sig.)
Beta (Sig.)
Age 0.01 (.26) Female 0.04 (.00) 0.07 (.00) Log Household Income Single 0.02 (.00) Divorced/Separated 0.03 (.00) Education 0.01 (.35) Health Problems Safety Count on Others Community Satisfaction Hours Worked Yesterday Job Fit Hours Worked Yesterday, Job Fit Interaction Term
0.02 (.00) 0.03 (.00) 0.06 (.00) 0.02 (.01) 0.03 (.00) 0.01 (.33) 0.08 (.00) 0.06 (.00) 0.07 (.00) 0.13 (.00)
0.02 (.01) 0.03 (.00) 0.06 (.00) 0.02 (.00) 0.03 (.00) 0.01 (.33) 0.08 (.00) 0.06 (.00) 0.07 (.00) 0.13 (.00) 0.05 (.00)
0.01 (.07) 0.03 (.00) 0.04 (.00) 0.02 (.01) 0.03 (.00) 0.01 (.15) 0.08 (.00) 0.05 (.00) 0.06 (.00) 0.11 (.00) 0.05 (.00) 0.14 (.00)
Beta (Sig.) 0.01 (.07) 0.03 (.00) 0.04 (.00) 0.02 (.01) 0.03 (.00) 0.01 (.15) 0.08 (.00) 0.05 (.00) 0.06 (.00) 0.11 (.00) 0.05 (.00) 0.13 (.00) 0.03 (.00)
R R Square R Square Change F Change Sig. F Change
0.09 0.01 0.01 39.39 0.00
0.22 0.05 0.04 265.01 0.00
0.22 0.05 0.00 68.98 0.00
0.26 0.07 0.02 496.65 0.00
0.26 0.07 0.00 15.93 0.00
2. Significant Effects by Region (Beta, Sig.)
421 Standardized Coefficients:
Africa
Hours Worked Yesterday 0.02 (.10) Job Fit 0.10 (.00) Hours Worked Yesterday, 0.04 (.00) Job Fit Interaction
East Asia 0.10 (.00) 0.06 (.04) 0.02 (.51)
Eastern Europe, FSU 0.09 (.00) 0.14 (.00) 0.05 (.00)
Latin America 0.02 (.10) 0.17 (.00) 0.00 (.82)
Western Europe, North America, Australia 0.14 (.00) 0.13 (.00) 0.05 (.00)
South Asia
Southern Europe
0.01 (.57) 0.10 (.00) 0.11 (.00) 0.21 (.00) 0.03 (.04) 0.03 (.44)
422
Section III: Differences in the Social Context of Well-Being
TABLE 13.5 Summary of Standardized (Beta) Main Effects and Interactions for Daily Experience Items
More Days Like Yesterday Enjoyment Smile/Laughter Well-Rested Treated with Respect Proud of Something Learn or Interest Choose What Do with Time Worry Sadness Stress Boredom Depression Anger
Hours Workeda
Job Fita
Interactiona
0.09* 0.06* 0.05* 0.20* 0.05* 0.05* 0.02* 0.18* 0.05* 0.00 0.11* 0.02* 0.01 0.06*
0.15* 0.15* 0.12* 0.12* 0.13* 0.13* 0.10* 0.15* 0.10* 0.10* 0.09* 0.12* 0.10* 0.08*
0.05* 0.05* 0.03* 0.03* 0.05* 0.03* 0.03* 0.04* 0.01 0.01 0.02* 0.02* 0.02* 0.02*
a = based on hierarchical regression analysis, with hours worked, job fit, & interaction term entered after age, marital status, income, gender, and other subjective domains (health, community, relationships, & safety) *= p<.01
experience/affect items (wanting to have more days like yesterday, enjoyment, choice, respect, pride, and less boredom, worry, sadness, and depression). The interaction between job fit and hours worked was particularly strong for ‘‘wanting to have more days like yesterday,’’ ‘‘enjoyment,’’ ‘‘being treated with respect,’’ and ‘‘being able to choose what you do with your time.’’ The interactive effect of job fit and hours worked is more consistent and salient for positive than for negative experiences. For people with high job fit, additional hours do not severely deter positive experiences such as enjoyment, being treated with respect, and choice. For people with low job fit, additional hours deter positive daily experiences, through increasingly less enjoyment, more disrespect, and lack of control over time. As hours worked increased, it is the decline in positive experience that is most predominant for those with low job fit. Figures 13.4 through 13.10 show, descriptively, the interactive nature of these relationships for specific items. People who worked nine or more hours show declining positive experiences as hours worked increased for people with low job fit. People with high job fit still maintained relatively high positive experiences with increasing hours of work, but still had substantial levels of stress, and positive experiences declined with hours worked even for those with high job fit, but at a much slower rate.
Work and Well-Being Around the World
Low Job Fit 100 90 80 70 60 50 40 30 20 10 0
76
74
75
75
423
High Job Fit
69
67
66 55
54
0 Hours
51
1-4 Hours
48
5-7 Hours
45
8 Hours
45
9 Hours
39
10 Hours
35
11-12 Hours
31
13+ Hours
Figure 13.4 Would you like to have more days just like yesterday (percent that answered yes)? Note: Figure presents unstandardized marginal means on percentage of wanting to have more days like yesterday adjusted based on age, gender, log income, marital status, and subjective domains (health, community, relationships, and safety).
Low Job Fit 100 90 80 70 60 50 40 30 20 10 0
81
83
58
60
0 Hours
1-4 Hours
82
54
5-7 Hours
High Job Fit
82
80
55
54
8 Hours
9 Hours
80
79
48
47
46
10 Hours
11-12 Hours
13+ Hours
75
Figure 13.5 Enjoyment (percent that answered yes)? Note: Figure presents unstandardized marginal means on percentage of respondents reporting enjoyment adjusted based on age, gender, log income, marital status, and subjective domains (health, community, relationships, and safety).
Discussion
In this chapter, we have investigated the relationship between hours worked and job fit and their impact on well-being, from various perspectives, based on data from wave 1 (2006) of the Gallup World Poll. High perceived job fit explains approximately 1.5 steps on the ladder of life, after controlling
424
Section III: Differences in the Social Context of Well-Being
Low Job Fit 100 90 80 70 60 50 40 30 20 10 0
75
58
73
67
66
High Job Fit
59
56
52
48
56 46 38
0 Hours
1-4 Hours
5-7 Hours
8 Hours
35
9 Hours
32
30
29
10 Hours
11-12 Hours
13+ Hours
Figure 13.6 Did you feel well rested yesterday (percent that answered yes)? Note: Figure presents unstandardized marginal means on percentage of respondents reporting feeling well rested adjusted based on age, gender, log income, marital status, and subjective domains (health, community, relationships, and safety).
Low Job Fit 100 90 80 70 60 50 40 30 20 10 0
91
73
0 Hours
88
87
89
68
69
High Job Fit 89
86
63
62
9 Hours
10 Hours
87
86
78
1-4 Hours
5-7 Hours
8 Hours
66
11-12 Hours
62
13+ Hours
Figure 13.7 Were you treated with respect all day yesterday (percent that answered yes)? Note: Figure presents unstandardized marginal means on percentage of respondents reporting being treated with respect adjusted based on age, gender, log income, marital status, and subjective domains (health, community, relationships, and safety).
for demographic differences and other subjective life domains (see Figure 13.1). Number of hours worked per week explains approximately one-half step on the ladder of life. This finding alone points to the importance of the subjective experience of work as respondents evaluate their overall lives, after controlling for demographics and subjective domains. Prior studies show substantial relationships between the subjective experience of
Work and Well-Being Around the World
Low Job Fit 100 90 80 70 60 50 40 30 20 10 0
65
70
49
49
0 Hours
1-4 Hours
72
425
High Job Fit
70
72
72
73
77
46
45
46
47
46
8 Hours
9 Hours
10 Hours
11-12 Hours
13+ Hours
54
5-7 Hours
Figure 13.8 Are you proud of something you did yesterday (percent that answered yes)? Note: Figure presents unstandardized marginal means on percentage of respondents reporting being proud of something they did adjusted based on age, gender, log income, marital status, and subjective domains (health, community, relationships, and safety).
Low Job Fit 100 90 80 70 60 50 40 30 20 10 0
79
55
76
72
66
High Job Fit
64
62
59
39
37
57
58 50 42
37
26 0 Hours
1-4 Hours
5-7 Hours
8 Hours
9 Hours
10 Hours
11-12 Hours
13+ Hours
Figure 13.9 Were you able to choose what you did with your time yesterday (percent that answered yes)? Note: Figure presents unstandardized marginal means on percentage of respondents reporting being able to choose what they did with their time adjusted based on age, gender, log income, marital status, and subjective domains (health, community, relationships, and safety).
work and objective outcomes such as productivity, employee turnover, and profit. Perceived job fit was a robust predictor of life evaluation across regions. Optimum well-being (life evaluation) was observed for those working between 35 and 44 hours per week, with high perceived job fit.
426
Section III: Differences in the Social Context of Well-Being
Low Job Fit 100 90 80 70 60 50 40 30 20 10 0
41 34
34
High Job Fit
41
43
36
32
9 Hours
10 Hours
26 21
24
27
29
0 Hours
1-4 Hours
5-7 Hours
8 Hours
45
48
37
39
11-12 Hours
13+ Hours
Figure 13.10 Stress (percent that answered yes)? Note: Figure presents unstandardized marginal means on percentage of respondents reporting stress adjusted based on age, gender, log income, marital status, and subjective domains (health, community, relationships, and safety).
Perhaps the most telling finding uncovered here is that as hours worked increase for any given day, job fit becomes a more important factor in determining positive experiences. Job fit also predicts negative experience, but the interaction with hours worked is less strong. It appears some negative experience or affect (stress, for instance) is an outcome of more work, in general. We observed that stress is significantly higher for those with low job fit, but increases as number of hours worked increase, for those with both high and low job fit. The same can be said for feeling well rested. As work hours increase, feeling well rested decreases for both those with high and low job fit, but those with high job fit report being significantly better rested for similar hours worked. Pride increases as hours at work increases per day, for those with high job fit, but declines as hours of work increase above 5–7 hours for those with low job fit. As hours worked increases, feeling respected stays quite high, for those with high job fit, but declines for those with low job fit. Ability to choose what you do with your time declines as hours of work are added, but does so at a slower rate for those with high job fit. It remains to be more fully examined in future research, but it is possible that positive affect stimulated by high job fit serves as a buffer against the normal negative stressors that come with work, in general. Pressman & Cohen (2005), for instance, report consistent findings linking positive affect and health, lower morbidity and symptoms of pain, possibly indicating healthprotective physiological responses from positive affect. But more research is needed to refine the methodologies and combine the findings from individual studies into meta-analyses.
Work and Well-Being Around the World
427
Despite the broad coverage in this data set, we were limited in our data in some parts of the world. For instance, the limited income data in the Middle East led us to remove the region from these analyses. Due to the timing of the field periods, the United States and several Western European countries had missing data for the weekly work time variable, but we had great coverage on time spent working on a typical day. Future data sets will allow us to study the United States and Western European countries in more detail on the hours per week variable. We found it interesting that hours worked in a week negatively impacts well-being in every region except Africa. But daily hours worked negatively impact positive experience in all regions, whereas hours worked in a day do not significantly impact negative experience in Africa, Latin America, and South Asia. These are regions that include countries where work may be less plentiful and be seen as a luxury, rather than a burden. The implications from the general findings may have significant value. First, it is clear that perceived job fit is an important component of well-being for working people. This research extends the importance of the concept of job fit beyond traditionally studied outcomes, such as productivity. When one considers the possible indirect relationships between affect at work and the spillover effect away from work, the implications may stretch beyond the immediate workplace. In organizations, selection systems certainly are one avenue to improving job fit. But this has less applicability to the selfsubsistence situation, or to people working outside the organizational setting. Our hope is that this research will help those thinking about policies consider the interactive nature of individual differences and the importance that these differences play in collaborative work settings. As economic progress is made in various regions of the world, with the progress should come job choice. This may be one of the most desirable natural outcomes of economic development. That is, that people within societies can more readily choose the type of work that best suits their talents. And in situations where more than the standard forty-hour work week is deemed necessary, the impact of job fit on well-being should be considered. In any event, we should stop short of claiming that those working long hours are always in grim misery. These data suggest that the positive experiences that come from being in a good job might buffer some of the negatives that naturally come with work, which may have indirect effects on health. In follow-up analyses by gender, we found number of hours worked yesterday more greatly impacted negative daily experience for males than females. This finding was particularly salient for East Asia and Southern Europe. Future research should look at gender and age differences and how they manifest themselves in the context of what we have learned here, and within specific regions of the world.
428
Section III: Differences in the Social Context of Well-Being
References Avolio, B. J., & Sosik, J. J. (1999). A life-span framework for assessing the impact of work on white-collar workers. In S. L. Willis & J. D. Reid (Eds.), Life in the middle: Psychological and social development in middle age (pp. 251–274). San Diego, CA: Academic Press. Bouchard, T. J., Jr. (1997). Genetic influence on mental abilities, personality, vocational interests, and work attitudes. In C. L. Cooper & I. T. Robertson (Eds.), International Review of Industrial and Organizational Psychology, 12, 373–395. New York: John Wiley. Bouchard, T. J., Lykken, D. T., McGue, M., Segal, N. L., & Tellegen, A. (1990). Sources of human psychological differences: The Minnesota study of twins reared apart. Science, 250, 223–228. Clifton, D. O., & Harter, J. K. (2003). Strengths investment. In K. S. Cameron, J. E. Dutton, & R. E. Quinn (Eds.), Positive organizational scholarship (pp. 111–121). San Francisco: Berrett-Koehler. Clifton, D. O., Hollingsworth, F. L., & Hall, W. E. (1952). A projective technique for measuring positive and negative attitudes towards people in a real-life situation. Journal of Educational Psychology, 43, 273–283. Crouter, A. C., Bumpus, M. F., Head, M. R., & McHale, S. M.. (2001). Implications of overwork and overload for the quality of men’s family relationships. Journal of Marriage and Family, 63(2), 404–416. May 2001. Csikszentmihalyi, M. (1997). Finding flow: The psychology of engagement with everyday life. New York: Basic Books. Deaton, Angus (2008). Income, health, and well-being around the world: Evidence from the Gallup World Poll. Journal of Economic Perspectives, 22(2), 53–72. Dembe, A. E., Erickson, J. B., Delbos, R. G., & Banks, S. M. (2005). The impact of overtime and long work hours on occupational injuries and illnesses: New evidence from the United States, Occupational Environmental Medicine, 62, 588–597. Diener, E. (2005). Guidelines for national indicators of subjective well-being and illbeing. SINET, November 2005: 4–6. Dillman, D. A., Phelps, G., Tortora, R., Swift, K., Kohrell, J., & Berck, J. (2001). Response rate and measurement differences in mixed mode surveys using mail, telephone, interactive voice response, and internet. Research Paper. Pullman, WA: Washington State University. Edwards, J. R. (1991). Person–job fit: a conceptual integration, literature review, and methodological critique. In C. L. Cooper & I. T. Robertson (Eds.), International Review of Industrial and Organizational Psychology, 6, 283–357. New York: Wiley. Edwards, J. R., Caplan, R. D., & Van Harrison, R. (1998). Person–environment fit theory: Conceptual foundations, empirical evidence, and directions for future research. In C. L. Cooper (Ed.), Theories of organizational stress (pp. 29–67). New York: Oxford University Press. French, J. R. P., Caplan, R. D., & Van Harrison, R. (1982). The mechanisms of job stress and strain. New York: Wiley. Gallup (2007). The state of global well-being. New York: Gallup Press. Golden, L., Wiens-Tuers, B. (2005). Mandatory overtime work in the United States: Who, where, and what? Labor Studies Journal, 30, 1–25.
Work and Well-Being Around the World
429
Harter, J. K., Schmidt, F. L., & Keyes, C. L. M. (2002). Well-being in the workplace and its relationship to business outcomes: A review of the Gallup studies. In C. L. Keyes & J. Haidt (Eds.), Flourishing: The positive person and the good life (pp. 202–224). Washington, DC. American Psychological Association. Harter, J. K., Schmidt, F. L., Killham, E. A., & Asplund, J. W. (2006). Q12 Meta-Analysis. Technical Paper. Omaha, NE: The Gallup Organization. Harter, J. K., Schmidt, F. L., & Hayes, T. L. (2002). Business-unit-level relationship between employee satisfaction, employee engagement, and business outcomes: A meta-analysis. Journal of Applied Psychology, 87(2), 268–279. Harter, J. K., Hayes, T. L., & Schmidt, F. L. (2004, January). Meta-analytic predictive validity of Gallup Selection Research Instruments (SRI). Omaha, NE: The Gallup Organization. Isen, A. M. (1987). Positive affect, cognitive processes, and social behavior. In Berkowitz (Ed.), Advances in experimental social psychology, vol. 20 (pp.203–253). San Diego, CA: Academic Press. Judge, T. A. & Colquitt, J. A. (2004). Organizational justice and stress: The mediating role of work–family conflict. Journal of Applied Psychology, 89(3), 395–404. Judge, T. A., & Illies, R. (2004). Affect and job satisfaction: A study of their relationship at work and at home. Journal of Applied Psychology. 89(4), 661–673. Judge, T. A., Thoresen, C. J., Bono, J. E., & Patton, G. K. (2001). The job satisfaction– job performance relationship: A qualitative and quantitative review. Psychological Bulletin, 127(3), 376–407. Kahneman, D., Diener, E., & Schwarz, N. (Eds.) (1998). Well-being: The foundations of hedonic psychology. New York: Russell Sage Foundation. Kahneman, D., & Riis, J. (2005). Living and thinking about it: Two perspectives on life. In F. Huppert, N. Baylis, & B. Kaverne (Eds.), The Science of Well-being: Integrating Neurobiology, Psychology, and Social Science. Oxford, UK: Oxford University Press. Kivimaki, M., Ferrie, J., Brunner, E., Head, J., Shipley, M., Vehtera, J., & Marmot, M. (2005). Justice at work and reduced risk of coronary heart disease among employees. Archives of Internal Medicine, 165, 2245–2251. Krueger, A., & Kahneman, D. (2006). Developments in the measurement of subjective well-being. Journal of Economic Perspectives, 20, 3–24. Schmidt, F. L., and Hunter, J. E. (1998). The validity and utility of selection methods in personnel psychology: Practical and theoretical implications of 85 years of research findings. Psychological Bulletin, 124(2), 262–274. Schmidt, F. L., & Rader, M. (1999). Exploring the boundary conditions for interview validity: Meta-analytic validity findings for a new interview type. Personnel Psychology, 52, 445–464. Sparks, K., Cooper, C., Fried, Y., & Shirom, A. (1997). The effects of hours of work on health: A meta-analytic review. Journal of Occupational and Organizational Psychology, 70(4), 391– 408. Dec. 1997. Spurgeon, A., Harrington, M. J., & Cooper, C. L. (1997). Health and safety problems associated with long working hours: a review of the current position. Occupational Environmental Medicine, 54, 367–75. Maslow, A. H. (1943). A theory of human motivation. Psychological Review, 50, 394–395. Pressman, S. D., & Cohen, S. (2005). Does positive affect influence health? Psychological Bulletin, 131(6), 925–971.
430
Section III: Differences in the Social Context of Well-Being
Warr, P. (2003). Well-being and the workplace. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The foundations of hedonic psychology. New York: Russell Sage Foundation. Weston, R., Gray, M., Qu, L., Stanton, D. (2004). Long work hours and the well-being of fathers and their families, Labor and Demography, 0405007, EconWPA. Wrzesniewski, A., McCauley, C., Rozin, P., & Schwartz, B. (1997). Jobs, careers, and callings: People’s relations to their work. Journal of Research in Personality, 31, 21–33.
Appendix TABLE 13.A
Overall Study Variable Descriptives and Intercorrelations
Correlations
Minimum N
Age
Hours House-hold Worked Income (InterDivorced/ YesEducation Single Married Separated terday Gender national $)
431
Age
37994
1
Gender
38157
.03**
1
Log 30204 Household Income (International $)
.05**
.05**
Hours Worked Per Week
Community Health Prob- Count on Satisfaction Job Fit lems Others
NegaPositive tive Ladder Exper- Experience Safety Present ience
1
Education
37945
.11**
.00
Single
37573
.48**
.04**
.03**
.06**
Married
37573
.31**
.05**
.11**
.03**
.80** 1
Divorced/ Separated
37573
.23**
.15**
.12**
.06**
.20** .41**
1
Hours Worked Yesterday
34066
.03**
.08**
.03**
.00
.05**
.06**
.01*
Hours Worked Per Week
30204 0.01
.11**
.05**
.04**
.02**
.03**
.02**
.34**
1 1
1
.35**
1
(continued )
TABLE 13.A
(Continued) Correlations
Minimum N
Age
Hours House-hold Worked Income (InterDivorced/ YesEducation Single Married Separated terday Gender national $)
432
Job Fit
36821
.05**
.13**
.05**
.05**
.06**
Health Problems
37796
.16**
.03**
.08**
.07**
.06**
.02**
Count on Others
37561
.08**
.02**
.10**
.08**
Community Satisfaction
32550
.04**
.02**
.03**
.02**
.04**
Safety
37281 0.01
.17**
.00
.03**
0.01
Ladder Present
37871
.06**
.02**
.26**
.17**
.02**
Positive Experience
38115
.01**
.02**
.11**
.06**
.01*
Negative Experience
38029
.02**
.06*
.09**
.03**
**
0.01
.01*
0.01
Hours Worked Per Week .01*
Community Health Prob- Count on Satisfaction Job Fit lems Others 1
.07**
.03**
.02** .05** 1
.04**
.02**
.02**
.04**
.01*
.00
.02**
.03**
.01**
.03**
.06** .04**
.02**
.13**
.02**
.07**
.02**
.02**
.18** .09**
.14**
.11**
.05** 1
.01**
.03**
.13**
.04**
.26** .06**
.11**
.15**
.07**
.02** .05**
.05**
.07**
.10**
.14**
.04** .01*
0.01
.11** .08** .21** .03**
.03** .19**
Correlation is significant at the 0.01 level (2-tailed) Correlation is significant at the 0.05 level (2-tailed) Notes: Correlations calculated after first standardizing variables within each country. Sample sizes range from 30204 to 54028.
*
NegaPositive tive Ladder Exper- Experience Safety Present ience
.10**
1 .07**
1 1
.19**
1
.09** .18** .40**
1
TABLE 13.B
Mean Hours Worked by Country Hours Worked Yesterday
Region
Country
Africa
Angola Benin Botswana Burkina Faso Burundi Cameroon Chad Ethiopia Ghana Kenya Madagascar Malawi Mali Mauritania Mozambique Niger Nigeria Rwanda Senegal Sierra Leone South Africa Tanzania Togo Uganda Zambia Zimbabwe
7.2 6.6 6.6 6.8 7.5 6.2 6.6 8.1 6.0 7.2 5.3 6.3 6.4 6.2 6.0 6.6 6.8 5.1 7.8 7.3 6.9 7.5 6.7 6.8 6.3 5.3
East Asia
Hong Kong Japan Singapore South Korea Taiwan
6.0 5.3 7.4 6.4 6.6
Armenia Azerbaijan Belarus Czech Republic Estonia Georgia Hungary Kazakhstan Kyrgyzstan Latvia Lithuania Moldova
7.2 7.5 8.8 7.0 5.9 5.5 7.9 7.9 6.0 6.6 7.3 8.5
E. Europe, FSU
Hours Worked Per Week 45.8 53.0 49.0 48.9 49.9 44.8 45.9 51.0 45.7 44.6 40.1 46.6 50.9 46.9 43.9 44.6 44.6 35.3 51.8 46.8 41.4 52.9 48.5 47.1 41.3 42.6 45.1 45.5 47.4 46.6 48.1 43.9 44.9 45.7 42.9 39.6 48.1 48.7 42.8 44.2 50.1 (continued )
433
TABLE 13.B
(Continued)
Region
Hours Worked Yesterday
Country Poland Romania Russia Slovakia Slovenia Tajikistan Ukraine Uzbekistan
7.1 8.0 6.0 6.5 6.6 7.6 6.5 7.4
Latin America
Argentina Bolivia Brazil Chile Colombia Costa Rica Cuba Dominican Republic Ecuador El Salvador Guatemala Honduras Mexico Nicaragua Panama Paraguay Peru Puerto Rico Uruguay Venezuela
6.3 6.7 6.3 7.3 6.7 6.7 6.4 5.7 7.1 6.0 6.9 7.2 7.4 6.0 5.8 6.5 7.2 5.9 6.4 6.4
W. Europe, North America, Australia
Australia Austria Belgium Canada Denmark Finland France Germany Ireland Netherlands New Zealand Norway Sweden Switzerland United Kingdom United States
5.7 6.4 5.1 5.8 4.1 5.7 4.1 5.9 5.4 4.3 5.2 5.2 3.0 6.2 5.5 6.1
Hours Worked Per Week 50.5 45.5 43.7 42.5 45.2 48.6 44.7 50.3 52.1 45.2 49.5 53.1 52.8 48.0 52.0 55.7 50.6 58.4 49.4 48.7 55.1 49.0 56.4 53.9 41.4 51.1 45.0 43.1 41.0
39.4 39.4
38.6 37.1 38.7 41.1 38.5
(continued )
434
TABLE 13.B
(Continued)
South Asia
Southern Europe
Bangladesh Cambodia India Indonesia Laos Malaysia Myanmar Nepal Sri Lanka Thailand Vietnam
9.2 5.5 7.5 6.4 8.5 7.4 6.3 7.6 7.5 7.9 7.0
Cyprus Greece Israel Italy Portugal Spain
3.5 6.2 6.4 5.0 6.5 4.9
435
56.7 45.1 46.4 53.1 48.1 49.1 54.2 45.9 55.4 54.1 42.8 48.6
46.0
Chapter 14 Work, Jobs, and Well-Being Across the Millennium Andrew E. Clark Paris School of Economics and IZA
The hope of pleasure in the work itself: how strange that hope must seem. —William Morris, ‘‘Useful Work versus Useless Toil,’’ 1884
Subjective well-being has arguably been moving into the mainstream in economics over the past ten years. One measure of the success of this insertion is the number of published articles in the domain. An ECONLIT search for journal articles with either ‘‘Happiness,’’ ‘‘Life Satisfaction,’’ ‘‘Well-being,’’ or ‘‘Job Satisfaction’’ in the title, described in Clark et al. (2008b), identifies 614 published articles between 1960 and 2006. Of these, 363 (59%) have been published since 2000. We have thus moved (in economics) from a pre-2000 period where one ‘‘well-being’’ article (by this count) was published every two months to a new regime where more than one is published per week. A number of these papers have addressed policy issues. Sometimes these concerned fairly specific problems, such as the rationality of smokers’ choices (Gruber & Mullainathan, 2006) or the compensating differential for aircraft noise (van Praag & Baarsma, 2005). A wider, separate literature has broadly asked the question, ‘‘What makes a good life?’’ Contributions along these lines include Di Tella & MacCulloch (2006), Layard (2005), and Kahneman et al. (2004). This paper aims to contribute to this policy debate, but more particularly in the world of work. There have now been a number of papers using different data sets across different countries to answer the question: What matters in life? When individuals rank life domains such as income, family, 436
Work, Jobs, and Well-Being Across the Millennium
437
marriage, leisure, housing, job, friendship, and health, the category ‘‘job’’ often comes towards the top of this ranking. An early contribution is Glenn and Weaver (1981), and a recent careful econometric analysis is found in Ferrer-i-Carbonell & van Praag (2004). In addition, individuals spend a great deal of time at work: arguably almost more than they spend on ‘‘doing’’ any other particular thing (apart from perhaps watching television: see Benesch et al., 2006). In the context of well-being at work, I ask three specific questions about Organization for Economic Co-operation and Development (OECD) labor markets over the past twenty years, although the answers will concentrate on only one of them:
• How important is it to have a job (especially when you want one)? Or: How bad is unemployment?
• What job characteristics are most important for job satisfaction? And how have these characteristics been changing over the past twenty years in OECD countries? • Is there anything particular about self-employment in OECD countries? The approach taken here will be mostly subjective. To find out about workers’ jobs and lives, and what it is that they like and dislike about them, I will use the answers given by individuals to questions explicitly asking them to evaluate these same jobs and lives. Specifically, the first question will mostly appeal to a measure of overall satisfaction to evaluate the harm from unemployment, while the second and third questions will be addressed using overall job satisfaction information. Although it is probably fair to say that there is no unanimity on the usefulness of subjective well-being information in economics, there is likely now greater willingness to pay attention to such measures than in the past. A mini-industry of validation studies has used panel data to show that what people say today is a strong predictor of what they will do in the future: links have been established between current well-being and future life expectancy, morbidity, productivity, quits, absenteeism, unemployment duration, and marriage duration. At the same time, a flourishing literature in psychology has examined the links between measures of job satisfaction or employee engagement, on one hand, and firm performance on the other (where this latter includes profitability, productivity, turnover, and absenteeism): see, for example, the meta-analyses in Harter et al. (2002 and 2006) and Judge et al. (2001). Were individuals’ responses purely idiosyncratic, then they could not be compared to each other, and none of the above relationships would pertain.
438
Section III: Differences in the Social Context of Well-Being
I address the second and third of the above questions using the three separate waves of the International Social Survey Programme (ISSP) that cover the labor market and work orientations. However, I will not be able to deal with the first question in this way. Here I would like to evaluate the effect of employment, and unemployment, on some measure of psychological well-being. Unfortunately, the ISSP, for all of its advantages, does not include a measure of life satisfaction or overall psychological functioning. To establish the relationship between work and well-being, I turn to European Community Household Panel (ECHP) and British Household Panel Survey (BHPS) data, both of which include well-being measures that are asked of both the employed and unemployed. This chapter is structured as follows. The first section considers the simple relationship between employment, unemployment, and subjective well-being, and sets out the policy implications that result. Section 2 recalls that employment is nonetheless not a homogeneous state, and uses repeated ISSP data to track the movements in job quality in OECD countries from 1989 to 2005. Section 3 then distinguishes self-employment from employment, and asks: If the self-employed are so satisfied, what are the rest of us doing still being employees? Last, Section 4 concludes.
1. Working and Well-Being
We might imagine that one of the most important aspects of the labor market in terms of well-being is whether individuals are able to find a job, assuming that they want one. The relationship between individual wellbeing and unemployment (along with that with income) has been something of a mainstay in the literature. The very consistent result in empirical research is that unemployment is strongly negatively correlated with various measures of well-being. Figure 14.1 illustrates with data from the ECHP1, from 1994 to 2001, where average satisfaction with work or main activity (on a 1–6 scale, with 6 being the most satisfied) is plotted against labor force status. The self-employed have satisfaction scores that are only little different from those of the employed in this data set (of which, more below). Of most interest here is the well-being of the unemployed compared to those who are in work. The former report satisfaction scores that are, on average, two points lower than those for the employed and self-employed; this is a very large difference on a six-point scale. This negative relationship persists, with little change in its size, in multivariate regression analyses that control for the level of income as well as a variety of other explanatory variables (for example, Clark & Oswald, 1994; and Winkelmann & Winkelmann, 1998).
Work, Jobs, and Well-Being Across the Millennium
439
4.5
Satisfaction
4
3.5
3
2.5
2 Self-employed Employed Unemployed Labor Force Status
NLF
Figure 14.1 Satisfaction with Work or Main Activity and Labor Force Status. Source: ECHP, 1994–2001. Satisfaction measured on a one to six scale.
As such, the main effect of unemployment on well-being would seem to be non-pecuniary: it would take a great deal of money indeed to make the unemployed as happy as the employed. The satisfaction of the unemployed is lower than that of the employed in Figure 14.1. Can we therefore immediately conclude, for policy purposes, that ceteris paribus, putting the unemployed into work will raise ‘‘national well-being’’? There are three reasons to pause for thought at this juncture.
A) It is not unemployment that makes people unhappy, it is the unhappy who become unemployed
In this reverse causality argument, putting the unemployed back into work may have little or no effect on their satisfaction: it is their inbuilt personality, or something else, that is causing their lower level of well-being, independently of their labor market status. This argument is behind the well-known idea of individual set-points in well-being: two well-known contributions are Diener et al. (2005) and Fujita & Diener (2005). Panel data, where we follow the same individual over time, allow us to observe the same individuals as they find and lose jobs. If it is indeed unemployment that causes lower well-being, then the latter will fall sharply upon the individual’s entry into unemployment. However, if the low
440
Section III: Differences in the Social Context of Well-Being
well-being observed amongst the unemployed is caused by the overrepresentation there of individuals with low well-being set-points (who wouldn’t have much higher well-being scores were they employed), then panel analysis, which is within-subject, will only pick up a small or null correlation between unemployment and well-being. In practice, panel regression analysis consistently finds that wellbeing does rise when the unemployed find a job (and falls when the employed lose their jobs). While there is probably some truth to the argument that unhappiness causes unemployment, this composition or shift-share effect does not seem to entirely explain the unemployment well-being gap.
B) It depends when we put the unemployed back in work
What if individuals ‘‘get used’’ to unemployment? A growing literature in economics and psychology tackles the question of whether individuals adapt over time to new circumstances. In our context, this implies that the longer-run effects of unemployment will be milder than their short-run effects, and in the case of complete adaptation, the long-term unemployed will be just as happy as the employed. The policy conclusion is, then, that we should concentrate our fire on the short-term unemployed, where the well-being returns from employment are largest. Clark et al. (2008a) look for evidence of habituation in twenty waves of German Socio-Economic Panel (GSOEP) data, considering marriage, divorce, widowhood, birth of child, layoff, and unemployment. They carry out a fixed-effect econometric analysis using panel data, which effectively follows the same individual over time before, during and after the event in question. The dependent variable is overall life satisfaction. The results for unemployment, for men and women, are shown in Figure 14.2. These show the movements in life satisfaction for the employed before they move into unemployment (which occurs at time t = 0, illustrated by the vertical line), and then their life satisfaction when unemployed. Note that these satisfaction movements are traced out only for those who remain unemployed, which seems a natural way to address the issue of adaptation. The results show that there is little evidence of adaptation to unemployment for men: those who have been unemployed for three years or more report just as low levels of satisfaction as those who are in the first year of unemployment. The picture is more mixed for women, although again the second and third years of unemployment do seem to be just as bad in life satisfaction terms as the first year of unemployment.
Work, Jobs, and Well-Being Across the Millennium
441
Unemployment (Males) 0.6 0.4 0.2 0.0 –0.2 –0.4 –0.6 –0.8 Significant at 1%
–1.0
Significant at 5%
–1.2
Significant at 10%
–1.4 –4
–3
–2
–1 0 1 2 3 No. of years before and after the event
4
5
4
5
Unemployment (Females) 0.6 0.4 0.2 0.0 –0.2 –0.4 –0.6 –0.8 –1.0 –1.2 –1.4 –4
–3
–2
–1 0 1 2 3 No. of years before and after the event
Figure 14.2 Unemployment and Life Satisfaction over Time in German Panel Data. Source: Clark et al. (2008a).
C) It depends which unemployed we put back in work
The previous paragraph discussed potential adaptation to unemployment. This can usefully be thought of as a comparison to myself in the past: what I think of my current situation depends on the past situations that I have
442
Section III: Differences in the Social Context of Well-Being
experienced. Adaptation thus concerns intrapersonal intertemporal comparisons (with respect to the same person, and over time). A second kind of comparison is interpersonal but atemporal: I compare my current situation to that of other people around me at the same point in time. A lively literature across the social sciences has found evidence in favor of such comparisons with respect to income (Blanchflower & Oswald, 2004; Clark & Oswald, 1996; and Luttmer, 2005, to name just three). Social comparisons in other areas have also been investigated, such as health (Powdthavee, 2009) and weight (Blanchflower et al., 2009; and Clark & Etile´, 2008). Social comparisons with respect to unemployment imply that my own unemployment will hurt less the more unemployment there is around, broadly speaking. We should then concentrate on moving back into work those who live in low-unemployment regions (as they suffer the most), effectively creating unemployment ghettos. Note that although this is anti-egalitarian with respect to unemployment, it is egalitarian with respect to subjective well-being. There is empirical evidence of such unemployment comparisons. Clark (2003) used the first seven waves of the BHPS to show that unemployment does indeed seem to hurt less in high-unemployment regions. The graph in Figure 14.3 illustrates the main results. The well-being score here is the 12item version of the General Health Questionnaire (the GHQ-12). This consists of twelve individual questions, covering, for example, trouble sleeping, belief in self-worth, enjoying day-to-day activities, and playing a useful role. All of these questions are answered on a one-to-four scale. Two of the answers are negative and two are positive. For example, with respect to trouble sleeping, the positive answers are ‘‘Not at all’’ and ‘‘No more than usual,’’ and the negative answers are ‘‘Rather more than usual’’ and ‘‘Much more than usual.’’ The ‘‘Caseness’’ measure of GHQ adds up the number of questions that are answered using one of the negative responses. We here invert this score, producing an index on a scale of 0 to 12, where higher numbers mean better psychological functioning. The precise form of the GHQ questions and their possible responses is available on the BHPS website (http://www.iser.essex.ac.uk/survey/bhps). Figure 14.3 shows the bivariate relationship between this Caseness score and regional unemployment. The vertical axis measures the average difference in GHQ scores between the employed and the unemployed: this therefore shows how much worse off the average unemployed person is than the average employed person (or the average psychological cost of unemployment). This figure has been calculated per wave (7) and per standard British region (11: as listed in the note underneath the graph), producing 77 data points in all. The figure on the horizontal axis shows the Labour Survey Force (LFS) unemployment rate, again by region and by
Work, Jobs, and Well-Being Across the Millennium
443
EA95
3
NW96
WA92 NT97 EM91 YH91
2
RS91
SW92 EM93WA93 YH93 WM95 NT95 RS95 SW91 SW91 SW93 EA94 WA96 GL91 RS93 RS96 WM91 WA95WA91 RS95 W A91 NW94 MW91 WA91 EA91 YH92 NW97 RSA4 SC91 MW91 EM95YH96 SC94 GL96 EA97 NT96 NT91 SC96 NW93 YH94 WM92 SC97 GL92 SW96 WM93 GL95 GL97 EM97 SC93 EM92 YM92 NW92 EA93 WM97 SW97 SC95 NT92 SW95 EM94
GHQ Difference
RS97
1
EM96 EA92
GL93 GL94
0
EA96
WM96
NT93 NT95
WA94 WM94
NT94
-1 5
10 Regional Unemployment Rate
15
Figure 14.3 The Well-Being Gap between those in Work and the Unemployed (GHQE-GHQU) and Regional Unemployment Rates. Key: GL = Greater London, RS = Rest of the South East, SW = South West, EA = East Anglia, EM = East Midlands, WM = West Midlands, NW = North West, YH = Yorkshire and Humberside, NT = North, WA = Wales, SC = Scotland. Source: Clark (2003).
year. The scatterplot suggests that the psychological cost of unemployment is indeed lower in regions where the unemployment rate is higher.2 The raw correlation between the two series is –0.32, significant at better than the one percent level. This analysis can be formalized by running multivariate regressions of individual GHQ scores on own unemployment, regional unemployment, and the interaction between the two, as well as a host of other standard socio-demographic control variables. The results of such regressions are summarized in Table 14.1, which presents the predicted well-being levels of individuals in different labor market situations. The actual figure refers to the predicted probability of having a ‘‘high’’ GHQ level (of 11 or 12 on the 0 to 12 scale). The top panel of Table 14.1 refers to the relationship between well-being and regional unemployment. The first line refers to an
444
Section III: Differences in the Social Context of Well-Being
TABLE 14.1 Well-Being and a) Regional Unemployment, and b) Partner’s Unemployment Labour Market Situation Employed and 5% Unemployment Region Employed and 10% Unemployment Region Unemployed and 5% Unemployment Region Unemployed and 10% Unemployment Region Both Individual and Partner Employed Individual Employed and Partner Unemployed Individual Unemployed and Partner Employed Both Individual and Partner Unemployed
Predicted Probability of High GHQ (%) 60.2 57.6 30.8 38.4 57.9 55.6 30.9 40.0
Notes: These calculations are based on the regression results using BHPS data in Clark (2003). High GHQ means a score of 11 or 12 on the 0 to 12 GHQ scale.
individual in work (E) in a region with a five percent unemployment rate, who is predicted to have a 60 percent probability of having a high level of GHQ. Moving this employed individual to a higher-unemployment (10%) region slightly reduces this probability. This relationship between GHQ and regional unemployment is reversed for the unemployed. An unemployed individual in a region with a 5 percent unemployment rate is predicted to have a 31 percent chance of having high well-being; moving this unemployed individual to a region with higher unemployment actually increases their well-being (to a figure of 38%). This result concords with findings on suicide and para-suicide rates by the unemployed, which are highest in low-unemployment regions (see Platt & Tansella, 1992).3 The bottom panel of Table 14.1 repeats this analysis, but this time for a much-tighter reference group: the individual’s spouse. The empirical analysis here covers couples who are both active in the labor market. There are then four possible couple labor-market outcomes, given by the combinations of employment and unemployment between the individual and her spouse. In the first line the probability that a worker with an employed spouse will report high well-being is 58 percent; this probability falls slightly, to 56 percent for an employed individual with an unemployed spouse. However, again, the worst situation is not when the individual and her partner are both unemployed, as perhaps might have been imagined, but rather when the individual is unemployed and her partner works. The unemployed are therefore unhappy, and not only because it is the unhappy who are more at risk of unemployment. However, the unemployed are not necessarily all equally unhappy. In particular, social context seems
Work, Jobs, and Well-Being Across the Millennium
445
to matter. The stronger is the social norm of employment, the less well the unemployed do, psychologically, as pointed out by Stutzer and Lalive (2004). Higher levels of regional unemployment may make those who are already unemployed better off. At the limit, being unemployed in a very high unemployment area may not be much worse in well-being terms than being employed in the same region. While empirical evidence does point in favor of such a social-norm effect, the average effect of unemployment on the unemployed remains substantial and negative: in the data sets that social scientists typically analyze, unemployment remains one of most damaging events that we can identify. This suggests that while social norms may well apply, they mostly operate by changing the unemployment experience from something awful to something slightly less awful. In addition, we have found only little evidence of any habituation to unemployment. Employment then continues to be a key element of individual and societal well-being for those who wish to work. This is not to say, however, that all jobs are the same. The following section uses repeated cross-section data from the ISSP to examine various job characteristics, and how they have changed over time. In particular, we can use these data (collected in 1989, 1997, and 2005) to evaluate the hypothesis that job quality has fallen secularly over the past twenty years.
2. Changing Job Quality over the Millennium
This section is concerned with the general issue of job quality and how this has changed over time. Here I am able to appeal to repeated cross-section data from three waves of the ISSP (see http://www.issp.org/). Many different countries are present in the ISSP data sets, with the number of countries tending to grow over time. The analysis on job quality I present here is based on the OECD countries found in the three ‘‘Work Orientations’’ waves of the ISSP, which contain a great deal of both objective and subjective cross-country information about job quality: these took place in 1989, 1997, and 2005. Table 14.2 shows the distribution of OECD countries, together with the number of employees interviewed, in these three ISSP waves.4 The ISSP Work Orientations modules contain information on both job values (what workers think is important) and job outcomes (what they actually get). With respect to the former, the importance of eight different job characteristics is recorded, using five rankings from ‘‘Not at all important’’ to ‘‘Very important’’: High income; Flexible working hours; Good
446
Section III: Differences in the Social Context of Well-Being
TABLE 14.2 Number of employees interviewed in OECD countries: ISSP Module on Work Orientations. 1989, 1997 and 2005 1989
1997
2005
West Germany Great Britain USA Hungary Netherlands Italy Norway Sweden Czech Republic Poland New Zealand Canada Japan Spain France Portugal Denmark Switzerland
622 675 838 641 681 586 1305
648 545 800 626 1018 475 1366 793 526 564 695 546 607 387 700 843 600 1727
531 469 961 437
812 520 428 556 1084 1012 1092 662
Total
5348
13466
10943
846 866 667
opportunities for advancement; Job security; Interesting job; Allows to work independently; Allows to help other people; and Useful to society. Very likely due to data restrictions, there has been a tendency in the empirical literature to summarize job quality by simple measures of earnings and hours (and perhaps job security). The answers to these job values questions will allow us to see if other, more neglected, work domains might be equally (or even more) important. The results across the three ISSP waves are presented in Table 14.3. This table has five columns. The first three present job value information in 1989, 1997, and 2005 for employees in those five OECD countries that participated in all three ISSP waves (West Germany, Great Britain, the United States, Hungary, and Norway: see Table 14.2). The last two columns present analogous figures for the 15 countries that were present in both ISSP waves, 1997 and 2005. Job characteristics are ranked from ‘‘Not at all important’’ to ‘‘Very Important.’’ Note that this is an absolute rather than a relative ranking: it is possible to class all eight characteristics as very important. Table 14.3 shows, separately for men and women, the (weighted) percentage describing each job characteristic as ‘‘Very important.’’ The asterisks show whether the changes from one ISSP wave to the next are significant.
Work, Jobs, and Well-Being Across the Millennium
TABLE 14.3
447
Job Values ISSP 1989, 1997 and 2005 Job Values: Percentage Saying ‘‘Very Important’’ WOMEN
High Income Flexible Working Hours Good Opportunities for Advancement Job Security Interesting Job Allows to Work Independently Allows to Help Other People Useful to Society
1989
1997
2005
1997
2005
20.5 19.5 23.0 61.4 48.2 28.7 22.7* 25.4
19.9 18.7 20.6 62.4 49.7 31.0 25.5** 24.9**
21.1 17.7 21.7 61.2 50.1 28.7 29.5 31.2
18.9** 18.4** 18.8** 58.0 52.7* 31.7 27.2** 24.5**
23.7 21.4 22.7 58.2 55.2 32.2 30.9 29.5
1989
1997
2005
1997
2005
24.4 14.1 22.4* 57.1 45.6 31.6 15.4 22.1**
22.8 13.7** 19.5** 58.6 47.0 30.8* 15.3** 17.1**
24.7 17.6 24.5 59.5 48.2 34.0 24.5 26.0
20.7** 15.0** 18.7** 54.9 50.1* 32.1 18.2* 20.3**
26.3 19.8 23.2 53.2 50.9 33.9 22.7 23.7
MEN
High Income Flexible Working Hours Good Opportunities for Advancement Job Security Interesting Job Allows to Work Independently Allows to Help Other People Useful to Society
Notes: Weighted Data; ** (*) = significant difference by year at the one (five) per cent level.
The results are remarkably consistent between men and women. There is some movement in recent years towards the increasing importance of high incomes, flexible working hours, and promotion opportunities. The (muchdiscussed) aspect of job security is the most likely to be described as very important, but it is of interest that the percentage doing so has barely changed between the different waves of the ISSP, arguing against any sharp increase in concern over this aspect of the job. The percentage saying that independence is very important has increased a little (for men), but the largest movements come with respect to arguably the most difficult to measure aspects of a job: whether it is helpful and whether it is useful. It can be countered that these latter figures should be taken with a pinch of salt, as there is likely a social desirability imperative for respondents to agree that these categories are important. Even so, the percentage saying that having a helpful job is very important is less than half as large as that saying that job security is very important; and in any case, we are here looking at changes in these percentages over time.5
448
Section III: Differences in the Social Context of Well-Being
The conclusion from Table 14.3 is that men are overall similar to women in terms of what they find important in their job, and that there is broad stability in terms of what employees find important in a job. Perhaps most importantly for the current exercise, any evaluation of job quality based uniquely on income and hours risks missing out on many job domains that workers value. We now move on to consider movements in actual job outcomes from 1989 to 2005 in OECD countries. We identify six broad classes of job outcomes:
• • • • • •
Pay; Hours of work; Future prospects (promotion and job security); How hard, stressful, or dangerous the job is; Job content: interest, prestige, and independence; and Interpersonal relationships.
These job outcomes are mostly measured using subjective self-reports from workers, although we do have information on monthly net earnings and weekly hours of work. The details regarding the construction of the indices that are used here are found in Appendix 14.A1. Last, as a summary measure of all of the aspects that workers appreciate or dislike about their job, some of which we explicitly measure here and some of which we do not, the ISSP asks each employee ‘‘How satisfied are you in your (main) job?’’ Answers are on a one-to-seven scale, where one means completely dissatisfied, and seven means completely satisfied. The distribution of job satisfaction over the three waves of the ISSP is shown in Appendix 14.A2. The ISSP subjective job outcome information is summarized in Table 14.4. The different domains covered here are of course not exhaustive, but they do reflect a number of job characteristics that are not consistently found in other surveys. Any other unmeasured job aspects will be picked up in the summary measure in the last row: overall job satisfaction. Table 14.4 shows the (weighted) percentage of employees in the ISSP data with the job outcome in question. Separate figures are presented for men and women, and for 1989, 1997, and 2005. Table 14.4 shows that (subjective) income rose, while the desire to reduce hours of work rose for women (in the fifteen OECD countries surveyed between 1997 and 2005) but stayed broadly the same for men. Promotion opportunities rose for both sexes, especially between 1997 and 2005. Job security is of great interest, judging by the amount of ink that has been spilled. Reported job security fell sharply from 1989 to 1997, but has then made a recovery. This recovery has been almost total for women, but
Work, Jobs, and Well-Being Across the Millennium
TABLE 14.4
449
Job Outcomes
ISSP 1989, 1997 and 2005 Job Outcomes: Percentage Reporting the Characteristic in Question WOMEN
Income is high Prefer to spend less time in their job Prefer to spend more time in their job Opportunities for advancement are high Job is secure Hard work Good job content Good relations at work High job satisfaction
1989
1997
2005
1997
2005
15.3 35.4** 13.7** 18.2
15.5 42.4** 9.6* 17.6**
16.4 37.0 11.9 23.4
16.0** 39.3* 10.2 17.7**
19.0 41.4 10.3 22.3
72.7** 37.5** 42.3* 68.4 39.5
66.8* 42.4 45.9 66.8 37.4**
70.7 41.1 49.3 67.6 45.9
63.6** 40.2 43.7* 66.8 41.1
66.4 40.9 45.9 65.6 42.0
1989
1997
2005
1997
2005
25.8* 37.4** 12.2 23.0*
23.6** 41.5** 10.9 20.2**
27.6 36.7 12.2 26.5
24.2** 41.0 8.9 21.8**
29.5 40.6 8.8 27.5
70.1** 54.2 41.6 67.3 38.2
63.5 52.1 40.2* 64.6** 35.9**
64.0 55.2 43.9 69.4 44.7
61.2* 50.6 41.5 65.8 40.3**
63.4 51.4 41.6 66.4 43.3
MEN
Income is high Prefer to spend less time in their job Prefer to spend more time in their job Opportunities for advancement are high Job is secure Hard work Good job content Good relations at work High job satisfaction
Notes: Weighted Data; ** (*) = significant difference by year at the one (five) per cent level. See Appendix A1 for the definition of the ISSP job outcome variables.
only partial for men. Towards the bottom of the tables, job content has improved for women, and relations at work have improved for men. Concentrating on the right-hand panel of Table 14.4, the overall picture looks fairly rosy between 1997 and 2005: everything that has changed significantly has done so for the better. This is reflected in the last line in the percentage reporting high job satisfaction (very or completely satisfied on the one-to-seven scale), which has risen significantly by three percentage points for men, but only insignificantly so for women. Table 14.4 is based on the raw data, describing the experience of the average person in the economy; however, this average person may have changed over time. As such, movements in job outcomes can reflect composition effects (such as the aging of the work force, with older
450
Section III: Differences in the Social Context of Well-Being
workers being more satisfied than younger workers), rather than the changing nature of jobs. The key question is then whether job satisfaction and other measures of job quality have risen because of changes in the type of people doing jobs, or changes in the kind of jobs they do. Policy in this area is arguably about changing the latter rather than the former (although the former will undoubtedly respond to changes in the latter). We address this issue in a standard way, appealing to regression analysis with individual demographic controls, as well as country and year dummies. These latter will pick up changes in job quality over the three waves of the ISSP surveys, conditional on the other right-hand side variables. These regressions will reveal whether job quality has evolved over time, holding the structure of the work force constant in terms of variables such as sex, age, and education. In the regression analysis, we will use two separate specifications: one without and one with earnings and hours. The latter analysis therefore asks whether an employee who kept the same hours of work and real earnings from 1989, to 1997, to 2005 would consider their job to have become more or less attractive over time. While this may sound like an artificial exercise, it does help isolate the role of changing earnings and hours in explaining movements in job quality. While much work has insisted on the role of income in determining job quality, hours of work have been relatively neglected. Yet, as Harter & Arora (2008) show, using both a measure of life satisfaction (the Cantril Ladder) and experienced wellbeing (feelings and emotions over the past 24 hours), hours are an important component of subjective well-being in most regions of the world. Table 14.5 shows the regression results for overall job satisfaction. The table is divided up into four columns. The two left-hand side columns consider changes from 1989 to 2005, and are thus estimated on the five countries that appear in all three waves. The two right-hand side columns analyze changes from 1997 to 2005 only, and are estimated for a far larger sample of countries (see Table 14.2). Columns 1 and 3 do not control for earnings and hours as explanatory variables, while columns 2 and 4 do. The estimated coefficients show that men are less satisfied with their jobs than women, ceteris paribus,6 and that older employees and the married are more satisfied. A number of the country dummy variables are significant: employees in Hungary, Japan, and France are relatively miserable; the ‘‘winners’’ are workers in Denmark (the omitted category), Switzerland, and the United States. What interests us perhaps the most in Table 14.5 are the estimated coefficients on the 1997 and 2005 dummy variables: these reveal whether the ‘‘average’’ employee found their job more satisfying in later waves than in 1989. In the left-hand panel, which shows the three-wave evolution, job satisfaction fell from 1989 to 1997, but had more than recovered by 2005. When earnings and hours are
TABLE 14.5
Overall Job Satisfaction Regressions. ISSP 1989, 1997 and 2005 1989–2005 Standard
2005 1997 Male 30 to 44 45 to 65 Married Years of Education
0.064* (0.027) 0.080** (0.025) 0.066** (0.021) 0.030 (0.028) 0.123** (0.030) 0.094** (0.023) 0.001 (0.004)
Earnings ($000) per month Hours per week West Germany Great Britain USA Hungary
0.040 (0.032) 0.069* (0.032) 0.144** (0.029) 0.355** (0.032)
1997–2005
With earnings and hours 0.027 (0.030) 0.079** (0.029) 0.136** (0.025) 0.008 (0.031) 0.066* (0.033) 0.073** (0.025) 0.007 (0.004) 0.103** (0.015) 0.001 (0.001) 0.050 (0.035) 0.063 (0.034) 0.133** (0.031) 0.236** (0.047)
Norway Sweden Czech Republic New Zealand Canada Japan Spain France
Standard
With earnings and hours
0.044** (0.015)
0.026 (0.017)
0.023 (0.015) 0.026 (0.021) 0.123** (0.022) 0.129** (0.016) 0.006** (0.002)
0.066** (0.017) 0.001 (0.023) 0.079** (0.024) 0.099** (0.017) 0.000 (0.002) 0.089** (0.009) 0.001 (0.001) 0.313** (0.044) 0.432** (0.045) 0.229** (0.040) 0.619** (0.049) 0.389** (0.038) 0.415** (0.040) 0.533** (0.047) 0.440** (0.049) 0.450** (0.046) 0.771** (0.047) 0.278** (0.053) 0.578** (0.039)
0.269** (0.042) 0.390** (0.043) 0.166** (0.038) 0.610** (0.043) 0.343** (0.036) 0.399** (0.038) 0.493** (0.042) 0.379** (0.039) 0.393** (0.044) 0.738** (0.044) 0.256** (0.045) 0.568** (0.038)
(continued )
451
452
Section III: Differences in the Social Context of Well-Being
TABLE 14.5
(Continued) 1989–2005 Standard
With earnings and hours
Portugal Switzerland Observations Log-Likelihood Log-Likelihood at zero
10554 15245.09 15413.98
8853 12791.12 12924.14
1997–2005 Standard
0.316** (0.038) 0.018 (0.036) 21045 30706.47 31134.28
With earnings and hours 0.345** (0.042) 0.117** (0.039) 17834 25944.07 26374.64
Notes: Standard errors in parentheses. * significant at 5%; ** significant at 1%.
controlled for, satisfaction in 2005 is the same as in 1989 (in the five countries for which we can carry out this calculation). Part of rising job satisfaction, then, results from higher income and lower hours of work (this is why column 1 differs from column 2; and column 3 differs from column 4).7 Previous work on the first two Work Orientations modules of the ISSP (Clark, 2005b) had shown falling job satisfaction in OECD countries across the 1990s, and had suggested that hard work was partly behind this movement. Offer (2006, p. 244) suggests that demanding and challenging work was behind a longer-run downward movement in job satisfaction in the United Kingdom and the United States. Green (2006) is a careful analysis of movements in job characteristics, pointing to greater work intensity and less worker discretion as a cause of flat if not declining job satisfaction in a number of countries. The new data in the third wave of the ISSP provide some evidence of a turning point in this decline,8 and it will be of great interest to track this evolution through both continuing panel studies such as the BHPS and the German SOEP, and the putative fourth wave of the ISSP Work Orientations module, due perhaps eight years after the third wave in 2005. Tables 14.6A and 14.6B repeat the regression analysis in Table 14.5, but changing the dependent variable from overall job satisfaction to the seven ‘‘domain’’ measures of job outcomes presented in Table 14.4 and Appendix 14.A1. This produces a lot of numbers. For ease of representation, and to keep the size of the tables down, only the estimated coefficients on the ‘‘1997’’ and ‘‘2005’’ wave dummies are shown for each of the regressions, which do however all include the same control variables as in Table 14.5. To set the scene, the first lines of Table 14.6A and 14.6B reproduce the
TABLE 14.6A 1989–2005
Estimated Changes over Time in Various Job Outcomes. ISSP Estimated Coefficients on ‘‘1997’’ and ‘‘2005’’ 1989–2005
Job satisfaction Income is High Hours Preferences Opportunities for advancement are high Job is secure Hard work Good job content Good relations at work
1997
2005
0.080** (0.025) 0.112** (0.025) 0.086** (0.025) 0.044 (0.025) 0.205** (0.026) 0.123** (0.028) 0.027 (0.027) 0.027 (0.030)
0.064* (0.027) 0.111** (0.027) 0.070** (0.027) 0.060* (0.027) 0.184** (0.027) 0.166** (0.027) 0.023 (0.028) 0.026 (0.032)
1997–2005 2005 0.044** (0.015) 0.080** (0.015) 0.002 (0.015) 0.111** (0.015) 0.021 (0.015) 0.028 (0.016) 0.013 (0.016) 0.035* (0.018)
Notes: Standard errors in parentheses. * significant at 5%; ** significant at 1%. Hours preferences: workers would prefer to work more hours, fewer hours, or the same hours. Fewer hours is coded as 1, the same hours is coded as 2, and more hours is coded as 3. The regressions include the same control variables as those reported in Table 14.5.
TABLE 14.6B Estimated Changes over Time in Various Job Outcome Measures, Controlling for Earnings and Hours of Work. ISSP 1989–2005. Estimated Coefficients on ‘‘1997’’ and ‘‘2005’’ 1989–2005
Job satisfaction Income is High Hours Preferences Opportunities for advancement are high Job is secure
1997
2005
0.079** (0.029) 0.090** (0.029) 0.090** (0.030) 0.029 (0.029) 0.162** (0.030)
0.027 (0.030) 0.207** (0.031) 0.056 (0.031) 0.038 (0.031) 0.167** (0.031)
1997–2005 2005 0.026 (0.017) 0.033 (0.018) 0.019 (0.017) 0.075** (0.017) 0.012 (0.017) (continued )
453
454
Section III: Differences in the Social Context of Well-Being
TABLE 14.6B
(Continued)
Estimated Coefficients on ‘‘1997’’ and ‘‘2005’’ 1989–2005
Hard work Good job content Good relations at work
1997
2005
0.088** (0.033) 0.024 (0.030) 0.041 (0.035)
0.197** (0.031) 0.056 (0.032) 0.045 (0.037)
1997–2005 2005 0.074** (0.018) 0.009 (0.018) 0.022 (0.020)
Notes: Standard errors in parentheses. * significant at 5%; ** significant at 1%. Hours preferences: workers would prefer to work more hours, fewer hours, or the same hours. Fewer hours is coded as 1, the same hours is coded as 2, and more hours is coded as 3. The regressions include the same control variables as those reported in Table 14.5.
estimated wave coefficients found in Table 14.5’s overall job satisfaction regressions, showing the U-shaped pattern in job satisfaction between 1989 and 2005. The remaining lines summarize the changes in the other measures of job outcomes over the different ISSP waves. In Table 14.6A, the last column shows how different measures of job quality have changed in 15 OECD countries between 1997 and 2005, without controlling for earnings or hours of work. The results here indicate that income and promotion opportunities have both improved. On the contrary, ‘‘hard work’’ has become more prevalent, and ‘‘job content’’ has become less good, although neither estimate is significant at the five percent level. The overall net effect, as in Table 14.5 and the top line of Table 14.6A, is higher job satisfaction in 2005 than in 1997. Table 14.6B repeats this analysis, but this time controlling for hours and income. Table 14.6A asked the question whether the average worker, given their demographic characteristics, has become more satisfied over time. Table 14.6B asks whether the average worker, given their demographic characteristics, and the same level of real net earnings and hours of work, has become more satisfied over time. As income has risen and hours of work have fallen in general across OECD countries over the time period analyzed here, we expect the overall positive effects outlined in Table 14.6A to be more muted in Table 14.6B, and this is indeed the case.9 The positive trend on ‘‘income is high’’ in the last column of Table 14.6A becomes negative in the last column of Table 14.6B. Equally, the higher overall job satisfaction in 2005 than in 1997 in Table 14.6A disappears in Table 14.6B. In that sense, the improvements in job satisfaction between 1997 and 2005 can be entirely explained by movements in real earnings and hours of work.
Work, Jobs, and Well-Being Across the Millennium
455
One obvious question that can be asked here is the extent to which movements in job quality are cyclical. To investigate, we added measures of GDP growth, the unemployment rate and inflation to the regressions that are summarized in Tables 6A and 6B. The errors in these new regressions are clustered at the country-year level. While it is probably wise to take the results as merely indicative, seeing as they are identified off of only a very few country-year points, the overall conclusion is that adding these macro variables broadly changes only slightly the qualitative results. The macro variables do explain the positive estimate on ‘‘2005’’ in the top-right corner of Table 14.6A. However, we already suspected that this was driven by higher incomes (which is why the corresponding estimate in Table 14.6B is insignificant). With respect to the different job domains, hours preferences, job security, and hard work all seem to be significantly correlated with the economic cycle. Greater GDP growth increases job satisfaction and promotion opportunities, but also increases hard work and is associated with less good job content. Job security is reduced by higher unemployment, unsurprisingly. These results do suggest some role for aggregate variables in determining individual wellbeing at work, and this is probably a subject that warrants further research. Table 14.7 presents separate results by country. These are perhaps less useful, mainly due to the small sample sizes in most countries. It is of
TABLE 14.7 Change in Overall Job Satisfaction by Country. ISSP 1989–1997–2005 Country West Germany Great Britain USA Hungary Norway Sweden Czech Republic New Zealand Canada Japan Spain France Portugal Denmark Switzerland
1989–1997
1989–2005
1997–2005
(10%) 0 0
þ (10%) 0 0 þ 0
þ þ þ (10%) þ 0 0 0 0 þ (10%) 0 (10%) 0 þ
Note: This table summarises the results from single-country estimation of the regression reported in Table 14.5. Significant rises in job satisfaction are indicated by a ‘‘þ’’ and falls in job satisfaction by a ‘‘’’.
456
Section III: Differences in the Social Context of Well-Being
interest to note that Hungary, despite its low ‘‘job quality’’ ranking in OECD countries, has shown improvement during both 1989–2005 and 1997–2005. There has been, if anything, something more of a trend towards better quality jobs in Anglo-Saxon countries (Germany, United States, Canada, Great Britain) than elsewhere. This chapter so far has detailed movements in different job outcome variables. What it has not yet done is show how important these separate outcomes are for individual job satisfaction. Table 14.8 remedies this oversight. Here, overall job satisfaction is regressed on the various individual job quality measures (used in Tables 14.6A and 14.6B), as well as standard demographics and year and country dummies. The first column refers to the five countries observed over all three ISSP waves, and the second to the fifteen countries present in both 1997 and 2005. The estimated coefficients show that all of the domain variables are indeed correlated with overall job satisfaction; as such any changes in these domains will feed through to overall job quality (as measured here by overall job satisfaction). It is however not straightforward to use the results in Table 14.8 to draw up a ranking of the different right-hand side variables. While research on the ranking of different life domains (health, job, marriage, etc., as in
TABLE 14.8 Overall Job Satisfaction and Job Quality Components (ISSP 1989, 1997 and 2005)
High Income Want to Spend Less Time In Job Want to Spend More Time In Job Good Promotion Opportunities Job Secure Hard Work Good Job Content Good Relations at Work N Log Likelihood Log Likelihood at zero Pseudo-R2
1989–2005
1997–2005
0.182** (0.014) 0.315** (0.026) 0.068 (0.036) 0.177** (0.013) 0.108** (0.012) 0.092** (0.012) 0.282** (0.011) 0.515** (0.019)
0.149** (0.009) 0.294** (0.018) 0.070** (0.027) 0.158** (0.009) 0.079** (0.008) 0.112** (0.008) 0.283** (0.008) 0.481** (0.013)
8427 10478.70 12179.60 0.140
17696 22580.31 25996.41 0.131
Work, Jobs, and Well-Being Across the Millennium
457
Ferrer-i-Carbonell & van Praag, 2004) is able to draw on standardized questions (‘‘How satisfied are you with your X on a scale of one to seven’’), we here have no such standardized scale for the explanatory variables we can use to calculate the relative contributions of high income and job relations, for example, or of job content and hard work. Even so, we can be sure that all of them are significant correlates of overall job satisfaction.
3. The Changing Value of Self-Employment
This last topic emphasises the specific role of self-employment in determining job quality. The self-employed are something of an enigma, as they arguably do worse on many job domains. Specifically, they typically earn less but work more hours, are arguably more insecure and face greater risks. On the plus side, they do certainly enjoy more autonomy, but enjoy relatively less social contact (people don’t like being with their co-workers very much, with their boss even less, but worst of all is being on your own: see Kahneman et al., 2004). Despite these apparent handicaps, the selfemployed typically report higher levels of overall job satisfaction than do the employed, at least in OECD countries. A natural interpretation of this satisfaction differential is that there are unmeasured aspects of jobs for which the self-employed score far higher than the employed. If this is so, the question then arises of why more of the employed do not become self-employed, given that the latter seems to be a preferable status on the labor market. This phenomenon is illustrated in Table 14.9. The left-hand panel of this table shows the change over time in the percentage of the ISSP sample who are self-employed (expressed as a percentage of all of those who are in work). As the ISSP samples are fairly small (see Table 14.2), so is the number of self-employed per country per wave, and these percentages should probably best be taken as illustrative. Although the percentage of self-employed has been rising in some countries in the ISSP data, the broad picture is of a small fall in the self-employment rate. This is consistent with the analysis of comparable LFS data in OECD (2005). The right-hand side panel of Table 14.9 shows the evolution of the percentage of respondents who, in response to the question: ‘‘Suppose you were working and could choose between different kinds of jobs. Which of the following would you personally choose?’’ replied ‘‘Being self-employed’’ rather than ‘‘Being an employee.’’ This percentage has also been falling across the three waves of the ISSP. However, what is most interesting perhaps is that the percentage who express a preference for selfemployment is systematically three or four times higher than the actual
458
Section III: Differences in the Social Context of Well-Being
TABLE 14.9
Self-Employment in the ISSP 1989–1997–2005
West Germany Great Britain USA Hungary Norway Sweden Czech Republic New Zealand Canada Japan Spain France Portugal Denmark Switzerland
Percentage of Working who are Self-Employed
Percentage of Working who Prefer Self-Employment to Employment
1989
1997
2005
1989
1997
2005
11.0% 11.7% 12.1% 5.9% 5.1%
10.2% 15.2% 13.4% 14.5% 9.8% 10.7% 10.6% 9.2% 15.2% 16.8% 3.4% 8.9% 23.6% 6.5% 12.1%
10.4% 12.9% 13.3% 9.0% 10.9% 10.3% 14.9% 15.1% 8.6% 11.4% 14.3% 8.4% 14.1% 8.5% 10.1%
51.4% 49.6% 63.5% 42.2% 26.6%
61.7% 46.2% 72.3% 58.8% 27.5% 38.0% 42.8% 63.4% 58.7% 42.7% 42.9% 42.7% 76.3% 26.1% 65.6%
44.3% 48.7% 64.4% 39.1% 28.4% 31.8% 30.7% 55.0% 55.6% 33.4% 33.9% 40.6% 51.8% 28.4% 47.2%
self-employment rate. Taking Table 14.9 at face value, there are substantial numbers of people who would prefer to be self-employed but who are currently employees. The correlation between self-employment and overall job satisfaction is detailed in Table 14.10. This shows the estimated coefficients on various self-employment variables that have been added to our ‘‘standard’’ overall job satisfaction regressions detailed in Table 14.5. The results in columns 1 and 3 show that, in both the long and the short data set (1989–1997–2005
TABLE 14.10 2005
Self-Employment and Overall Job Satisfaction. ISSP 1989–1997–
Self-Employment Self-Employment*1997 Self-Employment*2005
1989–2005
1989–2005
1997–2005
1997–2005
0.327** (0.034)
0.377** (0.062) 0.062 (0.085) 0.076 (0.081)
0.352** (0.023)
0.302** (0.032)
0.105* (0.046)
* = Significantly different from zero at the five per cent level; ** = Significantly different from zero at the one per cent level.
Work, Jobs, and Well-Being Across the Millennium
459
and 1997–2005, respectively), the self-employed are significantly more satisfied with their jobs. Columns 2 and 4 then add an interaction between self-employment and the year dummies to see if this positive relationship between self-employment and well-being at work has changed between 1989 and 2005. There is no significant evidence of any such evolution in the five countries making up the long data set between 1989 and 2005. However, for the 15 countries observed between 1997 and 2005, the interaction between self-employment and 2005 is positive and significant: in this larger sample of countries, self-employment is not only good, but is even getting better. This might be thought to pose something of a problem. If job satisfaction is higher when one is self-employed, and increasingly so, why is the number of self-employed falling in parallel? One interpretation of the job satisfaction differential is in terms of matching on the labor market. Some people may really like autonomy but not dislike risk that much: they end up self-employed. Others like autonomy less, and really dislike risk: they end up employed. It is easy to parameterize utility functions such that those who choose self-employment are more satisfied than those who choose employment.10 However, in this matching or sorting story, the employed do not want to become self-employed: not only are the employed less satisfied than the self-employed, they would be even less satisfied than they are now if we forced them to change from employment to self-employment. The matching story, then, does well in explaining why the job satisfaction of the self-employed might be higher than that of the employed in equilibrium. However, it does a much less good job of explaining the righthand side of Table 14.9: the employed shouldn’t want to be self-employed (because they have freely chosen employment), yet that is what it looks like they would prefer. One interpretation that does fit is in terms of life satisfaction vs. job satisfaction. When individuals answer the preference question, they may well have only job satisfaction in mind. However, their actual choice of labor force status will be determined by life satisfaction, which might be lower for the self-employed. Working for yourself is typically associated with greater decision-making, and Helliwell & Huang (2009) have shown that this decision-making is positively correlated with job satisfaction but negatively correlated with life satisfaction. A second alternative reading of Table 14.9 is that there are barriers to entry to self-employment: some people really want to be self-employed but cannot. These barriers are often imagined in terms of difficulties in raising the necessary capital to start one’s own business. This has been demonstrated by careful work showing that individuals are more likely to become self-employed after they have received a windfall gain (inheritance or
460
Section III: Differences in the Social Context of Well-Being
lottery win): see Blanchflower & Oswald (1998). A consistent reading of Tables 14.9 and 14.10 is then that the barriers to self-employment entry have become greater over time. This would seem worthy of further research, both via subjective and objective data. Specifically, it would seem important to carry out careful country-by-country analysis to try and establish in which countries the ‘‘returns’’ to self-employment have risen, and see whether this ties in with what we know about differences in access to capital between countries.
4. Conclusion
This chapter has analyzed repeated cross-section data from a number of OECD countries to consider movements in some of the relationships between work and well-being. There are three main results. First, analysis of European data repeats the simple message that there is a wide gulf in well-being between employment and unemployment, which does not seem to result from reverse causality (whereby the intrinsically unhappy are more likely to end up unemployed). There is no evidence that individuals adapt to unemployment, which starts bad and stays bad. However, some work has uncovered empirical evidence consistent with social comparisons in unemployment: unemployment hurts less the more there is of it around. If this is the case, then greater equality in social welfare will go hand-inhand with greater inequality in unemployment. However, despite any social comparison effect, the average well-being gap between the employed and the unemployed remains very large. For the kind of unemployment rates that have prevailed in OECD countries over recent years, social comparison effects are arguably marginal compared to the sheer size of the psychological well-being impact of not having a job. ISSP data from 1989, 1997, and 2005 show that job values are quite stable over time, although there is some evidence of workers’ giving increasing importance to ‘‘useful’’ and ‘‘helpful’’ jobs. Regarding job outcomes, job quality mostly fell between 1989 and 1997. However, the analysis of the latest 2005 wave of data reveals that job quality bounced back between 1997 and 2005. Overall job satisfaction in 2005 was actually higher than it was in 1989. Most of the different components of job quality (income, hours, promotion, relations at work, and job security) have improved in line with overall job satisfaction between 1997 and 2005. There are exceptions, though: the percentage of employees who report that they work hard rose over this period, and a measure of good job content (picking up the extent to which employees think that their job is interesting, useful, helpful, and provides independence) fell over the same period,
Work, Jobs, and Well-Being Across the Millennium
461
although neither estimate is significant. This latter movement is particularly interesting as ‘‘job content’’ picks up the job aspects that workers report as becoming increasingly important over recent years. Last, the rate of self-employment has been falling gently in ISSP data; even so, three to four times as many people say they would prefer to be selfemployed than are actually self-employed. In job satisfaction analysis in OECD countries, the self-employed are more satisfied than are employees. Moreover, this job satisfaction ‘‘premium’’ from self-employment seems to be rising. One consistent interpretation of the above is that the barriers to self-employment have been growing. Most generally, the analysis of repeated cross-section surveys across a variety of countries provides fascinating information on the evolution of the quality of workers’ lives. While the drop in job quality between 1989 and 1997 was more than reversed in 2005, more work needs to be carried out to consider exactly how this reversal of misfortune came about. In addition, the current work has mostly considered aggregate analysis across many different countries. Yet the movements in job quality are far from uniform: some countries have shown sharp improvements, while others have trodden water or even stagnated. Countries probably differ in many ways that are amenable to policy, including their education and employment institutions (see Davoine et al., 2008; and Gallie, 2007). The identification of just why some countries have done so well will surely help us identify ways in which what remains an essential element of so many people’s lives can be improved. Notes 1. See http://epp.eurostat.ec.europa.eu/portal/page/portal/microdata/echp for details of the ECHP data. 2. Similar results have been reported using German (Clark et al., 2008c) and South African (Powdthavee, 2007) data. 3. We might imagine that this reflects some kind of shift-share phenomenon: the unemployed in high-unemployment regions include a broad section of the population, whereas the few unemployed in low-unemployment regions may include a greater proportion of ‘‘difficult’’ people with lower GHQ levels. This automatically explains why the employed-unemployed gap in GHQ is smaller in highunemployment regions. To test this, I look at the GHQ score of those moving into unemployment in low- and high-unemployment regions. If shift-share holds, then those moving into unemployment in high-unemployment regions should have higher GHQ (before they became unemployed) than those moving into unemployment in low-unemployment regions. There is no evidence of this in the BHPS data. 4. The ISSP samples were mostly stratified, and designed to be representative of adults (aged 18 or over) living in non-institutional accommodation. The mode of administration was most often face-to-face. There are a number of differences between countries in this respect. Details regarding the questionnaire, sampling,
462
5. 6. 7.
8. 9.
10.
Section III: Differences in the Social Context of Well-Being and data collection are available in the Study Monitoring Report for each ISSP wave. That for 2005, for example, is available via the following web page: http:// www.gesis.org/en/services/data/survey-data/issp/modules-study-overview/workorientations/2005/. It could of course be countered that social desirability has increased over time also. We have no way of knowing this. See Clark (1997) & Sousa-Poza & Sousa-Poza (2000). The estimated coefficient on income is positive and significant, while that on hours of work is insignificant. Harter & Arora (2008) find a negative correlation between hours and their measures of both evaluative and experienced well-being. One explanation for the difference in the effect of hours is that we here consider job satisfaction, whereas Harter and Arora analyze global measures of individual well-being. They also find that the negative effect of hours of work is mitigated by the extent of ‘‘job fit’’ (defined as job satisfaction plus a measure of having the opportunity to do what they do best every day). The introduction of this kind of heterogeneity is a useful avenue for future research on job quality. Gallie (2005) had already noted a plateau in one component of job quality, work pressure, in European Union countries between 1996 and 2001. There is a potential issue regarding the quality of the earnings information in the ISSP. This is one reason for carrying out the analysis of job quality both with and without earnings. Comparing the evolution of earnings in the ISSP data with OECD data casts some doubt on the reported earnings movements in France, Spain, and the United States. Re-estimating the key regressions (in Tables 14.5 and 14.6) without these countries did not change the qualitative conclusions regarding changing job quality over time. An example is given in Section 4.3 of Clark & Senik (2006).
References Benesch, C., Frey, B. S., & Stutzer, A. (2006). TV channels, self control and happiness. University of Zurich, mimeograph. Blanchflower, D. G., & Oswald, A. J. (1998). What makes an entrepreneur? Journal of Labor Economics, 16, 26–60. Blanchflower, D.G., & Oswald, A. J. (2004). Well-being over time in Britain and the U.S.A. Journal of Public Economics, 88, 1359–1386. Blanchflower, D.G., Oswald, A.J., & Van Landeghem, B. (2009). Imitative obesity and relative utility. Journal of the European Economic Association, 7, 528–538. Clark, A. E. (1997). Job satisfaction and gender: Why are women so happy at work? Labour Economics, 4, 341–372. Clark, A. E. (2003). Unemployment as a social norm: Psychological evidence from panel data. Journal of Labor Economics, 21, 323–351. Clark, A. E. (2005a). What makes a good job? Evidence from OECD countries. In S. Bazen, C. Lucifora, & W. Salverda (Eds.), Job quality and employer behaviour. Basingstoke, UK: Palgrave Macmillan. Clark, A. E. (2005b). ‘‘Your money or your life: Changing job quality in OECD countries.’’ British Journal of Industrial Relations, 43, 377–400.
Work, Jobs, and Well-Being Across the Millennium
463
Clark, A. E., Diener, E., Georgellis, Y., & Lucas, R. (2008a). Lags and leads in life satisfaction: A test of the baseline hypothesis. Economic Journal, 118, F222–F243. Clark, A. E., & Etile´, F. (2008). Happy house: Spousal weight and individual wellbeing. Paris School of Economics, mimeograph. Clark, A. E., Frijters, P., & Shields, M. (2008b). Relative income, happiness and utility: An explanation for the Easterlin Paradox and other puzzles. Journal of Economic Literature, 46, 95–144. Clark, A. E., Knabe, A., & Ra¨tzel, S. (2008c). Unemployment as a social norm in Germany. Paris School of Economics, Discussion Paper 2008–2045. Clark, A. E., & Oswald, A. J. (1996). Satisfaction and comparison income. Journal of Public Economics, 61, 359–381. Clark, A. E., & Senik, C. (2006). The (unexpected) structure of ‘‘rents’’ on the French and British labour markets. Journal of Socio-Economics, 35, 180–196. Davoine, L., Erhel, C., & Guergoat-Larivie`re, M. (2008). Monitoring quality in work: European employment strategy indicators and beyond. International Labor Review, 147, 163–198. Di Tella, R., & MacCulloch, R. (2006). Some uses of happiness data in economics. Journal of Economic Perspectives, 20, 25–46. Diener, E., Lucas, R., & Scollon, C. (2006). Beyond the hedonic treadmill. American Psychologist, 61, 305–314. Fujita, F., & Diener, E. (2005). Life satisfaction set point: Stability and change. Journal of Personality and Social Psychology, 88, 158–164. Gallie, D. (2005). Work pressure in Europe 1996–2001: Trends and determinants. British Journal of Industrial Relations, 43, 351–375. Gallie, D. (Ed.). (2007). Employment regimes and the quality of work. Oxford, UK: Oxford University Press. Glenn, N., & Weaver, C. (1981). The contribution of marital happiness to global happiness. Journal of Marriage and the Family, 43, 161–168. Green, F. (2006). Demanding work: The paradox of job quality in the affluent economy. Princeton, NJ: Princeton University Press. Gruber, J., & Mullainathan, S. (2005). Do cigarette taxes make smokers happier? Advances in Economic Analysis & Policy, 5, Article 4. Harter, J., & Arora, R. (2008). The impact of time spent working and job fit on wellbeing around the world. The Gallup Organization, mimeograph. Harter, J., Hayes, T., & Schmidt, F. (2002). Business-unit-level relationship between employee satisfaction, employee engagement, and business outcomes: A metaanalysis. Journal of Applied Psychology, 87, 268–279. Harter, J., Schmidt, F., Killham, E., & Asplund, J. (2006). Q12 Meta-Analysis. The Gallup Organization, mimeograph. Helliwell, J. F., & Huang, H. (2009). How’s the job? Well-being and social capital in the workplace. Industrial and Labor Relations Review, forthcoming. Judge, T., Thoresen, C., Bono, J., & Patton, G. (2001). The job satisfaction–job performance relationship: A quantitative and qualitative review. Psychological Bulletin, 127, 376–407. Kahneman, D., Krueger, A., Schkade, D., Schwarz, N., & Stone, A. (2004). Toward national well-being accounts. American Economic Review, 94, 429–434. Layard, R. (2005). Happiness: Lessons from a new science. London: Penguin. Luttmer, E. (2005). Neighbors as negatives: Relative earnings and well-being. Quarterly Journal of Economics, 120, 963–1002.
464
Section III: Differences in the Social Context of Well-Being
Organization for Economic Co-operation and Development (2005), OECD Factbook. Offer, A. (2006). The Challenge of affluence. Oxford, UK: Oxford University Press. Platt, S. M. R., & Tansella, M. (1992). Suicide and unemployment in Italy: Description, analysis and interpretation of recent trends. Social Science & Medicine, 34, 1191–1201. Powdthavee, N. (2007). Are there geographical variations in the psychological cost of unemployment in South Africa? Social Indicators Research, 80, 629–652. Powdthavee, N. (2009). Ill-health as a household norm: Evidence from other peoples’ health problems. Social Science and Medicine, 68, 251–259. Sousa-Poza, A., & Sousa-Poza, A. A. (2000). Taking another look at the gender/jobsatisfaction paradox. Kyklos, 53, 135–152. Stutzer, A., & Lalive, R. (2004). The role of social norms in job searching and subjective well-being. Journal of the European Economic Association, 2, 696–719. van Praag, B. M. S., & Baarsma, B. E. (2005). Using happiness surveys to value intangibles: The case of airport noise. Economic Journal, 115, 224–246. van Praag, B., & Ferrer-i-Carbonell, A. (2004). Happiness quantified. Oxford, UK: Oxford University Press.
Appendix 14.A1 ISSP Variable Definitions
1) Pay
Objective measure: Respondent’s monthly net earnings, converted to U.S. dollars using Purchasing Power Parities from the OECD (http://www.oecd.org/dataoecd/61/56/39653523.xls). All figures are expressed in real 1989 values by deflating for U.S. Consumer Price Index (CPI) inflation (ftp:// ftp.bls.gov/pub/special.requests/cpi/cpiai.txt). The following countries have their gross earnings converted to net: & &
&
1989—G.B., U.S.A., Norway; 1997—G.B., U.S.A., Norway, Sweden, New Zealand, Italy, Canada, Japan, France, Denmark, Switzerland 2005—G.B., U.S.A., Norway, Sweden, New Zealand, Canada, Japan, France, Denmark, Switzerland.
These conversions are carried out using the OECD tax database (http://stats. oecd.org/OECDStat_Metadata/ShowMetadata.ashx?Dataset=AWCOMP_ OLD&ShowOnWeb=true&Lang=en). This provides data for 1997 to 2004. The 1997 tax rates were used to convert the 1989 and 1997 ISSP data from gross to net as necessary, and those from 2004 to convert the 2005 ISSP data. Subjective measure: Income is high. ‘‘My income is high’’—strongly agree or agree. 2) Hours of work
Objective measure: Weekly hours of work. Subjective measure: Would like to spend less or more time in job. ‘‘Suppose you could change the way you spend your time, spending more time on 465
466
Section III: Differences in the Social Context of Well-Being
some things and less time on others. Which of the things on the following list would you like to spend more time on, which would you like to spend less time on, and which would you like to spend the same amount of time on as now?’’ – A bit less time or much less time in a paid job (overwork variable) – A bit more time or much more time in a paid job (underwork variable) 3) Future prospects—promotion and job security
Opportunities for advancement are high: ‘‘My opportunities for advancement are high - strongly agree or agree.’’ Job secure. ‘‘My job is secure - strongly agree or agree.’’ 4) How difficult is the job?
Hard Work. ‘‘Based on answers to the following four questions. How often do you: – – – –
come home from work exhausted? have to do hard physical work? find your work stressful? work in dangerous conditions?’’
All of which are coded as: 1. 2. 3. 4. 5.
Always Often Sometimes Hardly ever Never
Cronbach’s alpha statistic is a way of evaluating the reliability of an additive scale created over a number of items. It measures the correlation between the scale and the underlying factor. The alpha statistic over these four elements is 0.62. Dichotomous variables were created, with 1 representing Always, Often or Sometimes, and 0 representing Hardly ever or Never. The sum of these four dummies (analogous to the Caseness scale of individual well-being in psychology) counts the number of ‘‘bad’’ job outcomes with respect to difficulty. The value zero corresponds to no bad outcomes, and four to jobs that are at least sometimes unpleasant on all of the criteria above. A dummy variable was created for workers reporting three or more such bad outcomes. The stress at work question was not asked in the United States in 1997.
Work, Jobs, and Well-Being Across the Millennium
467
5) Job content: interest, prestige and independence
Good job content. Based on answers to the following four questions. – – – –
My job is interesting In my job I can help other people My job is useful to society I can work independently
All of these are coded as: 1. 2. 3. 4. 5.
Strongly agree Agree Neither agree nor disagree Disagree Strongly disagree
Cronbach’s alpha over these four elements is 0.69. Dichotomous variables were created, with 1 representing Strongly Agree or Agree. The sum of these four variables is a measure of good job content. A dummy variable was created for workers reporting positive job content on all four aspects.
6) Interpersonal relationships
Good Relations at Work. Based on answers to the following two questions: – Relations at the respondent’s workplace: Between management and employees – Relations at the respondent’s workplace: Between workmates / colleagues Both of these are coded as: 1. 2. 3. 4. 5.
Very good Quite good Neither good nor bad Quite bad Very bad
Cronbach’s alpha over these two elements is 0.66. A dummy variable was created for those reporting Very Good or Quite Good relations with both management and with colleagues.
Appendix 14.A2 ISSP Distribution of Job Satisfaction
Job Satisfaction 1 2 3 4 5 6 7
Countries in 1989, 1997 and 2005 (5) 0.74 1.4 4.31 11.59 41.97 27.64 12.36
468
Countries in 1997 and 2005 (15) 0.73 1.47 4.74 11.85 39.39 29.43 12.39
Index
Acquired immunodeficiency syndrome (AIDS), 105–36 death from, 123–24 infection rates in Africa, 105–6 mortality from, 114 ACT. See Activation Activation (ACT), 21–22, 24–25 Adams, John, 38–39 Adaptation, 30, 220, 237, 441–42 v. basic needs, 235–37, 236t estimating differential, across rich and poor individuals, 229, 231 to health gains, 273–76 to life circumstances, 321–22n2 for scale, 279n8 v. social comparisons, 151 to unemployment, 440, 460 Adaptation effects, 224, 226, 238n11 formal hypothesis used to test for, 222 Adaptation to income, 217 basic needs and, 226–37 GWP cross-country data (2005), 232–37 happiness and, in 16 European countries (1975–2002), 229–32, 230t happiness and, in West Germany (1985–2000), 226–29, 227t income-happiness relationship and, 237 individual-level panel data, 226–29 pooled cross country-time series data, 229–32
Advanced countries income comparisons in, 139 relative income in, 151–52 time-series in, 140, 146, 161 Affect, 3–4, 334. See also Hedonic level of affect; Judgment-affect dimension; Negative affect; Positive affect as component of well-being, 129 items, 44, 51 Affect, measures of, 108–9. See also specific measures of affect life evaluation by age group/sex, HIV prevalence, and, 111f, 112 mortality on, effects of, 124 Affect-as-information theory, 58 Affect balance, xiii, 6, 9–10, 336 percent of variance, 8–9 scores for life satisfaction and happiness, 8 Affective activation, measures of, 21 Affective appraisal, 334 Affective component, 329 Affective experience, 19 dimensions of, 21, 21f life evaluation and, 333 measurement of, 17 Affective forecasting, 30
469
470
Index
Africa, 201. See also specific parts of Africa averaged coefficients from country regressions on SWB of knowing family member who died and other controls, 120t, 122 country fixed-effect regressions on SWB of knowing family member who died and other controls, 120t, 122 data and regressions, 201 data sources and dates, 216t differences in SWB between people who report having lost family member and those who do not in, 116, 118–19, 118t education and, 122–23 foreign aid to, 128 fraction of people who report knowing someone who died of various conditions, 114–15, 115t fraction of people who report losing immediate family member in last twelve months in, 116, 117t gender and age in, effect of, 123 GWP data from (2006 and 2007), 105–8, 116, 130 healthcare, 126–28 HIV infection rates in, 105–6 HIV prevalence across, variability in, 110 income equivalent of losing an immediate family member in, 123–24 income v. disease in, importance attached to, 125–26 mean SWB in, 206–7t public policy in, 125 regression results, 211t social connections in, 273 value of life and SWB in, 126–30 work hours in, 411, 427 Africa, well-being in HIV infection and, 108–14 mortality and, 114–26 Afrobarometer surveys, 125–27 Age effect in Africa, 123 life evaluation, 305 patterns of, 309–10
Age group, 407 difference in HIV infection for, 110 HIV prevalence, life evaluation, enjoyment, smiling, sadness, and depression, by sex and, (separately for low- and high-HIV countries), 111f, 112 life evaluation by, 110, 112 life satisfaction and, 84 simulations from model with DIF: percent distribution of global life satisfaction by, 93–94t AIDS. See Acquired immunodeficiency syndrome AIS. See HIV/AIDS Indicator Survey American Institute of Public Opinion, 141 Anger, 407–8 Annas, J., 129 Argentina data and regressions, 198–99 mean SWB, 202t regression results, 208t survey coverage in, 175 Aristotle, 36–37, 40–41 Arora, R., 25 Arousal, 21–22 Asia. See also specific places in Asia data and regressions, 200–201 data sources and dates, 215–16t mean SWB, 205–6t regression results, 210–11t Asianbarometer, 168–69 Aspirations bias, 266 unfulfilled, 334 Aspirations paradox, 265–71 relative income, inequality, and, 267–71 Atheists, 380–83 Austria, happiness over time in, 155f Barrington-Leigh, C., 43, 53–54 Basic needs, xvi, 218, 220, 237, 322n11 v. adaptation, 235–37, 236t adaptation to income and, 226–37 gratification of, 323–33 Becker, G. S., 219–20, 346n10 Behaviorism, ix Belgium, happiness over time in, 155f
Index Belief systems, 352, 361, 384. See also Atheists; Communism; Religion Bentham, J., 329, 339 BHPS. See British Household Panel Survey Bias aspirations, 266 focusing, 170–71 memory, 49–50 optimism, 266 personality-driven, 318 positivity, 50–51 questioning, 170–71 Blanchflower, D. G., 71, 110, 112, 231 Boredom, 407–8 Brazil data and regressions, 199 mean SWB, 203t regression results, 208t Brickman, P., 220 British Household Panel Survey (BHPS), 438 Broaden and build theory, 338–39 Buddhist monks, 385, 386n10 Cabinet Office of Japan, 196 Canada, 151–52 Cantril, H., 3–4, 43, 342, 344 Cantril Ladder of Life, 3–4, 13, 43–44, 51, 117–18, 225, 232, 234, 320–21, 413 cross-national differences in mean levels of, 53–54 v. GDP, 235–37, 236t ladder framing of, 292–93, 313 percent of variance, 8–9 responses, comparing SWL and, 293–96, 295f, 313 scores by log GDP per capita, 12f scores for life satisfaction and happiness, 8 value after death in family, 119 Cantril Ladder of Life questions, 129, 192n1, 291, 294 income and, relationship between, 250 CentERpanel, 72–73, 99n2 Central America, 311–12 Chattopadhyay, S., 260–61, 265 Chen, E. E., 41
471
Children, 28–29 Chile, 263 data and regressions, 199 GDP per capita in, recent growth rates of, 190–91 Honduras and, comparing rich and poor persons in, 270–71, 270f between Japan and, differences in global life satisfaction, 47 mean SWB, 203t regression results, 209t survey coverage in, 175 China, 168–69, 270, 366, 383 calculations of time-series changes in life satisfaction in, 181 data and regressions, 200 GDP per capita in, recent growth rates of, 190–91 GWS, 196 mean SWB, 205t regression results, 210t religiosity in, 375, 377 survey coverage in, 175, 196 urban effect in, 264 between U.S. and, differences on SWLS, 48 Cholera, 114–15 Clark, A., 399, 442 Clifton, D., 402 Coates, D., 220 Cognitive comparison, 334 Cognitive component, 329 Cohen, S., 426 Collectivist nations, 61 signatures of well-being in, 55–56 Columbus, Ohio average DIFMAX in, 20–21 comparison of well-being in, and Rennes, France, 28t structure and content of well-being in, 17–31 Common preferences, 292 Communism, 360, 366, 374–75, 380–81, 383. See also Ex-Communist countries Comparison effects, 248, 264. See also Social comparison Components and time-frame of well-being, 14, 37t
472
Index
Contentment, 329, 334 cultural variables of wants and, 333 happiness and, 335–36, 335t life appraisal and, 330–31 Contextual effects, 308 Conveniences, 6, 9, 13 Correlates of well-being, individual level, 10t Corruption, 278, 311 index, 51–52 as mediation factor, 271–73 prevalence of perceived, 306–7, 310 Crime, 278 victim, well-being effect of, 273 Cross-section, 190 actual changes in SWB and, patterns, 183–84 comparisons, 166 GDP per capita and financial satisfaction WVS, 187, 187f, 189 GDP per capita and life satisfaction WVS, 186–87, 186f GDP per capita and SWB, relation of, 179–80 regression relationships between full-period actual change in SWB and change predicted from WVS, 186t for SWB, 321 time-series and, patterns between happiness and income, 218–19 Cultural relativity, 345 Culture, 311 Current Population Survey, 73 Czech Republic, 366 Davidson, R. J., 386n10 Day Reconstruction Method (DRM), 17, 45, 52 Death calculating compensation for, of relative, 106–7 effects from different causes of, 124 fraction of people who report, of acquaintance, 15t, 114–15 fraction of people who report, of immediate family member in last twelve months in, 116, 117t
from HIV/AIDS, 123–24 in immediate family, 107, 116, 117t, 118t, 119, 122–24, 129 monetary compensation for, 129 Deaton, A., 71, 108, 122, 140, 153–54, 248, 256, 258, 260, 278n4 Decision making, 58, 459 Democracy, 354, 364 Democratization, 171, 354 Demographic, 90 variables, 409–10 Demographic and Health Surveys (DHS), 105–6, 130 Denmark, 385 employees in, 450 happiness over time in, 155f SWB in, 361, 364 Depression, 24–25, 108–9, 117–18, 407–8 associations with, 36 effects of mortality on, 124 HIV and, 112–13 markers, 23–24 probability of, 119 by sex and age group, 111f, 112 Deprivation effects economic, 320 relative, 267 social, 320 Determinants of well-being, 18f, 19 Developing countries, 260. See also specific countries happiness trends in, 166–93 health in, 273–74 income-happiness relationship in, 249–50, 271 mean SWB in, 202–7t measurement errors for, 256 relationship between long-term economic growth and growth rate of SWB, 189–91 social context in, 320 SWB regression results in, 208–11t SWB trends in, 176–78, 356 Development, 364 SWB and, relationship between, 351–52
Index DHS. See Demographic and Health Surveys Diener, C., 53 Diener, E., ix, 3, 16, 22, 25, 36, 41–42, 43, 50, 53, 57, 250 Diener, M., 53 DIF. See Differential Item Functioning Differential Item Functioning (DIF), 20–21, 47–48 adjustments, 87 comparing self-reported happiness across two countries in case of, 85f, 86 simulations from model with, 93–94t, 96, 97–98t DIFMAX. See Experienced happiness DIFMAX (episode), 20 Disease. See also specific diseases importance attached to, in Africa, 125–26 self-reported well-being and, 106–8 Di Tella, R., 228, 264–65 Domain(s). See also specific domains life satisfaction by, 73–77 life satisfaction by, self-reports on, 74t ordered probits for global life satisfaction against satisfaction with specific, 77t satisfaction, 50 Dominican Republic, 7 DRM. See Day Reconstruction Method Earnings, 450 East Asia(s’) v. North America, 53 PA and, 58 work in, 427 Easterlin, R., xv, 4, 167, 217, 238n2, 238n5, 247, 257–58, 353 Easterlin paradox, xv, 4, 70, 231, 247–48, 267, 271, 276–78, 353, 360 debates over, 248–50 ECHP. See European Community Household Panel Economic(s) deprivation, 320 indicators, 61 life satisfaction, sources of, 146
473
model to calculate counterfactual distributions of life satisfaction, 87–88, 99 SWB in, 436 Economic development, 353–55, 384 religion and SWB, 370–71, 370t SWB and, 357, 360–61 Economic growth. See also Happinesseconomic growth relationship happiness immunity to country level, 261–62t life satisfaction and, relationship between, 258, 259f, 260 relationship between growth rate of SWB and long-term, in developing countries, 189–91 SWB and, 184–89 Economies, transitional, 257–58 effects of, 277 Education, 24–25, 54, 90, 222, 224, 257, 266, 277, 407 Africa and, 122–23 HIV and, 119 satisfaction, 256 Emmons, R. A., 22 Emotions experience of, 6, 55–56 expression of, 345n3 HIV knowledge and, 113, 134t HIV prevalence and, 112–13, 112t HIV risk and, 113, 136t language and, 35 life evaluation and, measure of, x Employees, 450 interviewed in OECD countries, 445, 446t Employment effect of, 438 v. self-employment, 459 status, 222, 224 unemployed and, well-being gap between, 442–43, 443f v. unemployment, 460, 461n3 Enjoyment, 108–9, 117–18, 407–8 effects of mortality on, 124 probability of, 119 by sex and age group, 111f, 112 EQ5D, 279n13 ESM. See Experience Sampling Method
474
Index
Eudaimonia, 34, 36–37, 40–41 Eudaimonic well-being, 37 Eurobarometer, 196, 250 cumulative surveys (1975–2002), 223–25, 229–32 survey series definitions, 244–45 Europe. See also specific places in Europe effects of inequality in, 269 happiness and adaptation to income in, (1975–2002), 229–32, 230t longevity in, 275–76 pooled cross country-time series data summary, 223–25, 224t SWB, cross-country differences in, 71 work in South, 427 Europe, Eastern, 170, 374–75 democratization and happiness in, 171, 257–58 religiosity in, 375 Europe, West country-level mean life satisfaction in, 160t happiness over time in, 154, 155–60f, 157f European Americans, 58 European Community Household Panel (ECHP), 438 European Values Study, 354–55 Evaluated well-being, Net Affect and, 255f Evaluation, 21, 86, 88 Evaluative well-being, 403 Ex-Communist countries, 352, 357, 366 relationship of life satisfaction and religiosity in, 381–83, 382f religiosity in, 369, 374–75 SWB in, 378 trajectories of happiness and life satisfaction in, 359f, 360 Experienced happiness (DIFMAX), 17–18, 22, 31 average, 20–21 correlation between marriage and, 25–26 v. life satisfaction, 24 Experienced well-being, 403 Experience Sampling Method (ESM), 17 Face-to-face surveys, 6, 108, 403 Faith, 360–63 religious, 271–72
Family death, 107, 116, 117t, 118t, 119, 120–21t, 122–24, 129 income, 319–20 income equivalent of losing immediate, 123–24 NA/PA and death in, 107 SWB after death in, 116, 118–19, 118t in U.S., 73–75 Family domain, 76 evaluating independent impact of, 83–84 U.S. v. N.L. satisfaction in, 73–75 vignettes covering, 78–79 Ferrer-i-Carbonell, A., 76, 100n5 Financial satisfaction, 55, 168. See also Subjective well-being actual and predicted full-period change in, 185t actual change in, and that predicted from change in GDP per capita, 184, 184f annual growth rates of GDP per capita and, in specified countries, 178t, 184, 186 bias questioning, 170–71 on GDP per capita, regression of, 180f, 184, 186 GDP per capita and, actual and predicted annual rate of change in, 188f GDP per capita and, WVS cross-section, 187, 187f, 189 life satisfaction and, annual rate of change in, 176–77, 176f in U.S., 146–48, 147t Finland, happiness over time in, 156f Fleurbaey, M., 127–28 Focal activities, 20 Focusing bias, 170–71 Food adequacy, 312–13, 318 Food inadequacy, 310 income and, 305–6 Foodnet, 322n6, 322n12 France. See also Rennes, France employees in, 450 happiness over time in, 156f Fredrickson, B. L., 338–39
Index Freedom, 310, 318, 337 happiness and, 352–61 as mediation factor, 271–73 personal, 307, 311, 322n2 political, 54, 61 Frey, B., 152, 167–68 Gallup Organization, x, 5–6 Gallup Survey (China), 196 Gallup World Poll (GWP), x, 5–6, 25, 51, 192n1, 248–49, 257–58, 294, 313, 398, 403 adaptation to income cross-country data (2005), 232–37 African data from (2006–2007), 105–8, 116, 130 cross country data (2005), 225–26, 232–37 cross-country variance of, 318 distribution of responses across questions, 283t, 284–85f estimates of required income, 152 first wave, 292 international shares of total variance of individual responses from first three waves, (2006–2008), xi–xiii, xiif means and standard deviations of key variables for waves, 11t philosophy and definitions of, 245–46 question phrasing, 283t, 326–27 rank of nations (2007), 53 region definitions, 405–6t sub-Saharan Africa module of, (2006), 114 summary statistics for (2006), 225t survey variables sorted by country, descriptive statistics of, 286t variables, 250–51 wave 1 and wave 2, 5–6 Gaulier, G., 128 GDP. See Gross domestic product Gender, 407 effect of, in Africa, 123 equality, 354 life satisfaction and, 90 General Health Questionnaire (GHQ), 442–44 employed-unemployed gap in, 461n3
475
General Social Survey (GSS), 142, 145–46, 163n13, 231 German Socio-Economic Panel (GSOEP), 152, 219, 231, 264–65, 440 definitions, 243–44 individual-level data panel (1984–2000), 221–23, 226–29 summary statistics for, (1985–2000), 222t Germany, 100n5, 221, 243 East, 366 happiness over time in, 156f home owners v. tenants, 237 Germany, West adaptation to income and happiness in, (1985–2000), 226–29, 227t GDP per capita in, 149f income-happiness relationship in, 148–51, 160–62 life satisfaction in, 149f, 160 regressions for explaining individual life satisfaction in, 150t GHQ. See General Health Questionnaire Gilmour, R., 41 Global life satisfaction, 47, 51–52, 59–60, 76. See also Vignettes, global life satisfaction determinants of, 96 model of, 89–92 model simulations, 92–96 ordered probits for, against satisfaction with specific domains, 77t self assessment of, 89–90, 89t simulations from model with DIF: percent distribution of, by age group, 93–94t simulations from model with DIF: percent distribution of, by income group, 96, 97–98t thresholds of estimated equation for, 92t Global reports of well-being, 45–52, 45f, 59–60 item functioning in, 47–48 limitations to, 62 memory bias in, 49–50 number/scale-use in, 47 online and, distinctions between, 57 positivity bias in, 50–51 reference-group effect, 51–52 self-presentation in, 48
476
Index
GNP. See Gross national product Goals, pursuit of, 56 God surveys, importance of, 366, 367–69t, 369 changes in mean scores on, (1981–2007), 394–97t life satisfaction and, 378–80, 379f Graham, C., 249, 260–61, 263–65, 269–70 Greece, happiness over time in, 157f Greed effect, 263 Gross domestic product (GDP), 54 v. Cantril Ladder of Life, 235–37, 236t computing long-term growth rate of, 172–73 growth rates, 238n9, 455 job satisfaction and, 455 ladder score by log, per capita, 12f life expectancy and, 274f per capita, 7 per capita in U.S. over time, 141f per capita in West Germany over time, 149f regression of, 12 SWB and, 364 SWB and, short-term positive relationship between, 173–74 in U.S., 184 Gross domestic product (GDP) per capita financial satisfaction and, actual and predicted annual rate of change in, 188f financial satisfaction and, annual growth rates in specified countries of, 178t, 184, 186 financial satisfaction and, WVS crosssection, 187, 187f, 189 financial satisfaction and that predicted from change in, actual change in, 184, 184f financial satisfaction on, regression of, 180f, 184, 186 growth rates between 1960 and 2005, 239n18 growth rates of, in China, Chile, and South Korea, 190–91 v. happiness, 226, 232, 233f, 234–35 life satisfaction and, actual and predicted annual rate of change in, 188f
life satisfaction and, annual growth rates in specified countries of, 177t life satisfaction and, in Portugal (1985–2003), 232, 232f life satisfaction and, relationship between, 258, 259f, 260 life satisfaction and, WVS cross-section, 186–87, 186f life satisfaction and that predicted from change in, actual change in, 183, 183f life satisfaction on, regression of, 179f, 184, 186 log of, and happiness, 234–35f, 235 SWB, international cross-sectional relation of, 179–80 SWB and, regression relationships between annual growth rates of, 189t SWB and different types of society, 365f, 366 Gross national product (GNP), 353, 364, 366 per capita v. SWB, 218f GSOEP. See German Socio-Economic Panel GSS. See General Social Survey Gunatilaka, R., 264 GWP. See Gallup World Poll Habituation, 220 Hagerty, M., 4, 167, 353 Haisken-De New, J., 228 Happiness, 3, 9–10, 34, 278n2, 355. See also Experienced happiness; Income-happiness relationship; Subjective well-being adaptation to income and, in 16 European countries (1975–2002), 229–32, 230t adaptation to income and, in West Germany (1985–2000), 226–29, 227t atheists and, 380–83 comparing self-reported, across two countries in case of DIF, 85f, 86 consequences of, 338–39 contentment and, 335–36, 335t cross-cultural definitions and aspects of, 40–42 degree of, 343
Index democratization and, 171 desirability of, cross-cultural variation in, 41–42, 57–58 dictionary definitions of, 38, 58–59 distinctions of, 16–17 ‘‘don’t know’’ response to survey questions about, 331–32, 331t in Eastern Europe, 171, 257–58 economic growth, immunity to country level, 261–62t equation, 221–24 equation for U.S. couples, 151 factors responsible for trends in, 4 faith and, 360–63 family member and, loss of, 129 freedom and, 352–61 v. GDP per capita, 226, 232, 233f, 234–35 GDP per capita and, log of real, 234–35f, 235 health gains and, 273–76 hedonic level of affect and, 335–36, 335t income and, per capita (1990s), 249f income effects on, formal hypothesis used to test for, 223 income growth and average, 139–40 income in U.S. and, log of real, 143t inequality and, 142 judgment-affect dimensions of, 13 ladder and Affect Balance scores, 8 levels of nations (1981–2007), 354–56 life expectancy and, 339 measure v. measures of well-being, xi in nations around 2000, 344t perceived sources of, 341–42, 342t permanent effects on, 30 relative, 267 religion and, 369–71, 378–80 routes to, 383–84 societal requirements for, 336–37 SWB based on reported life satisfaction and, 362–64t theories of, 335–36, 335t trajectories of life satisfaction and in four types of societies, 358–59f, 358–60 U.S. use of term, over time, 38–39 in U.S., 147f, 160, 167
477
in U.S., by income quintile, 144f in U.S., financial satisfaction and, 147f in U.S. over time, (1950–2006), 141, 142f in U.S., regressions for explaining individual, 145t value-reality gap in nation and, 335 variations in, 345 Western Europe, over time in, 154, 155–60f, 154 Happiness, average, 343 by societal characteristics, 337t Happiness, concept of, 329, 332t cross-cultural, 58–59 Happiness, universality of, 328, 343–44 conditions and, 336–38 consequences of happiness and, 338–39 life appraisal and, 330–36 seeking happiness and, 339–43 sub-questions, 329 Happiness-economic growth relationship in developing countries, 166–93 time-series study of, 167–68 Happy peasants and frustrated achievers paradox, 263–65, 277, 360–61 Harris, A., 43, 53–54 Harter, J., 25, 255 Health in developing countries, 273–74 happiness and, 273–76 income and trade-offs between, 127–28 monetary evaluation of, 128 Healthcare, 126–28 Health domain, 275 evaluating independent impact of, 83–84 status, 81 U.S. v. N.L. satisfaction in, 73–76 vignettes covering, 78–79 Health satisfaction, 260 aspirations paradox and, 265 across Latin America, 265–66, 274–75 wealth and, 266 Hedonic balance, 55–56, 61 Hedonic level of affect, 329 extroversion and, 347n23 gratification of universal needs and, 332–36
478
Index
Hedonic level of affect (Continued ) happiness and, 335–36, 335t life appraisal and, 330 Hedonic psychology, ix Hedonic treadmill, 238n5, 352 Hedonic well-being, 403–4 Helliwell, J., 251, 268–69, 272 Helliwell, J. F., 43, 53–54, 110, 112 Hemenway, D., 152–53 Hepatitis, 114–15 Heuristic questions, 18f, 19 Hierarchical Linear Modeling (HLM), 355 HIV. See Human immunodeficiency virus HIV/AIDS Indicator Survey (AIS), 130 HLM. See Hierarchical Linear Modeling Home ownership, 223, 229, 237 Honduras, 270–71, 270f Howell, C., 253–54 Howell, R., 253–54 Huang, H., 43, 53–54 Human behavior, 339 Human immunodeficiency virus (HIV) death from, 123–24 depression and, 112–13 education and, 119 life evaluation and, link between, 108 SWB and, (means by African country), 109–10, 109t HIV infection difference in, for sex/age group, 110 rates, 105–6, 114 well-being in Africa and, 108–14 HIV knowledge, 108–9, 113, 130–31 emotions and, 113, 134t life evaluation and, 113, 133t HIV prevalence, 108–9, 130–31 emotions and, 112–13, 112t life evaluation, enjoyment, smiling, sadness, and depression, by age group and sex (separately for lowand high-HIV countries), 111f, 112 life evaluation and, 112t variability in, across African countries, 110 HIV risk, 130–31 emotions and, 113, 136t life evaluation and, 113, 135t perceived, 108–9 Hun, C., 264
Hungary, 173–74 employees in, 450 Hunger reduction, 125 Identification problem, 88 ILO. See International Labor Organization Income, 9, 24–25, 90, 407. See also Adaptation to income actual, 148f advanced countries, comparisons in, 139 in Africa, importance attached to, 125–26 coefficient on, 462n7 contextual effects, 308 cyclical component of, 154 effects on happiness, formal hypothesis used to test for, 223 family, 319–20 family member, equivalent of losing immediate, 123–24 food inadequacy and, 305–6 happiness and per capita, (1990s), 249f health and trade-offs between, 127–28 initial, 238n9 job satisfaction and, 454 Ladder of Life questions and, relationship between, 250 in Latin America, reference group, 267–68 in life satisfaction, role of, 70–99, 85 logarithm, 122–23 permanent v. transitory, 229 satisfaction, 220 trend, 154 U.S. and N.L., as variable for comparing, 91–92 in U.S., inequality over time, 143f value of life, elasticity of, 126 variable, 252, 320 well-being and, 4 Income, absolute, 146, 150, 248 effect of, 161 Income, comparator, 146, 161 negative effect of, 151 Income, log, 252 coefficients of, in N.L., 100n8 coefficients of, in U.S., 100n8 in U.S., by income quintile, 144f in U.S., happiness and, 143t
Index Income, log of household, 310, 312–13 life satisfaction and, 305 Income, relative, 161–62 in advanced countries, 151–52 effects for Latin America, 267–68, 268t hypothetical questions for, 152–53 inequality and, 267–71 perceived, 146 Income, required, 152, 161 actual and, 148f Income change habituation to, 220 with well-being measures, correlations of, 11–13 Income domain U.S. v. N.L. satisfaction in, 73–75 vignettes covering, 78–79 Income group, simulations from model with DIF: percent distribution of global life satisfaction by, 96, 97–98t Income growth, 14 average happiness and, 139–40 v. status gains, 264–65 Income-happiness relationship, 238n2, 247–58, 276–78 adaptation to income and, 237 altering factors of, 249–50 within countries, 248, 257, 263–64 country selection issues, 257–58 cross-country analysis of, 252–53, 263–64 cross-section and time-series patterns between, 218–19 in developing countries, 249–50, 271 Latin America, question phrasing within and across, 251–52, 254t mediating factors in, 271–73 rich v. poor countries, 256 in West Germany, 148–51, 161–62 Income-happiness relationship in U.S., 140–48, 160–62 aggregate data (1950-now), 141–44 individual data (1972-now), 145–46 India, 168–69 data and regressions, 200 mean SWB, 206t regression results, 211t survey coverage in, 175
479
Individualistic nations, 61 signatures of well-being in, 55–56 Inequality, 277, 357 effects of, 269 happiness and, 142 relative income and, 267–71 in U.S., effects of, 269 in U.S., income, 143f Inflation, 455 Inglehart, R., 4, 12–14, 54, 167, 249, 353 International Labor Organization (ILO), 399 International Social Survey Programme (ISSP), 438 job satisfaction, distribution of, 468 self-employment in, 458–60, 458t variable definitions, 465–67 Work Orientations modules, 445, 446t, 452 Internet surveys, 72 Iraq, 386n5 religiosity in, 375 Ireland, happiness over time in, 157f IRT. See Item Response Theory ISSP. See International Social Survey Programme Italy, happiness over time in, 157f Item functioning, 47–48 Item Response Theory (IRT), 47–48 Jackson, Andrew, 39 Janoff-Bullman, R., 220 Japan, 36, 154, 168–69, 196 calculations of time-series changes in life satisfaction in, 181 between Chile and, differences in global life satisfaction, 47 data and regressions, 200–201 employees in, 450 GDP per capita, 184 mean SWB, 206t regression results, 210t survey questions, 196 between U.S. and, differences in global life satisfaction, 47, 51–52 Japan 1 (1958–1969), 169 Japan 2 (1970–1991), 169 Japan 3 (1992–2007), 169 Japan 4 (1981–2005), 169
480
Index
Jefferson, Thomas, 39 Jepson, C., 238n11 Job. See also Work content, 454, 460–61 security, 447, 455 values, 447–48, 447t, 460 Job characteristics movement in, 452 ranks, 446–47 Job domain, 455 evaluating independent impact of, 83–84 U.S. v. N.L. satisfaction in, 73–75, 452 vignettes covering, 78–79 Job fit, 398, 400–402, 407, 410, 462n7 hierarchical regression analysis with dependent variable, 409, 414–15t, 417–18t, 420–21t life evaluation for workers with high and low, by hours worked per week, 413, 413f negative experience and, 426 negative experience for workers with high/low, 416, 419, 419f perceived, 427 positive experience for workers with high/low, 416, 416f well-being and, 411, 412t, 413f work hours, and life evaluation by region, 408–9, 408–9t work hours and, interaction between, 422, 423–26f, 431–32t Job outcomes, 448–50, 449t estimated changes over time, 452, 453t estimated changes over time with controls, 452, 453–54t movements in, 456 for 1989–2005, 448 Job quality change in, over millennium, 445–57 job satisfaction and, 456–57, 456t measure of, 450, 454 movements in, 455 self-employment and, 457 Job satisfaction, 220, 398–99, 401, 407, 452, 460 composition effects, 449–50 by country (1989–1997–2005), 455–56, 455t
distribution, 448 factors of movement in, 452 GDP and, 455 ISSP distribution of, 468 job quality and, 456–57, 456t labor force status and, 438, 439f life satisfaction and, 459 mood and, 398–99 real earning and work hours, 454 regressions, 450, 451–52t, 452 self-employment and, 457–60, 457t work hours and, 454 Judgment, 3–4, 13 Judgment-affect dimension, 3–4, 9–10 of happiness, material variable, and life satisfaction, 13 Kahneman, D., ix, 3, 43 Kharas, H., 125 King, L., 57 Kitayama, S., 41 Knight, J., 264 Krueger, A. B., 24, 43 Labor force status, 457 job satisfaction and, 438, 439f Labour Survey Force (LFS), 442–43 Ladder. See Cantril Ladder of Life Latin America, 168, 385 cell phones in, 272 data and regressions, 198–200 data sources and dates, 212–14t effects of inequality in, 269 friendship in, 271 health satisfaction across, 265–66, 274–75 income-happiness relationship question phrasing within and across, 251–52, 254t income reference group in, 267–68 mean SWB, 202–5t regression results, 208–10t relative income effects for, 267–68, 268t religiosity in, 369 SWB in, 364, 366, 369 work in, 427 Latinobarometer, 168, 172, 198, 260–61 questions, 192n6, 195–96
Index Laugher, 407–8 Lay theory of well-being, 23 LFS. See Labour Survey Force Life circumstances, 312 adaptation to, 321–22n2 correlates of, 22–25 international differences in, 292 life evaluation and, 307 relationship of, to well-being measure, 23–24, 23t Life evaluation, xiii affective experience and, 333 age effect, 305 age patterns in, 110, 112 as component of well-being, 129 death in family and, 107 HIV and, link between, 108 HIV knowledge and, 113, 133t HIV prevalence, enjoyment, smiling, sadness, and depression, by age group and sex (separately for lowand high-HIV countries), 111f, 112 HIV prevalence and, 112t HIV risk and, 113, 135t individual-level variance of, 297, 305 international differences in, 292, 307 job fit, work hours, and, by region, 408–9, 408–9t life circumstances and, 307 measure of, 107, 109 mood or emotion and, measure of, x across regions, 410–11 on a scale 0–10, 6 social context and, 312–20 for workers with high and low job fit by hours worked per week, 413, 413f Life expectancy, 318 GDP and, 274f happiness and, 339 ‘‘Life in Nation’’ survey, 168–69 questions, 196 Life satisfaction, 4, 9–10, 18, 168, 278n2, 355. See also Global life satisfaction; Subjective well-being actual and predicted full-period change in, for selected countries, 182t actual change in, and that predicted from change in GDP per capita, 183, 183f age and, 84
481
annual growth rates of GDP per capita and, in specified countries, 177t bias questioning, 170–71 country-by-country modeling of, 308–12 democratization and, 171 distinctions of, 16–17 distribution of coefficient values from within-country regressions, 309–10, 309–10f by domain, 73–77 economic growth and, relationship between, 258, 259f, 260 economic model to calculate counterfactual distributions of, 87–88, 99 economic sources of, 146 education and, 256 emotional experiences and, relationship between, 6, 55–56 in ex-Communist countries, 359f, 360, 381–83, 382f v. experienced happiness, 24 financial satisfaction and, annual rate of change in, 176–77, 176f framing questions, 250 on GDP per capita, regression of, 179f, 184, 186 GDP per capita and, actual and predicted annual rate of change in, 188f GDP per capita and, in Portugal (1985–2003), 232, 232f GDP per capita and, relationship between, 258, 259f, 260 GDP per capita and, WVS cross-section, 186–87, 186f gender differences in, 90 God and, importance of, 378–80, 379f GWP questions on, 6 income and, log of household, 305 income in, role of, 70–99, 85 individual, estimation form for analysis of, 308–9 job satisfaction and, 459 judgment-affect dimensions of, 13 ladder and Affect Balance scores, 8 mood and, 19 national mean, high-impact event and, 54 non-financial determinants of, 297
482
Index
Life satisfaction (Continued ) predictors of, 23 religion and, 369–71 religion and, in rich and poor countries, 381–82, 381–82f SAT, as represented by, 22 self-reports on, with domains of life, 74t in South America, 311–12 SWB based on reported happiness and, 362–64t time-series behavior of, 154 time-series calculations of changes in, in China and Japan, 181 time-series movement in, 176–77 trajectories of happiness and, in four types of societies, 358–59f, 358–60 two-level analysis of, 296–308 unemployment and, 440, 441f between U.S. and N.L., differences in, 72–99 valuation components, 272f in West Europe, country-level mean, 160t in West Germany, 149f, 161 in West Germany, regressions for explaining individual, 150t Living standards, 337–38 effect of rising, 235 Loewenstein, G., 238n11 Lora, E., 258, 260, 265 Louie, J. Y., 41 Lu, L., 41 Lucas, R. E., 3 Luttmer, E. F. P., 151 Lutz, C., 42 Luxembourg, happiness over time in, 158f Lyubomirsky, S., 57 MacCulloch, R., 228 Madison, James, 39 Malaria, 114–15 Marital status, 28, 83, 90, 99, 222, 224, 305, 338, 407 correlation between, and DIFMAX, 25–26 Markon, 40–41 Marsella, J. A., 36 Marxist ideology, 374 Material well-being, 12–13 Matsumoto, D., 311
McMahon, D. M., 41 Meaning in Life Scale, 44 Mean SWB in developing countries, 202–7t differences in, across cultures, 53–54 levels of, 61 Measurement errors, 35, 43–46, 59–60 for developing countries, 256 Measures of well-being, 117–18. See also specific measures of well-being degree of sensitivity of various, 61 distribution of responses to alternative, 294, 295f, 296 happiness measure v., xi with income change, correlations of, 11–13 at individual level, inter-correlations of, 9t inter-correlations of, 9t internationally comparable, x multiple-item, 35 relationship of, to life circumstances, 23–24, 23t single-item, 35 subjective and objective, 266–67 two-level global and regional equations based on individual responses, 296–308 using multiple, xi Memory bias, 49–50 Men job outcomes, 448–50, 449t job values for, 447–48, 447t Methodological issues, 43–46, 59 Mexico data and regressions, 199 mean SWB, 204t regression results, 209t survey coverage in, 175 Miao, F. F., 41 Middle aged, global vignettes for, 80t, 81, 83 Midlife in the United States (MIDUS), 57 MIDUS. See Midlife in the United States Millennium Development Goals, 125 Modernization, 351, 361, 384–85 Møller, Valerie, 172 Monroe, James, 39 Monthly survey (MS), 73, 99n2
Index Mood, 330 function of, 332–33 job satisfaction and, 398–99 life evaluation and, measure of, x life satisfaction and, 19 Morris, W., 436 Mortality acceptance of high levels of, 12i7 from AIDS, 114 effects of, on affect measures, 124 rates, UNAIDS (2008) estimates of, 116–17 well-being in Africa and, 114–26 MS. See Monthly survey Multiple-item measures, 35 NA. See Negative affect National Opinion Research Center (NORC), 141 National Survey of Families and Households, 151 Need theory, 342–43 Negative affect (NA), 58 death in family and, 107 Negative experience as dependent variable, 420–21t job fit and, 426 summary of standardized main effects and interaction for daily, 419, 422, 422t for workers with high/low job fit, 416, 419, 419f in workplace, 407–8 Net Affect, 20 evaluated well-being and, 255f Netherlands (N.L.) coefficients of log income in, 100n8 happiness over time in, 158f health domain, 73–76, 275 income domain satisfaction in, 73–75 life satisfaction in, v. U.S., 72–99 responses in, v. U.S., 96, 98 Ng, W., 25 Nigeria, 168 mean SWB, 206t regression results, 211t survey coverage in, 175 NORC. See National Opinion Research Center Nordic countries, 366, 384–85
483
North America. See also specific places in North America v. East Asia, 53 Norway, 386n6 happiness over time in, 158f Number/scale-use, 47 Nussbaum, M. C., 40–41 Objective reality, 59–60 OECD. See Organization for Economic Cooperation and Development OED. See Oxford English Dictionary OLS. See Ordinary least squares Online reports of well-being, 45–46, 45f, 52, 59 global and, distinctions between, 57 Optimism bias, 266 Ordinary least squares (OLS), 175, 222, 355 Organization for Economic Cooperation and Development (OECD), 251, 306, 311, 437, 445 number of employees interviewed in, countries, 445, 446t Oswald, A. J., 71, 110, 112, 126, 231 Oxford English Dictionary (OED), 38 PA. See Positive affect PANAS. See Positive and Negative Affect Schedule Paradox of unhappy growth, 258–63, 260t, 277, 279n11, 360–61 Partnership status, 95–96 Pattern of Human Concerns, 342–43 The Pattern of Human Concerns (Cantril), 344 PCA. See Principal components analysis Penn World Tables, 308 Personality, 30, 228–29, 319, 338 driven bias, 318 influence of, 347n23 Person-environment fit, theory of, 399 Peru data and regressions, 199 happy peasants and frustrated achievers paradox in, 263–65 mean SWB, 204t regression results, 209t Pettinato, S., 249, 263–65, 269–70
484
Index
Pleasure, 340 Poland, 366, 386n8 Polio, 114–15 Pollak, R., 219–20 Portugal happiness over time in, 159f life satisfaction and GDP per capita, (1985–2003), 232, 232f Positive affect (PA), 58, 239n15 Positive and Negative Affect Schedule (PANAS), 44 Positive experience as dependent variable, 417–18t summary of standardized main effects and interaction for daily, 419, 422, 422t for workers with high/low job fit, 416, 416f in workplace, 407–8 Positivity bias, 50–51 Poverty reduction, 125 Powdthavee, N., 126 PPP. See Purchasing Power Parity Pressman, S. D., 426 Pride, 61, 407–8 Primary sampling units (PSU), 108, 403 Principal components analysis (PCA), 252 Principles of Morals and Legislation (Bentham), 339 PSU. See Primary sampling units Psychological Well-Being Scale, 44 Public policy in Africa, 125 directing, 107 priorities, 125–27 Purchasing Power Parity (PPP), 252 Question framing, 250–51, 253–54, 267–68, 276–77 bias, 170–71 role of, 256 Question phrasing, 255 income-happiness relationship, within and across Latin America, 254t in survey questions, 283t Quincy, James, 39 RAND American Life Panel, 72–73 Random-digit-dial (RDD), 6, 245, 403
Rayo, L., 219–20, 346n10 RDD. See Random-digit-dial Reagan, Ronald, 39 Reference-group effect, 51–52 Religion, 361, 369–70 changes in emphasis (1981–2007), 376, 377f economic development and SWB, 370–71, 370t in ex-Communist countries, 369, 374–75, 381–83, 382f on happiness, impact of, 369–71, 378–80 importance of God, 366, 367–69t, 369 life satisfaction and, 369–71 life satisfaction and, in rich and poor countries, 381–82, 381–82f SWB and, 352, 371, 372–74t, 374 SWB and, among rich and poor countries, 375, 376f, 377 Religious faith, 271–72 Rennes, France average DIFMAX in, 20–21 comparison of well-being in, and Columbus, Ohio, 28t structure and content of well-being in, 17–31 Reported well-being, 278n2, 279n12 Research on well-being, checklist for cross-cultural, 60t Respect, 407–8 Response consistency, 86–87, 88, 91 Response equation, 87–88 Retired, global vignettes for, 80t, 81 Russell, J. A., 35 Russia happy peasants and frustrated achievers paradox in, 263–65 religiosity in, 377 SWB in, 357–58 Rwanda, 386n5 religiosity in, 375 Sadness, 108–9, 117–18, 407–8 effects of mortality on, 124 probability of, 119 by sex and age group, (separately for low- and high-HIV countries), 111f, 112
Index Satisfaction with life (SWL), 291, 320–21 Ladder responses and, comparing, 293–96, 295f, 313 WVS data, 292–94 Satisfaction with Life Scale (SWLS), 44, 59 positivity bias and, 50–51 responses, 47 U.S. v. China on, 48 Schkade, D. A., 24, 43 Schwartz, N., ix, 43 Self actualization, 400 Self-Anchoring Striving Scale, xvii. See Cantril Ladder of Life Self-determination, 56 Self-employment, 438 v. employment, 459 in ISSP, 458–60, 458t job quality and, 457 job satisfaction and, 457–60, 457t rate of, 461 value of, 457–60 Self-esteem, 51, 55–56, 61 Self-expression, 354 Self-presentation, 48 Self-reports, 96, 239n15 frequency distribution of, 86 inadequacies of, 130 on life satisfaction by domain, 74t v. vignettes, 88 Seppala, E., 41 Set-point theory, 357 Shao, L., 50 Single-item measures, 35 Slovenia, 366 Smallpox, 114–16 Smiling, 108–9, 117–18, 407–8 by age group and sex (separately for low- and high-HIV countries), 111f, 112 effects of mortality on, 124 probability of, 119 Smith, H., 50 Social capital as mediation factor, 271–73 OECD definition of, 311 Social comparison v. adaptation, 151 theory, 353 unemployment, with respect to, 441–42
485
Social connection, measure of, 306 Social contacts domain, 76 evaluating independent impact of, 83–84 U.S. v. N.L. satisfaction in, 73–75 vignettes covering, 78–79 Social context, xiii in developing countries, 320 importance of, compensating differentials, 316–17t, 319 importance of, macro-level estimates, 314–15t, 318 life evaluation and, 312–20 measuring quality and nature of, 294 variables, 318–19, 320 Social deprivation, 320 Social equality, 54 Social indicators, 61 Social rank, 337–38 Social support, 56 Social tolerance, 354, 364 Societal conditions average happiness by, 337t for happiness, 336–37 Solnick, S., 152–53 South Africa, 168 crime in, 273 mean SWB, 207t regression results, 211t survey coverage in, 175 South Africa Quality of Life Trends Study, 168, 171 questions, 197 South America, life satisfaction in, 311–12 South Asia, 427 South Korea, 168–69 data and regressions, 201 GDP per capita in, recent growth rates of, 190–91 happiness trends in, 173–74 Soviet successors, 366, 374–75 Spain, happiness over time in, 159f Standard of living, perceived, 260 Statistics Netherlands, 72–73 Status gains v. income gains, 264–65 Stevenson, B., 4, 140, 142, 153–54, 166–67, 173–75, 248, 258, 260, 278n4
486
Index
Stone, A. A., 43 Strengths theory, 402 Stress, 24–25, 407–8 Structure of well-being, 17–18 Stutzer, A., 152, 167–68 Subjective Happiness Scale, 44 Subjective well-being (SWB), 16–17, 34, 167–68, 250. See also Mean SWB African country fixed-effect regressions on, of knowing family member who died and other controls, 120t, 122 averaged coefficients from African country regressions on, of knowing family member who died and other controls, 121t, 122 computing long-term growth rates and short-term fluctuations in, 172–73 consequences of, 57–58, 61 correlates of, across cultures, 54–56 cross-cultural differences in, 71 cross-national differences in levels of, 361, 364 cross-sectional explanation for, 321 cross-sectional patterns and, actual changes in, 183–84 demographic status, 119 developing countries, regression results in, 208–11t developing countries, trends in, 176– 78, 356 in developing countries, relationship between growth rate of long-term economic growth and, 189–91 development and, relationship between, 351–52 differences in, between people who report having lost a family member and those who do not, 116, 118–19, 118t economic development and, 357, 360–61 economic growth and, 184–89 in economics, 436 equations based on global samples, 297, 298–301t, 305 equations based on samples separated by world region, 302–4t Europe, cross-country differences in, 71
in ex-Communist countries, 378 in 52 countries, 389–93t formula for non-WVS scale and WVS scale, 174 GDP and, 364 GDP and, short-term positive relationship between, 173–74 GDP per capita and, regression relationships between annual growth rates of, 189t GDP per capita and different types of society, 365f, 366 GDP per capita to, international cross-sectional relation of, 179–80 v. GNP per capita, 218f happiness and life satisfaction, based on reported, 362–64t HIV and, (means by African country), 109–10, 109t in Latin America, 364, 366, 369 predicted and actual trends in, 179–84 regression relationships between full-period actual change in, and change predicted from WVS cross-section, 186t religion among rich and poor countries and, 375, 376f, 377 religion and, 352, 371, 372–74t, 374 religion and economic development, 370–71, 370t in Russia, 357–58 SWB and, 352 U.S., cross-country differences in, 71 U.S. post World War II, 217 value of life and, in Africa, 126–30 WVS regression relationships between full-period actual change in, 186t Subjective well-being (SWB) index, 355 change on, 356f Sub-Saharan Africa, 108–9 death of family in, 119 module of GWP (2006), 114 transitional economies, effects of, 277 value of life in, 106, 126 Suh, E. M., 3, 50 Suicide rate, 52
Index Survey variables descriptive statistics of, sorted by country, 287t sorted by country, descriptive statistics of, 286t SWB. See Subjective well-being Sweden, happiness over time in, 159f Switzerland, employees in, 450 SWL. See Satisfaction with life SWLS. See Satisfaction with Life Scale Tanaka-Matsumi, J., 36 Terminology, xi Time-series, xv–xvi, 139, 153–54 adaptation to income, pooled cross country data, 229–32 in advanced countries, 140, 146, 161 behavior of life satisfaction, 154 calculations of, changes in life satisfaction in China and Japan, 181 cross-country, data summary, 224t cross-section and, patterns between happiness and income, 218–19 happiness-economic growth relationship, study of, 167–68 life satisfaction, movement in, 176–77 Time-use, xiv, 9, 17 analyses of, 25–26 comparison of, for women with and without a mate, 25t, 26 experienced affect for a weekday and, 27t mood and, 19 Tortora, R. D., 125 Trust questions, 306 Tsai, J. L., 41 Tuberculosis (TB), 114–15 Turkey, 168–69 data and regressions, 201 mean SWB, 206t Ubel, P. A., 238n11 Uchida, Y., 41 U.K. See United Kingdom UNAIDS mortality rates, (2008), 116–17 Unemployment, 239n13, 264, 455 adaptation to, 440, 460 effect of, 438 v. employment, 460, 461n3
487
employment and, well-being gap between, 442–43, 443f individual well-being and, 438–40 life satisfaction and, 440, 441f partner unemployment and, well-being and, 444, 444t regional, 445 social comparison with respect to, 441–42 United Kingdom (U.K.), 100n5 happiness over time in, 160f job satisfaction in, 452 value of a statistical life, 126 well-being over time in, 71 United States (U.S.). See also Incomehappiness relationship in U.S.; specific places in United States average happiness and average log income in, by income quintile, 144f between China and, differences on SWLS, 48 employees in, 450 family and social contacts domain in, 73–75 financial satisfaction in, 146–48, 147t GDP per capita in, 184 GDP per capita over time in, 141f happiness and log of real income in, 143t happiness equation for couples, 151 happiness in, 147f, 161, 167 happiness in, by income quintile, 144f happiness in, regressions for explaining individual, 145t happiness over time in, (1950–2006), 141, 142f happiness over time in, use of term, 38–40 health standards in, 73–76, 275 income domain satisfaction in, 73–75 income in, coefficients of log, 100n8 income inequality over time in, 143f inequality in, effects of, 269 between Japan and, differences in global life satisfaction, 47, 51–52 job satisfaction in, 73–75, 452 life satisfaction in, v. N.L., 72–99 post World War II SWB in, 217 responses v. N.L., 96, 98
488
Index
United States (U.S.) (Continued ) SWB, cross-country differences in, 71 value of a statistical life, 126 Universal strivings, 340 Urry, H. L., 44–46 Value of life in Africa, SWB and, 126–30 income elasticity of, 126 in sub-Saharan Africa, 106, 126 Van Praag, B. M., 76, 100n5, 152 Veenhoven, R., 4, 7, 53–54, 57, 167, 218, 353 Venezuela data and regressions, 199–200 mean SWB, 205t regression results, 210t Vietnam, 366, 383 religiosity in, 375 Vignette(s), 96, 99, 100nn6–7 description of, 78–79 effect of, descriptions on evaluation, 82–83t, 83–84 equivalence, 88, 91 evaluations, 88 v. self-reports, 88 theory of, 85–86 Vignettes, global life satisfaction, 78–79, 102–4 for middle aged, 80t, 81, 83 responses to, 79–85 for retired, 80t, 81 variation in, 79t, 80–81 for youth, 80t, 81 Vitality Scale, 44 Washington, George, 38 Wathieu, L., 219–20 Wealth, health satisfaction and, 266 Wealth index, 252 descriptive statistics of assets used to generate, 287t Webster’s Unabridged Dictionary (WUD), definitions of happiness, 38 Welfare, 128, 247 Well-being, 35–37 cross-cultural similarities and differences in, 40–43 historical changes in, 37–40
Whyte, M., 264 Wolfers, J., 4, 140, 142, 153–54, 166–67, 173–74, 175, 248, 258, 260, 278n4 Women comparison of time-use for, with and without mates, 25t, 26 job outcomes, 448–50, 449t job values for, 447–48, 447t in Rennes, France v. Columbus, Ohio, 19, 22–23 Work, 398. See also Job attitudes, 401–2 negative experience at, 407–8, 416, 419, 419f overtime, 400 positive experience at, 407–8, 416, 416f quantity and quality of, 399–400 related variables, 407 subjective experience and objective outcomes, 424–25 well-being and, 437–45, 460 Work hours, 84, 398–400, 419, 426, 450, 462n7 in Africa, 411, 427 average, per week by region, 410 hierarchical regression analysis with dependent variable, 409, 414–15t, 417–18t, 420–21t job fit, and life evaluation by region, 408–9, 408–9t job fit and, interaction between, 422, 423–26f, 431–32t job satisfaction and, 454 life evaluation for workers with high and low job fit by, 413, 413f mean, by country, 433–35t preferences, 455 well-being and, 411, 412t, 413 Work Orientations modules (ISSP), 445, 446t, 452 World Bank, 169 World Database on Happiness, 7, 329–30, 334 World Values Survey (WVS), x, 167, 169, 171, 198, 250, 353 five waves of, (1981–1984, 1989–1993, 1994–1999, 1999–2004, 2005–2007), 168
Index GDP per capita and financial satisfaction, cross-section, 187, 187f, 189 GDP per capita and life satisfaction, cross-section, 186–87, 186f 1981–2006, 249 questions, 195 regression relationships between full-period actual change in SWB and change predicted from, cross-section, 186t response scale, 174
489
shift in survey coverage, 175 SWL data, 292–94 WUD. See Webster’s Unabridged Dictionary WVS. See World Values Survey (WVS) Youth, 28–29 global vignettes for, 80t, 81 Yuutsu, 36 Zimbabwe, SWB in, 361, 364
