The Price of Virtue
There is no possibility of making a numerical estimate of the total scale of philanthropic action...
25 downloads
639 Views
880KB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
The Price of Virtue
There is no possibility of making a numerical estimate of the total scale of philanthropic action. Nor indeed would much be gained by attempting this. Individual philanthropic agencies differ from one another so widely that they cannot be made the subject of useful statistical summary. W. Beveridge, Voluntary Action (George Allen and Unwin, London, 1948).
The Price of Virtue The Economic Value of the Charitable Sector
Vivien Foster The World Bank, Washington, DC, USA and CSERGE, University College London, UK
Susana Mourato T.H. Huxley School of Environment, Earth Sciences and Engineering, Imperial College London and CSERGE, University College London, UK
David Pearce CSERGE, University College London, UK
Ece Özdemirog˘lu Economics for the Environmental Consultancy Ltd (EFTEC), London, UK
Edward Elgar Cheltenham, UK • Northampton, MA, USA
© Vivien Foster, Susana Mourato, David Pearce, Ece Özdemirog˘lu 2001 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 or photocopying, recording, or otherwise without the prior permission of the publisher. Published by Edward Elgar Publishing Limited Glensanda House Montpelier Parade Cheltenham Glos GL50 1UA UK Edward Elgar Publishing, Inc. 136 West Street Suite 202 Northampton Massachusetts 01060 USA
A catalogue record for this book is available from the British Library Library of Congress Cataloguing-in-Publication Data The price of virtue : the economic value of the charitable sector / Vivien Foster … [et al.]. Includes bibliographical references and index. 1. Charities—Economic aspects—Evaluation. 2. Voluntarism—Economic aspects—Evaluation. 3. Fund raising—Evaluation. I. Foster, Vivien, 1968– HV48.P75 2001 338.4'33617—dc21 00–062287 ISBN 1 84064 485 0 (cased) Typeset by Manton Typesetters, Louth, Lincolnshire, UK. Printed and bound in Great Britain by Bookcraft (Bath) Ltd.
Contents List of figures List of tables A note on contributions Acknowledgements Preface
vi viii x xi xii
PART I MEASURING THE ECONOMIC VALUE OF THE CHARITABLE SECTOR 1 2 3 4 5
Conceptual foundations The benefits of charities to the general public The benefits of charities to users: the homeless The benefits of volunteering The aggregate benefits of the charitable sector: summary
PART II 6 7 8
3 19 72 101 114
CAPTURING THE ECONOMIC VALUE OF CHARITIES
Providing fiscal incentives for giving Choosing fundraising methods Targeting donors
123 147 167
PART III POLICY AND SOCIAL CAPITAL 9
On social capital
193
10 Conclusions and policy implications
204
References Index
209 219
v
List of figures 1.1 1.2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 3.10 3.11 3.12 4.1 4.2 6.1 6.2
Social benefits from charities Social costs of charities Importance of charities Attitudes towards donating Attitudes towards charitable donations: scope Attitudes towards charitable donations: foresight Attitudes towards charitable donations: commitment Attitudes towards volunteering: motive Alternatives to volunteering WTP distribution in CVALL and CVHH Sources of income Current expenditure breakdown Percentage frequency distribution of current weekly income and expenditure Duration of homelessness Percentage of respondents using general hostel facilities Percentage of respondents using counselling and support services Satisfaction with the quality of general hostel facilities and services Satisfaction with the quality of hostel counselling and support services Frequency distribution of housing preferences across different time horizons Answers to the question ‘Would you like this idea to actually happen?’ Distribution of WTA compensation amounts Predicted expenditure breakdown after the proposed scenario Breakdown of differences between valuation methodologies for males in 1993 Breakdown of differences between valuation methodologies for females in 1993 Percentage frequency distribution of tax price Percentage frequency distribution of efficiency price vi
12 17 33 34 38 39 40 44 46 50 80 81 81 83 84 85 86 87 88 89 92 96 111 111 134 136
List of figures
6.3 7.1 7.2 8.1 8.2 8.3
Percentage frequency distribution of time-value price Percentage frequency distribution of average donations for different fundraising methods Marginal effect of an approach on the probability of a gift against number of approaches Positioning of some of the UK’s major environmental groups in characteristics space Fitted probability of group membership against continuous explanatory variables Marginal cost–revenue ratio against optimal number of targeting variables for the ‘AND’ strategy
vii
137 157 164 173 178 184
List of tables 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 2.10 2.11 2.12 2.13 2.14 2.15 2.16 2.17 2.18 2.19 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8
Bid vectors for CVALL and CVHH Payment ladder used in the general public survey (CV version) and choices of a hypothetical respondent Sample question used in the general public survey – CR version Socioeconomic characteristics of different subsamples Comparison of participation rates in different forms of donations Summary of existing donations of time and money Correlation between different motives for monetary donations Correlation between different approaches to monetary donations Correlations between different motives for volunteering Percentage of respondents in each response category Yearly WTP estimates for different models and uncertainty assumptions Yearly WTP for CVALL and for CVHH from payment ladder Distribution of total WTP across different areas of charitable activity Description of explanatory variables Valuation functions for CVALL and CVHH: payment ladder results Results of contingent ranking version Summary of debriefing questions across subsamples A comparison of WTP estimates across models Relative importance of different charitable subsectors Definition of services provided by different types of hostels List and characteristics of selected hostels Summary statistics of selected socioeconomic variables Housing history Correlation between attitudes towards the proposed scenario and attitudes towards hostel services WTA compensation to give up hostel services Description of explanatory variables Valuation function: WTA compensation for hostel closure viii
23 24 26 29 31 32 36 41 45 47 48 50 52 54 56 59 60 62 64 77 78 79 82 90 91 93 94
3.9 4.1 4.2 4.3
4.4 5.1 5.2 6.1 6.2 6.3 6.4 6.5 6.6 6.7 7.1 7.2 7.3 7.4 7.5 8.1 8.2 8.3 8.4 8.5
List of tables
ix
Reasons for not answering the valuation question Summary of Heckman selectivity models used to fit net hourly wage Estimated value of volunteer time using the opportunity cost approach IGS volunteer categories against equivalent NES occupational categories with corresponding average male and female hourly wage rates for 1993 Estimated value of volunteer time using the replacement cost approach The net social value of housing and homelessness charities, 1997 The net social value of all charities, 1997 Summary of literature on price and income elasticities of giving Contrast between tax treatment of charitable donations in the USA and the UK Summary of literature on price and income elasticities of volunteering Monthly philanthropic participation rates against different price variables Mean monthly donations against different price variables Heckman selectivity models for giving and volunteering Price and income elasticities for different types of philanthropic activity Cross-tabulation of fundraising approaches and giving behaviour Summary statistics for different fundraising methods Ordinary least squares regression models for the number of fundraising approaches Selectivity models for all philanthropic gifts with controlling for fundraising effort Estimated marginal effects of fundraising A comparative profile of some of the UK’s major environmental groups Environmental group membership patterns from BSAS Summary of coefficient estimates for logit models of group membership P and N against number of variables used in targeting Overlap ratios for target population segments between groups for the ‘AND’ strategy
97 104 105
107 108 117 118 126 127 130 138 139 142 144 154 155 159 162 163 172 174 177 183 188
A note on contributions Chapters 1–5 are the result of teamwork between EFTEC (Economics for the Environment Consultancy) and CSERGE (Centre for Social and Economic Research on the Global Environment) on a project funded by the Charities Aid Foundation. Contributing authors are Vivien Foster, Susana Mourato, Ece Özdemirog˘lu, Steve Dobson, David Pearce and Tannis Hett. Chapters 6–8 have been prepared by Vivien Foster under an award from the UK Economic and Social Research Council (ESRC). Vivien wishes to thank the ESRC for their support. Chapters 9–10 were drafted by David Pearce, Vivien Foster and Susana Mourato.
x
Acknowledgements Our overwhelming debt is to Cathy Pharoah, the Head of Research at the Charities Aid Foundation, and Michael Brophy, Chief Executive of the Charities Aid Foundation. It was an idea of Michael Brophy’s that sparked off the partnership between CSERGE, EFTEC and the CAF. Michael was convinced that there had to be a better way of determining the social value of the voluntary sector. Over a London lunch several of us rashly agreed that what we had been doing for years in the sphere of environmental economics could be applied to the voluntary sector. We would have an answer in months. Several years later this book has emerged, much of it containing the results of the work that was funded by the CAF. We are grateful to the CAF for that sponsorship, but more grateful still for the trust Cathy and Michael had in us to do the work. We are aware that what we had done is partly controversial, but science makes no forward moves without controversy. A major part of the work has been contributed by Vivien Foster and Part II is adapted from Vivien’s doctoral thesis for the Department of Economics at University College London. Finally, we wish to thank Steve Dobson, who was involved in the initial stages of the CAF work when he was at EFTEC. He helped shape the ideas and ‘models’ that we subsequently applied. VF, SM, DWP, EÖ London, August 1999
xi
Preface Economists have paid some, but not much, attention to the charitable sector in today’s economies. Also known as the voluntary or ‘non-profit’ sector, the activity of giving time and money for the benefit of other people, the environment and cultural assets has grown to such proportions that it effectively forces detailed scrutiny. Giving takes many forms, and this presents problems of defining just what the voluntary sector is. A major international research project – the Johns Hopkins Comparative Non-profit Sector Project – classifies voluntary organizations according to the following characteristics: formal activity in the sense of having rules, self-governing and independent of government (though not financially), acting primarily as a non-business, not distributing profits, and being voluntary in terms of donations of time or money or both (Kendall and Knapp, 1995). On this basis, the voluntary sector accounts for around 2 per cent of total employment in Italy and Japan, 4 per cent in the UK and France, and 6 per cent in the USA (Kendall, 1996). Across eight countries (UK, USA, Sweden, France, Germany, Hungary, Italy and Japan) the operating expenditures of the non-profit sector accounted for an average of 4.6 per cent of those countries’ GNP in 1990, an absolute magnitude of $614 billion (Salamon et al., 1995). By any yardstick, the voluntary sector is huge, and it is growing. The problem with prevailing measures of the size of the non-profit sector is that they do not measure the true ‘social value’ of the sector. Social value must somehow reflect the output of the sector. Yet the sum of donations and grants does not measure output; it measures input, that is, the cost of supplying charitable services. In fact it does not even fully measure inputs, since the value of volunteer time is not taken into account. Similarly, contributions to GNP are not measured by social value but, again, by the costs of supplying the services. This approach to GNP measurement is familiar: many economic activities are not bought and sold in the marketplace, so there are no ‘revenues’ to observe. In such circumstances it is commonplace to measure the contribution to GNP in terms of the costs of providing the service, as with public education and public health services. But the resulting measures are imperfect and potentially misleading. One aim of the current volume, then, is to pursue the idea of measuring the economic value of the charitable sector by looking at measures based on output, not input. Put another way, we ask the question: what are people willing to pay for the services provided by the charitable sector? xii
Preface
xiii
To our knowledge, the answers to this question, reported in Chapters 2–6 of this volume, represent the first attempt ever to measure the value of the charitable sector in terms of willingness to pay. In case this looks like a straw man, inventing a measure that has no particular rationale, willingness to pay is precisely the measure that is used to measure the output of the marketed sector of the economy. Willingness to pay (hereafter WTP) reflects individuals’ preferences for a good or service, whether that be the contents of a supermarket trolley, the conservation of a historic building, wildlife preservation, or the provision of care for the aged. We therefore treat charitable services just like any other economic good. The interest lies in the fact that those services are not directly marketed, and hence we have to resort to techniques of ‘non-market valuation’ to find out the WTP for them. Non-market valuation techniques involve discovering what people would be willing to pay if only there were a market. Broadly, two techniques are involved. The first looks for existing markets and asks if they embody in some way the value of the associated good or service we are interested in. Suppose the problem is the economic value of cleaning up air pollution. We do not buy and sell pollution, but we do buy and sell houses and we know that house prices reflect the neighbourhood amenities surrounding those houses, including the quality of the air. This ‘revealed preference’ approach, then, looks for an associated, or complementary, market and estimates WTP from observations in that market. The second technique is familiar to anyone who has ever carried out, or been the subject of, market research: we ask people for their WTP. Sophisticated questionnaires are constructed with the aim of either asking directly for WTP (what are you willing to pay?), or asking whether respondents are willing to pay a particular price (are you willing to pay X? – yes/no), or asking for individuals’ rankings of alternative options where there is a link to the cost of providing the option. In the last case, WTP is inferred rather than stated directly. These contingent valuation and contingent ranking approaches have become very powerful in recent years and are widely used in environmental economics and health economics. This is the first time they have been applied to charitable services. Because the detail of the valuation study is extensive, we have presented a summary of the results of the study in Chapter 5. Chapters 1–4 explain the analytical foundations for the study and the questionnaire results. Three sources of economic value are identified: first, the WTP of the general public to ensure that charitable services continue to be provided; second, the WTP of the beneficiaries of the services to maintain those services; third, the benefits to volunteers from the opportunities provided by charities. A questionnaire-based approach is used to establish the WTP of the general public. But we treat the beneficiaries’ WTP separately. Because there are so many different groups of beneficiaries, we cannot possibly survey them
xiv
The price of virtue
all. We deliberately chose a group of users where we would anticipate serious potential problems of using our valuation techniques: the homeless. Our reasoning was that if our approaches worked in this complex and sensitive case, they would probably work in many others. We make no apology for treating charitable services like other economic commodities, but we fully appreciate that some people will find the approach questionable. For example, the very reason we have a voluntary sector appears to be precisely because the market system does not provide those services. Why does the market system fail? It fails because many of the goods and services do not have apparent markets – recall the air pollution example. It also fails because market systems operate through prices and those prices may well exceed the ability to pay of the vulnerable groups who are the very targets of many charities. But great care needs to be exercised when using these failures of the marketplace to criticize the economic valuation approach. First, we have already seen that there are often markets in the non-market service or benefit: the example of the housing market for air pollution can be extended. Historic houses and archaeological sites have implicit market value because people spend money travelling to see them. Their costs of travel provide a clue to finding their ‘price’. Second, even if markets are absent, it does not follow that people would not be willing to pay if there were a market. If we can find that WTP, then it may be possible to ‘capture’ at least part of it by establishing an institution that translates the WTP into actual payments. Charges for entry to parks, conservation areas, cathedrals and so on are all examples of ‘capture’. Third, charitable services are rarely provided by charities alone. Care of the elderly is provided by charities, by local government and by the private sector. There are potentially comparable markets out there. But this rationale has to be treated with caution. The users of charitable services are sensitive to who provides them, and there are distinct preferences for provision by one agent rather than another. Additionally, while the service may appear superficially the same, there are often marked qualitative differences. None the less, there is substitutability across a range of charitable services. Fourth, as Chapter 5 concludes, charities compete for scarce resources. Money received from donations could have been used elsewhere. Money received from government grants could have been used to provide other public services. It is important therefore to ensure that charities are the best way of providing those services. But we cannot know that unless we know what the ratio of their output is for every pound or dollar they receive. It may be more efficient to provide a service through local government rather than via a charity, or vice versa. Cost-effectiveness indicators exist, but they have limited validity once it is recognized that outputs vary in the range of benefits
Preface
xv
they provide and in the quality of service provided. We will not know more about comparative efficiency until we have some broader calculus of costeffectiveness which, at the moment, does not exist on any widespread scale. We hope our work begins the process of achieving that. Finally, the techniques we have used are familiar in the world of environmental economics. When environmental economics began there were concerns that the environment was being turned into ‘commodity’, and that this was illicit because the environment is ‘beyond price’, and somehow not to be brought within the measuring rod of money. Some critics still argue that. But the simple logic of opportunity cost – that whatever we do uses resources that could have produced some benefit elsewhere – remains. Anyone who argues that costs and benefits are irrelevant to social decision-making has to explain how cost and benefit can be ignored. All kinds of moral arguments can be invoked for doing so, but the fact that charitable services use resources to provide those services remains. This always means that those resources could have been used to provide some other service, which may just as easily serve some moral purpose. Morality is not irrelevant, but it must account adequately for opportunity cost. Part I of the book is concerned with demonstrating the economic value of charities. Part II focuses on the next logical question: if people are willing to pay more than they actually pay for charities, how can this extra WTP, this surplus, be captured and turned into flows of income for charities? Chapter 6 addresses the question of the role that government fiscal policy can play in stimulating the flow of resources to the charitable sector. It has long been argued that donations can be far more effective if they are ‘taxefficient’, that is, more donations are generated if tax allowances on giving are provided. While it may seem obvious that giving tax incentives should increase donations, this is not borne out in practice. A higher rate of marginal tax in a context where there are tax allowances should increase giving because more tax is written off for each pound or dollar given. But higher tax rates also mean less after-tax income, so there is an income effect, which depresses giving. The two forces work in opposite directions. The empirical evidence reported in Chapter 6 indicates that the net effect of tax incentives on giving is positive for the UK: giving is higher in the presence of tax incentives than it would otherwise be. However, tax incentives do not provide a very large stimulus to giving. Indeed, the extra donations stimulated by fiscal incentives are not large enough to offset the associated loss of tax revenues to the Treasury. A possible reason for the limited impact of tax incentives is the relatively restrictive scope of mechanisms for tax-efficient giving in the UK. In practice it is the efforts of fundraisers, rather than any tax incentives, which provide the greatest stimulus for philanthropic giving. Chapter 7 ex-
xvi
The price of virtue
amines the relative efficacy of different fundraising methods, in particular those which involve direct face-to-face contact between fundraiser and potential donor versus those which rely on remote methods such as telephone, television, mailshots and so forth. The evidence shows that people are significantly more likely to give when they are approached face to face. However, the average size of donations received tends to be larger via remote fundraising methods. Overall, face-to-face methods present charities with the highest return to each fundraising approach. However, people’s generosity diminishes markedly as they are approached an increasing number of times. Chapter 8 considers the extent to which charities should target the population from which donations are sought. It is well known that many charities do this. But what exactly is the return to the charities from this targeting? Chapter 8 finds that targeting does indeed elicit more donations than ‘scattergun’ approaches, although some targeting variables are shown to be much more effective than others. But it also shows that larger charities have less to gain from targeting than smaller charities. Since targeting involves reducing the size of the population that is ‘trawled’ for donations, the effect of targeting for large charities is to lose out from the reduction of scale that comes with targeting and to gain from the targeted population. There is a trade-off. For small charities, the scale effect is less important than the targeting effect: they stand to gain most from targeting. Overall, Chapter 8 offers the elements of a theory of optimal fundraising. Part III raises an issue for discussion: can our measure of the economic value of charities be used as a measure of social capital? There is now a substantial interest in the notion of ‘social capital’ – the ‘glue’ that holds society together and without which there is mistrust and social enmity that interferes with the smooth workings of the economy. Indicators of social capital are scarce. There are numerous negative indicators – measures of inequality, corruption and social breakdown such as crime, divorce and family disruption – but there are few positive indicators based on broad questionnaires about feelings of trust, or on measures of civil and political liberties. Yet the most obvious indicator of the extent to which human beings care for each other, the relationship of giving, has been ignored in the socialcapital literature. We offer the observation that, in measuring the ‘output’ of the charitable sector, we are contributing to that measure of social capital. Chapter 9 raises the possibility of using the value of the charitable sector as at least a component part of the definition of social capital, a departure from the existing, and rapidly growing, literature on the subject. Chapter 10 completes the book with some summary conclusions on the size of the charitable sector and on the mechanisms for capturing the surplus value we claim we have identified.
PART I
Measuring the Economic Value of the Charitable Sector
1. Conceptual foundations 1.1
PURPOSE OF THE BOOK
The charitable sector uses inputs such as land, capital and paid and unpaid labour to produce outputs such as care of the elderly, shelter for the homeless, environmental amenity and so on. In this respect the sector is just like any other productive activity in that it transforms resources into something that society values. However, unlike many other productive sectors, the goods and services produced by charities are not openly traded on the market. When goods and services are bought and sold on the market, the price at which they are traded provides a real indication of the value consumers place on the corresponding outputs. In the case of charities, services are usually provided at zero, or highly subsidized, cost to specific beneficiary groups who typically lack the means to secure the services in any other way. The price therefore does not provide any indication of the value which beneficiaries place on the service provided, making it difficult to put a monetary figure on the output of the charitable sector. The absence of any measure for the value of the services provided by charities is problematic for a number of reasons. First, as already noted above, if output could be measured, it would enable an economic calculation of the value added of charitable organizations, thereby providing a more accurate economic measure of their size. This would permit them to be integrated into the national accounts on an equal and consistent basis with the private sector. Second, there is a great deal of interest in the concept of social capital, the ‘glue’ that holds society together – such as family and other bonds, the assumption of trust in contractual relationships and socially responsible behaviour. From the economist’s standpoint, one reason for this interest is that societies where there are strong social bonds – good endowments of social capital – may develop faster than those where there is less social capital (Fukuyama, 1995; Putman, 1993). However, measuring social capital is problematic. Some insight into an indicator of social capital may be obtained by considering the size of the charitable sector relative to the market sectors of an economy. Participation in volunteer work is a sign of social concern. It 3
4
Measuring the economic value of charities
should be noted, however, that there is a risk that it might equally well be argued that the bigger the charitable sector, the more evidence there is of breakdown of social capital. This is because charitable services are often remedial, for example drug rehabilitation or working with young offenders. Third, without a measure of output, there is no satisfactory measure of the efficiency of the sector. Cost-effectiveness measures such as cost per unit of service provided might be appropriate, but even this requires some idea of output such as lives saved, illness avoided and so on. Cost-effectiveness measures are valuable, but they only permit answers to a limited set of questions. Broader measures of efficiency, or real output (in terms of value added) divided by input, should throw some light on whether voluntary agencies are more or less efficient than private or government agents at providing social services. In the longer term, it might enable some comparison of the relative efficiency of the charities themselves. Fourth, output measures require a focus on the users of charitable services, the customers, whether as visitors to a National Trust property or as homeless people. Adopting non-market valuation procedures serves as a reminder that the important social groups influencing the structure and size of the voluntary sector are the users themselves, whether the general public, the target group, or both. Finally, and of immediate policy relevance, a true analysis of charitable output serves to highlight the role that government funding should play in the financing of charities, an issue that will be taken up further in Chapter 5. The purpose of this book is to develop and illustrate a methodology for valuing the output of the charitable sector. This introductory chapter lays the conceptual foundations for this valuation exercise. The chapter begins by reviewing existing approaches to measuring the scale of charitable activity, and explains why these fall short of providing a value for the outputs of the sector. Thereafter, the methodological foundations for a new approach to measuring the value of the output of the charitable sector are derived from standard welfare economics. The discussion goes on to explain how this approach can be implemented by adapting a number of valuation techniques, originally developed in the environmental economics field, to the charitable context. These methodologies permit the estimation of monetary values for goods and services which are not traded in the marketplace. Finally, the remainder of Part I of this book illustrates a complete application of the proposed methodology.
Conceptual foundations
1.2
5
EXISTING INDICATORS OF THE SIZE OF THE CHARITABLE SECTOR
To motivate the need for a new methodology to measure the value of the output of the charitable sector, this section reviews a number of measures that are currently used to measure the significance of the charitable sector and explains why they cannot be regarded as reliable measures of output. 1.2.1 Income Measures The total operating income to a charity or group of charities from all sources can be summarized as private earned income plus private giving plus income from government (Kendall and Knapp, 1995). Kendall and Knapp (1995) find that the total operating income of the ‘broad voluntary sector’ in the UK is £29.5 billion and of the ‘narrow voluntary sector’ £12.3 billion. The broad voluntary sector is defined by these authors as comprising all organizations which meet the core definition criteria, which are that they are (1) formal, (2) independent of government and self-governing, (3) not profit-distributing and primarily non-business, and (4) voluntary. The narrow definition is a subset of the above organizations, excluding those which are regarded as too exclusive, not independent enough, insufficiently altruistic or not particularly oriented to providing public benefits. Amongst the organizations excluded from this narrow definition are those providing recreation, for example sports clubs, primary, secondary and higher education, trade unions and business associations. Total operating incomes of the top 500 fundraising charities are also published annually by the Charities Aid Foundation (the latest at the time of writing is CAF, 1999). It is important to recognize that, though informative, such statistics do not give us any clear indication of the value of charities to society. There is no certainty that a charity with a higher operating income yields greater social benefits, since this will depend on how effectively the income is used in the provision of services. Indeed, such statistics merely reflect the ability of charities to raise funds from all sources, but not the social value of the use of such funds. Although fundraising is a very important aspect of charities’ activities, the interest here is the ability of charities to add to the value of their income by increasing social welfare. None the less, it could be argued that income provides a reasonable proxy for the social value of a charity, inasmuch as the income of charities derives from voluntary contributions which in and of themselves reflect the social value which the different donor groups place on the services provided by charities. However, there are a number of problems with this argument. A more detailed consideration of the economic context of charitable contributions from each of
6
Measuring the economic value of charities
the different sources, undertaken below, illustrates that the link between income and social value is more tenuous than may at first sight appear. Private earned income Kendall and Knapp (1995) show that private earned (commercial) income can be separated into four main subcategories. These are user fees, sale of goods, membership dues and investment income. Since membership dues are in many cases a form of donation, and investment income is interest on past income, attention here focuses on user fees and sale of goods. As noted above, charities rarely charge users full cost rates for the services they provide, since they are usually targeting disadvantaged groups who for various reasons may be unable to pay full market rates for the services provided. Consequently, user fees are likely to underestimate the value of the services provided to beneficiaries. The issue of the income from sales is even more problematic. When buying a good from a charity shop or a ticket to a charity fundraising event, for example, the consumer will be valuing both the good and the output of the charity in taking the decision to buy. In this sense, the revenue from the sale overestimates the value which the purchaser places on the services provided by the charity. A further complication arises because, in some cases, the goods and services sold by charities for fundraising purposes represent donations on the part of third parties. For example, consider a situation in which a charity shop sells a donated second-hand coat for a price of £10. In this case, the coat represents a donation in kind reflecting the value that its original owner places on the activities of the charity. However, the value of the second-hand coat to the original owner may be very different from the price at which the charity succeeds in selling on the item to the eventual purchaser. Hence the sale value that appears in the charity accounts is not a reliable indication of the value that the original donor places on the services provided by charities. For all of these reasons, it is very difficult to draw any conclusions about the social value of the output of charities from data on sales revenues. In the case of user fees, revenues may understate the value of the services received by the beneficiaries, while, in the case of sales of goods, revenues are likely to overestimate the value that the purchasers of those goods place on the services provided by charities. Private giving Kendall and Knapp (1995) categorize private giving into three main forms: individuals, companies and grant-making trusts. It seems likely that such donations reflect to some degree donors’ perceptions of the social value of the services provided by charities. However, there are several reasons to believe that private giving may not be an accurate reflection of social value.
Conceptual foundations
7
First, there is a well-known tendency for people to free-ride on the generosity of others. That is to say that many people who value the output of charities may none the less fail to give because they believe that the services will be financed anyway by other donors even if they do not give themselves. Thus free-riders seek to benefit from the services provided by charities without incurring any of the costs. In some cases, free-riders may avoid giving altogether, while, in others, they may simply give less than the true value that they place on the service. Laboratory experiments suggest that free-riding may depress the voluntary contributions by as much as 50 per cent relative to their full potential value (Foster, 1999). Consequently, voluntary giving may seriously underestimate the full value that people place on charitable services. The issue of free-riding by donors will be addressed in some depth in Chapter 2. Second, a significant component of charitable giving is in kind, most notably by volunteers who give time rather than money, but also by private companies which may donate their services or private individuals who sometimes give second-hand goods. Estimates of charities’ income rarely attempt to place a value on donations in kind, largely because of the methodological difficulties of valuing such donations. This omission is particularly serious in the case of volunteering. For example, using average hourly wage rates to value volunteer time, the Volunteer Centre UK (1995) finds an economic value for volunteering of £41 billion. While Foster (1999) shows that, even taking the conservative approach of valuing volunteer time at the minimum wage (that is currently £3.50 per hour), the aggregate value of time volunteered each year in the UK (at £3.5bn for 1992) exceeds the aggregate value of private monetary donations (at £2.9bn for 1992). A detailed discussion of alternative methods for valuing volunteer time is provided in Chapter 4. Third, charitable giving entails certain transaction costs which may dissuade donors, particularly where relatively small gifts are concerned. Examples of transaction costs include identifying suitable recipient charities, making telephone calls, setting up covenants, writing cheques and so forth. The problem of transaction costs, together with the strategies that charities may use to overcome them, will be further explored in Chapter 7. Fourth, charitable giving may not always be a reflection of the value that donors place on the activities of the charity. It may instead be dictated by social pressures or norms. For example, Salamon et al. (1995) identify the problem of philanthropic paternalism, whereby the priorities of the voluntary sector come ‘to be shaped by the preferences not of the community as a whole but of its wealthy members’ due to their greater ability to donate. The question of motivations for philanthropic giving will be investigated more fully in Chapter 2. For all of these reasons, private giving is unlikely to provide an accurate estimate of the full value that society places on the services provided by
8
Measuring the economic value of charities
charities. The problems of free-riding, donations in kind and transaction costs suggest that donations will underestimate the value of charitable services, while the problem of motivations for giving suggests that donations could be an overestimate. Income from government In general, voluntary contributions will fail to provide the socially optimal level of funds due to the free-rider problem. Therefore, it could be argued that the addition of government funds could theoretically be regarded as an attempt to redress the shortfall of voluntary donations to charitable organizations relative to their true social value. In practice, however, government expenditure decisions are rarely made on the basis of these sorts of criteria. Government assistance is most likely to go to those causes that have majority support due to the constraints of the political system. This would not ensure that funds are allocated to those organizations yielding greatest social value, as different individuals may place very different values on different outputs. 1.2.2 Measures of Expenditure In addition to income information, the Charities Aid Foundation also publishes figures giving the total expenditure of the top 500 charities and how this is divided amongst charitable expenditure, fundraising, administration and general expenditure. Charitable expenditure means expenditure directly on the provision of charities’ services, or in other words the cost of producing these services. It is thus a measure of the part of income which is spent directly on providing benefits to the charity’s target group. As in the case of income, expenditure measures are unlikely to provide an accurate reflection of the value of charitable output. First, there is no reason to expect that the value of charities’ output should be commensurate with the costs of production. Indeed, to the extent that charities add value to the resources they use, the value of outputs should exceed the cost of the associated inputs, although the value of output could conceivably be less than expenditure if resources are being used inefficiently. A charity that spends more than another does not necessarily produce more output; it may simply be less efficient. Second, expenditures exclude the contribution of volunteers who, by definition, are not paid for the services they provide. In this sense, published expenditure figures are not even an accurate reflection of the use of resources by the charitable sector. This omission creates a significant distortion in making comparisons between charities. Thus one charity could spend considerably less than another but produce much more output if it is relying to a greater extent on volunteers than on paid employees.
Conceptual foundations
1.3
9
MEASURING THE ECONOMIC VALUE OF THE CHARITABLE SECTOR
The previous section explained why traditional income and expenditure measures of the scale of charitable activity fail to provide a reliable measure of the value of the goods and services which charities produce. The purpose of this section is to develop an alternative approach which meets this objective. The proposed approach builds upon the foundations of standard welfare economics, and takes advantage of techniques developed in the field of environmental economics in order to permit the application of these concepts to the charitable sector. 1.3.1 Foundations in Welfare Economics Welfare economics is concerned with the level of well-being or welfare of society. Economic theory assumes that human well-being is determined by the fulfilment of people’s preferences. A benefit is defined as anything that increases human well-being and a cost as anything that reduces human wellbeing. The intensity of a person’s preferences can be measured by their maximum willingness to pay (WTP) for a benefit (or for the avoidance of a cost) or their minimum willingness to accept (WTA) compensation for tolerating a cost (or foregoing a benefit). These measures correspond to the value which a person places on changes in the quantity or quality of a good or service. This is because they reflect the extent to which that person is willing to sacrifice alternative uses of their limited income in order to secure that particular outcome. The well-being of society in turn is assumed to depend on the well-being or utility of the individuals comprising society. A social welfare function provides a way of aggregating the well-being of individuals to obtain the overall well-being of society. The choice of welfare function is a subjective one and will reflect ethical judgements about the relative importance of the well-being of different individuals. A generalized social welfare function can be stated as follows: n
W (U1 , U2 , …, Un ) = ∑ aiUi , 1
where U is the utility level of individual i from 1 to n and Σ indicates summation. The multipliers a1, …, an indicate the weight accorded to each individual’s well-being in the overall welfare of the society. These multipliers have sometimes been used to give more weight to the well-being of individuals with relatively low income, in order to address issues of equity in the
10
Measuring the economic value of charities
evaluation of social welfare. However, this practice has fallen into disuse. This is partly in response to the subjectivity entailed in specifying such distributional weights. Moreover, it also reflects the view that income distribution is best dealt with as a separate issue in its own right, and not brought into welfare analysis of other issues. Consequently, a popular form of the social welfare function is to set a1 = a2 = … = an = 1. The resulting Benthamite utility function is a simple sum of individuals’ utility level, and thus effectively treats the well-being of each person in society as equally important. This implies that society is indifferent or unprejudiced between similar increases in well-being for different individuals; that is, a unit of well-being is equally valued regardless of to whom it accrues. One of the main uses of social welfare functions is to measure changes in the welfare of society. The value of a welfare change, if we apply the Bethamite welfare function, will be equal to the sum of the changes in the utilities of individuals: Change in total social welfare = ∆W = Σ∆Ui = total social benefits – total social costs This change in total social welfare can also be referred to as the net social or economic value. One of the fundamental decision rules in economics is that activities should only be undertaken when their net social value is positive. Following the terminology developed above, the total social benefits of a charity’s output are the sum of all the benefits that accrue to members of society as a result of the existence of the charity, which they would not otherwise receive. The provision of charitable services also entails substantial costs. For example, the donations and grants that charities receive could alternatively be spent elsewhere in the economy to provide other social benefits. These foregone alternative benefits (or opportunity costs) represent the total social costs of providing charitable services. Therefore the net social value of charities is equal to the total social benefits minus the total social costs, or in other words the difference between the overall level of social welfare with and without the existence of the charity. To summarize: ● ●
●
Total social value (TSV) of charities = total social benefits that accrue to individuals from the existence of the charity. Total social cost (TSC) of charities = total foregone benefits (or opportunity costs) from alternative uses of resources devoted to the provision of charitable services. Net social value (NSV) of charities = total social benefits – total social costs of the charity.
Conceptual foundations
11
1.3.2 Total Social Value of Charities The analysis above indicates that to value the contribution that charities make to social welfare, all the benefits that accrue to different members of society as a result of the existence of the charity must be taken into account. This requires the identification of the various groups of society who benefit from the services provided by charities. In general terms, society can be divided into two groups of beneficiaries: the target group (or the direct users) and the rest of the society (or the indirect users). Consider first the benefits to the target group of a charity, that is to say those people who actually use the charity’s services. Direct recipients could include, for example, handicapped people, the homeless, cancer sufferers and so on. In many cases, the target group will be well defined and selected by the charity. However, in other cases, the direct users or beneficiaries may simply be those members of society who choose to take advantage of the services offered by the charity, for example visitors to a museum or National Trust site. Next there are the benefits to the rest of society, meaning those who are not direct users of the charities’ services. These indirect benefits take a variety of forms. First, there are altruistic benefits. People may derive benefits simply from observing or knowing that a charity’s services are provided to the target group, because they feel altruistically towards that group. It is important to note that altruistic preferences may sometimes be paternalistic, in the sense that they do not necessarily coincide with the perceived self-interest of the target group. For example, people may think that it is best for the homeless to stay in overnight shelters even when the homeless themselves may prefer to sleep rough on the streets. Second, there are external benefits. Some of the services provided by charities provide spin-off benefits to those who do not form part of the target group. For example, the existence of a charity providing youth clubs may have a positive effect on the local area in terms of reduced crime and more peaceful streets, over and above the direct benefits to the young people themselves. Local residents will consequently benefit from the existence of the charity, even if they do not feel altruistically disposed to the young people of the neighbourhood. Third, there are option benefits. Even though people may not be direct beneficiaries of charities’ services at the present time, it is possible that they might require such services in the future. For example, someone who is perfectly well at the present time might in the future become severely depressed and require the services of the Samaritans. A person in this situation may value the existence of the Samaritans as an insurance policy against their own future need.
12
Measuring the economic value of charities
Fourth, there are ‘warm-glow’ benefits. Such benefits are due to what Andreoni (1990) terms impure altruism, whereby donors gain satisfaction purely from the knowledge of their own act of generosity as donors over and above any altruistic sentiments they have towards their beneficiaries. The ‘warm glow’ may be a private feeling of moral well-being, or may reflect the social recognition or status attached to being a donor. Fifth, those who volunteer may obtain private benefits in the form of moral satisfaction, social contacts, work experience, acquisition of new skills or opening up of career opportunities. In fact, in some cases, volunteers may even be members of the charities’ target group and have therapeutic or rehabilitative reasons for volunteering.
CHARITIES
Indirect benefits
Warm glow, altruistic, option, external benefits
Donors
Direct benefits
Volunteer, altruistic, external benefits
Volunteers
External, altruistic benefits
Target groups
Rest of society Figure 1.1
Social benefits from charities
Figure 1.1 provides a graphical representation of this typology of the total social benefits from the charitable sector. In order to estimate the total social benefits of the charitable sector according to the framework developed above, it is necessary to aggregate the change in individuals’ well-being that results from the provision of charitable services compared with a situation where charities do not exist. Regarding the benefits to volunteers, these can be estimated in terms of the opportunity cost of the time that the volunteer gives up to the charity. The
Conceptual foundations
13
reason is that volunteers could be expected to give time up until the point where the marginal benefit of an hour of volunteering is equal to the marginal benefit spent in the volunteer’s next most preferred activity. This argument and the procedure for valuing volunteer benefits on this basis is further developed in Chapter 4. In the case of the target group and the wider beneficiaries in the rest of society, a lower-bound estimate of welfare already exists in the form of the actual fees paid for charitable services as well as the donations voluntarily contributed. In these cases, the problem lies in estimating the additional welfare that is not captured through these actual payments. There are two possible methods for measuring this additional well-being. The first is to consider the maximum amount of money that the individual is willing to pay to avoid the loss of the charities’ services (their WTP). The second is to estimate the minimum amount of money that the individual is willing to accept in compensation for the loss of the charities’ services (their WTA). In both cases, the change in the person’s well-being as a result of the existence of charities is being valued in terms of the equivalent amount of money that would need to be paid by (WTP) or given to (WTA) the individual in order to bring about a similar change in their well-being. In the context of charitable services, WTP will not always be a very appropriate measure of welfare. The reason is that WTP clearly depends upon ability to pay, and charitable services are often provided precisely because the target groups are not able to pay for the basic services that they require. This is particularly true of charities operating in the social services sector. Thus, in such circumstances, the WTP of the target groups will be negligible even though the benefits they obtain from the service may be very large. Consequently, it can be argued that, for the target group of many charities, the correct measure of value is not WTP for the provision of the service, but the WTA compensation for the loss of the service. Freeman (1993) discusses the choice of the correct welfare measure and shows that it depends on whether there is an implied property right in the status quo. He argues that WTA compensation is an appropriate welfare measure in circumstances where the beneficiaries are perceived to have a right to the services they receive. This would appear to be applicable in the case of charities, in that they provide basic services to which the beneficiaries could be argued to have a basic right. As far as indirect benefits to the rest of society are concerned, the status quo is that many people do pay for the provision of charitable services, whether directly through donations or indirectly through government grants. The implication is that WTP is the appropriate welfare measure for these indirect benefits.
14
Measuring the economic value of charities
WTP is nothing more than the demand for a good or service. When a good or service is provided in a market context, it is relatively easy to estimate this demand curve and thus the total WTP for the associated good or service, which is equivalent to the area underneath the demand curve. Even when this does not prove to be possible, the market price at least gives a lower bound for each individual purchaser’s WTP, as it is clear that if an individual’s WTP is lower than the market price, they will choose not to purchase. In the case of charitable services it is not so straightforward to estimate WTP, since these services do not tend to be traded on the market, but rather provided at zero or highly subsidized cost. Furthermore, there is often also a public-good component in charities’ services, in that some members of society receive benefits from the provision of such services, whether or not they contribute to their provision. However, these difficulties are already well known to economists working in environmental and health economics, and a variety of techniques has been developed for estimating WTP in the absence of direct market prices. The principal methods that may be used to value the services provided by charities in the absence of markets are briefly described below. They can be divided into two broad categories: the indirect approaches and the direct approaches. The first category of ‘indirect approaches’ aims to find some indirect link between the services provided by charities and services provided in the marketplace. Where such a link exists, the market price can be used to approximate the value of the charitable activity. These methods include the private sector equivalent, the opportunity cost approach and the production function approach. Private sector equivalent In many cases charities provide services that are also provided in the private sector. Examples include care for the elderly, hostels for the homeless, and education and health services. In these contexts, the rationale for charitable provision is that the charity aims to provide coverage to a target group that the private sector would not find attractive to serve. Where this situation arises it may be possible to use the prices of the equivalent services charged in the private sector as a means of valuing the services provided by the charity. An important assumption underlying this approach is that there are no differences in the quality of services provided by charities and by the private sector. However, this assumption may not always be appropriate, thereby invalidating the use of this method.
Conceptual foundations
15
Opportunity cost approach In a more limited number of cases, it may be possible to determine the value of a charity’s services by estimating the opportunity cost of not providing those services. For example, some day-care services enable the carers to go to work. Thus these services could be valued in terms of the extra earnings that the carers are able to obtain by going to work. However, clearly this valuation would not take into account the emotional relief that a carer may experience from being able to make use of a day-care centre. Production function approach Charitable services often boil down to the production of certain outputs whose economic values are already known, either because they are traded in the marketplace or because they have already been estimated in other contexts. For example, some charities’ principal output may be to save lives or to extend life expectancy. There is already an extensive literature in the health economics field on the economic value of additional life-years. If the relationship between the charities’ activities and the life expectancy outcomes could be accurately estimated, it would then be possible to apply the corresponding economic values to obtain a monetary measure of benefits. Where it is not possible to make any link between the services provided by charities and other services provided in the marketplace, a second category of ‘direct approaches’ may be used. These techniques, also known as stated preference techniques – which include contingent valuation, choice modelling techniques and related variants – involve using a questionnaire within which respondents are asked directly to reveal their WTP or WTA. Contingent valuation method The contingent valuation method (CV) is a survey-based technique. By means of an appropriately designed questionnaire, a hypothetical market is described where the good in question can be traded (Mitchell and Carson, 1989). This contingent market defines the good itself, the institutional context in which it would be provided and the way it would be financed. A random sample of people are then asked directly to express their maximum WTP or minimum WTA for a hypothetical change in the level of the provision of the good. A critical assumption underlying this method is that respondents’ behaviour in the hypothetical market will be no different than it would have been in an equivalent real situation. The empirical evidence on this point is mixed, with some studies finding that hypothetical WTP exceeds real WTP (see Foster et al., 1997). Furthermore, a number of factors may systematically bias respondents’ answers. These factors are not specific to CV but are common to most surveybased techniques and are mostly attributable to survey design and implementation problems. Mitchell and Carson (1989) provide an extensive review. It is not
16
Measuring the economic value of charities
straightforward to assess the validity of the estimates produced by CV studies for the obvious reason that actual WTP is unobserved. Nevertheless, certain aspects of validity can be tested by indirect means (see EFTEC, 2001). Choice modelling techniques Choice modelling, on the other hand, applies to a family of survey-based methods that model preferences for bundles of characteristics of goods and isolate the value of individual product characteristics typically supplied in combination with one another. There are many variations of choice modelling; however, they all share the same basic characteristics. Respondents are given a choice set comprising two or more alternatives. For example, the choice set may comprise three different types of hospice services. Each alternative has a certain number of characteristics, but differs in terms of the level of those characteristics. For example, the characteristics might be the cost of using the hospice services, the quality of the rooms provided and the availability of support services. Some of the hospices in the choice set may be more expensive than others but offer higher-quality rooms and/or a wider range of services. Respondents are asked to express their preferences over the items contained in the choice set. This might be by choosing their most preferred alternative (choice experiments), ranking the alternatives in order of preference (contingent ranking), or simply giving each alternative a score on some predetermined scale (contingent rating). With the aid of statistical techniques, it is possible to infer WTP or WTA for the characteristics presented from the choices reported in the survey. 1.3.3 Total Social Cost of Charities Total social costs are the opportunity costs of the services provided by the charity, that is the value of the total social benefits foregone by allocating resources to the charitable sector as opposed to their best possible alternative uses. As with the total social benefits, the total social costs can be broken down into a number of distinct categories. First, if the target group has to pay a price for each service that the charity provides, then the corresponding opportunity cost is the total value of the (maximum) benefits that they would gain from using the money (and possibly the time) spent on the charity’s services on alternative activities, in the absence of the charity. This can be captured by the value of the fee revenue paid. Second, the opportunity cost to donors of the charity’s provision of services is the total value of the (maximum) benefits that they could gain from spending their donation on alternative goods and services, in the absence of the charity. This can be measured in terms of the total financial value of donations.
Conceptual foundations
17
Third, in the case of volunteers, one way of valuing the resources absorbed by the use of volunteer time is to consider the costs of replacing volunteers with paid employees on the open market. This approach is explained in greater depth in Chapter 4. Fourth, if the government also contributes grant finance to the charity, then there is an additional opportunity cost of the charity’s provision of services. This is equal to the (maximum) value of social benefits that could be gained from reallocating the grant to some alternative area of public expenditure, in the absence of the charity. The opportunity cost of government grants is not straightforward to measure either, since it would depend on the alternative uses to which such funds were put. Thus, in common with standard practice in cost–benefit analysis, the financial value of the grant is used as an approximation to the opportunity cost. Figure 1.2 provides a graphic representation of this typology of the total social cost of the charitable sector. Rest of society Donors
Volunteers
Government
Target groups
Donations
Time, skills
Grants
Fees, price
CHARITIES Figure 1.2
Social costs of charities
1.3.4 Net Social Value of Charities In conclusion, this introductory chapter has argued that traditional measures of the size of the charitable sector – such as the income and expenditure of charities – do not provide any meaningful indication of the value of the output of charities. In order to solve this problem it is necessary to measure the net social value of charitable activities, that is the difference between the social benefits generated by charities and the opportunity cost of the resources that they absorb.
18
Measuring the economic value of charities
The services provided by charities are beneficial not only to the immediate target group; they also generate a range of indirect benefits for society as a whole. These benefits are not straightforward to value because they are not traded in the marketplace like conventional goods. None the less, using a range of techniques already tried and tested in the environmental and health economics literatures, it is possible to put a monetary value on these benefits. Resources used by charities have an opportunity cost in that they can no longer be put to work in alternative uses. These resources include fees paid by direct users, gifts made by donors, time given by volunteers and grants from the public sector. By and large, these social costs are more straightforward to measure than the social benefits. The net social value of the charitable sector is nothing but the difference between these benefits and costs. The remaining chapters of Part I undertake the steps outlined here in greater detail for one particular example, that of charities working in the field of housing and homelessness. Chapter 2 applies contingent valuation and choice modelling techniques to the problem of measuring indirect benefits to society at large of charities in housing and other sectors. Chapter 3 uses contingent valuation techniques to value the benefits to users of hostels for the homeless in London, using a welfare measure based on willingness to accept compensation. Chapter 4 values the opportunity cost of volunteer time. All these elements are brought together in Chapter 5, which performs a calculation of net social value.
2. The benefits of charities to the general public 2.1
INTRODUCTION
A significant number of attempts have been made to estimate the size of the voluntary sector, as measured by employment, income and expenditure. However, these studies provide only incomplete measures of the value of the sector as they focus on the costs rather than the benefits, that is, the value added of the sector. The experiment reported in this chapter uses data from a study that, for the first time, applied stated preference techniques to the valuation of the output of the charitable sector in the UK, with special reference to the housing and homelessness charities. Broadly speaking, charities can be seen as providing use value to a relatively small group of beneficiaries and non-use, option and indirect use values to society at large. This chapter deals with the estimation of the latter type of values from the perspective of the general public. The purpose of investigating the value of charities to the general public in the UK is twofold. The first objective is to measure the benefits which the charitable sector provides to society at large, over and above the benefits received directly by the target groups. As mentioned in Chapter 1, these benefits could potentially be motivated by a number of considerations. People may benefit indirectly from the charities’ activities, for example by a reduction in the number of rough sleepers they encounter on city streets. Or they may feel that the charity provides a safety net, which could be of use to them at some future date, for example research into a cure for cancer. Alternatively, for moral reasons, people may simply value the fact that the basic needs of others are taken care of by charitable organizations. The second objective is to measure the benefits to society at large which are specifically associated with the activities undertaken by housing and homelessness charities. A survey of the direct beneficiaries of these charities has also been undertaken in a separate investigation (see Chapter 3) in order to uncover the direct use values of these philanthropic activities. The summation of the two values obtained from each of the surveys permits an estimation of the overall social value of such charities (see Chapter 5). 19
20
Measuring the economic value of charities
To carry out these objectives, two stated preference techniques were used in parallel: contingent valuation (Mitchell and Carson, 1989) and contingent ranking (Beggs et al., 1981). These techniques have been widely used to value public goods and services in the areas of environment, culture and health. However, they have rarely, if ever, been applied to the valuation of charitable sector services. In order to apply the stated preference approach to the problem of estimating the social value of the charitable sector, it is necessary to construct a scenario which will enable people to think about how much the sector is worth to them. In the context of the present study, this was done by describing a hypothetical situation where one or more sectors of charitable activity were threatened with shutdown due to a funding crisis. The respondents were then told that the government could prevent the shutdown by giving the charities an emergency grant, but that this could only be achieved at the expense of raising everybody’s tax bill. Following on from this, various means were used to establish how much (if any) additional tax the respondent was willing to pay to fund the emergency grant. It was stressed that this payment would be over and above any existing donations that the respondent was already making. In theory, a scenario of this kind should be able to elicit from respondents the desired value for the charitable sector. By using taxation as the means of payment and presenting respondents with the prospect of losing the charitable sector altogether, the scenario should avoid the free-rider problems associated with voluntary donations. By asking for willingness to pay over and above existing contributions, the survey should induce respondents to reveal the extent to which their current contributions understate their true valuation of the sector. This chapter summarizes the design characteristics of the general public survey and presents the results.
2.2
SURVEY DESIGN
For the purposes of the survey, the UK charitable sector was classified into four subsectors: (1) housing and homelessness charities, which provide emergency short-term accommodation, counselling and support services for homeless people in hostels and night shelters, for example Shelter, Crisis, Centrepoint and the Salvation Army; (2) social services charities, which work to improve the lives of particularly needy groups of people such as the elderly, the physically and mentally handicapped, disabled people such as the blind and the deaf, and children from troubled backgrounds, for example Age Concern, Dr Barnardo’s and the NSPCC; (3) health and medical research
Benefits of charities to the public
21
charities, which fund scientific research into presently incurable diseases as well as providing hospice care for the terminally ill, for example the Imperial Cancer Research Fund, the Multiple Sclerosis Society and the British Heart Foundation; and (4) a residual category comprising charities active in the areas of culture, environment and overseas aid, for example Oxfam, Save the Children Fund, Friends of the Earth, the RSPCA and the National Trust. The overall objectives of the general public survey necessarily required the separate identification of housing and homelessness charities. The social services and health categories were chosen because they represent a large proportion of overall charitable activity. The final category was necessarily something of a catch-all. However, overseas aid, environment and culture were explicitly identified as the major charitable activities in this subcategory. Educational and religious charities, understood to refer to those that exist to benefit schools and churches, as well as political parties were explicitly excluded from the survey. The survey of the general public consisted of a combined contingent valuation (CV) and contingent ranking (CR) split-sample experiment. In particular, three different versions of the survey were administered to different subsamples of the population: 1.
2.
3.
Contingent valuation of housing and homelessness charities (CVHH) (sample size: 282): this version elicited the value of housing and homelessness charities only using a CV approach; Contingent valuation of all four charitable sectors (CVALL) (sample size: 279): this version elicited the value of all the four sectors of charitable activity described above using a CV approach; and Contingent ranking (CR) (sample size: 290): this version permits an elicitation of the value of the whole charitable sector and of each of the four subsectors separately by means of a CR approach.
In a preliminary section, common to all versions, respondents were introduced to each of the four charitable sectors described above and asked to think about their relative importance. Information was then collected on each respondent’s current donations of time and money to each of these areas of philanthropic activity. Owing to the free-rider problem, the magnitude of existing contributions can only be regarded as a lower bound on people’s valuation of charities. As mentioned before, the purpose of the survey was to establish how much more people would actually be willing to pay over and above their existing donations. This was determined by creating a hypothetical scenario in which all or some of the charities faced the prospect of shutting down as a result of a shortfall in funding. Respondents’ WTP to prevent this eventuality was then
22
Measuring the economic value of charities
elicited using increases in taxation as a payment vehicle. In this context, the WTP measure is an equivalent variation welfare measure, that is, the maximum amount respondents are willing to pay to avoid the closure of one or more sectors of charitable activity. An example of the hypothetical scenario used in the CVHH version is presented below. There were only very minor adjustments to the wording across the different versions of the questionnaire. Please imagine that, due to a financial crisis, all the charities in the country dealing with housing and homelessness were facing the prospect of shutting down for a whole year. (Don’t worry, this is definitely not going to happen! But, even so, we would like you to think about how you would feel about this if it were to happen.) Clearly, if these charities were to shut down, many of the people whom they currently help would be left with no alternative but to sleep rough on the streets. Now suppose that the government was considering making an emergency grant to these charities, so as to prevent them from having to shut down. The only way in which the government could fund this emergency grant would be by raising the taxes we all pay – for example, taxes on income and the sale of goods.
Following the description of the hypothetical scenario, part of the sample received a dichotomous choice CV treatment, while another part received a CR treatment. The next sections present a detailed description of the valuation questions presented in both the CV (CVALL and CVHH) and CR surveys. Technical details of the theoretical and statistical procedures involved in the estimation of WTP measures from these elicitation methods can be found in the Statistical Appendix to this chapter. 2.2.1 Contingent Valuation The CVALL subsample received a double-bounded dichotomous choice CV question (Hanemann et al., 1991) which dealt with the potential closure of all charities in all four of the sectors identified; while the CVHH subsample received a double-bounded dichotomous choice CV question which dealt with the potential closure of all charities operating in the housing and homelessness sector only. In particular, respondents were asked whether or not they would be willing to pay a specific tax amount £X, to which they might answer ‘yes’ or ‘no’. The tax amount £X was varied across respondents. In addition, they were asked a follow-up payment question that depended on the response to the first tax level: if the respondent accepted the initial bid, they were asked to pay a higher bid; if the answer to the first tax level was ‘no’ then the respondent was presented with a lower amount. This procedure is known in the literature as a ‘double-bounded dichotomous choice elicitation’. As an illustration be-
Benefits of charities to the public
23
low is the exact wording of the initial valuation question used in the CVHH version. Suppose that the funding of the emergency grant would cost everyone, including you personally, £X each month throughout the coming year – which would add up to £12X over the year. This would be over and above any contributions of time and money you already make towards charities dealing with housing and homelessness. Would you be prepared to pay this extra amount to prevent these charities from shutting down? Please think carefully about how much you can really afford and where the additional money would come from and try to be as realistic as possible.
The actual amounts of money used in the question varied from £0.20 to £15.00 per month in CVHH and from £0.20 to £25.00 per month in CVALL. Respondents were also reminded of the yearly implications of a monthly payment. Table 2.1 shows the tax vectors that were used in the dichotomous choice approach. The bid levels were chosen based on the results of the pilot open-ended valuation questions. Table 2.1
Bid vectors for CVALL and CVHH (£ per month and £ per year) All charities (CVALL)
Housing and homelessness charities (CVHH)
Bid levels Low M A I II III IV V
Initial M A
0.2 2.4 0.5 6 0.5 6 1 12 1 12 3 36 3 36 6 72 6 72 12 144
Notes:
Bid levels High M A
N
Low M A
1 3 6 12 25
54 58 56 57 53
0.2 2.4 0.5 6 1 12 0.5 6 1 12 2.5 30 1 12 2.5 30 5 60 2.5 30 5 60 10 120 5 60 10 120 15 180
12 36 72 144 300
Initial M A
High M A
N
56 56 54 58 58
M: monthly; A: annually; N: number of respondents; I–V: bid vectors.
Following the recommendations of the NOAA Panel (Arrow et al., 1993), respondents were also offered a ‘don’t know’ option after each valuation question. This implies that there were three possible answers to each bid level presented: ‘yes’, ‘no’ and ‘don’t know’. Wang (1997) has recently shown how such answers can be incorporated into the choice model for the case of single-bounded dichotomous choice. In this chapter, the two extreme cases are reported: the case where ‘don’t knows’ are treated as a rejection of the bid level (providing a lower bound on true WTP) and the case where ‘don’t
24
Measuring the economic value of charities
knows’ are treated as an acceptance of the bid level (providing an upper bound on true WTP). As an internal test of consistency, after the dichotomous choice questions, respondents were asked another WTP question, this time under the openTable 2.2
£/month nothing 10p 25p 50p 75p £1.00 £1.50 £2.00 £2.50 £3.00 £3.50 £4.00 £4.50 £5.00 £6.00 £7.00 £8.00 £9.00 £10.00 £12.50 £15.00 £17.50 £20 £25 £30 £40 £45 £50 £75 £100 Over £100
Payment ladder used in the general public survey (CV version) and choices of a hypothetical respondent (£/year equivalent)
✔ or ✘
(nothing) (£1.20) (£3) (£6) (£9) (£12) (£18) (£24) (£30) (£36) (£42) (£48) (£54) (£60) (£72) (£84) (£96) (£108) (£120) (£150) (£180) (£210) (£240) (£300) (£360) (£480) (£540) (£600) (£900) (£1200) (over £1200)
✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔ ✔
✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘ ✘
Benefits of charities to the public
25
ended format. Open-ended questions can be unaided (‘what is your maximum WTP?’) or aided, whereby respondents are presented with a card containing several money amounts (a ‘payment card’) and are asked to choose the figure that best approximates their maximum WTP. Given that the latter approach is thought to simplify the valuation task for the respondent, a version of the payment card was adopted. Specifically, respondents were presented with a ladder of values identical to that shown in Table 2.2. The ladder was completed by asking respondents to begin with the lowest values and put a tick against those amounts that they were ‘almost certain that they would be willing to pay’. Subsequently respondents were asked to turn to the highest values and put a cross against those amounts that they were ‘almost certain that they would not be willing to pay’. Thus the example given in Table 2.2 indicates that the respondent was almost certain that she would be willing to pay as much as £3 per month and equally certain that she would not be willing to pay as much as £6 per month. Between those two values, the respondent was unable to mark either a tick or a cross, thereby indicating that WTP was uncertain over this range. The results obtained from the ladder complement those obtained from the dichotomous choice questioning and are particularly valuable in revealing the degree of confidence that respondents had in stating their WTP. In the CVALL version of the questionnaire, following the payment ladder, respondents were asked to apportion their overall WTP between the four different sectors of philanthropic activity considered. This exercise provides an alternative ‘top-down’ welfare measure for the services provided by the charities dealing with housing and homelessness issues which can be used as a point of contrast with the single sector welfare measure administered separately to the CVHH subsample. 2.2.2 Contingent Ranking In this version of the survey, respondents were asked to rank in order of preference a series of scenarios which involved different tax increments and different patterns of closure of charitable sector activities. The valuation question was described along the following lines: The amount of additional tax that you would pay will depend on which charities and how many charities are rescued from shutdown. The following showcard lists a number of rescue options. Option 1 is where all of the 4 types of charities shutdown and you pay NO extra money to rescue any of them. Option 2 is where you pay an extra £2.50 each month for a whole year (which would add up to £30 over the year) to rescue the social services charities ONLY.
26
Measuring the economic value of charities Option 3 is where you pay an extra £5 each month for a whole year (which would add up to £60 over the year) to rescue the health and medical research charities ONLY. If you were faced with a choice between these three options, which one would you choose? Please think carefully about whether you can really afford the extra payment, and where the additional money would come from. Remember that this would be over and above any contributions of time and money you already make towards these charities. If you were faced with a choice between the two remaining options, which one would you choose?
Each respondent was given three different sets of options to rank. The actual amounts of money used in these questions varied across different versions of the questionnaire, but were chosen to lie in the same range as those offered in the CV questions. An illustrative example of one of the CR choice sets is provided in Table 2.3. These questions were administered as a sequential choice of the most preferred alternative, first from the full set of three options and then from the set of two options remaining after the firstbest option had been removed. This meant that respondents were effectively required to provide a complete ranking of all the options in the choice set. Table 2.3
Sample question used in the general public survey – CR version Option 1 Option 2 Option 3 ✘ ✘ ✘ ✘
✔ ✘ ✘ ✘
✔ ✔ ✔ ✔
none
20p
50p
(none)
(£2.40)
(£6)
Housing and homelessness charities Social services charities Health and medical research charities Overseas aid, environment and culture Additional payment each month for a year (Corresponding annual amount) Notes:
✘: charity shut down and ✔: charity kept running.
Three design features of the CR version were identified as critical in allowing rigorous comparisons to be made with the parallel CV. First, all choice sets should include an alternative which involves no additional tax payment but which entails the closure of all of the charities (the corresponding situation in the CV version is a ‘no’ answer to the proposed bid level). Second, respondents receiving the CR version should be faced both with options which avoided the closure of a single sector of charitable
Benefits of charities to the public
27
activity, such as the housing and homelessness sector, and with options which avoided the closure of all charities. This would permit the estimation of welfare measures equivalent to those defined for the CV version. Third, the same set of bid levels that were attached to the two versions of the CV questions should be attached to the corresponding alternatives in the CR choice set. The need to meet these three criteria for full comparability with CV necessarily entailed a departure from the conventional principles of fractional factorial design (Louviere, 1988) for the choice of the set of alternatives to be presented in the questionnaire. In this context it is important to note that the purpose of fractional factorial design is to ensure that ranking alternatives are orthogonal, because this is a requirement of the traditional monotonic analysis of variance applied to this kind of experimental data. When analysis is undertaken in a logit regression framework, as in the present case, such orthogonality continues to be desirable but is no longer essential for the method to work satisfactorily. Thus, in order to construct the chosen three-element CR choice sets, the following procedure was adopted. First, the six alternative patterns of charity closure of interest were identified: all charities are shut down, each of the four subsectors alone is maintained, and all four of the subsectors together are maintained. Second, it was established that in order for each possible pair-wise combination to be presented to respondents alongside the zero payment baseline, ten different triplets had to be presented in the survey (given the constraint that the baseline must appear in each of the triplets, there are C25 = 10 different pairs that can be formed from the remaining five alternatives). Third, the bid levels used in the dichotomous choice CV were allocated to the options appearing in the triplets so that each bid level appeared at least once. This involved some repetition of the triplets, bringing the total number up to 18. Finally, the 18 triplets were grouped into six sets of three to be administered to different subsamples of the CR version. The series of three were constructed in such a way that each respondent faced a wide range of bid levels and one alternative involving the preservation of all of the charity sectors. Thus, in all of these ways, the CR questions were designed to mimic the CV questions to the greatest possible degree, so that the only difference between the two was the method by which the response was elicited. 2.2.3 Overview The main objective of the experimental design described in this section was to obtain valid and reliable estimates of the value that the general public attaches to the charitable sector in the UK in general and to the charities
28
Measuring the economic value of charities
dealing with housing and homelessness issues in particular. But the survey also had two secondary objectives. The first was to test for robustness of results to the method. The literature on economic valuation indicates that the results obtained can be highly sensitive to the actual method chosen to elicit WTP from the survey population. Since none of the methods can unambiguously be demonstrated to give the ‘right’ answer, it is consequently desirable to use a variety of different methods in parallel. This makes it possible to examine how robust the results are to the methodology used. Where substantial variations across methods arise, these can be used to place bounds on true WTP. The experimental design adopted permits a systematic comparison of results obtained from dichotomous choice CV and CR. The second was to test for sensitivity of respondents to scope. The literature on economic valuation indicates that respondents are often not very sensitive to the scope of what is being valued. In this context, scope refers to whether respondents are being asked to value all charitable activities or just one particular subsector of those activities, such as housing and homelessness. In particular, the expectation is that, when asked to value housing and homelessness charities alone, people will be willing to pay considerably more than if they are asked to value all charities and then state what proportion of that value is attached to the housing and homelessness subsector. By including these two different approaches in different versions of the questionnaire, it is possible to examine to what extent this problem is impinging on the results and to investigate whether, as has been suggested in the literature (Hanley et al., 1998), CR avoids the problem of ‘part–whole bias’ which has been documented in the case of CV. Assuming that all other differences between questions have been adequately controlled for, the experimental design adopted makes it possible to investigate these secondary objectives. The survey instrument was extensively piloted. The main survey took place in the spring and summer of 1997 and covered a random sample of the UK population composed of 851 respondents across 20 sampling points. The survey was administered by means of a 30-minute face-to-face interview that took place in the respondent’s home. Just under half of the interviews were conducted after the death of Diana, Princess of Wales. A number of factors about this event were important for the study including the depth of the public reaction, the princess’s high profile role in the charitable sector and the volume of donations into her special memorial fund. Since the dates of individual interviews were recorded, it was possible to examine to what extent this event had a significant impact on attitudes towards charities and, in particular, WTP for their preservation.
Benefits of charities to the public
2.3
29
THE SURVEY RESULTS
2.3.1 Socioeconomic Characteristics Table 2.4 undertakes a comparison of each of the three subsamples receiving different versions of the questionnaire. The table states the mean values of each variable. Analysis of the figures in Table 2.4 reveals that there are no significant differences between the various subsamples. Thus any differences in estimation arising from each of the valuation methods employed can be attributed to the method itself rather than to any idiosyncrasies of the sample population to which it was applied. Table 2.4
Socioeconomic characteristics of different subsamples CVHH CVALL
Males (%) Average age (years) Socioeconomic group (%) AB C1 C2 DE Married Children (%) Pre-school age School age Age completed full-time education (years) Employment (%) Full-time Part-time Unemployed Monthly income (£)
CR
Overall
41 43.5
51 45.7
46 44.3
46 44.5
20 28 25 28 60
18 27 24 31 65
18 28 27 27 66
18 27 26 29 64
24 64 16.6
20 64 16.2
26 62 16.5
23 63 16.5
46 18 4 945.6
41 16 3 997.2
49 17 3 950.4
45 17 3 965.3
Notes: CVHH: contingent valuation for housing and homelessness charities; CVALL: contingent valuation for all four charitable sectors; CR: contingent ranking.
2.3.2 Current Donations The questionnaire collected information on existing donations of time and money made by respondents to the four charitable subsectors of interest.
30
Measuring the economic value of charities
These data are useful for two reasons: first, when combined with the results of the valuation question they permit the estimation of an overall value for the charitable sector; second, they can be used to check for consistency with the results of earlier studies of philanthropic behaviour, in particular the Individual Giving Survey (IGS) (Halfpenny et al., 1992, 1993, 1994). Although it was not the purpose of the present survey to collect detailed information on modes of charitable giving, a list of possible ways of giving was used in the questionnaire in order to help respondents to remember how much they had actually given. This list was loosely based on the classification of payment modes used in the IGS. For the purposes of comparison, Table 2.5 gives the proportion of respondents who stated that they had given by each method, both for this survey and for the last year of the IGS. The broad pattern of giving is fairly close across the two surveys and is not significantly different for most comparable classifications. However, the survey does appear to report substantially lower levels of participation than the IGS for collection boxes, sponsorship and raffles. A substantially higher level of participation is recorded in the case of credit cards which benefit charities. Turning to the actual contributions made, Table 2.6 summarizes monthly and annual averages for cash donations and hours volunteered to the four charitable subsectors. As a point of comparison, the equivalent figures from the 1993 IGS are also quoted, adjusted to 1997 prices where relevant. The mean annual donation and hours volunteered is less than 12 times the mean monthly donation and hours volunteered. In fact, the annual values are only about 60 per cent of the grossed-up monthly values. There are two possible explanations for this. First, people may have found it difficult to recall over an annual period and thus may not have been able to record all the contributions they actually made. Second, people may have mistakenly attributed to the past month contributions which they actually made further back in the past; this phenomenon is known as telescoping. Unfortunately, it is not possible to distinguish between these two hypotheses; indeed, it is likely that elements of both could be occurring. The mean monthly cash donations and hours volunteered which emerge from the present survey are comparable to those obtained from the IGS, although significantly lower with their confidence intervals not quite overlapping. On average, people contribute most money to health and medical research charities, followed by overseas aid, environment and culture, then social services and finally housing and homelessness charities. However, many of the confidence intervals for donations to each of these subsectors overlap, indicating that the differences are not statistically significant. In particular, the category of overseas aid, environment and culture has a very wide confidence interval, indicating that behaviour varies considerably across individuals.
31
Note:
confidence intervals in parentheses.
Non-tax-efficient (a) philanthropic ● TV, telephone, mail or newspaper appeal ● collection box in street, shop, pub, at door, work ● sponsoring somebody ● donating via a church or school ● donating to a church or school (b) purchases ● goods from a charity shop, catalogue or sale ● attending a fundraising event ● raffle or lottery ticket (excluding National Lottery) ● using a credit card which benefits charity ● subscription to a charitable organization ● The Big Issue (c) other
0.230 (0.204–0.256) 0.065 (0.050–0.080) 0.302 (0.274–0.330) 0.007 (0.002–0.012) 0.045 (0.032–0.058)
0.202 (0.175–0.229) 0.054 (0.039–0.069) 0.182 (0.156–0.208) 0.025 (0.015–0.035) 0.025 (0.015–0.035) 0.085 (0.066–0.104) 0.066 (0.049–0.083)
0.012 (0.005–0.019)
0.061 (0.046–0.076) 0.607 (0.577–0.637) 0.224 (0.198–0.250) 0.137 (0.116–0.158)
0.096 (0.078–0.114)
IGS 1993
0.043 (0.029–0.057) 0.417 (0.384–0.450) 0.135 (0.112–0.158) 0.052 (0.037–0.067) 0.074 (0.056–0.092)
0.054 (0.039–0.069) 0.019 (0.010–0.028)
Current survey
Comparison of participation rates in different forms of donations
Tax-efficient ● donating by covenants, gift aid, payroll ● giving from a CAF account
Table 2.5
32
Note:
confidence intervals in parentheses.
IGS 1993
Total
Overseas aid, environment and culture
Health and medical research
Social services
0.70 (0.50–0.90) 1.14 (0.89–1.4) 2.36 (1.69–3.03) 1.49 (0.63–2.35) 6.32 (4.73–7.91) 9.51 (8.58–10.44)
Monthly 5.56 (4.50–6.60) 9.26 (7.38–11.14) 15.10 (11.92–18.28) 11.47 (5.41–17.53) 42.98 (34.9–51.1) n/a
Yearly
Cash donations (£)
Summary of existing donations of time and money
Housing and homelessness
Table 2.6
0,07 (–0.01–0.15) 0,41 (0.33–0.49) 0,20 (0.10–0.30) 0,21 (0.03–0.39) 1,48 (1.03–1.93) 4,07 (3,19–4,55)
Monthly
0,34 (0.11–0.58) 5,33 (3.04–7.62) 2,43 (1.51–3.35) 3,01 (1.01–5.01) 11,43 (8.35–14.5) n/a
Yearly
Hours volunteered (hrs, mins)
Benefits of charities to the public
33
On average, people contribute most time to social services charities, followed by overseas aid, environment and culture, then health and medical research, and finally housing and homelessness charities. However, as before, these differences are not always significantly different in the statistical sense. 2.3.3 Attitudes towards Donations The preliminary section of the survey contained a considerable number of attitudinal questions which were intended to make respondents explore their personal thoughts on philanthropic issues as a preparation for responding to the valuation question. In addition, these questions were designed to reveal as much as possible about the underlying motives for philanthropic donation, so as to aid in the interpretation of the valuation responses. The opening attitudinal question asked respondents about the degree of importance of the contribution which charitable organizations made to society. Figure 2.1 indicates that over 70 per cent of the sample thought that charities were either ‘important’ or ‘very important’. However, it is worth noting that these attitudes may be biased upwards by the fact that respondents already knew that charities formed the primary focus of the survey.
not important at all of some importance important very important 0
0.1
0.2
0.3
0.4
0.5
0.6
‘How important a contribution do you think charities make to our society?’ Figure 2.1
Importance of charities
A large number of attitudinal questions were posed with respect to monetary donations and can be grouped into those aimed at uncovering the motives for giving and those aimed at understanding the process of giving. Figure 2.2 presents the results for the questions relating to underlying motive. They are organized along a spectrum from primarily selfish to increasingly altruistic motives.
34
Measuring the economic value of charities
strongly agree agree neutral disagree strongly disagree 0 (a)
0.1
0.2
0.3
0.4
0.5
0.6
‘I often give because I feel too embarrassed to say “No” when someone asks’
strongly agree agree neutral disagree strongly disagree 0 (b)
0.1
0.2
0.3
0.4
0.5
0.6
‘I give to charities because I like the feeling of being generous’
strongly agree agree neutral disagree strongly disagree 0 (c)
0.1
0.2
0.3
0.4
0.5
0.6
‘I give to charities because I or my family may personally benefit from them at some stage’
Benefits of charities to the public
35
strongly agree agree neutral disagree strongly disagree 0 (d)
0.1
0.2
0.3
0.4
0.5
0.6
‘I give to charities because they help to create a better society for everyone by reducing the level of social problems’
strongly agree agree neutral disagree strongly disagree 0 (e)
0.1
0.2
0.3
0.4
0.5
0.6
‘I give to charities because I want to support the good causes for which they work’
Figure 2.2
Attitudes towards donating
On one end of the spectrum is an embarrassment motivation: a person is ashamed to appear to be ungenerous and gives simply to escape from an embarrassing situation. Figure 2.2(a) indicates that the majority of respondents seemed to resent this characterization of their philanthropic motives, with over 60 per cent disagreeing or strongly disagreeing. Nevertheless, a substantial minority of 25 per cent admitted that embarrassment was often a determining factor in their decision to give. Another type of selfish motivation for charitable giving is what has become known in the literature as warm glow (Kahneman and Knetsch, 1992): a person enjoys the feeling of their own generosity. Figure 2.2(b) reveals a similar
36
Measuring the economic value of charities
pattern to the preceding case, with nearly 60 per cent of respondents denying that they exhibit this tendency, but over 20 per cent admitting that they do. Moving along the spectrum we find option value: a person sees their charitable donation as a kind of insurance premium to finance a safety net service from which they or their relatives may personally benefit at some future stage. As shown in Figure 2.2(c), the question of option value split the sample more or less down the middle: 40 per cent identified with this motivation and 40 per cent rejected it. Indirect use values may also motivate philanthropic giving: a person feels that they benefit in an indirect way from the charity’s services because they serve to strengthen the social fabric. Figure 2.2(d) shows that there was a strong tendency to identify with this motivation, which met with agreement from over 60 per cent of the sample. Finally, at the altruistic end of the spectrum of motivations we find existence value: a person thinks that services provided by charities are worthwhile for their own sake, irrespective of any personal spin-offs they may generate. The pattern of results revealed in Figure 2.2(e) is very similar to the preceding case but with an even higher degree of identification, corresponding to over 80 per cent of the sample. Overall, these results indicate that people tend to view their philanthropic donations in strongly altruistic terms. None the less, there is a significant subset of the population which gives for reasons more closely aligned with self-interest. It is interesting to enquire to what extent these different motivations overlap at the level of individual respondents. Table 2.7 reports the correlation coefficients between each pair of attitudinal variables and reveals a number of interesting points (none of the correlations reported is particularly high in absolute terms, although most are statistically significant). There is a particularly strong correlation (0.33) between people motivated by embarrassment and people motivated by the warm glow. Indeed, 72 per Table 2.7
Correlation between different motives for monetary donations
Embarrassment Warm glow Option value Indirect use Existence value
Embarrassment
Warm glow
Option value
Indirect use
Existence value
1 0.33 0.14 0.08 –0.06
1 0.27 0.23 0.10
1 0.22 0.23
1 0.55
1
Benefits of charities to the public
37
cent of the sample consistently either agreed or disagreed with both of these statements. Similarly, there is an even stronger correlation (0.55) between those motivated by indirect use concerns and those motivated by existence value. Indeed, 86 per cent of the sample consistently agreed or disagreed with both of these statements. There is a very weak negative correlation (–0.06) between those motivated by embarrassment and those motivated by existence value considerations. Indeed, only 32 per cent of respondents consistently agreed or disagreed with both statements. Hence there seems to be a consistent thread in respondents’ answers suggestive of the existence of two broadly distinct groups of respondents: a majority that is mainly driven by altruistic motivations and a minority that is motivated primarily by more selfish considerations. The survey also examined the process of giving. Three distinct areas were investigated: the scope of the donations, that is the extent to which giving was focused on one particular subsector or spread widely across different areas of charitable activity; the level of foresight, that is the extent to which giving was spontaneous or carefully planned in advance; and the degree of commitment, that is the extent to which giving was regular over time or tended to be simply on a one-off basis. The responses to the issue of scope are contained in Figure 2.3. Parts (a) and (b) contain responses to what are essentially identical questions but worded in different ways: one as a positive statement and the other as a negative statement. The two statements do not appear contiguously in the questionnaire. This is done to ensure that responses are meaningful and consistent through the questionnaire and are not merely an artefact of the way the statements are expressed. In fact the pattern of responses to the two questions is almost identical, with over 50 per cent agreeing or strongly agreeing that they are best characterized as focused givers. However, a substantial minority (in excess of 25 per cent) responded that they tended to give very widely. Finally, Figure 2.3(c) presents the results of a question as to whether people really cared what kind of good cause they were giving to. Interestingly, this question divided the sample more or less down the middle, with just over 40 per cent agreeing or strongly agreeing and just under 40 per cent disagreeing or strongly disagreeing. The responses to the issue of foresight are contained in Figure 2.4. Once again two versions of the question are used to test for sensitivity to wording. Although the overall pattern of responses is skewed towards agreement with these statements in both cases, that is, revealing a tendency towards spontaneous giving, the second statement seems to elicit a substantially higher rate of agreement than the first, just under 70 per cent versus just under 50 per cent.
38
Measuring the economic value of charities
strongly agree agree neutral disagree strongly disagree 0
(a)
0.1
0.2
0.3
0.4
0.5
0.6
‘I tend to give to one or two favourite charities or a specific area of charitable activity’ strongly agree agree neutral disagree strongly disagree 0
(b)
0.1
0.2
0.3
0.4
0.5
0.6
‘I don’t tend to give to a wide range of different charities’ strongly agree agree neutral disagree strongly disagree 0
(c)
0.1
0.2
0.3
0.4
0.5
0.6
‘When I give, I don’t really mind what I’m giving to as long as it’s some kind of good cause’
Figure 2.3
Attitudes towards charitable donations: scope
Benefits of charities to the public
39
strongly agree agree neutral disagree strongly disagree 0 (a)
0.1
0.2
0.3
0.4
0.5
0.6
‘I give money spontaneously when I am approached by someone’
strongly agree agree neutral disagree strongly disagree 0 (b)
0.1
0.2
0.3
0.4
0.5
0.6
‘I don’t plan my charitable giving carefully in advance’
Figure 2.4
Attitudes towards charitable donations: foresight
The responses to the issue of commitment are contained in Figure 2.5. In this case both versions of the question are positive. Thus a consistent response would require that a person who agreed with the first statement would disagree with the second. The figures show that, on average, this is indeed the case. Just under 60 per cent of the sample identify themselves as uncommitted givers by disagreeing with the first statement and agreeing with the second, while the converse pattern of responses identifies just under 30 per cent of the sample as committed givers. Once again it is interesting to consider the interrelationship between these different types of giving habits, as shown in Table 2.8. The results indicate that the different questions relating to scope have the expected mutual correlations. The value is 0.20 for responses to the questions
40
Measuring the economic value of charities
strongly agree agree neutral disagree strongly disagree 0 (a)
0.1
0.2
0.3
0.4
0.5
0.6
‘I tend to support a particular cause with regular donations over a period of time’
strongly agree agree neutral disagree strongly disagree 0 (b)
0.1
0.2
0.3
0.4
0.5
0.6
‘I tend to make one-off donations with no particular long-term commitment’
Figure 2.5
Attitudes towards charitable donations: commitment
addressing the range of charities to which a person gives. However, both these questions are negatively correlated with the question indicating that people don’t really mind what good cause they’re giving to. None of these correlations is particularly high in absolute terms. The different questions relating to foresight and commitment also have the expected mutual correlations even though, in absolute terms, the values are not particularly high. The correlation is 0.17 for responses to the questions about giving spontaneously and not planning one’s giving, while the value is –0.11 for responses to the questions about making one-off donations and supporting causes regularly over time.
41
1 0.20 –0.05 –0.07 –0.01 0.37 0.07
Scope (a)
1 –0.11 –0.21 0.15 0.03 –0.01
Scope (b)
1 0.41 0.05 0.10 0.11
Scope (c)
1 0.17 0.06 0.13
Foresight (a)
Correlation between different approaches to monetary donations
Scope (a) Scope (b) Scope (c) Foresight (a) Foresight (b) Commitment (a) Commitment (b)
Table 2.8
1 –0.18 0.20
Foresight (b)
1 –0.11
Commitment (a)
1
Commitment (b)
42
Measuring the economic value of charities
There is a high correlation (0.41) between those who tend to give spontaneously and those who are completely indifferent about which good cause they are supporting. There is also a high correlation (0.37) between those who tend to give to favourite charities and those who give regularly over time. Indeed, 62 per cent of respondents consistently agreed or disagreed with both of these statements. Putting together the results of the attitudinal questions on motive and those on the process of donating uncovers some further interesting associations. The tendency to give spontaneously is quite strongly associated both with emotional motivations for giving (a correlation of 0.31 with the embarrassment motive and 0.23 with the warm-glow motive) and with more altruistic motives for giving (a correlation of 0.36 for the indirect use motive and 0.27 for the existence value motive). Furthermore, there is quite a strong correlation (0.37) between those who tend to concentrate their giving on a few favourite charities and those who identify themselves as being motivated to give out of regard for the associated good cause. By and large these results conform to prior expectations regarding the motivations behind support of charitable activities in the light of previous findings reported elsewhere in the literature. While no single motivation stands out as the most important factor driving respondents’ attitudes, as many considerations seem to play a role in individual attitudes, by and large the more altruistic group of motivations seems to play a fundamental role in explaining people’s attitudes towards charitable donations. As for the process of giving, typically it is neither carefully planned nor implies a long-term commitment by donors and it tends to be focused on a number of favourite charities. 2.3.4 Attitudes towards Volunteering In parallel to the investigation of the motivations behind monetary donations, the survey also collected information on the motives behind volunteering. Figure 2.6 summarizes the results. Once again the statements are organized along a spectrum from selfish to altruistic motives. However, given the different nature of volunteering, the motivational categories do not exactly match up with those used for monetary donations. At the selfish end of the spectrum of motivations lies personal enjoyment: a person volunteers because the very act of volunteering is enjoyable or personally advantageous in some way. Figure 2.6(a–c) explores three aspects of this: volunteering as a way of occupying free time, volunteering as a way of making friends and volunteering as a way of gaining skills. The responses to all of these questions split the sample to a considerable extent. Just over 40 per cent agree that it is a good way of occupying free time, just over 50 per
Benefits of charities to the public
43
cent agree that it is a good way of making friends, and nearly 50 per cent agree that it is a good way of acquiring skills. The expectation of some personal benefit may also provide the rationale for volunteering: a person volunteers because their relatives are among the direct beneficiary group of the services provided. This is not unlike the option value statement posed for monetary donations. As with the option value question discussed above, Figure 2.6(d) shows that the sample of volunteers is fairly evenly split between those who acknowledge some kind of personal benefit and those who do not. The former are in a slight majority at just under 50 per cent. At the other end of the spectrum of motivations lie altruistic considerations: a person volunteers because he values the ultimate good cause which the charity works towards. Figure 2.6(e) shows that the vast majority of people see their volunteering activity in these terms. As in the case of donations, over 80 per cent of the sample identify with this altruistic motivation. In order to shed light on the overlap between motivations for individual respondents, Table 2.9 reports the correlation coefficients between each pair of attitudinal variables. In contrast to the results reported for monetary donations in Table 2.7, there are strong positive associations between all the different motivations, indicating that many of these considerations tend to play a role for many individuals as opposed to some individuals being driven primarily by some motivations and others by different ones. A final attitudinal question asked volunteers to think about what would be the most likely alternative use of their time if they no longer had the opportunity to volunteer. This makes it possible to establish what is being sacrificed by the volunteer in order to give time to the charity and thus sheds some light on the issue of how much the volunteer values their donation of time. Economists have often approached the valuation of volunteer time in terms of the net hourly wage that the volunteer could have earned had they been working instead of volunteering (see Chapter 4). Clearly, this is only legitimate to the extent that people are genuinely substituting between volunteering and paid employment. However, the results of the attitudinal question presented in Figure 2.7 indicate that only 20 per cent of volunteers see themselves as giving up opportunities for paid employment in order to volunteer. Hence the traditional approach of valuing volunteer time in terms of the net hourly wage is only valid for a small subset of the volunteering population. Indeed, Figure 2.7 indicates that an equal proportion of just over 20 per cent of volunteers state that they would not have anything in particular to do if they could not volunteer. This suggests that this group of people place no value whatsoever (or perhaps even a negative value) on the time which they give to charities as volunteers. By far the largest category of volunteers are giving up time that they would otherwise devote to domestic work (about 40 per cent) or to recreation (about
44
Measuring the economic value of charities
strongly agree agree neutral disagree strongly disagree 0 (a)
0.1
0.2
0.3
0.4
0.5
0.6
0.5
0.6
0.5
0.6
‘Volunteering is a good way of occupying my free time’
strongly agree agree neutral disagree strongly disagree 0 (b)
0.1
0.2
0.3
0.4
‘I find it enables me to meet people and make friends’
strongly agree agree neutral disagree strongly disagree 0 (c)
0.1
0.2
0.3
0.4
‘I think of volunteering as gaining valuable skills and experience’
Benefits of charities to the public
45
strongly agree agree neutral disagree strongly disagree 0 (d)
0.1
0.2
0.3
0.4
0.5
0.6
‘I think of volunteering as helping my family and friends’
strongly agree agree neutral disagree strongly disagree 0 (e)
0.1
0.2
0.3
0.4
0.5
0.6
‘I find I enjoy contributing to a good cause’
Figure 2.6 Table 2.9
Attitudes towards volunteering: motive Correlations between different motives for volunteering
Free time Make friends Gain skills Help family Good cause
Free time
Make friends
Gain skills
Help family
Good cause
1 0.19 0.34 0.37 0.26
1 0.31 0.42 0.44
1 0.47 0.33
1 0.60
1
46
Measuring the economic value of charities
Paid employment 20%
Recreation 14%
Nothing in particular 21%
Other 4%
Home responsibilities 41%
‘If for any reason you no longer had the opportunity to volunteer, what would you be most likely to do with your time instead?’ Figure 2.7
Alternatives to volunteering
14 per cent). These uses of time could be expected to have a positive value to respondents but one which does not necessarily bear any relationship to the net hourly wage that they could potentially earn in employment. 2.3.5 Valuation Results This section describes the estimation of the value of all charities and the value of the subset of the housing and homelessness charities obtained from both the CV and the CR versions of the general public survey. Contingent valuation results As described in Section 2.2, two types of elicitation questions were asked in the CV versions of the questionnaire, for both CVALL and CVHH: first, the dichotomous choice approach was used, whereby respondents were asked whether they would be willing to pay a sequence of predetermined amounts of money in taxes (the bid levels) that varied across people, to which they only had to respond ‘yes’, ‘no’ or ‘don’t know’; and second, respondents were presented with the payment ladder depicted in Table 2.2 and asked to choose the amount that best described their maximum WTP. The results of each approach will be considered in turn. Table 2.10 describes the percentage of respondents that accepted both bids (YY), accepted the first and rejected the second (YN), rejected the first bid but accepted the second (NY) or rejected both bids (NN), for each of the five bid vectors in Table 2.1. Inspection of Table 2.10 shows that, in the case of all charities and of the housing and homelessness charities, only 8 per cent and
Benefits of charities to the public
47
Table 2.10 Percentage of respondents in each response category NN
NY
YN
YY
All Housing All Housing All Housing All Housing charities charities charities charities charities charities charities charities I II III IV V
31 26 32 42 58
25 32 46 47 53
6 3 14 14 15
4 14 11 15 16
13 28 31 23 19
14 20 21 29 17
50 43 23 21 8
57 34 22 9 14
Notes: NN: rejected both bids; NY: rejected the first bid accepted the second; YN: accepted the first bid, rejected the second; YY: accepted both bids; I–V: bid vectors as in Table 2.1.
14 per cent of respondents respectively accepted the highest bid offered (CVALL: £25 per month, corresponding to £300 per year; CVHH: £15 per month, corresponding to £180 per year). As expected, the percentages are low since these were the highest proposed tax levels. The data summarized in Table 2.10 do not provide a direct WTP value for the services provided by charities. They only clarify whether a respondent’s WTP is above or below certain bid levels. However, using appropriate statistical models and assuming particular probabilistic distributions for the WTP, it is possible to infer the average WTP of the sample either using only the results from the first valuation question (the single-bounded approach) or using both valuation questions (the double-bounded approach). The latter approach was followed. Further details can be found in the Statistical Appendix at the end of this chapter. In the surveys of all charities and of the housing charities, 15 per cent and 18 per cent of respondents respectively were uncertain about being prepared to pay the specified bid levels and answered ‘don’t know’ to the valuation question. As such it was necessary to make assumptions about the most probable direction of this type of answers. In the calculations that follow, we chose the two uncertainty assumptions: we treated all ‘don’t know’ answers as ‘no’ (the lower-bound approach) and also as ‘yes’ (the upper-bound approach). Thus true WTP can be said to lie between these two boundaries. Table 2.11 displays both approaches for all charities and for housing charities alone. Yearly bid levels were used in the estimation. The statistical models depicted in Table 2.11 do not identify WTP directly, but estimate the probability of accepting to pay a particular tax level as a function of that yearly tax level. As such the coefficient on the tax level measures its impact on the probability of acceptance. All the tax coefficients
48
–771.19 278
Log-L N
Notes:
5.153 –16.203
£63.21 £49.84–£76.58
–414.47 278
0.6755 –1.0687 7.042 –17.937
£33.80 £21.81–£45.79
–372.80 282
0.4300 –1.2723
coeff. 4.789 –14.124
t-stat.
‘Don’t know’ = ‘No’
£50.57 £40.38–£60.77
–412.56 282
0.7120 –1.4079
coeff.
7.478 –15.649
t-stat.
‘Don’t know’ = ‘Yes’
Housing and homelessness charities
Model specification = double-bounded probit; CI = confidence interval calculated using the Krinsky and Robb (1986) procedure.
Yearly WTP £46.90 95% CI £31.38–£62.41
0.4603 –0.9816
Constant Tax level
t-stat.
coeff.
coeff.
t-stat.
‘Don’t know’ = ‘Yes’
‘Don’t know’ = ‘No’
All charities
Table 2.11 Yearly WTP estimates for different models and uncertainty assumptions
Benefits of charities to the public
49
are negative as expected: the higher the bid level presented to respondents, the lower the probability of acceptance. The last two rows in Table 2.11 show the mean yearly WTP for all charities and for the housing and homelessness charities respectively and their respective confidence intervals, estimated under both uncertainty treatments (see the Statistical Appendix to this chapter for details of the estimation procedure used). WTP for all charitable services (over and above the current level of donations) ranges from £47 to £63 per person per year, while WTP for the housing and homelessness charities alone is in the interval of £34 to £51 per person per year. As expected, the upper-bound approach of treating ‘don’t know’ as ‘yes’ yields higher estimates than the corresponding lower-bound treatment where uncertain answers are treated as refusals. However, due to the imprecision associated with the estimates (which can be judged by the magnitude of the confidence intervals around the mean), these boundary values are actually not statistically different from each other. Another way of putting this is that the stated confidence intervals around the mean are overlapping. Finally, whatever the uncertainty assumption taken, WTP for CVHH is smaller than that for CVALL. However, in none of the cases are the differences between WTP amounts statistically significant. This lends support to the insensitivity to scope hypothesis discussed initially: when valuing one subsector on its own, people tend to overestimate its value. In our case, the WTP for the housing and homelessness sector, estimated independently, does not seem to be significantly different from the WTP for all charitable sectors. Attention now turns to the results obtained from the second type of valuation questions used: the payment ladder described in Table 2.2, for both the CVALL and CVHH versions of the questionnaire. It should be noted that these results are likely to be correlated with the estimates obtained in the dichotomous choice elicitation, since the same respondents were asked both valuation questions. The yearly WTP results from the payment ladder for all charities and for the housing and homelessness charities alone are displayed in Table 2.12. The average maximum yearly WTP is £49 for all charities and £41 for the housing and homelessness charities: the two values are not, however, statistically different. As already indicated by the dichotomous choice results, there seems to be considerable insensitivity to the scope of the charitable sector being analysed. Median WTP values for CVALL (£24) and CVHH (£18) are considerably lower than their respective means, which signals the presence of outliers (unrealistically high bids), as can be confirmed by looking at the distribution of WTP amounts depicted in Figure 2.8. Both distributions are quite similar – generally downward sloping as expected with some outliers at the far end of the right-hand tail.
50
Measuring the economic value of charities
Table 2.12 Yearly WTP for CVALL and for CVHH from payment ladder Housing and homelessness charities
All charities Maximum WTP (✔)
Minimum not WTP (✘)
Maximum WTP (✔)
Minimum not WTP (✘)
Mean £49.21 £66.87 £41.22 £56.94 95% confidence £40.00–£58.41 £54.81–£78.92 £33.56–£48.88 £47.23–£66.65 interval Median £24 £36 £18 £30 Minimum £0 £1.2 £0 £1.2 Maximum £600 £900 £540 £600 N 272 266 275 274
90 80 70 60 50 40 30 20 10 0
All charities
Figure 2.8
401–600
201–400
176–200
151–175
126–150
100–125
76–100
51–75
26–50
1–25
0
Housing charities
WTP distribution in CVALL and CVHH
Comparing the WTP values directly obtained from the payment ladder with the indirectly estimated WTP amounts from the dichotomous choice (‘yes’ or ‘no’) procedure for all charities and for the housing charities alone (Tables 2.11 and 2.12), it can be seen that the payment ladder values are much closer and statistically not different from the dichotomous choice estimates under the lower-bound uncertainty assumption, that is, when ‘don’t know’ answers are treated as ‘no’. In contrast, the payment ladder values are much lower than the upper-bound estimates from the dichotomous choice
Benefits of charities to the public
51
models, which suggests that the latter may be biased upwards and not a true reflection of people’s preferences. Further inspection of these results shows that the upper-bound dichotomous choice estimates fall within respondents’ uncertainty range, as revealed by the payment ladder, that is, the difference between what respondents are willing to pay for sure (the ticks) and the amounts they are sure they would not pay (the crosses). These results indicate that respondents seem to have been reasonably consistent when answering the dichotomous choice and the payment ladder valuation questions. The WTP estimates seem to be robust to the elicitation method chosen when the right uncertainty assumption is made, that is, the lower-bound approach. The ladder WTP results match the dichotomous choice estimates when ‘don’t know’ answers are treated as ‘no’, which suggests that, in the latter, respondents only accepted to pay a given amount when they were almost sure they actually would do it. It is interesting to note that the number of respondents expressing uncertainty in the payment ladder (blank values between the last tick and the first cross) and in the dichotomous choice questions (‘don’t know’ answers) is remarkably similar: 18 per cent in both approaches for CVHH; and 22 per cent for ladder versus 15 per cent for dichotomous choice for CVALL. In the valuation of all charities questionnaires (CVALL), when they had completed the payment ladder elicitation procedure, respondents were also asked how many pence (out of every pound in their global WTP amount for all charities) they thought should be given to the four charitable sectors of interest. Table 2.13 shows the results. This procedure, whereby the total value of a good is first estimated and then reallocated into subcomponents of that good, is sometimes called the ‘top-down’ approach. As expected, the top-down estimation procedure yields significantly lower values for the subcomponent areas of the charitable sector than if these latter values were directly estimated separately. This is apparent by looking at the value of the housing and homelessness sector for which we have both the WTP derived from the top-down approach (£13 per year in Table 2.13) and the directly estimated value, either from the payment ladder (£41 per year in Table 2.12) or from the dichotomous choice approach (£34 to £51 per year in Table 2.11). These large differences suggest that due to the degree of insensitivity to scope exhibited by respondents, the top-down approach may be preferable for estimating the value of individual subsectors which are part of a larger sector: first, estimate the total value of the charitable sector; then allocate that value across subsectors. Otherwise, when faced with a subsector on its own (as the housing and homelessness charities), respondents tend to offer their whole available budget for all charities to the particular subsector they are being asked to value. Arguably this phenomenon would take place also with the other types of charities considered (social services, health and
52
Weighting Estimated mean WTP (£/year) 95% Confidence interval N 0.24 £13.10 £10.5–£15.7 216
Housing and homelessness 0.26 £18.09 £13.3–£22.9 216
Social services 0.37 £19.37 £15.5–£23.3 216
Health and medical research
Table 2.13 Distribution of total WTP across different areas of charitable activity
0.13 £8.21 £6.4–£10.1 216
Overseas aid, environment and culture
1 £58.78 £47.4–£69.8 216
Total WTP (£/year)
Benefits of charities to the public
53
medical research and overseas aid, environment and culture), although our survey did not test for them explicitly. It should be noted that the total WTP for all charities drawn from the payment ladder as described in Table 2.13 (£59 per year) is different from that depicted in Table 2.12 (£49 per year). This is due to non-response to the top-down approach question. Different numbers of respondents were used to calculate the value in both cases: the top-down estimates in Table 2.13 are based on 216 respondents vis-à-vis the 272 answers on Table 2.12. The validity of the payment ladder WTP results was assessed using valuation functions which estimate the impact of possible explanatory variables on WTP. The valuation functions are obtained by regressing the maximum WTP amounts (the last tick on the ladder) on a set of economic, attitudinal and other explanatory variables. It is then possible to investigate whether these variables affect WTP in the manner predicted by theory. For example economic theory suggests that WTP should vary positively with income, if charitable services are considered to be normal goods. Table 2.14 describes the sociological, economic, behavioural and attitudinal variables thought to influence WTP that were included in the valuation functions. The same type of analysis was conducted on the dichotomous choice estimates with similar results (further details can be found in EFTEC, 1997). The results of the valuation functions for CVALL and CVHH are illustrated in Table 2.15. The regression method used was ordinary least squares (OLS – see the Statistical Appendix for further details). Two model specifications were tested: in the first, personal income is included as a possible economic explanatory variable, and in the second, income is excluded from the set of economic explanatory variables and included instead are two proxies for income (dummy variables reflecting full-time employment status and socioeconomic population segment AB). The reason for this dual estimation procedure lies in the high proportion of income non-response (46 per cent), which is not uncommon in household surveys. Overall, the regressions with income perform better than the equations without income but with income proxies, with an explanatory power of 58 per cent (CVALL) and 42 per cent (CVHH) vis-à-vis 33 per cent (CVALL) and 25 per cent (CVHH). These levels of explanatory power (R2) are relatively high for cross-sectional data contingent valuation studies (where the minimum acceptable threshold is considered to be as low as 15 per cent: Mitchell and Carson, 1989). Conforming to prior expectations, both income and the level of current donations are found to be strong determinants of WTP: the richer one is and the higher the current level of philanthropic giving, the higher the WTP for all charities in general and for the housing charities in particular. In the alternative specification with the income proxies the population segment AB
54
Measuring the economic value of charities
Table 2.14 Description of explanatory variables Sociological variables Sex Age Education Economic variables Income Full-time job Segment AB Cash donations Attitudinal variables Charities’ importance Housing charities importance Existence values
Indirect use values
Option values
Type of donation Diverse donations Tax-efficient Charity credit card Appeals Merchandise Big Issue Beggars Other variables Tax level Volunteering Diana London
1 – male; 0 – female Mid-point of interval Age at which full-time education was completed Adjusted mid-point of interval (individual, monthly) 1 – employed full-time; 0 – otherwise 1 – segment AB; 0 – otherwise £ donated in the previous year 3 – very important, to 0 – not important at all 1 – most important charity area, to 4 – least important charity area 2 – if strongly agrees that support for charities is due to existence related motives, to (–2) – if strongly disagrees 2 – if strongly agrees that support for charities is due to indirect use related motives, to (–2) – if strongly disagrees 2 – if strongly agrees that support for charities is due to option related motives, to (–2) – if strongly disagrees 2 – if donations are focused on a small number of charities; (–2) – if donations are wide-ranging 1 – if donations are tax efficient; 0 – if otherwise 1 – if uses a charity credit card; 0 – if otherwise 1 – if gives in response to an appeal; 0 – if otherwise 1 – if buys charity merchandise; 0 – if otherwise 1 – if buys The Big Issue; 0 – if otherwise 1 – if gives directly to beggars and others; 0 – otherwise Initial tax or bid level presented (in yearly terms) 1 – if volunteered in the previous year; 0 – if not 1 – if questionnaire was done after the death of Diana, Princess of Wales; 0 – if before Sampling point: 1 – London; 0 – elsewhere
Benefits of charities to the public
55
is a positive and significant determinant of giving as expected, with current donations remaining significant only for the housing and homelessness regression. Regarding attitudinal variables, existence value motivations are found to positively influence WTP for all charities, albeit only at the 10 per cent level. Interestingly, in the case of housing charities, indirect use benefits seem to be behind donations. The results suggest that those who receive indirect benefits from the services provided by these charities, for example keeping the homeless away from the streets and thus providing a better social atmosphere, are more likely to pay for such services. In CVHH, the coefficient of the variable that reflects the importance of housing and homelessness charities in particular, vis-à-vis other areas of charitable giving, is not statistically significant, even though it has the expected negative sign, meaning that the higher the rank obtained by this charitable area (1, 2, 3 or 4, with 1 being the highest), the higher the probability of accepting paying for it. Those who think charitable work is important in general are found to have a higher WTP. The type of donation is seen to have a significant impact on the WTP for all charities when income is not included as a regressor but less so when income is part of the specification. Respondents who donate to a wide variety of charities rather than channelling their donations to particular causes (diverse donations) and who respond to TV, mail, telephone or newspaper appeals are found to value charitable services more highly. In contrast, those who use a charity credit card or buy charity merchandise tend to give less. The impact of this type of variable on the WTP for the housing and homelessness sector is limited. Other regressors specifically related to housing and homelessness services such as giving directly to beggars in the street or buying The Big Issue are not statistically significant even though they have the expected positive signs. Socioeconomic variables such as gender and age have a significant impact on the WTP for all charities. The results show that men are more willing to pay than women (which contradicts the stylized fact that women are more generous than men, as uncovered in various other valuation studies), as are younger people. Education has a positive though insignificant influence. None of these variables is significant in explaining WTP for housing and homelessness charities. Interestingly, the death of Diana, Princess of Wales, does not seem to have a significant impact on actual WTP levels, whatever the specification, although the variable coefficient has a positive sign. Finally, the tax level coefficient is positive and significant in nearly all OLS regressions. This is an important finding as it indicates that the initial bid level (presented in the dichotomous choice elicitation questions) posi-
56
Constant Sociological variables Sex Age Education Economic variables Income Full-time job Segment AB Cash donations Attitudinal variables Charities’ import. Housing char. imp. Indirect use value Option values Existence values –0.24 2.00 –2.29 0.82 2.32 – – 2.43 0.47 –0.39 – – 1.65
21.03 –0.77 1.89
0.02 – – 0.46
3.78 –1.81 – – 10.99
t-stat.
–12.36
Coeff.
With income
3.03 –7.52 – – 9.36
– 18.91 25.05 0.05
20.80 –0.63 0.85
37.50
Coeff.
0.56 –1.72 – – 1.80
– 2.45 1.85 0.87
2.32 –2.06 0.43
0.92
t-stat.
Without income
All charities (CVALL)
12.92 –6.61 16.43 2.31 –
0.03 – – 0.33
1.33 0.03 3.96
–95.04
Coeff.
1.67 –1.09 3.30 0.65 –
2.51 – – 3.64
0.13 0.07 0.97
–1.23
t-stat.
With income
12.08 –2.42 13.13 3.62 –
– 13.37 25.01 0.213
10.62 0.21 2.52
–77.24
Coeff.
2.22 –0.59 2.92 1.03 –
– 1.54 2.18 2.09
1.37 0.65 0.70
–1.11
t-stat.
Without income
Housing and homelessness charities (CVHH)
Table 2.15 Valuation functions for CVALL and CVHH: payment ladder results
57
–1.35 – –0.73 0.66 –2.36 – – 2.44 0.92 – 0.19
–8.05 – –24.71 10.90 –23.61 – –
2.99 11.58 – 1.83
0.58 113
Type of donation Diverse donations Tax-efficient Charity credit card Appeals Merchandise Big Issue Beggars Other variables Tax level Volunteering London Diana
R2 N 0.33 196
1.59 41.67 – 3.03
–16.71 – –59.60 20.27 –16.53 – – 1.57 3.16 – 0.34
–3.30 – –2.20 1.91 –2.17 – –
0.42 111
2.76 – –21.32 4.39
– –49.84 – – – 5.72 13.23 1.88 – –1.41 0.33
– –1.82 – – – 0.35 0.57
0.25 199
3.50 – –6.91 1.63
– –18.15 – – – 14.74 20.63 2.73 – –0.66 0.16
– –1.03 – – – 1.14 1.30
58
Measuring the economic value of charities
tively influences the WTP outcome derived from the payment ladder. This confirms the correlation hypothesis previously mentioned: the WTP amounts chosen on the ladder are not independent of the preceding dichotomous choice questions. This results from the fact that the same sample was confronted with both sets of questions, rather than having a split-sample context. This within-sample approach permits a test of internal consistency, which was achieved by respondents: the WTP values estimated both from the payment ladder and the ‘yes’/‘no’ questions are statistically indistinguishable (under the lower-bound uncertainty assumption). However, the significant coefficient of the tax level in the regressions of Table 2.15 suggests that payment ladder results may suffer some sort of anchoring bias, where the bid levels presented in the dichotomous choice procedure affect the WTP statement. Contingent ranking results In order to derive WTP estimates from the ranking data it is necessary to make use of a statistical model known as the rank-ordered logit. This statistical model estimates coefficients which indicate the sign and magnitude of the effect of each particular aspect of the options offered on the ranking accorded to that option. Further details can be found in the Statistical Appendix to this chapter. Amongst the explanatory variables used in the model are the various attributes of the options presented to respondents (see Table 2.3). Each of these variables shows whether or not the respective subsector is shut down under any particular option. In addition, the ‘overall’ variable represents an interaction of all four subsectors in combination, that is, it indicates whether the whole charitable sector shuts down. The last attribute (tax bill) represents the monetary cost associated with any particular option. The results from this model are reported in Table 2.16. A positive coefficient on a variable indicates that the respective subsector is positively valued. The negative coefficient on the overall variable indicates that the overall value for all four sectors in combination is smaller than the sum of the values of each sector when presented individually. This is consistent with the insensitivity to scope phenomenon that was also apparent in the contingent valuation version. The tax bill associated with the options has a negative coefficient, indicating that the higher the cost of an option, the less likely people are to select that option. On the basis of these coefficient results it is possible to calculate the WTP for each sector individually and for all sectors in combination (see Statistical Appendix). The results in Table 2.16 show that the health and medical research sector has the highest annual WTP for any individual charitable sector at £149.95 on average. This is followed by social services with £99.16 on
Benefits of charities to the public
59
Table 2.16 Results of contingent ranking version Model coefficient (confidence interval) Housing and homelessness Social services Health and medical research Overseas aid, environment and culture Overall Tax bill
N Log-likelihood Note:
0.1598 (0.0669–0.2527) 0.4498 (0.2658–0.6338) 0.6800 (0.4642–0.8958) 0.3012 (0.0985–0.5038) –0.7017 (–1.1494– –0.2541) –0.0046 (–0.0061– –0.0031)
t-statistic 1.72 4.79 6.18 2.91 –3.07 –5.92
Annual WTP (confidence interval) £35.75 (–£4.21–£75.71) £99.16 (£51.36–£146.96) £149.95 (£96.74–£203.16) £66.14 (£20.10–£112.18) £195.52 (£132.00–£259.04) n/a
702 –1230.81
Confidence intervals estimated with the Krinsky and Robb (1986) approach.
average, overseas aid, environment and culture with £66.14 on average and housing and homelessness with £35.75 on average. However, the confidence intervals on all the estimated values are very wide. Consequently the WTP for housing and homelessness charities is not significantly different from zero. Moreover, the valuation of the housing and homelessness, social services, and overseas aid, environment and culture subsectors are not significantly different from each other. The value of the health and medical research subsector is not significantly different from the social services and overseas aid, environment and culture subsectors. Neither is it significantly different from the estimated overall value accorded to all charities, which at £195 appears to be unrealistically high. Debriefing questions Table 2.17 summarizes the responses given to the common debriefing questions which were administered on each version of the questionnaire. These responses should be regarded as indicative rather than conclusive evidence. This is because respondents were faced with a list of precoded reasons why they were or were not willing to pay for charitable services and were asked to choose only one of these reasons, while, typically, WTP responses are the end result of a number of different (maybe even contradictory) motivations, as the attitudinal part of the survey shows. Still, they illustrate some of the potential caveats associated with valuation surveys.
60
Measuring the economic value of charities
Table 2.17 Summary of debriefing questions across subsamples (percentage of respondents aggreeing with the statement) CVHH
CVALL
CR
Reasons for being willing to pay: I think that the work done by these charities is worth the extra contribution.
36.2 64.5 51.0 (29.5–42.9) (57.6–71.4) (43.9–58.1)
I think this is a very important issue, I am not sure if I could pay this amount but I wish I could.
30.6 35.5 48.9 (24.1–37.1) (28.6–42.4) (41.8–56.0)
I think charities in general are a very important sector. My answer reflects my concern for charities in general and not simply for those working on housing and homelessness.
33.2 (26.5–39.9)
n.a.
n.a.
7.6 (2.1–13.1)
5.6 (0.3–10.9)
Reasons for not being willing to pay: I do not care that much about these charities, 5.6 and would rather spend the money on other (0.7–10.5) things that are more important to me. I could not afford the additional amount of tax which was being asked of me.
23.6 22.8 18.1 (14.8–32.4) (14.2–31.4) (9.1–27.1)
I think I already make enough voluntary donations to charities as it is, and that other people ought to pay more.
9.0 (3.1–14.9)
3.3 (–0.4–7.0)
6.9 (1.0–12.8)
I think I already pay enough tax as it is.
13.5 (6.4–20.6)
12.0 (5.3–18.7)
18.1 (9.1–27.1)
I think people should be left to make their 44.9 50.0 44.4 own contributions if they want to rather than (34.5–55.3) (39.8–60.2) (32.8–56.0) being forced to contribute through tax. I did not find the idea very convincing.
Note:
3.4 (–0.3–7.1)
4.3 (0.2–8.4)
6.9 (1.0–12.8)
confidence intervals in parentheses.
By and large, there are no significant differences between the responses given for each version of the questionnaire, as can be seen from the fact that the respective confidence intervals always overlap. In the surveys, people were asked the reasons behind their WTP. This information can be used to distinguish between those with a genuine WTP and those who may simply be using this question to express a positive attitude towards charities. The results indicate that a sizeable minority of at
Benefits of charities to the public
61
least a third fell into this latter category. For the CVHH version of the questionnaire a further third claim to be expressing their valuation of all charities rather than simply the housing and homelessness sector as requested in the questionnaire (which goes some way to explain the insensitivity to scope found). For the CVALL two-thirds of respondents confirm that they are expressing a genuine WTP, whereas for CR the proportion is substantially lower, at about one-half. Respondents were also asked for the reasons why they were not willing to pay. This question attempts to distinguish between those who are genuinely unwilling to pay and those who may simply be registering a ‘protest’ of some kind against the questionnaire. The first three of the reasons identified (which account for about a third of the sample) constitute a genuine unwillingness to pay. The most common of these reasons is inability to afford the contributions. The last three of the reasons identified (which account for about two-thirds of the sample) are more indicative of people expressing hostility towards the scenario proposed. The most common of these reasons is a rejection of the idea of compulsory contributions through taxation. Reassuringly, only a very small proportion of respondents (3.4 per cent–6.9 per cent) protested on the grounds that they found the questionnaire unconvincing.
2.4
DISCUSSION AND CONCLUSIONS
The discussion thus far has presented results from a wide range of methods for assessing WTP to preserve charitable organizations in the UK. The purpose of this section is to bring together the main results in a way that permits comparisons to be made and thereby to assess which results should be used for the purposes of valuing the sector. Table 2.18 summarizes the results obtained for the valuation of all charities and the housing and homelessness subsector, on the basis of four different methods incorporated in the survey: double-bounded dichotomous choice, payment ladder, top-down allocation and contingent ranking. Inspection of the table reveals a high degree of consistency in the results obtained from the various permutations of CV: the double-bounded model and the payment ladder. These produce results in the range £47–£49 per person per year for all charities, and £34–£41 per person per year for housing and homelessness charities. The results obtained from the CR are not significantly different from those obtained from the CV in the case of the housing and homelessness charities. However, the CR gives a value for all charities which is about four times as large as that obtained from CV. Furthermore, the confidence intervals on the CR estimates are substantially wider than those for the CV estimates, sug-
62
Measuring the economic value of charities
Table 2.18 A comparison of WTP estimates across models (£ per person per year)
Elicitation method
All charities
Housing and homelessness charities
Dichotomous choice (with ‘don’t know’ = ‘no’)
£46.90 (£31.38–£62.41)
£33.80 (£21.81–£45.79)
Payment ladder
£49.21 (£40.00–£58.41)
£41.22 (£33.56–£48.88)
Top-down allocation (from payment ladder)
£58.78 (£47.4–£69.8)
£13.10 (£10.5–£15.7)
Contingent ranking
£195.52 (£132.00–£259.04)
£35.75 (–£4.21–£75.71)
Actual donations
£42.98 (£34.90–£51.10)
£5.56 (£4.50–£6.60)
Note:
confidence intervals in parentheses.
gesting a higher degree of uncertainty among respondents faced with the former valuation method. The above results suggest that the divergence between the CV and CR results is not so much attributable to the differences between the methods as to the fact that in the latter survey respondents were taken through a bottomup route to the valuation of all charities (that is to say they were presented with a series of scenarios that focused only on the preservation of a single subsector before they were presented with a scenario that preserved the whole sector). This explains the very close correspondence of results with the CV question, where respondents focused only on preserving the housing and homelessness sector. However, it is in contrast to the CV, where respondents were asked to focus on their WTP to preserve all sectors from the outset. This latter approach produces the more credible and conservative estimate. Lastly, there is a large divergence between top-down and bottom-up approaches to valuation. This is evident when the results obtained from CVHH (about £35) are compared with the results obtained from allocating a share of the result obtained from CVALL to the housing and homelessness subsector (about £13). The way to interpret this divergence is that people would be prepared to pay a lot more to prevent the closure of the housing and homelessness sector (if this were the only sector threatened with closure) than they would be prepared to pay if all sectors were threatened with closure.
Benefits of charities to the public
63
For policy purposes, the most conservative estimates of the value of the charitable sector in the UK, over and above current donations, are therefore £47 per person per year for the whole philanthropic sector and £13 per person per year for the housing and homelessness subsector. These results have important practical implications. Since CR and CV valuation approaches were found to be perceived differently by respondents, in each particular circumstance, care must be exercised to choose the most appropriate methodology. The research findings suggest a number of relevant guidelines for practitioners. When the policy objective is to evaluate a set of changes or policies, then the suggested approach is direct CV on the inclusive good. CR is likely to produce a series of single policy evaluations that, if summed up, may seriously overestimate the value of the whole set. When the question of interest involves evaluating a single isolated change or policy, then both direct CV and indirect CR approaches could be used. CR has the added advantage of being able to produce values for several of these changes simultaneously. Finally, if the aim is to uncover the value of a good or policy that is embedded in a more inclusive good or policy, whose other components may also be expected to vary, then the top-down approach seems to have considerable advantages and to produce more robust results. Table 2.18 also includes an estimate of the survey respondents’ actual level of donations. Inspection of the table shows that even the lowest estimates of WTP (over and above current charitable giving) obtained from the valuation survey are substantially greater than current levels of annual donations. This is an important result and is suggestive of the extent of free-riding that exists with voluntary donations to charities. Taking into account the most conservative WTP estimates, both for all charities and for the housing and homelessness subsector, respondents state that they are willing to pay more than double their current level of annual donations in order to prevent the closure (about £47 on top of the current £43 per person per year for all charities and about £13 on top of the current £6 per person per year for housing and homelessness charities). As most stated preference surveys, our study collected various types of evidence, not just monetary values, as to the relative importance that people place on each of the charitable subsectors considered. There are four ways in which this can be measured: first, by looking at attitudinal responses: at the beginning of the questionnaire respondents were asked to rank the four charitable subsectors according to their relative importance; second, using the CV responses: at the end of the CV section, respondents were asked to allocate their total WTP to preserve all charities among the four different charitable subsectors; third, using the CR responses: during the CR exercise, respondents implicitly revealed the relative importance they attached to each charitable subsector by the way in which they ranked alternatives which preserved different areas of charitable activity; and lastly, by analysing actual
64
Measuring the economic value of charities
Table 2.19 Relative importance of different charitable subsectors Attitudinal Score
CV allocation CR weighting
Current donations
Rank
Score
Rank
Score
Rank
Score
Rank
Health and medical 1.77 research (0.03)
1st
37.2 (1.44)
1st
0.68 (0.11)
1st
2.36 (0.34)
1st
Social services
2.29 (0.03)
2nd
26.4 (1.09)
2nd
0.45 (0.09)
2nd
1.14 (0.13)
3rd
Housing and homelessness Overseas aid, environment and culture
2.30 (0.03) 3.56 (0.03)
3rd
23.7 (0.99) 17.4 (0.84)
3rd
0.16 (0.09) 0.30 (0.10)
4th
0.7 (0.10) 1.49 (0.44)
4th
4th
4th
3rd
2nd
donations data: during the first section of the questionnaire, information was collected about the amounts of money that people were currently giving to each of the different charitable subsectors. Table 2.19 summarizes this information. A number of interesting points emerge. Although all four methods put health and medical research charities as the sector to which people attach the greatest importance, there is significant disparity in the relative rankings awarded to the remaining three sectors depending on the valuation method used. The results obtained from the CR method provide the closest match with the pattern which shows up in actual donations. In common with actual donations, the CR method puts the health and medical research sector first, the housing and homelessness sector last and only differs on the relative ranking of the social services, and overseas aid, environment and culture charities. However, in both cases the confidence intervals for these two intermediate sectors overlap, suggesting that it is more accurate to think of these sectors as tying rank under both methods. Interestingly, the CV method gives exactly the same ranking as that which emerges from the attitudinal questions. This makes sense inasmuch as both approaches require respondents to think explicitly about the relative importance of the sectors. This is in contrast to the other two methods, where the rankings emerge implicitly from the decisions which people make when faced with an opportunity to give. With a view to estimating the total benefits provided by the charitable sector in the UK, further analysis of the results presented in this chapter will be undertaken in Chapter 5, incorporating information on the value attached by users to the housing and homelessness charities (Chapter 3) and a discussion of the value of volunteering (Chapter 4).
Benefits of charities to the public
65
STATISTICAL APPENDIX This appendix briefly presents an overview of the theoretical and statistical framework used to obtain WTP welfare measures from the experimental survey design used in this chapter. We will examine, in turn, the CV and the CR approaches. Contingent Valuation Approach There are basically two types of data from CV studies: data from open-ended or payment ladder questions and data from dichotomous choice questions. The open-ended format directly elicits an individual’s maximum WTP while the dichotomous choice format presents a monetary amount – the bid level – or a sequence of amounts to the respondent and asks for a ‘yes’ or ‘no’ vote on the WTP for each bid. Dichotomous choice questions do not elicit individual maximum WTP directly but intervals where it lies. Therefore, the only way to obtain a mean WTP value from the data is through a statistical model linking the money amounts offered to people’s responses. Several variations of both these question formats were used to elicit individual WTP for avoiding closure of the charitable sector in the UK. Accordingly, different econometric specifications were tested to analyse the results from the various types of questions. This section presents a summary of the theoretical and statistical models that were used to analyse the CV survey data. Double-bounded dichotomous choice elicitation Since the mid-1980s, dichotomous choice elicitation procedures have become the most popular way of obtaining information about individual WTP for environmental resources and services in CV studies. The method stems from the seminal work of Bishop and Heberlein (1979) and was subsequently endorsed by NOAA’s Blue Ribbon Panel in 1993 (Arrow et al., 1993). As noted above, the discrete choice format consists of take-it-orleave-it questions: respondents are asked whether or not they are willing to pay a certain amount of money (varied across subsamples) for a change in the good or service in question. Using the observable yes/no answers to the payment question it is possible to infer the distribution of the underlying (unobservable) WTP and form a statistic of interest such as the mean or median WTP. The models that can be used to infer people’s maximum WTP from binary choice questions are qualitative response models and survival models. The main advantage of using discrete response formats is the fact that it may be easier for respondents to answer ‘yes’ or ‘no’ to a specific bid level
66
Measuring the economic value of charities
than to come up with a value for the change of interest. Under certain circumstances it is also an incentive-compatible format (Carson et al., 1997). Related disadvantages of the dichotomous choice method are as follows: the answers are less informative than open-ended alternatives since they give only a discrete indication of a person’s WTP; it is statistically more burdensome to analyse qualitative answers than numerical ones and it is necessary to specify a parametric distribution of the WTP or, equivalently, of the indirect utility function to estimate the mean WTP; the choice of bids must be done carefully in order to span the range of respondents’ true valuations; and these formats are found systematically to result in higher mean WTP amounts than open-ended questions, arguably because of a so-called ‘yea-saying’ bias. This chapter makes use of the double-bounded variant of the dichotomous choice approach (Hanemann et al., 1991), where respondents are faced with two valuation questions. The first question asks whether respondents are prepared to pay a fixed sum of money for some improvement (the bid level varying across different subsamples). Subsequently, they are faced with a follow-up payment question that depends on the response to the first bid level: if the respondent accepted the initial bid they are asked another WTP question with a higher bid; if the answer to the first bid level is ‘no’, then the respondent is presented with a lower bid (with the bid amounts varying across subsamples). In order to estimate a monetary welfare measure from dichotomous choice data it is necessary to employ some microeconomic model of choice. The random utility model approach, suggested by Hanemann (1984), provides the theoretical choice framework (an alternative approach was developed by Cameron, 1998). Assume that individuals have indirect utility functions of the form: V = U (P, Y , X, C 0 )
(2.1)
where P is a vector of prices, Y is income, X a vector of individual characteristics and C0 a vector of public goods, such as the services provided by the charitable sector. Hanemann (1984) explains the individual yes/no answer to the WTP question in terms of a random utility model. Since some of the components of the utility function are not observed by the researcher, equation (2.1) can be rewritten as: V = U (P, Y , X, C 0 ) + ε
(2.2)
where again P is a vector of prices, Y is income, X a vector of individual characteristics, C0 a vector of public goods, including charitable services, and ε is the error term reflecting unobserved taste components.
Benefits of charities to the public
67
In this new random framework, consider a possible shutdown of the charitable sector that reduces the vector of public goods from C0 to C1. Then, C1 < C0 and: U (P, Y , X, C 0 ) + ε 0 > U (P, Y , X, C1 ) + ε1
(2.3)
In a dichotomous choice context, the individual is faced with the choice of saying ‘yes’ or ‘no’ to avoid the welfare decrease arising from the charities’ shutdown at a cost of B. The bid level B is varied across different subsamples. The respondent accepts to pay this price if: U (P, Y − B, X, C 0 ) + ε 0 ≥ U (P, Y , X, C1 ) + ε1
(2.4)
U (P, Y − B, X, C 0 ) − U (P, Y , X, C1 ) ≥ ε1 − ε 0
(2.5)
∆U ≥ η
(2.6)
where ∆U is the utility difference U0 – U1 and η = ε1 – ε0. Hence the probability of accepting the price is given by: P( ‘yes’) = P( ∆, U ≥ η) = P( η ≤ ∆U) = Fη ( ∆U )
(2.7)
where Fη is the cumulative distribution function (cdf) of η, whose functional form depends on the distribution chosen by the researcher. Typical cdfs include the standard normal, logistic, log-normal, log-logistic and Weibull distributions. Hanemann and Kanninen (1996) provide an extensive overview. Since we are dealing with binary choices (a ‘yes’ or ‘no’ answer to a particular bid level) the data can be analysed in the framework of parametric probability models. These models can be estimated by maximum likelihood techniques (Greene, 1997). Assuming a linear-in-income utility specification, the log-likelihood function for the double-bounded dichotomous choice model is described by equation (2.8): N
log L = ∑ {yyi log[ Fη ( a − bHBi )] + yni log[ Fn ( a − bIBi ) i =1
− Fη ( a − bHBi )] + nyi log[ Fη ( a − bLBi ) − Fη ( a − bIBi )] + nni log{1 − Fη ( a − bLBi )]}
(2.8)
where yy, yn, ny and nn are dummy variables corresponding to the four possible response pairs (yes/yes, yes/no, no/yes and no/no); LB, IB and HB
68
Measuring the economic value of charities
correspond to the lower, initial and high bid levels, respectively; B is the bid level offered to the individual; a and b the regression coefficients; and Fη is the cdf assumed by the researcher. Once the unknown parameters in (2.8) have been estimated, the mean WTP from dichotomous choice data can be determined. Hanemann (1984) and Hanemann and Kaninnen (1996) show that, for the linear utility model referred to above, and on the assumption of a zero mean symmetrically distributed error term, mean WTP can be calculated as a ratio of coefficients as given by expression (2.9). This welfare measure corresponds to a Hicksian equivalent variation. WTP = − a / b
(2.9)
Although the asymptotic distribution of the maximum likelihood estimator for the parameters a and b is known, the asymptotic distribution of the maximum likelihood estimator of the welfare measure is not, since mean WTP is a non-linear function of the parameter vector. One way of obtaining confidence intervals for this measure is by means of the procedure developed by Krinsky and Robb (1986). This technique simulates the asymptotic distribution of the coefficients by taking repeated random draws from the multivariate normal distribution defined by the coefficient estimates and their associated covariance matrix. These are used to generate an empirical distribution for the welfare measure, and the associated confidence intervals can then be computed. The previous analysis can be readily generalized in the presence of more explanatory covariates. However, the statistics that are usually of interest to the researcher, the population (unconditional) mean or median, can be as easily estimated by the marginal methods described above as by a conditional approach that first estimates conditional mean WTP as a function of covariates and then finds its average with respect to an estimate of the density of the covariates (McFadden, 1994). Payment ladder elicitation The charities survey also included payment ladder WTP questions as followups to the dichotomous choice elicitation mechanism. These questions consisted of presenting respondents with a payment card and asking them to identify with a tick the amounts they would be willing to pay to avoid the closure of the charitable sector and crossing the amounts they were sure they would not pay. The main advantage of these data is that they are statistically easy to manipulate, providing more information than dichotomous choice questions, that is, exact WTP values and not WTP intervals, so that very few assumptions are needed to estimate the mean WTP.
Benefits of charities to the public
69
A possible disadvantage of payment cards is that respondents may find it more difficult to come up with an answer. Assigning meaningful monetary values to sometimes complex environmental changes without some kind of external assistance can be a very difficult task to complete in the short time the CV interview lasts. A common way of explaining estimated WTP values from payment card data is by modelling them as a function of possible explanatory factors. Equation (2.10) illustrates what has been called in the contingent valuation literature ‘valuation function’ or ‘bid curves’: WTPi = f (X i )
(2.10)
where WTPi is the reported WTP for individual i and Xi a vector of explanatory variables thought to influence individual valuation. These regressors may be socioeconomic characteristics of interest (such as education and income) or variables reflecting general attitudes towards the charitable sector, for example. The most common specification of the valuation function assumes a linear relationship between the regressors: WTPi = β‘ X i + ε I ε i ~ N (0, σ 2 )
(2.11)
where β is a vector of unknown parameters reflecting the impact of changes in a given explanatory variable on WTP estimates and εi is a random error term reflecting factors affecting utility that the researcher is unable to observe. εi is assumed to be normally distributed with zero mean and constant variance. Valuation functions using WTP estimates resulting from open-ended or payment ladder questions may be modelled by a simple classical ordinary least squares (OLS) regression. The results from the payment ladder elicitation procedure reported in this chapter were modelled using this procedure. Note that these data can also be modelled using interval data procedures as maximum WTP can be interpreted to be an unobserved amount lying in between the highest amount ticked in the payment ladder and the next amount up. Contingent Ranking Approach The CR method (Beggs et al., 1981) is part of a number of stated preference techniques originally designed by marketing practitioners to isolate the value of individual product characteristics or attributes typically supplied in combina-
70
Measuring the economic value of charities
tion with one another. These techniques, also survey-based, provide a natural way of analysing multidimensionality of goods and services and situations where trade-offs between product attributes are of particular interest. Stated preference techniques such as contingent ranking are gaining increased popularity among valuation practitioners (Laureau and Rae, 1985; Adamowicz et al., 1994; Johnson and Desvousges, 1997; Foster and Mourato, 2000; Atkinson et al., 2000). As noted by Hanley et al. (1998), these methods share with the dichotomous choice approach a common theoretical framework in the random utility model as well as a common basis of empirical analysis in limited dependent variable econometrics. Their presumed advantage lies in their ability to identify part-worths for different components of the change in question. Other benefits which have been claimed for these methods are the avoidance of the anchoring problem associated with dichotomous choice (Adamowicz, 1995) and of the part–whole bias problem which arises in CV more generally (Hanley et al., 1998). On the other hand, these techniques tend to impose a considerable cognitive burden upon respondents, increasing the likelihood of unreliable and even inconsistent choices. In a CR experiment, respondents are asked to rank a number of alternatives, each one consisting of a combination of attributes and prices, set at varying levels. From the ordinal rankings, the monetary welfare change associated with each attribute can be indirectly calculated. In the exercise presented in this chapter, respondents were asked to rank according to their preferences three hypothetical options implying closure of one or more charitable subsectors, each different option implying a different tax burden. The attributes of the choice were therefore the four charitable subsectors of interest (described in the chapter) and the tax level. Each attribute was presented at different levels: the charitable sectors could either be ‘open’ or ‘closed’, while seven different tax levels were considered. As in the dichotomous choice approach, the random utility model provides the economic theory framework for analysing the data from the CR exercise. According to this framework, respondents will select the option that maximizes their utility or satisfaction. Since the researcher does not observe all the determinants of individual choice, the utility function for each respondent i can be decomposed into two parts: a deterministic element, which is a linear index of the attributes (X) of the j different alternative options in the choice set; and a stochastic element (ε) which represents unobservable influences on individual choice. This specification is shown in equation (2.12). Vij = bXij + ε ij
(2.12)
Under the assumption of an independently and identically distributed random error (εij) with a Weibull distribution, Beggs et al. (1981) developed a
Benefits of charities to the public
71
rank-order logit model capable of using all the information contained in a survey where alternatives are fully ranked by respondents. It can be shown that the probability of any particular ranking of options for the charitable sector being made by individual i can be expressed as:
exp(bXij ) Pi (Vi1 > Vi 2 > Vi 3 ) = ∏ 3 j =1 ∑ exp(bXik ) k = j 2
(2.13)
The parameters of the utility function can be estimated by maximizing the log-likelihood function given in equation (2.14):
exp(bXij ) log L = ∑ log∏ 3 j =1 i =1 ∑ exp(bXik ) k = j N
2
(2.14)
After the parameters in (2.12) have been estimated by maximizing the loglikelihood function given in (2.14), the welfare measures of interest, that is, the WTP to avoid the closure of one or more sectors of charitable activity, can be readily calculated as the marginal rate of substitution between the relevant charitable subsector and the tax level (a ratio of model coefficients), as indicated in equation (2.15): ∂U j ∂Charitable sectorij WTP = − ∂U j ∂Taxij
(2.15)
Confidence intervals for the welfare measure can be calculated using the Krinsky and Robb (1986) procedure described above.
3. The benefits of charities to users: the homeless 3.1
INTRODUCTION
According to the broad definition of homelessness, a homeless person is anyone living in precarious, insecure or short-term accommodation, that is, in hostels, hotels, bed and breakfast (B&B), squatting, sleeping rough and hidden homelessness (those who sleep around friends and family). Although accurate statistical information is difficult to obtain, it is estimated that there were around 140 000 single homeless people in the UK in 1994 (personal communication, Shelter, 1997). Of those, around 270 were sleeping rough every night. Hostels are a necessary first step in the resettlement process of most homeless people. In London alone, there are about 26 000 hostel bed spaces in over 600 buildings (Resource Information Service, 1996). A common view of hostels is that they are large buildings, offering poor accommodation in dormitories, with regimes dominated by strict rules and regulations. However, while there are still some hostels that conform to this Victorian image, typically the reality is quite different. Today nearly 20 different types of hostels exist (Resource Information Service, 1996). The term ‘housing project’ is increasingly being used to describe many of the more recently established schemes run by the charitable sector. Apart from accommodation and food, hostels or housing projects also offer a range of support and counselling services that are needed by a large proportion of homeless people for problems related to drink, mental health, drugs or physical health, among others. This chapter summarizes a study which aimed to place an economic value on the services provided by hostels, that is, the housing and homelessness charities, to their direct beneficiaries, that is, the homeless people who use hostel services. This study was carried out as part of a wider project on valuing the output of the charitable sector in the UK. A parallel study was undertaken to estimate the value for society at large, that is, indirect users and non-users of the housing and homelessness subsector of charitable activity. This is discussed in Chapter 2. Together, the two studies permit a complete evaluation of the housing and homelessness charities, taking into account all stakeholders involved. 72
Benefits of charities to users
73
The price of hostel accommodation does not reflect the value of the services provided to users, as in the case of other private goods, first of all, because the current price of hostel accommodation is subsidized and thus does not reflect an equilibrium price between supply and demand, and second, because some of the support services provided by hostels, over and above food and accommodation (counselling, medical help, financial advice and so on) are probably not easily found or affordable by the homeless, outside the hostel circuit. Hence, for the target group of hostel residents, the benefits are expected to exceed the price currently paid for hostel services. In previous work, the value of this sector has been assessed by looking at the cost side, that is, the income received or the expenditures made by charitable organizations (Jurgen, 1988; Jencks, 1994; O’Flaherty, 1996). However, the income received by these charities through public donations of time and money and government grants is not the correct value of the services they provide. This is because: (i) given the voluntary nature of charitable giving, people tend to free-ride and to take for granted that donations by others will make up for their own lack of generosity and hence individual giving is suboptimal; (ii) government grants are not calculated in a manner that would compensate for the existence of free-riding behaviour. They are determined by political interests and lobbying forces and are not a result of an efficient choice. To overcome these problems, the approach introduced in Chapters 1 and 2 was used: in order to estimate the value of the housing and homelessness charities’ output we looked directly at the benefits they provide. This technique is known as contingent valuation (Mitchell and Carson, 1989). This chapter reports the results of a contingent valuation (CV) survey of hostel residents designed to evaluate the user benefits of hostels. To our knowledge, it is the first time ever that a minority of the population, living in precarious conditions and with a range of special support needs, has been administered an economic valuation survey. In this framework, to make respondents consider how much the sector was worth to them, the hypothetical scenario presented in the questionnaire was the following: due to a financial crisis all the hostels in the country were facing the prospect of shutting down for a whole year, leaving many people with no alternative but to sleep rough on the streets. A sample of homeless people were then asked how much compensation they would need to compensate them for the loss of hostel services over and above any state benefits they may already receive. This willingness to accept compensation (WTA) approach has also been applied in the environmental economics literature (for example Adamowicz et al., 1993; Shyamsundar and Kramer, 1996; Smith et al., 1997). While, in
74
Measuring the economic value of charities
theory, it should produce comparable results to the WTP approach (Randall and Stoll, 1980; Willig, 1976), in practice it has tended to give substantially larger values (Coursey et al., 1987; Cummings et al., 1986). Explanations for this phenomenon include income effects (Hanemann, 1991), loss aversion (Kahneman and Tversky, 1979) and strategic behaviour induced by the absence of a budget constraint. For all of these reasons, particular care must be taken in using a compensation-based measure. The chapter is organized as follows. Section 3.2 describes the survey design and Section 3.3 presents the survey results. Conclusions are presented in Section 3.4.
3.2
SURVEY DESIGN
The main objective of the CV survey was to elicit direct user benefits from housing and homelessness charities. Hostels dominate this subsector of charitable activity and its users are homeless people. Hence the CV interviews took place in hostels or associated day centres in London with a random sample of homeless people. The survey collected information about the reasons that led respondents to become homeless, their current use of hostel accommodation, the quality of the services provided by hostels and possible substitutes for hostel accommodation. Respondents were also asked about their current level of income and to estimate how much of that income was spent on accommodation and other expenditure categories. The valuation sections of the survey aimed to uncover, in money terms, the benefits received by the direct recipients of hostel services. For that purpose, as described above, respondents were faced with a hypothetical scenario – where all hostels would close down for a year – that was intended to make them consider carefully how much accommodation and other hostel services were worth to them. The wording was as follows: Hostels provide you with accommodation and a number of other support services. We would like to find out how much you value all of this. We will use the following question to try and find out how much the accommodation and services are worth to you. Imagine that, for some reason, all the hostels in the country had to shut down for a whole year and you had to find somewhere to stay. (Don’t worry, this is definitely not going to happen! But we would like you to think about what your life would be like if it did happen.) Suppose in order to compensate you for not being able to use the hostels anymore, you were given an additional cash payment each week, over and above what you said you receive at the moment.
Benefits of charities to users
75
The shutdown had to be on a national scale so as to eliminate the possibility of migrating to find alternative hostel accommodation elsewhere. It also had to last for a considerable period of time to oblige respondents to think about a formal alternative as opposed to just sleeping rough for a few weeks. However, arguably, the very dramatic nature of the shutdown described in the questionnaire could have undermined its credibility or caused undue anxiety to the hostel residents. For these reasons, the presentation encouraged respondents to regard the scenario as a thought experiment rather than as something that might actually happen. The WTA question was worded as follows: How much money would you have to receive each week during the whole year to give you the same quality of life as you have now (not better or worse, but just the same)?
The estimated WTA welfare measure is a compensating variation; further details of the theoretical framework behind welfare measures can be found in the Statistical Appendix to Chapter 2. After extensive pre-testing it was found that respondents seemed to be most comfortable with a simple open-ended elicitation framework, where they were directly asked for their required compensation amount. Open-ended WTA procedures may be problematic and result in respondents asking for unrealistically high amounts. This possibility was taken into account in the survey design and, to try to minimize the occurrence of overstated compensatory amounts, respondents were explicitly asked for subsidies that maintained (and not improved or reduced) their quality of life. To confirm that this was the case, two screening mechanisms were adopted. First, interviewers were instructed to repeat the valuation question to every respondent who asked for more than £500 per week, stressing that they should think of an amount that would afford them the same quality of life they have now, not a better one. Second, the sampled homeless were asked how they would spend the compensatory amount stated (accommodation, food and so on), were they to actually receive it. Then they were asked the following debriefing question: Taking into account what you have just said about how you would spend the money, do you think that your quality of life would be the same, better or worse than what you have at the moment?
If the answer was better or worse, the WTA question was repeated and respondents were urged to state a compensation amount that would leave them as well off as they were at the current moment. The valuation section concluded with a follow-up question for respondents who did not answer the WTA question, to try to uncover the motives behind the refusal.
76
Measuring the economic value of charities
The final part of the questionnaire collected basic socioeconomic information about the respondents, including gender, age, marital status, educational attainment, employment status and nationality. All interviews were conducted in person, on site at the housing establishments, and took about 20 minutes to complete. Due to the difficult nature of the work, the fieldwork was assigned to a firm specialized in conducting surveys of homeless people. The interviewers were extensively involved in the development of the questionnaire, and carefully briefed on the nature of the research methodology. After the questionnaire had been piloted in the field, the main survey began in June 1997 and ran over a period of six months. The sample was designed to cover the full range of accommodation options available to homeless people, which are extremely diverse. Table 3.1 identifies some of the main kinds of institutions and the types of services that they offer. From this range of alternatives, ten establishments were initially selected from the London Hostels Directory (Resource Information Service, 1996). Within each hostel, interviewers were instructed to select at random the specified quota from the register of current residents. In practice, it proved impossible to implement this sampling strategy. A number of the hostels in the original sample were unwilling to collaborate with the research so that alternative institutions had to be identified at short notice. Furthermore, some of the selected residents in self-contained accommodation refused to open their doors to the interviewers so that alternative respondents had to be found. As a result of these difficulties, out of the 200 interviews originally scheduled, only 150 were completed. Of these, 119 provided usable responses. In spite of the logistical problems and the lack of cooperation from hostels, the interviews that took place seemed to work well. A substantial majority of respondents (58 per cent) stated that they liked the questionnaire and found it interesting. Only 5 per cent of the sample considered the proposed idea to be unrealistic and fewer than 10 per cent found the questionnaire difficult to understand. These results indicate that the initial designing and pre-test phases were successful in producing an effective survey instrument that was credible and understandable. The characteristics of the participating 14 hostels are summarized in Table 3.2. These establishments account for 911 out of a total of some 26 200 beds in London. Although a wide coverage of the different types of institutions was accomplished, the final distribution cannot be regarded as fully representative. In particular, there is an under-representation of supportive projects and an over-representation of emergency night shelters. This problem was taken into account when presenting the valuation results.
Benefits of charities to users
Table 3.1
77
Definition of services provided by different types of hostels
Emergency night shelters
Low-support hostels
Semi-supportive projects
Housing schemes
Supportive projects
Traditional hostels
Specific needs projects
Provide direct access to short-term accommodation, on a nightly basis, for people with no alternative. Only basic accommodation and support are available. Patrons are expected to be out during the day. Offer accommodation to people in housing need, but with only very limited support. Many of these schemes are quite large and resettlement advice is not necessarily a main focus of their work. In many cases, the projects are staffed only by resident wardens. Aimed at those who want to live fairly independently, with some practical or personal support available. The main emphasis is usually on finding permanent housing and preparing residents for independent living. Most do not offer 24-hour staff cover. Organizations running a number of flats, bedsits or shared houses offering good-quality accommodation and little more support than help with benefit claims. Single-room, self-catering accommodation. Most accommodation is either intended as permanent or otherwise as a precursor to permanent accommodation. Offer medium- to long-term accommodation and place considerable emphasis on counselling, education and training with the aim of preparing residents for independent living. Most offer individual support (through a key worker system) and 24-hour staff cover. Large, long-established hostels, which are traditionally used by long-term homeless people. In some cases, accommodation is still in large domitories. Cater for specific individual situations such as ex-offender, drug addict, single mother. The type of accommodation and the level of support vary considerably.
78
25 29 _ _ _ 35 29 57 42 30
M/F F M/F M/F M/F M/F/C F M F M
£16
£10–£25 £24 £10
£6
£54 £5–£55
£7 £18
£20
£20 0 £6 0
Cost
D/F/H/O/R
B/F/H/O/R B/F/H/R B/F/H/R
B/F/H/O/R
B/O/R B/O/R
A/B/D/E/H/R E/F/H/O/R
B/E/F/H/O/R
D/F/H/O B/F/H/R A/B/D/E/F/H/R F/H/R
Support
6
16 18 10
4
1 9
5 12
30
15 10 6 8
N
33
60 166 54
22
74 31
12 82
163
27 27 110 50
Total
Notes: M=male; F=female; C=couples; Age: average age of hostel residents; Cost: approximate average cost required from residents in receipt of benefits, in £ per week; Support: A=alcohol; B=benefits; D=drugs; F=food; E=employment; H=health; O=referral or liaison with other agencies for additional support; R=resettlement; N: number of interviews conducted; Total: total number of bed spaces available.
18 19 35 42
Age
M/F M/F M M/F
Sex
List and characteristics of selected hostels
Emergency night shelters Centrepoint Shelter (Berwick Street) Lord Clyde Nightshelter St Mungo – Cedars Road Westminster Cathedral Nightshelter (Passage Day Centre) Low-support hostels Look Ahead – Aldgate Hostel Semi-supportive projects Opendoor – De Laune Street Providence Row – Daniel Gilbert House Housing schemes Single Homeless Project (Camden – Kings Cross Road) Single Homeless Project (Westminster) Supportive projects Thames Reach – Stamford St Hostel Traditional hostels English Churches – Queen Mary’s Hostel Salvation Army – Booth House Salvation Army – Hopetown Ex-offenders projects Opendoor – Warwick Road
Hostel
Table 3.2
Benefits of charities to users
3.3
79
THE SURVEY RESULTS
3.3.1 Socioeconomic Characteristics Table 3.3 provides a socioeconomic profile of the hostel residents interviewed. Both male and females were interviewed. The age range covered was wide, Table 3.3
Summary statistics of selected socioeconomic variables
Total number of individuals Demographic variables Males (%) Age (mean in years) Less than 18 years old (%) 18–25 years old (%) 26–39 years old (%) 40–59 years old (%) 60 years old or above (%) Single (%) Divorced/separated (%) UK citizen (%) Education Primary (%) Secondary without O levels/GCSEs (%) O levels/GCSEs (%) University (%) Age completed education (mean in years) Employment Looking for work (%) Sick/disabled (%) Student (%) Economic variables Income non-response (%) Weekly income: mean (s.e.) median range Expenditure non-response (%) Weekly expenditure: mean (s.e.) median range Note:
s.e.: standard error
150
65 36 6 23 33 28 10 69 21 81 6 41 31 14 17 44 19 11 5 £59.6 (£3.3) £48.2 £19.5–£300 9 £54.6 (£3.1) £47 £9–£300
80
Measuring the economic value of charities
varying from 16 to 88 years, with an average of 36. As expected, since few hostels cater for families, most respondents lived alone, being either single or divorced/separated (90 per cent). Nearly 20 per cent were foreigners, a percentage much higher than in the overall population. This substantiates the idea that many emigrants arrive in the UK in precarious conditions. Nearly 50 per cent of the sample did not complete their O levels/GCSEs but a surprising 14 per cent claimed to have university frequency. This could be explained by the fact that 11 per cent of the sample were students. The age at which full-time education was completed broadly ranged from 5 to 36, with an average of 17 years of age. However, this variable does not necessarily give the right picture in terms of educational attainment, as respondents may either have taken longer to complete lower levels of education or have had periods of interruption in their studies – the fact that 12 per cent reported having finished their education between the ages of 21 and 36 suggests that this is the case. It is interesting to note that only 5 per cent of respondents refused to reveal their income. This is substantially lower than typical income non-response rates for general population surveys, which are often as high as 25 per cent to 50 per cent. The average weekly income was around £60, consistent with the magnitude of the state benefits that constitute the major source of income of hostel users. As illustrated in Figure 3.1, only 6 per cent of the sample were receiving income from employment, with the remainder being almost solely reliant on government benefits (income support and job seeker’s allowance alone constitute a source of income for 70 per cent of the sample). Respondents were also asked to break down their weekly expenditure into a number of categories: the sum of all expenditures provided an alternative
Disability allowance 11%
Other benefits 13%
Income support 40% Note:
multiple sources of income are possible.
Figure 3.1
Sources of income
Begging 3% Wages 6%
Job seeker’s allowance 30%
Benefits of charities to users
Transport 8%
Personal items 9%
Recreation 4%
81
Others 8% Alcohol/drugs/ tobacco 20%
Hostel 30% Food 21% Figure 3.2
Current expenditure breakdown
measure of income. Some respondents had difficulty in recalling specific purchases and so the non-response rate is slightly higher than for income, at 9 per cent (Table 3.3). However, reassuringly, the average reported total expenditure per week (£55) is not statistically different from the reported income (£60). Figure 3.2 presents an overview of the expenditure allocation pattern. The largest proportion of respondent’s income (30 per cent) is allocated to
Income
70
Expenditure
Percentage of sample
60 50 40 30 20 10
Figure 3.3
>£400
£351–400
£301–350
£251–300
£201–250
£176–200
£151–175
£126–150
£101–125
£76–100
£51–75
£26–50
<£25
0
Percentage frequency distribution of current weekly income and expenditure
82
Measuring the economic value of charities
hostel accommodation and services (which, in many cases, include some meals). Additional expenditures for food and for alcohol, drugs and/or tobacco account for 21 per cent and 20 per cent of the available income, respectively. Figure 3.3 depicts the percentage distribution of current weekly income and expenditure levels. In both instances, there is a high concentration of cases in the £26–£50 interval. Both distributions match up quite closely. A number of introductory questions aimed to reveal respondents’ housing history. The responses are summarized in Table 3.4. Around 37 per cent of respondents said that the house of their parents (or other relatives) had provided the last place they thought of as a home, while 5 per cent were formerly Table 3.4
Housing history %
Which was the last place you thought of as home? Parents’/foster parents’ home Private rented accommodation Hostel Council or housing association tenancy Relatives’ home Own house/flat/bedsit Friends’ home B&B Sleeping rough Never had a home Special institution What was the main reason that you left home? Unemployment/money/accommodation problems Evicted/asked to leave/disputes with other residents Problems with parents Relationship problems with partner Wanted to leave the family home/to look for work Situation in another country Health reasons Family death Domestic violence/abuse Depression/loneliness Drink/drug problems Left institution
29 17 13 12 8 6 5 4 3 3 1 16 14 12 11 11 9 4 3 2 1 1 1
Benefits of charities to users
83
Time since lost old home
50 45 40 35 30 25 20 15 10 5 0
Time until new home expected
Figure 3.4
Duration of homelessness
>120 mths
61–120 mths
49–60 mths
37–48 mths
25–36 mths
13–24 mths
7–12 mths
Expected total duration of homelessness
<6 mths
Percentage of sample
living with friends (hidden homeless). For a further 29 per cent of the sample, their last home had been accommodation rented either from the private sector or from a local council or housing association, while 17 per cent considered a hostel or a B&B to be the last place thought of as home – which is suggestive of long-term homelessness. Only 6 per cent of respondents had a history of home ownership. For 3 per cent of the sample the street was considered to be their home, while a further 3 per cent declared never to have had a home. The largest single cause of homelessness appeared to be the breakdown of relationships with parents (12 per cent), partners (11 per cent) or neighbours (14 per cent), leading to voluntary departure in some cases or even eviction in others. The next most important cause was work-related, with 16 per cent abandoning homes as a result of financial hardship ensuing from unemployment and a further 11 per cent leaving home in search of a job. As many as 9 per cent of those interviewed were refugees fleeing the situation in another country. Figure 3.4 presents the frequency distribution for the duration of homelessness up until the time of the survey and the expected remaining period before a more permanent home is found. The figure indicates that about half of those interviewed had been homeless for less than a year, while two-thirds expected to have found longer-term accommodation within less than a year (in interpreting these results it is important to note that just over half of the respondents were unable to answer the question about the expected delay before more permanent accommodation would be found). The median ex-
84
Measuring the economic value of charities
pected overall duration of the spell of homelessness was two years, even though there were some homeless people in the sample with average duration of homelessness of almost five years. Seventy-five per cent of the sample expected to be resettled within a four-year period. There was, however, a significant group, representing just under 20 per cent of the sample, who expected to be homeless for more than ten years; that is, for all practical purposes, indefinitely. 3.3.2 Use of Hostel Services, Attitudes and Preferences The CV survey included a number of questions aimed at assessing the level of use of hostel facilities, the degree of satisfaction with these services and the preferred short-term and long-term accommodation alternatives. This section presents the results from these questions. Figures 3.5 and 3.6 show the level of use of general hostel facilities (food, laundry and washing services) and of counselling and support services for special needs (finding a home, money/benefits, training/finding work, general health/mental health problems, drug/alcohol problems). As depicted in Figure 3.5, the vast majority of respondents were making use of the basic services provided, such as meals (87 per cent), washing (99 per cent) and laundry (91 per cent). During the pre-test stages of the questionnaire, respondents repeatedly told interviewers that, although they were in precarious accommodation or sleeping rough, they still endeavoured to maintain their standards of cleanliness. This seemed to be an important issue for many people. Figure 3.6 shows that as many as 74 per cent of respondents used one or other of the counselling facilities available in hostels, which range from
99%
91%
87%
Laundry
Food
Percentage of sample
100 90 80 70 60 50 40 30 20 10 0
Washing Figure 3.5
Percentage of respondents using general hostel facilities
Benefits of charities to users
Figure 3.6
74% 51% 26%
25% 12%
Drug/alcohol problems
Health problems
Training/finding work
Money/benefits
Finding a home
17%
Mental health problems
51%
Total
Percentage of sample
100 90 80 70 60 50 40 30 20 10 0
85
Percentage of respondents using counselling and support services
support for health problems and substance abuse to advice on seeking employment and claiming benefits. As expected, the most popular type of assistance was finding long-term accommodation and claiming benefits, each used by 51 per cent of respondents. A quarter of the sample had support in finding a job and another quarter in dealing with general health problems; a further 17 per cent and 12 per cent sought counselling regarding drug or alcohol addiction and mental health problems, respectively. In general, the level of satisfaction with these services was relatively high. Figure 3.7 indicates that, in what concerns general facilities, only a fifth of the sample or less considered any of the services provided to be poor. Satisfaction was higher for issues such as opening hours and security and safety: 78 per cent and 76 per cent of the sample, respectively, considered them to be good. Service quality was considered relatively less good for meals and washing facilities, with only 34 per cent and 49 per cent considering them to be good, respectively. The most surprising results relate to views held on house rules and access hours. Almost all of the hostels where the interviews took place did not allow alcohol on their premises, and four of the selected hostels had strict opening and access hours, requiring residents to leave their bedrooms by 7 a.m., 8 a.m. or 9 a.m., and to return by 8 p.m., 8.30 p.m. or midnight, depending on
86
Measuring the economic value of charities
Staff attitudes Opening hours Hostel rules Privacy/space Security/safety Laundry Cleanliness Food Washing 0
20
40 60 Percentage of sample Good
Figure 3.7
Average
80
100
Poor
Satisfaction with the quality of general hostel facilities and services
the case. Despite such restrictions, Figure 3.7 shows that staff attitudes, opening hours and hostel rules are considered to be good by at least 60 per cent of the sample. As for those respondents who made use of counselling and support services, Figure 3.8 shows that over 60 per cent rated these services as being good, with the notable exception of help in finding a home, which was considered good by only 47 per cent of the sample. Furthermore, help in finding permanent accommodation and mental health counselling were thought to be of bad quality by about 30 per cent of the sample. Overall, these results indicate that hostel residents not only make use of the range of services that are provided by hostels in addition to bed spaces, but also consider these services to be generally of good quality. This is consistent with earlier literature which has found that 56 per cent of hostel residents were satisfied with the facilities provided (Randall and Brown, 1996). The implication is that support services seem to increase the well-being of their users and that as such their value should be significantly positive. It also suggests that
Benefits of charities to users
87
Training/finding work Money/benefits Finding a home Mental health problems Drug/alcohol problems Mental health 0
20
40 60 Percentage of sample Good
Figure 3.8
Average
80
100
Poor
Satisfaction with the quality of hostel counselling and support services
hostel users might prefer hostel accommodation vis-à-vis private sector equivalents (such as bed and breakfast accommodation), where support services are not available, on the assumption that they could be admitted into private accommodation and pay for it through their housing benefit. As regards accommodation preferences, the survey asked respondents to think about where they would be most likely to go in the short term (if they could not use the hostels for a couple of weeks), in the medium term (if hostels were shut down for an entire year), and in the long term (when their preferred housing choice might become available). The answers to these questions are illustrated in Figure 3.9 and reveal a striking variation in preferences across different time horizons as well as a remarkably high degree of unanimity among respondents. When faced with a sudden short-term shortage of hostel accommodation and no additional money, as many as 60 per cent of respondents would choose to sleep rough rather than seek paid accommodation or stay with family and friends. This finding indicates that private sector accommodation is not a viable alternative for many homeless people, probably due to inadequate financial resources and/or difficulties in claiming benefit. This explanation is reinforced by the fact that, for a further 32 per cent of respondents, staying with family or friends or in a squat seems to be the preferred solution to a short-term shortage of hostel accommodation.
88
Measuring the economic value of charities
Status quo Home ownership Private rental Institution Council B&B Family Friends Squat Sleeping rough 0
20
40 60 Percentage of sample
Long term
Figure 3.9
Medium term
80 Short term
Frequency distribution of housing preferences across different time horizons
On the other hand, when presented with a longer period of exclusion from hostel facilities and provided with the compensation payment described in the survey, some 55 per cent of respondents would opt to live in private rented accommodation. The next most popular alternative in this situation would be a B&B, which was selected by 16 per cent of the sample. This suggests that in the long term the informal arrangements selected for the short term would not be adequate for most residents and the financial compensation paid brings the private sector alternatives within reach. Nevertheless, one-fifth of the sample could not indicate an alternative form of accommodation. For many, a medium-term alternative to hostel accommodation may simply not exist in the private sector or in council estates. However, in the longer term the vast majority of the hostel residents (75 per cent) would prefer to obtain their permanent accommodation from the local council or housing association. The fact that this was not selected in the medium-term scenario suggests either that such accommodation is not available within that particular time horizon or possibly that respondents would
Benefits of charities to users
89
only be able to afford the private sector options with the compensation payment defined in the hypothetical scenario. Overall, these results show that respondents were realistic in assessing their future resettlement prospects over different time periods. 3.3.3 Valuation Results The first interesting finding that arose from the valuation section of the hostel users’ survey was the fact that the majority of respondents did not consider the idea of hostel closure followed by compensation, as presented in the CV hypothetical scenario, to be an attractive one. As illustrated in Figure 3.10, nearly 25 per cent of the sample disliked the idea, while a further 40 per cent actually claimed to have hated it. On the other hand, the figure also shows that about a quarter of respondents did like the idea. Would hate it Would not like it Don’t mind Would like it Would very much like it 0 Figure 3.10
10
20 30 40 Percentage of sample
50
Answers to the question ‘Would you like this idea to actually happen?’
Table 3.5 presents an analysis of the correlation between disliking the proposed scenario and considering the services provided by hostels to be of good quality. This relationship is expected to be positive, as the more an individual appreciates the services provided by hostels, the more likely he/ she is to dislike giving them up. The results indicate that, as expected, all the correlation coefficients are positive although generally not very high in absolute terms. The highest correlations are obtained for those who appreciate the quality of health-related support (0.43), the cleanliness of the premises (0.34), the food (0.29) and hostel rules (0.29). As described above, the WTA valuation question was open-ended, asking hostel residents to state their required compensation to forego hostel accom-
90
Table 3.5
Measuring the economic value of charities
Correlation between attitudes towards the proposed scenario and attitudes towards hostel services
Appreciating the quality of hostel services General facilities Food Washing facilities Laundry Cleanliness Privacy/space Security/safety Staff attitudes Opening hours Hostel rules Counselling services General health Mental health Drugs and alcohol Finding a home Money/benefits Training/finding work
Disliking the proposed scenario
0.29 0.18 0.15 0.34 0.20 0.09 0.14 0.08 0.29 0.43 0.23 0.07 0.22 0.24 0.20
modation for one year and still be as well off as before. Nearly 80 per cent of respondents were able to provide answers to the hypothetical valuation question. Table 3.6 presents the compensatory amounts required by hostel users to give up hostel accommodation and still maintain their initial welfare, while Figure 3.11 depicts the distribution of these WTA amounts. The overall (unweighted) mean WTA was found to be £157 per week (over and above current income), ranging between £7 and £1600, with a median of £125. These values are reasonable overall and indicate that, with few exceptions, respondents took the exercise seriously and did not react strategically to the valuation question by asking for extremely high compensations. Figure 3.11 also shows that the compensation amounts are concentrated on the lefthand side of the distribution with very few outliers (unrealistically high bids), as the small difference between the mean and median WTA already indicated. Table 3.6 also shows that the required compensation is higher for hostels providing more support and counselling services. For example, residents of low-support projects asked for an average compensation of £127 per week
91
Mean
£140.8 £127.4 £162.1 £147.9 £231.0 £210.6 £117.0 £157.4
Emergency night shelters Low-support projects Semi-supportive projects Housing schemes Supportive projects Traditional hostels Ex-offenders projects Total WTA £102.3–£179.3 £106.7–£148.1 £92.1–£232.1 £101–£194.8 £37–£425 £106.2–£315 £86.9–£147.1 £128–£186.8
Confidence interval £107.5 £120.6 £140.0 £140.0 £231.0 £150.0 £111.0 £125.0
Median £20 £20 £50 £50 £132 £7 £70 £7
Minimum
WTA compensation to give up hostel services (£ per person per week)
Hostel type
Table 3.6
£500 £230 £500 £250 £330 £1600 £170 £1600
Maximum
34 26 12 7 2 29 6 119
N
92
Measuring the economic value of charities
Percentage of sample
25 20 15 10 5 >£400
£351–400
£301–350
£251–300
£201–250
£176–200
£151–175
£126–150
£101–125
£76–100
£51–75
£26–50
<£25
0
Weekly WTA compensation per person (in £)
Figure 3.11
Distribution of WTA compensation amounts (in £ per person per week)
while users of traditional hostels and supportive projects required compensations above £200 per week. Aggregating hostels into two broad types, according to the level of support provided, gives the following average result: low support (night shelters, low support, housing schemes, exoffenders): £133 per person per week; and medium and high support (semi-supportive, traditional and supportive projects): £201 per person per week. The difference of £68 could be viewed as a rough lower-limit estimate of the value of the services provided by hostels over and above the provision of a bed space. This is consistent with the findings from the attitudinal part of the questionnaire reported above that suggested that respondents positively valued the range of services provided by hostels. However, due to the small number of interviews conducted in some of the hostels, the standard error of these estimates is quite high and the differences are not statistically significant. A regression analysis was performed on the WTA results to assess their theoretical validity, that is, to see whether a number of relevant explanatory variables manage to explain a significant proportion of the total variation in compensatory estimates in ways that are compatible with economic theory. Table 3.7 presents an overview of the variables that were selected as possible explanatory factors. These include sociological, economic and attitudinal variables and a number of dummy variables representing the type of hostel where the interviews took place.
Benefits of charities to users
Table 3.7
93
Description of explanatory variables
Sociological variables Sex Age Primary education Age education was completed Economic variables Expenditure per person Attitudinal variables Disliked the proposed idea Disliked the questionnaire Likes hostel rules Likes hostel counselling and support services Type of hostel (baseline: emergency nightshelters) Low support hostels Semi-supportive projects Housing schemes Supportive projects Traditional hostels Ex-offenders projects Other variables Looking for a job Would seek private rented accommodation after hostel closure
1 – male; 0 – female Average age in years 1 – if primary is highest level of education attained; 0 – if otherwise Age at which full-time education was completed Weekly average (in £) 1 – if disliked the idea; 0 – if otherwise 1 – if disliked the questionnaire; 0 – if otherwise 1 – if likes hostel rules; 0 – if otherwise 1 – if likes support services; 0 – if otherwise
1 – if low support hostel; 0 – if otherwise 1 – if semi-supportive project; 0 – if otherwise 1 – if housing scheme; 0 – if otherwise 1 – if supportive project; 0 – if otherwise 1 – if traditional hostel; 0 – if otherwise 1 – if ex-offenders project; 0 – if otherwise 1 – if looking for a job; 0 – if otherwise 1 – if seeking rented accommodation; 0 – if otherwise
The results from estimating the valuation function obtained by regressing the WTA amounts against the variables in Table 3.7 are reported in Table 3.8. The statistical method used was ordinary least squares (OLS) with correction for heteroskedasticity (more details about this approach can be found in the Statistical Appendix to Chapter 2). The best fit of the model was obtained by using the logarithm of WTA as the dependent variable. Overall, in spite of the small sample size, the regression manages to explain 36 per cent of the total variation, which is a good result for the type of cross-sectional data being analysed. As expected, weekly expenditure was found to be a strong positive determinant of the desired compensation amounts: the higher the expenditure level, the higher the compensation required. Expenditure is a good proxy for
94
Table 3.8
Measuring the economic value of charities
Valuation function: WTA compensation for hostel closure
Constant Sociological variables Sex Age Primary education Age education was completed Economic variables Expenditure Attitudinal variables Disliked the proposed idea Disliked the questionnaire Likes hostel rules Likes hostel counselling and support services Type of hostel (baseline: emergency night shelters) Low-support hostels Semi-supportive projects Housing schemes Supportive projects Traditional hostels Ex-offenders projects Other variables Looking for a job Would seek private rented accommodation after hostel closure R2 N
Coefficient
t-statistic
3.26
7.74
0.05 0.003 –0.54 –0.03
0.35 0.45 –2.13 –1.82
0.01
5.96
0.12 0.40 0.20 0.03
2.40 2.33 1.68 0.92
0.32 0.18 0.10 0.69 0.47 0.31
1.43 0.88 0.35 2.30 1.98 1.74
0.53 0.38
3.19 2.62
0.36 93
Note: Estimation method: OLS with correction for heteroscedasticity. Dependent variable: log(WTA).
income in this model and actually performs better, in statistical terms, than income itself (whose coefficient was also positive and significant in an alternative specification of the valuation function not reported here). Thus higher levels of income/expenditure are indicative of a higher quality of life and, arguably, of higher opportunity costs from losing hostel accommodation. This argument is intuitive but is also difficult to verify. Weak supporting evidence can be found by looking at the low but positive correlation between
Benefits of charities to users
95
expenditure and the modern housing projects included in the sample (0.12) and the small but negative correlation found with the more traditional hostel types (–0.01). Regarding sociological factors, those respondents who only have primary education are willing to accept less compensation than others with further educational degrees. Again this may be indicative of a lower quality of life that is reflected in lower WTA amounts. Sex and age do not have a statistically significant effect on WTA, and the age at which education is completed is only significant at the 10 per cent level. A number of dummy variables were included to represent the various different types of hostels where the interviews took place. Emergency night shelters were taken as the baseline: this means, for instance, that the coefficient of the supportive projects variable (0.6899) indicates that hostel residents in supportive projects required compensation amounts nearly £0.7 higher than the logarithm of the WTA amount of those using emergency night shelters. The prior expectation is that, the larger the level of support and counselling services offered, the higher the opportunity cost of hostel closure. The results reported in Table 3.8 confirm this expectation: all included hostel categories have positive coefficients, reflecting the fact that emergency hostels typically provide the lowest level of support services and thus are less valued. Supportive and traditional hostels have the highest coefficients, both significant at the 5 per cent level. Finally, regression results also show that respondents actively looking for a job seek higher compensations, as do those who would opt for renting private accommodation after the prospective hostel closure. Once more, this suggests that these individuals have higher opportunity costs and, possibly, a higher standard of living and hence require higher compensations for hostel closure, in order to be as well off as initially. By and large, these results conform to initial expectations regarding the influence of the included explanatory variables on compensation amounts and strike a positive note on the theoretical consistency of the survey results. Debriefing questions A number of debriefing questions were included in the questionnaire aimed at exploring the consistency of the reported compensation amounts and the reasons behind not answering the valuation question. Respondents who stated a positive WTA amount were asked to allocate the new hypothetical income they would have under the proposed scenario (current income plus compensation) among the same categories used in the question about current expenditure composition, described in Figure 3.2. The prospective expenditure breakdown question is useful to check whether respondents would roughly maintain their pattern of consumption, as would be
96
Measuring the economic value of charities
expected since the new income should only afford them a standard of living similar to their current one, or if instead they anticipate large changes in the composition of their expenditures which could be indicative of upwardbiased answers in the valuation question. Recreation 7%
Others 19%
Personal items 9%
Alcohol/drugs/ tobacco 18%
Transport 11% Accommodation 19%
Figure 3.12
Food 17%
Predicted expenditure breakdown after the proposed scenario
The results are depicted in Figure 3.12. Although with a lower importance in absolute terms than before, accommodation costs still constitute the largest proportion (19 per cent) of total expenditure, followed by alcohol/drugs/ tobacco (18 per cent) and food (17 per cent). This pattern is similar to the one depicted in Figure 3.2 but for the fact that the difference between accommodation costs and food and alcohol/drugs/tobacco is now much narrower: 1–2 percentage points versus 9–10 percentage points. Contributing to this smaller difference is the fact that, in most cases, hostels provide some meals while private accommodation does not, and hence the relative proportion of the budget spent on food would be expected to increase. The relative importance of transport, recreation and personal items under the hypothetical scenario is similar to the status quo. The major difference between the post-hostel closure and the current situation seems to lie in the residual category denominated other expenditures. Initially this corresponded to some 8 per cent of the budget while now 19 per cent fall in this category. This increase could be indicative of the fact that hostel users would need to make up for the lack of hostel general support services by incurring a number of other costs, for example laundry, counselling, rehabilitation programmes and so on. But since no further information was recorded, this argument is merely speculative. To make sure that respondents were stating compensation amounts that would enable them to achieve their current quality of life after the proposed change, the questionnaire included a debriefing question about whether they
Benefits of charities to users
97
felt their quality of life would improve, worsen or stay the same after the compensation had taken place. Respondents who expected their quality of life to change were given an opportunity to revise their stated compensations in order to arrive at an amount that would leave them as well off as they currently were. Only 11 per cent of the sample indicated that their welfare would change after the compensation; and when asked to revise their WTA amounts accordingly, only a third were able to do so. As such, there may be a small bias on the estimated compensation suggests amounts by the small minority of respondents who failed to revise their answers. But the small proportion of the sample that changed its initially stated compensation that the careful questionnaire wording may have helped in inducing truthful responses for most of the sample. Finally, those who failed to provide an answer to the WTA question were asked for the reasons behind their lack of response. In particular they were asked to agree or disagree with a number of related statements. Table 3.9 displays the results. Table 3.9
Reasons for not answering the valuation question Agree Disagree (%) (%) Number
I don’t like the idea of not being able to use hostels any more
53.9
46.2
26
0.0
100.0
24
64.3
35.7
28
I find it very difficult to manage my own money 12.5
87.5
24
Even with the money, I wouldn’t be accepted anywhere else I couldn’t find the services hostels offer anywhere else
I would end up spending the extra money on other things
0.0
100.0
24
I don’t think I would be given the amount of money I want
4.2
95.8
24
I did not find the idea very easy to understand or imagine
48.2
51.9
27
A total of 31 individuals did not answer the valuation question (21 per cent of the total sample). Sixty-four per cent said they could not find the type of
98
Measuring the economic value of charities
services provided by hostels elsewhere. It is interesting to note that these individuals chose not to answer the valuation question rather than to bid for very high compensatory amounts that would afford them the sort of special support they might need. A general dislike of the proposed idea (54 per cent) was the second most important reason that led to non-response, while 48 per cent found the hostel closure idea difficult to understand. Lastly, 13 per cent of hostel residents noted that they had difficulties managing their money, and would probably end up sleeping in the streets as a result. For this subgroup, the alternative scenario would probably never enable them to have a similar quality of life as the current one. None of the 31 non-response cases indicated that they didn’t answer because they didn’t value hostel services. Their motives reflected some type of protest against the proposed idea, an inability to understand the scenario or a feeling that hostels were just irreplaceable, which was translated into a nonresponse instead of a high compensation. By and large, these findings indicate that respondents took the valuation exercise seriously, providing answers that were coherent and accorded to previous expectations.
3.4
DISCUSSION AND CONCLUSIONS
Much has been written about the difficulties of applying economic valuation methods based on hypothetical markets in the context of developing countries, where the population of interest may not know how to read or write, where money may not be a common means of exchange and where markets may not be well developed. However, little research, if any, has been done on applying such techniques to minority groups in developed countries, living outside the prevailing culture and where many of the difficulties in conducting valuation studies encountered in apparently different contexts may strike a parallel. This chapter reported on the results of a novel application of economic valuation techniques to homeless people, with a view to finding out how much the services provided by charitable hostels was worth to them. In order to do that, a carefully designed CV questionnaire was administered to 150 hostel residents. The scenario proposed described a hypothetical hostel closure during one year accompanied by monetary compensation to users: respondents were asked for the minimum compensation required, in such a scenario, that would afford them the same quality of life as they enjoyed at present. A number of conclusions can be drawn from the survey. CV was successfully administered to a minority group (that of single homeless people) living
Benefits of charities to users
99
an unconventional lifestyle, without permanent accommodation, most without a job and with special support needs. Nevertheless, the majority of respondents not only answered the valuation question, but also described the whole questionnaire in emphatic terms. The valuation question was framed as a WTA compensation to forego the use of hostels. This format, as opposed to the more common WTP format, can give rise to strategic behaviour in the form of overbidding. However, in the present survey, careful design and pre-test of the questionnaire, and specifically of the wording of the scenario and the structure of the valuation section, ensured that the estimated WTA distribution was not skewed towards unrealistically high answers. The overall unweighted average WTA was found to be £157 per person per week. As expected, the average WTA varied across different types of hostels: hostels providing higher levels of support yielded higher-value estimates than those offering only basic services. This indicates that the range of additional support and counselling services for special needs offered by these charitable institutions has a positive value to its users. Moreover, considering the price of rentals in the London area, respondents could probably find accommodation for less than £157 per person per week. The positive difference would be an estimate of the value of hostel support services. The majority of respondents who refused to answer the valuation question did so because they thought they could not find the type of services offered by hostels in any other place. The significantly positive monetary values uncovered in the valuation section were supported by the results of the attitudinal questions included in the questionnaire. Nearly all respondents made use of hostel support, counselling and general services, over and above the simple bed space provided. And, most importantly, the large majority classified these additional services as being good, even being satisfied with unlikely items such as hostel rules and opening hours. As a result, when confronted with the closure scenario, accompanied by monetary compensation, more than 60 per cent of respondents openly expressed their dislike of such an idea. The estimated compensation amounts also performed well in terms of standard tests of coherence and validity. Main determinants of the requested compensation amounts were the type of hostel, the attitudes towards the proposed scenario and the questionnaire in general, and several indicators of the quality of life of hostel residents such as the current level of expenditure. After receiving the hypothetical compensation, respondents did not predict major changes in their pattern of consumption, leading us to believe that their stated WTA amounts would actually be just enough to afford them their present lifestyle. The rationale for the homeless survey was to provide an input into the estimation of the total value of the housing and homelessness charitable
100
Measuring the economic value of charities
sector, by uncovering the value of the services provided by hostels to its direct beneficiaries. Therefore, in Chapter 5, the results presented in this chapter will be further analysed, together with the findings from the survey of the general public reported in Chapter 2, and value of volunteering reported in Chapter 4.
4. The benefits of volunteering 4.1
INTRODUCTION
A question of enduring interest in voluntary sector research is the appropriate value to be accorded to volunteer time. The methodological difficulties associated with such a valuation exercise have been well documented, and it is perhaps for this reason that there have been comparatively few attempts actually to perform it. An important starting-point is the exercise conducted by the Volunteer Centre UK (1995), which took as a proxy for the value of volunteer time the national average for the gross hourly wage rate (£7.83 in 1993). Applying this figure to estimates of the total annual number of hours volunteered in the UK yielded a total value for volunteering of £24.5 billion in 1993. However, Knapp (1990) is critical of any approach based on ‘a blanket foregone wage for all volunteers’, arguing that ‘some are retired or unemployed, and others could be giving up a variety of forms of paid employment in order to participate’. Nevertheless, the authors of the 1995 study by the Volunteer Centre UK readily acknowledge that a simple approach of this nature is only ‘a taste of what could be achieved in a more detailed analysis’. The purpose of this chapter is to refine and extend the valuation approach applied by the Volunteer Centre UK (1995), by responding – at least in part – to some of the criticisms set out by Knapp (1990). One of the reasons why valuing volunteer time is such a complex issue is that there are three very different perspectives that can be taken on the problem, each with a particular validity of its own. First, one could take the perspective of the volunteer and ask to what extent they value the time that is being given up to volunteering. In other words, what is the opportunity cost of volunteer time to the volunteer? This perspective is a particularly important one in terms of the overall enterprise of this book, which is to measure the net social value of the charitable sector. Since volunteers could be expected to give time up until the point where the marginal benefit they experience from volunteering is equal to the marginal cost, knowing the opportunity cost of volunteer time provides a way of estimating the benefits that volunteers receive from this activity. 101
102
Measuring the economic value of charities
Second, one could take the perspective of the charity manager and ask to what extent volunteering enables charities to economize on employmentrelated costs. This perspective is also important as regards the overall objective of measuring the net social value of charities, since it permits an estimate of the resources that charities absorb in order to provide their services. The cost of replacing volunteers with paid employees is therefore an important component of the full social cost of the charitable sector, and one that has tended to be overlooked in traditional measures of charitable expenditure. Third, one could take the perspective of the charity’s beneficiaries and ask to what extent volunteering has a positive impact on their well-being. This perspective looks at the productivity of volunteers in terms of the output they produce. Each of these perspectives suggests a different methodological approach to the valuation of volunteer time, which will be developed in greater detail below. To the extent that is possible, the approaches developed will be illustrated using data on volunteering from the Individual Giving Survey (IGS) (Halfpenny et al., 1992, 1993, 1994). The IGS was an annual repeated cross-section survey of philanthropic behaviour in the UK funded by the Charities Aid Foundation, which ran from 1987 to 1993. Each year the survey covered about 1000 individuals who were selected so as to be representative of the UK population. A structured interview conducted in the respondent’s home was used to collect information about philanthropic activities during the month before the survey. In particular, respondents were asked to report the number of hours per month that they devoted to a wide range of volunteering activities.
4.2
AN OPPORTUNITY COST APPROACH
One way of thinking about volunteering is as a ‘donation in kind’. Whereas some people make financial contributions to charity, others may prefer to contribute their time. When people choose to volunteer, they are implicitly foregoing the opportunity to do something else, whether that be paid employment, working in the home, or taking leisure. Thus the cost of giving time to charity from the perspective of the volunteer is effectively the same as the value of time that could have been spent in the alternative activity. A key question in valuing volunteer time in accordance with this approach is whether people work under flexible employment contracts which allow them freely to substitute time and money at the margin (for example, the selfemployed or those able to work paid overtime at will). Where this is the case, people are at liberty to allocate their time between alternative uses with the
Benefits of volunteering
103
result that the value of a marginal hour assigned to any activity should be the same across uses, and in all cases equal to the net hourly wage from employment. In these circumstances, the hourly value of time allocated to volunteering (Vv) will be equivalent to the take-home (or net) hourly wage rate of the volunteer (WN) (Vv = WN). The reason is that volunteers will be able to adjust the number of hours they devote to charitable work up to the point where the benefits of a marginal hour of volunteering will be equal to the benefits of a marginal hour in employment. In this calculus, the benefits of volunteering will include not only the altruistic satisfaction of contributing to a good cause, but also any private benefits that the volunteer gains from the experience; such as social contacts or acquisition of skills. As an illustration of this approach, a net hourly wage can be estimated for volunteers recorded in the IGS (Halfpenny et al., 1992, 1993, 1994). Although the surveys do not collect data on wage rates, they do contain information on many of the individual characteristics (such as sex, age and education) which are important in determining wages. The presence of this information makes it possible to impute a net hourly wage to respondents in the IGS survey. In order to do so, it is necessary to have access to a complementary dataset containing information on the net hourly wage and on the same individual characteristics recorded in the original datasest. In this case, the Family Expenditure Survey (FES) was identified as a suitable complement to the IGS (Arellano and Meghir, 1992). Using the FES data, a regression model was estimated capable of predicting the net hourly wage on the basis of observable characteristics common to both datasets. One of the difficulties with estimating such an equation is that net wages are only observed for those respondents currently in employment. To get round this problem, a Heckman model is used. This model takes into account the fact that individuals select whether or not to join the labour market on the basis of their wage prospects. Thus the wages observed among those who work are likely to be greater than the wages that would have been earned by those who do not work had they chosen to do so. Accordingly, there are two components to the model: a wage equation which captures the relationship between net wages and respondent characteristics for those who work, and a participation equation which explains how observable characteristics influence the decision whether or not to enter the labour market (Heckman, 1979). Table 4.1 reports the coefficients and t-statistics for the two separate models estimated using male and female respondents in the FES. The participation equation indicates that being married has a major positive impact on labour force participation for men while, for women, the presence of pre-schoolaged children in the household substantially reduces the likelihood of observing
104
Table 4.1
Measuring the economic value of charities
Summary of Heckman selectivity models used to fit net hourly wage Males
Wage equation Age Age squared Age × education Education Education squared Greater London Southeast North of England Scotland or Wales Constant Participation equation Age Age squared Age × education Education Education squared Greater London Southeast North of England Scotland or Wales Single-parent status Living with partner Pre-school children Constant Correlation Variance Log-likelihood Observations
Females
Coefficient
t-statistic
Coefficient
t-statistic
–0.023 –0.002 0.003 –0.951 0.028 0.199 0.118 0.039 0.055 8.858
–2.382 –4.319 6.205 –3.701 3.716 8.075 6.207 2.172 2.513 4.024
–0.068 0.003 0.003 –0.726 0.021 0.272 0.070 0.034 0.051 7.716
–6.139 7.266 5.782 –2.664 2.645 10.517 3.391 1.790 2.205 3.296
0.190 –0.002 –0.003 2.325 –0.062 –0.078 0.090 –0.121 –0.145 0.020 0.399 –0.075 –23.206
15.086 –41.442 –3.989 5.849 –5.265 –2.049 2.937 –4.439 –4.381 0.496 16.892 –2.932 –6.910
0.208 –0.020 –0.004 1.034 –0.022 –0.057 0.060 –0.035 –0.099 –0.469 –0.038 –0.425 –12.902
15.898 –39.551 –6.018 2.841 –2.028 –1.607 2.110 –1.378 –3.186 –14.453 –1.665 –20.061 –4.146
–0.854 0.7737 –19 809.78 18 737
–0.865 0.7838 –19 120.78 20 634
labour force participation. For both genders, net wage levels show a strong positive correlation with age and education. There is also some evidence of regional wage differentials; particularly for the Greater London area. Almost
Benefits of volunteering
105
all variables show a very high degree of statistical significance, in excess of the 99 per cent confidence level. In order to test the performance of the model, predictions were calculated for the FES data and then compared with the original values of the net hourly wage. It was found that, on average, the fitted wage variable tends to underpredict the true variable by £0.87 per hour for males and £0.77 per hour for females. As might be expected, the range of variation exhibited by the fitted wage is considerably narrower than that found in the original wage variable, illustrating the difficulty of modelling outliers in the data. Overall, this performance is thought to be adequate for the present purposes, subject to the caveat of underprediction. By applying this model to the IGS data it is possible to impute the value of the net wage variable to the initial dataset. The resulting hourly values of volunteer time for males and females during the early 1990s are reported in Table 4.2. The approach produces estimates of over £4 per hour for females and under £6 per hour for males. As noted above, these imputed values are probably somewhat below the true net hourly wages for the IGS sample. Table 4.2
Males Females
Estimated value of volunteer time using the opportunity cost approach* (£1993/hour) 1991
1992
1993
5.80 4.13
5.89 4.40
5.51 4.32
Note: * Calculated for that subsample of the IGS for which it was feasible to impute the opportunity cost.
A possible application of this particular measure of the value of volunteers’ time would be in the calculation of matching grants. The level of such grants is usually set with reference to the total volume of monetary donations from the general public, with the result that charities receiving relatively high levels of volunteer support and relatively low levels of financial support lose out in the calculation of such grants. This situation could be rectified by using an opportunity-cost-based valuation of volunteer time. The reason why the approach is particularly appropriate to this context is that it captures the cost to the volunteer of contributing their time, and thus the size of the monetary donation that would be equivalent to the number of hours volunteered. Clearly, the main limitation with the approach described is that the model of fully flexible allocation of time between paid employment and other activities is unlikely to hold in many cases, for example, if people are employed on
106
Measuring the economic value of charities
fixed-time contracts with unpaid overtime, or if they are retired, unemployed or otherwise engaged. In these cases, the principle of valuing volunteer time at the opportunity cost of the next best alternative still holds. However, the measurement of this opportunity cost becomes much less straightforward than simply computing the net hourly wage. Instead, a more complex surveybased approach, requiring respondents to trade off money and alternative uses of time, would probably be required to ascertain the appropriate value.
4.3
A REPLACEMENT COST APPROACH
By relying on volunteers to assist in the organization and running of their programmes, charities may be able significantly to reduce the payroll costs associated with their operations. Thus an obvious way of valuing the contribution of volunteers from the perspective of the charity is to estimate the cost of replacing volunteers with paid staff. The relevant measure for the hourly value of volunteer time to the charity (VC) is no longer the net wage of the volunteer, but rather the full labour costs that would be associated with a substitute employee. These are equivalent to the gross hourly wage (WG), multiplied by a mark-up that reflects the margin of non-wage labour costs (λ). From this must be deducted the hourly financial costs which charities may incur in the course of managing and deploying volunteers but which would not arise in the case of paid employees (κ).
VC = WG (1 + λ ) − κ It is worth noting that the volunteer’s own wage in employment is not necessarily a good guide to the wage required to secure a substitute employee, in the sense that the activities performed by a volunteer may be very different from what they would do in paid employment. For example, if a merchant banker volunteers at a soup kitchen over the weekend, the wage earned by the merchant banker in their usual employment will be very different to the wage that the charity would have to pay to get an employee to do kitchen work. Therefore, in order to apply this approach, it is necessary to have information on the exact type of activity undertaken by the volunteer. Fortunately, the IGS is helpful in breaking down volunteer hours into a number of activitybased categories. By matching up these activities against occupational categories cited in the national New Earnings Survey (NES) (Department of Employment, 1991, 1992, 1993) it becomes possible to estimate the hourly wage corresponding to each activity. This process is conducted in Table 4.3, which shows considerable variation in the resulting hourly wage rate both
107
Equivalent NES occupational category
Managers and administrators (marketing and sales) Managers and administrators (officials of charities) Managers and administrators Professional occupations (social workers) Associate professional occupations (information officers) Clerical and secretarial occupations Plant and machine operatives (road transport) Personal service occupations Managers and administrators (officials of charities) Craft and related occupations Professional occupations (clergy)
Raising money Serving on a committee Organizing events Visiting people Providing information Doing secretarial work Providing transport Providing other services Representing a charity Making goods for sale Helping in a church
14.40 13.09 8.76 9.23 8.91 6.55 5.26 6.87 13.09 5.63 9.23
Males
10.19 10.01 6.26 8.06 7.92 5.86 4.57 4.93 10.01 4.20 8.06
Females
IGS volunteer categories against equivalent NES occupational categories with corresponding average male and female hourly wage rates for 1993 (£1993/hour)
IGS volunteer category
Table 4.3
108
Measuring the economic value of charities
between the sexes and across the different types of activities undertaken by volunteers. Not only is it necessary to identify the appropriate hourly wage, but this must be adjusted to take into account the additional payroll costs – such as National Insurance contributions – that would be incurred by the charity in employing such an individual. The national Labour Cost Survey (Janes and Roberts, 1990) is designed to track the proportion of total labour costs accounted for by wage and non-wage elements. For the years 1991–93, the survey records that wage costs accounted for just over 80 per cent of full labour costs. The resulting hourly values of volunteer time for males and females during the early 1990s are reported in Table 4.4. The approach produces estimates of over £8 per hour for females and around £12 per hour for males. Table 4.4
Males Females
Estimated value of volunteer time using the replacement cost approach (£1993/hour) 1991
1992
1993
11.13 7.35
11.64 8.64
12.23 9.10
A potential application for this particular measure of volunteer time would be to correct charity accounts for differences in the use of volunteers, thereby permitting more meaningful efficiency comparisons between charitable institutions. For example, take the case of two charities of a similar size providing the same type of service. Imagine that one of the charities is supported primarily through financial contributions, while the other is supported primarily through volunteering. Then the expenditure accounts of the two charities would look very different. Indeed, a superficial comparison between the two might lead to the conclusion that the volunteer-intensive charity was the more efficient, in the sense of incurring lower financial costs. One way of correcting for this distortion would be to adjust the expenditure figures of the volunteer-intensive charity by adding a shadow expenditure figure that reflected the total value of contributions of volunteer time according to the replacement cost method. The replacement cost approach is particularly appropriate in this context because it reflects the amount that charities would actually have to spend if volunteers were replaced with paid employees. Like the previous approach, this methodology also has a number of limitations. In particular, the procedure described relies upon two assumptions that are not necessarily supported in reality.
Benefits of volunteering
109
The first questionable assumption is that, if volunteers were no longer available, charities would choose to replace them with paid employees who would do exactly the same type and amount of work. For example, it may be that some of the types of work done by volunteers would not be worth doing if charities had to pay for them at full market rates. Alternatively, it may be that the provision of opportunities for volunteers is one of the objectives of the charity and thus that it would not make sense to replace volunteers with employees. The second questionable assumption is that the quality of work done by volunteers is no different to the quality of work done by paid employees. It seems likely that the relative quality of work performed by these two groups could differ in either direction. On the one hand, in some areas of activity, paid employees may perform their work with a higher degree of professionalism and therefore attain a higher level of productivity than volunteers. On the other hand, in other areas of activity (for example, visiting prisoners or befriending the desperate or the destitute), the very absence of a formal employment contract may make the interventions of a volunteer more valuable in the eyes of the beneficiary than those of a paid employee.
4.4
AN OUTPUT-BASED APPROACH
The third perspective on the value of volunteer time is perhaps the most important, and at the same time the most problematic to quantify. This approach begins with the assumption that the main purpose of volunteering is to provide a service to the charity’s intended beneficiaries. From this perspective, the hourly value of volunteering is essentially the value of the corresponding output to the charity beneficiary (VB). This in turn can be broken down into two components: the number of physical units of output (O) produced by the volunteer during the course of an hour (H), and the price (or monetary value to the beneficiary) of each physical unit of output produced (PB).
O VB = × PB H Quantifying the productivity of volunteer effort involves identifying the actual physical output produced per hour, whether that be cooking a certain number of meals for the elderly, or restoring a certain number of metres of footpath. Less straightforward is valuing each unit of this output, in the sense of estimating how much the service is worth to the beneficiary group. The studies presented in Chapters 2 and 3 of this volume illustrate a novel
110
Measuring the economic value of charities
attempt to do precisely this. However, there is little if any wider literature on this subject. Extending this third approach to the valuation of volunteer time therefore depends on further progress in the valuation of charitable output.
4.5
COMPARING APPROACHES
Since it is not possible to produce valuations based on the output approach, the comparisons will be limited to the two initial methodologies presented: the opportunity and replacement cost approaches. As a point of reference, the methodology adopted by the Volunteer Centre UK will also be considered. Comparing the results reported in Tables 4.2 and 4.4 indicates that the hourly values resulting from the replacement cost approach (at between £8 and £12 per hour) are approximately double the size of those resulting from the opportunity cost approach (at around £4 to £6 per hour). These can be compared to the corresponding values that arise from taking the national average for the gross hourly wage as suggested by the Volunteer Centre UK. This latter approach gives values of around £6 to £8 per hour, which lie exactly in between the values arising from the other two methodologies. It is pertinent to ask whether the differences that are in evidence between the three methodologies are merely attributable to the fact that one estimates a net wage, one a gross wage and the other a full labour cost. To investigate this issue, all three measures are converted into an hourly gross wage equivalent to see whether there are any underlying differences in the value of time associated with each of the approaches. The gross hourly wage equivalents for all three methodologies are presented in Figures 4.1 and 4.2, with the corresponding values for tax and non-wage costs reported alongside. The figures show that substantial differences in the hourly measure remain between the three methods, even after adjusting for tax and non-wage labour costs. Although the spread of the three estimates narrows considerably as a result of the adjustment, the initial ranking is preserved, with the opportunity cost of time continuing to give the lowest value, the replacement cost the highest value, and the Volunteer Centre methodology giving a value somewhere in between. The differences between the estimates are, of themselves, quite informative. The first point to note is that the replacement cost method still gives a gross hourly wage equivalent that is substantially higher than that obtained from the Volunteer Centre approach. This may well be a reflection of the fact that the charitable sector requires relatively highly skilled labour inputs, which would command a wage above the national average level if recruited from the marketplace.
Benefits of volunteering
111
Replacement cost Non-wage costs Tax costs Wage costs
Volunteer Centre
Opportunity cost
0 Figure 4.1
2
4
6
8
10
Breakdown of differences between valuation methodologies for males in 1993 (£1993/hour)
Replacement cost Non-wage costs Tax costs Wage costs
Volunteer Centre
Opportunity cost
0
Figure 4.2
2
4
6
8
10
Breakdown of differences between valuation methodologies for females in 1993 (£1993/hour)
The second point to note is that the opportunity cost method gives a gross hourly wage equivalent that is substantially lower than that for the replacement cost approach. There are two possible explanations for this. One is that, on average, people who volunteer have personal characteristics that may
112
Measuring the economic value of charities
prevent them from commanding an employment wage comparable to the national average. The other explanation could be that the difference is simply due to the fact that the imputation process described above underestimates the true value of the net hourly wage. It is difficult to distinguish between these two hypotheses, although the differences between equivalent gross hourly wages observed in the two methodologies are quite close to the size of the underprediction resulting from the use of the imputation model (of the order of £1).
4.6
FACILITATING FUTURE VALUATION
Clearly, these estimates could be further improved if additional data were collected. An important feature of the methodological framework developed above is that it points quite specifically to the kind of information that would need to be collected in future surveys of volunteering in order to permit a more sophisticated valuation exercise. In order to develop the opportunity cost approach, it would be necessary for volunteer surveys to pay greater attention to documenting alternative uses of the volunteer’s time. Key questions are what the volunteer would be doing if not volunteering, and how highly the volunteer would value time spent in this alternative activity. As part of this process, it is important to collect reliable information on the net hourly wage from employment. Of equal value are details of the volunteer’s employment contract which would help to establish the extent to which time in volunteering could be substituted at the margin for time in employment, and at what rates. As regards implementation of the replacement cost approach, the key question is to identify the type of services that the volunteer performs for the charity and to consider how these might be replaced by paid employees and at what cost. In undertaking these calculations, it is important to take into account the costs that the charity already incurs in deploying the volunteer. Finally, the application of the output-based approach would require substantial empirical advances in the economic analysis of the charitable sector. On the one hand, it would be necessary to improve understanding of the charitable production function, in other words the relationship between inputs of volunteer time and the output of philanthropic services. On the other hand, further research is needed on the valuation which charity beneficiaries place on the services delivered. A range of methodological alternatives for doing this are presented in Chapter 1 and one particular methodological approach is illustrated in Chapters 2 and 3. There is clearly considerable scope for further research on both of these questions.
Benefits of volunteering
113
The valuation figures presented in this chapter are therefore in no sense definitive. Their primary purpose has been to illustrate how it is possible to refine previous estimates of the value of volunteer time even on the basis of existing information, but more importantly to provide guidance for future valuation attempts.
5. The aggregate benefits of the charitable sector: summary 5.1
INTRODUCTION
A methodology for valuing the output of the charitable sector was developed in Chapter 1. The methodology requires an estimation of the net social value of the charitable sector, defined as the difference between total social benefits and total social costs. The main challenge of implementing this approach is to put monetary values on the benefits that charities generate for their immediate target groups, as well as for society at large. This task was undertaken in Chapter 2, where members of the general public were asked to state what additional contributions they would be willing to make over and above their existing donations in order to avoid losing the services provided by housing charities, among others, and in Chapter 3 for the beneficiaries of hostels for the homeless in London, who were asked to state their willingness to accept compensation for the loss of these services. Issues relating to the valuation of volunteer time were discussed in Chapter 4. Now that all of these missing components are in place, it is possible to proceed with the overall valuation exercise outlined in Chapter 1.
5.2
ACCOUNTING FRAMEWORK
5.2.1 Total Social Value As noted in Chapter 1, the total social value of charities is the sum of direct benefits to users, benefits to volunteers and indirect benefits to the general public. Total social value (TSV) = direct user benefits + benefits to the general public + benefits to volunteers Given that charity services are typically provided at zero or highly subsidized cost, the fees paid by users do not fully capture the benefits that this 114
Aggregate benefits of the charitable sector
115
group receives. To capture the full benefits to charity users it is necessary to have some measure of WTP (or WTA compensation) over and above any existing fee payments. As a result of the empirical work reported in Chapter 3, this measure is available for charities in the housing and homelessness sector, though not at present for other types of charities. Specifically for this case, it was found that WTA compensation for the loss of hostel services was £8164 per beneficiary per year. Turning to the general public, it was noted above that actual donations comprise cash gifts, profits from the sale of goods and investment income on accumulated past donations. The survey reported in Chapter 2 found that actual donations totalled £8 per year for housing and homelessness charities and £76 per year for charities overall. Due to the free-rider problem, actual donations are likely to underestimate the benefits that accrue. Once again, it is necessary to have some measure of WTP for charity services over and above existing donations. The empirical work reported in Chapter 2 provides an estimate of this value for all charity sectors. The most conservative measure of incremental WTP for housing and homelessness surveys derived from this survey was £13 per person per year, while for the charitable sector as a whole, the most conservative estimate lies in the range £47 to £58 per person per year. The benefits to volunteers can be estimated in terms of the opportunity cost of volunteer time, since, to the extent that volunteers behave rationally, they will give time up until the point where the marginal benefit of doing so is equal to the marginal cost. The survey of the general public reported in Chapter 2 also collected information on the average number of hours volunteered to the various different types of charities considered. However, owing to the absence of detailed information on the net hourly wage in employment, it was not possible to estimate the true opportunity cost of time. Hence an indicative value of £5 per hour was used. Applying this value to the average number of hours volunteered yields a total value of time volunteered of £7 per person per year for housing and homelessness charities and £89 per person per year for the charitable sector as a whole. The above can be summarized by the following formula, where AF represents the actual user fees paid to charities, IWTPdir is the incremental WTP (or WTA) of the direct beneficiaries, AD are the actual donations made to charities, IWTPgp is the incremental WTP of the general public, PRODPROF are the profits from the sale of goods, I is the interest on income invested by charities, and Bvol is the benefits to volunteers. Total social value (TSV) = AF + IWTPdir + AD + IWTPgp + PRODPROF + I + Bvol
116
Measuring the economic value of charities
5.2.2 Total Social Costs As indicated in Chapter 1, the costs of providing charitable services can be measured by the income they receive. This is summarized in the following formula, where AF are fees paid by users, AD are actual donations, PRODPROF are the profits from the sale of products, I is interest income from investment, Cvol are the costs of volunteering and Cgovt represents the cost of government grants. Most of these items can be derived directly from published charity accounts. A key exception is the cost of volunteer time, which according to the methodology developed in Chapter 4 should be valued in accordance with its replacement cost. Since it was not possible accurately to measure the replacement cost of volunteer time on the basis of the survey of the general public reported in Chapter 2, an indicative value of £5 per hour is once again used. Total social costs (TSC) = AF + AD + PRODPROF + I + Cvol + Cgovt 5.2.3 Net Social Value The net social value is the difference between total social value and total social costs that measures the net social value of the charities, the social ‘valued added’, stated in equation form below. As can be seen, all the expressions for fees and donations drop out since they appear in both equations, leaving the following reduced expression for net social value. Net social value is essentially the incremental WTP of donors and direct beneficiaries over and above the value of government grants, plus the net benefits to volunteers. Clearly, net social value could be positive or negative, depending on the relative value of the terms. Net social value (NSV) = IWTPdir + IWTPgp + Bvol – Cgovt – Cvol
5.3
ESTIMATING THE NET SOCIAL VALUE
Applying the above accounting framework to the values estimated in the preceding chapters, the net social value of the charitable sector can be calculated. For charities in the housing and homelessness sector it is possible to calculate the full net social value including the benefits to direct users. This calculation is reported in Table 5.1. For the charitable sector as a whole, the benefits to direct users are not available and thus can only be estimated. Consequently, the calculations presented in Table 5.2 are only preliminary, and as such the results can only be regarded as tentative and illustrative.
Aggregate benefits of the charitable sector
Table 5.1
117
The net social value of housing and homelessness charities, 1997
Element Total social value AF IWTPdir AD IWTPgp PRODPROF I Bvol Total social costs AF AD PRODPROF I Cvol Cgovt Net social value
Unit value
Relevant population
£884 £8164 £8 £13 n.a. n.a. £7
0.14m (target group) 0.14m (target group) 46.2m (adult population) 46.2m (adult population)
46.2m (adult population)
Total value £2.56bn £0.12bn* £1.14bn* £0.37bn* £0.60bn* n.a. n.a. £0.32bn* £0.84bn £0.12bn* £0.37bn* n.a. n.a. £0.32bn* £0.03bn† £1.72bn
Sources: figures marked * are from surveys undertaken as part of the research in Chapters 2– 4. Figures marked † are based on the ONS 1994–95 survey as reported in Pharoah (1997).
In the case of housing and homelessness charities, the results reported in Table 5.1 derive directly from the unit values reported above. The results show that by far the largest category of benefits is the incremental WTP of the direct beneficiaries; this reflects the high cost of housing and thus the high WTA compensation of the target group. It is interesting to note that the actual fee revenue is less than 10 per cent of the incremental WTP. In the case of the general public, incremental WTP is somewhat less than half the value of actual donations. The overall benefits to the general public at £0.97bn are considerably smaller than the overall benefits to the target group of £1.26bn. In the case of the charitable sector as a whole, the absence of information about the incremental WTP of beneficiaries meant that additional assumptions had to be made. In order to get around this problem, it was assumed that the ratio between the incremental WTP of the target group and that of the general population would be the same as that found for the housing and homelessness sector. On this basis, an estimate of the incremental WTP of direct users was obtained by applying a factor of 1.3 to the incremental WTP of the general population.
118
Measuring the economic value of charities
Table 5.2
The net social value of all charities, 1997
Element Total social value AF IWTPdir AD IWTPgp PRODPROF I Bvol Total social costs AF AD PRODPROF I Cvol Cgovt Net social value
Unit value
Relevant population
£76 £47–£58 n.a.
46.2m (adult population) 46.2m (adult population)
£89
46.2m (adult population)
Total value £20.47bn–£22.02bn £3.94bn £4.34bn–£5.36bn £3.52bn* £2.17bn–£2.68bn* n.a. £2.39bn† £4.11bn* £15.63bn £3.94bn £3.52bn* n.a. £2.39bn† £4.11bn* £1.67bn† £4.84bn–£6.39bn
Sources: figures marked * are from surveys presented in Chapters 2–4. Figures marked † are based on the ONS 1994–95 survey as reported in Pharoah (1997).
5.4
INTERPRETATION AND POLICY IMPLICATIONS
Although incomplete, the results in Tables 5.1 and 5.2 show that charities do indeed add value to the resources that they absorb. The exact level of value added must await further research, particularly of the direct users of charities, since only homeless direct beneficiaries were surveyed in this research. There is also a need for further work on the value of volunteer time. Notwithstanding these limitations, the results reported in Tables 5.1 and 5.2 tentatively suggest that the charitable sector as a whole adds value of around 35 per cent to total social costs (£21 billion/£16 billion), while the housing and homelessness charities add value by some 200 per cent (£2.6 billion/£0.8 billion). These figures can be thought of as demonstrating economic value. The issue for charities is to capture as much as possible of this full economic value, that is, the WTP over the fees charged to direct users, and the donations contributed by the general public. Table 5.2 suggests that the general public alone have an incremental WTP for the services of the charitable sector in the order of £2.5 billion per year, a sum that is not captured by charities.
Aggregate benefits of the charitable sector
119
It is interesting to compare the incremental WTP of the general public with the overall value of government grants to the charitable sector. Government grants represent one way of overcoming the free-rider problem by taking collective action to increase the flow of resources to charities. The results reported in Table 5.2 indicate that the current value of government grants at £1.7 billion falls short of incremental WTP, which stands at £2.5 billion. One possible conclusion from this comparison is that there is a case for increasing the size of the government grant to the charitable sector. However, government grants are not the only way of allocating additional resources to charities. Alternative methods include sharpening fiscal incentives for giving, or increasing charity fundraising efforts. Part II of this book turns to the problem of capturing economic value and examines the relative efficacy of grants, fiscal incentives and fundraising approaches.
PART II
Capturing the Economic Value of Charities
6. Providing fiscal incentives for giving 6.1
INTRODUCTION
In many countries, including the United Kingdom, charitable donations receive special tax-exempt status. The rationale for this policy is precisely to stimulate charitable giving by reducing the ‘price’ of making a donation, thereby counteracting the free-rider tendency and helping charities to capture their full economic value. Since the government contributes µ per cent of every £1 given to charity, where µ is the marginal rate of tax rate, it only costs £(1 – µ) to give £1 to charity. Clearly, the effectiveness of this policy depends on the ‘price’ elasticity of donors. A major focus of the US literature on giving and volunteering has been the estimation of price and income elasticities, particularly in the context of evaluating the efficacy of fiscal inducements towards philanthropic behaviour. Econometric studies have in general found that in the USA monetary donations are price-elastic with respect to tax incentives, even though in the case of volunteering the results are not so clear. At a time when the UK government has been reviewing its own – very different – tax treatment of the charitable sector (Banks and Tanner, 1998), the econometric literature gives very little guidance on the price-responsiveness of British philanthropists. The few British studies that exist either exclude the issue of price-responsiveness altogether (Banks and Tanner, 1997; Pharoah and Tanner, 1997) or are subject to data limitations which make it difficult to interpret the price effects obtained (Jones and Posnett, 1991a,b). Yet such estimates are critical in informing fiscal policy towards the charitable sector. Barrett (1991) shows how elasticity estimates can be used to test for the neutrality and treasury efficiency of fiscal policy. A priori, an increase in the marginal tax rate has an ambiguous effect on charitable donations since it will both reduce the price of making tax-efficient donations (thereby stimulating philanthropic gifts) and reduce the after-tax income of donors (thereby dampening their generosity). It can be shown that as long as the income and tax-price elasticities are equal in absolute magnitude the tax change will be ‘neutral’ in its impact on giving behaviour. If the tax-price elasticity exceeds the income elasticity in absolute terms, the tax rise will have a positive net effect on donations, and vice versa. 123
124
Capturing the economic value of charities
Another important concern in the design of fiscal policy is whether the extra donations induced by a tax incentive are greater or less than the resulting loss in government revenues. Barrett (1991) shows that if donations are tax-price-elastic, the induced flow of resources into the charitable sector more than offsets the reduction in tax revenues. This situation is described as treasury efficiency. Finally, if donations are found to be inelastic with respect to income, this would suggest that subsidizing donations may be a regressive way of achieving an increase in the aggregate supply of public goods. This chapter reports estimates of price and income elasticities for both monetary giving and volunteering in the UK, based on a new and hitherto unexploited source of information: the Individual Giving Survey (Halfpenny et al., 1992, 1993, 1994). The dataset is attractive in a number of ways. First, it is made up of repeated cross-sections of data across a period of substantial variation in the UK tax system, thereby incorporating variation in the taxprice variable which is not merely cross-sectional. Second, it contains very detailed information on gifts of time and money. Third, it records respondents’ perceptions of the price of giving as well as the objective value of the price itself, thereby making it possible to take into account the problem of imperfect information about prices. The results indicate that British philanthropists show significant responses to fiscal incentives, even though price elasticities are somewhat smaller than those found in the USA. Since tax-price elasticities are generally found to exceed income elasticities, an increase in the marginal rate of income tax can be expected to bring about a net increase in tax-efficient donations. However, the fact that donations are inelastic with respect to the tax price indicates that the induced increase in donations will not be large enough to offset the loss of revenue to the Treasury. The empirical analysis does not limit itself to considering conventional taxprice elasticities, but also examines the extent to which contributions are affected by two additional concepts of price. The first is the efficiency price or the rate at which charities absorb donations on administrative expenditures. The second is the price of time, that is the productivity differential between time in employment and time in volunteering. These wider price effects are also found to be significant in determining philanthropic behaviour. The remainder of the chapter is organized as follows. Section 6.2 provides an overview of the literature on the price-responsiveness of giving and volunteering behaviour both in the UK and North America. Section 6.3 introduces the dataset on which the analysis is based, while Section 6.4 explains how it was used to construct the relevant price variables. Section 6.5 reports the empirical results and calculates the corresponding price and income elasticities. Section 6.6 concludes.
Providing fiscal incentives for giving
125
A more extensive discussion of the material can be found in Foster (1999c).
6.2
LITERATURE OVERVIEW
6.2.1 Giving The existing literature on price and income elasticities of charitable giving is summarized in Table 6.1. The issue of charitable donations began to be studied empirically in the USA from the late 1960s onwards (Taussig, 1967; Schwartz, 1970; Feldstein and Clotfelter, 1976; Feldstein and Taylor, 1976; Boskin and Feldstein, 1977). A primary objective of the US literature has been to gauge the effect of taxation policy on philanthropic behaviour. Under US tax rules, summarized in Table 6.2, ‘itemized’ charitable donations listed on the annual tax return are deducted from overall income before the assessment of tax liability. Thus donating a dollar to a charitable cause only costs the donor (1 – µ) dollars of income, where µ is the marginal income tax band. The earliest US studies set about estimating the elasticity of donations to this tax price using time-series data from tax returns. However, it was soon noted that use of such data carried a number of significant drawbacks (Feldstein and Clotfelter, 1976). Particular problems were the partial non-identifiability of donations by low- and middle-income households (who may choose not to itemize their deductions), the possible over-reporting bias of donations by itemizers (in an attempt to minimize tax liability), and the limited availability of explanatory variables (particularly measures of permanent income and wealth). These problems were avoided by altering the basis of empirical work from tax-returns data to survey-based data (Clotfelter, 1985). Although early studies based on tax returns found relatively low tax-price elasticities (for example Taussig, 1967; Schwartz, 1970), a consensus was reached in later work that donations were tax-price-elastic, with most estimates lying in the range –1.0 to –1.5 in absolute value. Income elasticities, on the other hand, were invariably found to be below unity, with a typical value of 0.8. These estimates were derived for the most part from simple ordinary least squares (OLS) models of donations, often using logarithmic specifications. Two important innovations to be introduced during the 1980s were the use of panel data (Clotfelter, 1980) and the correction of possible endogeneity problems (Reece and Zieschang, 1985). Ironically, both of these refinements had the effect of depressing the resulting estimates of tax-price elasticity below unity and back towards the much lower levels found in the earliest studies (Taussig, 1967; Schwartz, 1970).
126
Tax returns Tax returns Survey Tax returns Survey Tax returns Survey Tax returns Survey Survey Survey Tax returns Survey Survey Tax returns Survey Individuals Groups Individuals Individuals Individuals Groups Individuals Individuals Individuals Individuals Individuals Individuals Individuals Individuals Individuals Individuals OLS OLS OLS OLS OLS OLS Tobit Fixed effects Tobit OLS Tobit Random effects Tobit Selectivity Selectivity Selectivity
Model specification Log-linear Log-linear Log-log Log-log Log-log Log-log Linear Log-log Linear Linear Linear Log-log Log-linear Log-linear Log-log Log-log
Functional form
0.2–0.4 0.4 –0.8 1.6 1.0–1.5 2.5 1.0–1.1 1.0 –1.4 0.6–1.1 0.1–0.2 2.8 0.4 1.1 1.6–1.8 n.a. n.a. n.a.
1.3–3.1 0.2–0.8 0.8 0.8 0.7 0.8–0.9 0.6–1.4 0.4–0.7 1.3–1.4 0.8 1.0 0.2 0.4 0.4–0.6 0.5–1.7 0.1–1.2
Tax-price Income elasticity elasticity
Notes: regarding data type, CS stands for cross-sectional and TS for time-series. Regarding model specification, OLS stands for ordinary least squares. Regarding tax-price elasticities, reported figures are absolute values; all estimates have the expected negative sign.
USA USA USA USA USA USA USA USA USA USA USA USA USA UK UK UK CS TS CS CS CS TS CS Panel CS CS CS Panel CS CS CS TS
1967 1970 1976 1965 1977 1978 1979 1980 1985 1985 1989 1991 1992 1991a 1991b 1997
Observations
Taussig Schwartz Feldstein and Clotfelter Feldstein and Taylor Boskin and Feldstein Abrams and Schmitz Reece Clotfelter Reece and Zieschang Schiff Kingma Barrett Brown and Lankford Jones and Posnett Jones and Posnett Banks and Tanner
Data source
Data type
Year
Authors
Country
Summary of literature on price and income elasticities of giving
Table 6.1
127
●
●
●
Non-itemizers are those who elect to stake a standard deduction against their gross income, rather than separately itemizing their deductible expenditures. This innovation was introduced in 1944 in an attempt to simplify income tax assessment. Effectively, non-itemization is an attractive option to all taxpayers whose deductible expenses sum to no more than the value of the ‘standard deduction’. It is thus more prevalent among low- and middle-income households. An important implication of this choice is that any charitable donations made by such households will not be subject to any tax incentive, facing instead a marginal price of unity. Itemizers, are those who elect to assess their allowable deductions by individually ‘itemizing’ on their tax returns all deductible expenditures incurred during the past tax year, rather than using the ‘standard deduction’. Effectively, ‘itemization’ is an attractive option to all taxpayers whose deductible expenses sum to more than the value of the ‘standard deduction’. It is thus more prevalent among high-income households. An important implication of this choice is that any charitable donations made by such households will not be subject to a tax incentive, facing a price equal to one minus the corresponding marginal income tax rate.
US taxpayers can be divided into two categories according to how they approach the calculation of their deductions:
the types of recipient organizations whose contributions qualify for deductability; the proportion of income given which can qualify for deductibility.
●
●
●
●
The Deed of Covenant scheme, which allows charities to reclaim tax paid on a regular annual gift committed over a period of at least four years. There are no upper or lower limits on the value of sums that can be covenanted. The payments must be accompanied by a written Deed of Covenant between the donor and the charity. The Payroll Giving scheme (in operation since 1987), which is operated by some employers and available only to their employees. The scheme allows people to have donations deducted directly from their pay before tax liability is assessed. There is an upper limit on the value of the donations (which was £1200 per annum in 1996/97). The Gift Aid scheme (in operation from October 1990), which allows charities to reclaim tax paid on lump-sum gifts in excess of a value threshold (which was £250 in 1996/97) with no upper limit. A Gift Aid form must be made out to the charity as an accompaniment to the net gift. The Millennium Gift Aid scheme (in temporary operation between July 1998 and December 2000), which allows charities to reclaim tax paid on lump-sum gifts in excess of a value threshold of £100. The scheme only applies to charities supporting education, health and poverty relief programmes in 80 specific developing countries.
Charitable donations are only eligible for tax relief if they are channelled through one of the four following schemes:
Federal income tax liability is assessed by subtracting allowable deductions from gross income. Allowable deductions include both business expenses and charitable donations. There are, however, certain restrictions regarding:
●
UK
Contrast between tax treatment of charitable donations in the USA and the UK
USA
Table 6.2
128
Capturing the economic value of charities
In the UK, the empirical analysis of charitable donations did not begin until the 1990s (Jones and Posnett, 1991a; Banks and Tanner, 1997, 1998; Pharoah and Tanner, 1997). These studies made use of data contained in the Family Expenditure Survey, which requires individuals to keep diary records of casual donations over the period of a fortnight, and also collects information on regular donations made by bank standing orders or payroll deductions. A further study by Jones and Posnett (1991b) makes use of Inland Revenue data on giving by covenant. Unlike some of the earlier US work, these UK studies give explicit consideration to the problem of modelling zero donations. Banks and Tanner (1997) identify three possible interpretations of zero donations, namely: a genuine preference for not making donations; infrequency of donations (leading to none being recorded during the survey period) or recording error (as a result of respondents failing to recall donations that were actually made). In the early US studies, this problem was dealt with by discarding zero observations or adding a small positive quantity to permit their inclusion within a logarithmic specification. Both of the UK studies considered here, as well as some of the later US studies, use limited dependent variable specifications to permit a more sophisticated incorporation of the zeros. Banks and Tanner (1997) distinguish between the Tobit specification and the selectivity model. The former models participation and expenditure as a single process and thus implicitly assumes that all zeros represent a genuine preference for not making donations. The latter allows participation and expenditure to be governed by separate processes, thereby providing greater flexibility in modelling zero donations. Both of the UK studies indicate a preference for the selectivity specification. The estimation of tax-price elasticities has not been such a prominent issue in the UK literature as in the US literature. This is largely a reflection of the differences in the tax treatment of charitable contributions between the two countries, as indicated in Table 6.2. Whereas in the USA ‘itemizers’ are at liberty to claim tax relief on all charitable gifts, in the UK tax relief is confined to a number of special modes of giving. These schemes are quite restrictive in the sense that they require either a long-term commitment to making regular contributions (covenants and payroll giving), or a relatively large lump-sum contribution (Gift Aid and Millennium Gift Aid). Perhaps as a consequence of this, tax-efficient giving remains a relatively small proportion of individual giving in the UK. For example, in 1992/3 taxefficient giving amounted to £1200m (Inland Revenue, 1998) or about 24 per cent of the estimated £5000m total value of charitable giving for the corresponding year (Halfpenny et al., 1994). All remaining donations can be described as non-tax-efficient in the sense that they do not attract any tax advantage.
Providing fiscal incentives for giving
129
None of the UK studies summarized in Table 6.1 presents estimates for price elasticities. Jones and Posnett (1991a,b) do in fact estimate significant price effects in both of their papers, but do not report price elasticies. In part, this reflects the authors’ concerns that the estimated price effects may be spurious due to limitations with the data. In the first study (Jones and Posnett, 1991a), which is based on the Family Expenditure Survey, the authors are surprised to find a significant price effect given relatively low participation in tax-efficient giving and consequently conjecture that the result may simply reflect collinearity beween tax-price and income variables. In the second study (Jones and Posnett, 1991b), which uses tax return data on giving by covenant, the authors claim that the significance of the tax-price variable is largely an artefact of the Inland Revenue system. The reason that they give for this is that only higher-rate taxpayers have any real incentive to report covenants to the Inland Revenue. As regards income elasticities, the estimates presented span the range of those which have been obtained for the USA. On the basis of their selectivity model Banks and Tanner (1997) find that while the income elasticity of participation is very low (0.1), the income elasticity of gifts among donors is relatively high (1.2). 6.2.2 Volunteering The existing literature on price and income elasticities of volunteering is summarized in Table 6.3. Largely owing to a paucity of suitable data, the empirical analysis of volunteering behaviour does not begin until considerably later than the research into monetary contributions. In the case of volunteering, the appropriate price variable is the opportunity cost of the time given to charity. A widely used measure for this is the net hourly wage rate, although strictly speaking this is only valid for those who are able freely to substitute time between work and leisure activities at the margin. Given the lumpy and restrictive nature of employment contracts, this is unlikely to be true except in a minority of cases. To the extent that there is substitutability or complementarity between giving and volunteering, the tax price of giving may become relevant as a cross-price variable. The earliest study by Dye (1980) confines its attention to estimating this cross-price effect which is found to be negative and significant and thus provides evidence of complementarity between gifts of time and money. A much more comprehensive treatment of the subject is provided in the seminal paper by Menchik and Weisbrod (1987). These authors estimate a cross-tax-price elasticity of –1.3, which is found to be substantially higher
130
USA
1997
1996
1995
Freeman
Day and Devlin
Knapp et al.
CS
CS
CS
CS
CS
CS
Survey
Survey
Survey
Survey
Survey
Survey
Data source
Individual
Individual
Individual
Individual
Individual
Individual
Observations
Selectivity
Selectivity
Probit and OLS
Tobit
Tobit
Probit and OLS
Model specification
Log-linear
Log-linear
Log-log
Log-linear
Linear
Log-log
Functional form
Insignificant
n.a.
Insignificant
n.a.
0.4
n.a.
Wage-price elasticity
0.1
n.a.
n.a.
1.1–2.1
1.3
0.6
Tax cross-price elasticity
n.a.
Insignificant
n.a.
n.a.
0.7
Insignificant
Income elasticity
Notes: regarding data type, CS stands for cross-sectional. Regarding model specification, OLS stands for ordinary least squares. Regarding tax-price elasticities, reported figures are absolute values; all estimates have the expected negative sign.
UK
Canada
USA
USA
Menchik 1987 and Weisbrod
Brown 1992 and Lankford
USA
1980
Dye
Country
Year
Data type
Summary of literature on price and income elasticities of volunteering
Study
Table 6.3
Providing fiscal incentives for giving
131
than the own-wage-price elasticity of –0.4. The negative sign of the crosstax-price elasticity once again suggests complementarity between giving and volunteering behaviour. They also estimate an income elasticity of 0.7 for volunteering, which lies in the range of results obtained for monetary gifts. Subsequent studies have failed to produce such clear-cut results. For example, Freeman (1997) finds positive and significant coefficients on income and the net hourly wage in a model of volunteer participation. Freeman argues that one explanation for this result is that hourly wage acts as a proxy for the productivity of volunteer time rather than capturing the opportunity cost of hours volunteered. He hypothesizes that the correct price variable for volunteering is not so much the net hourly wage as the ratio between the net hourly wage and the productivity of time in volunteering, since the lower this ratio, the more attractive it becomes to give time rather than money. Freeman constructs this variable on the basis of a subjective respondent assessment of the relative productivity of time in work and volunteering and in a second set of models finds a statistically significant relationship between hours volunteered and this revised price variable. There has only been one UK econometric study of volunteering behaviour to date (Knapp et al., 1995), making use of data collected by Volunteer Centre UK/Social and Community Planning Research survey (1991). Based on a selectivity model, the study finds that both income and the cross-tax-price of giving positively affect the probability of participation but do not influence the number of hours volunteered. The finding of a positive cross-tax-price effect indicates that giving and volunteering are substitutes, and is thus at odds with the findings of the US literature. As in the Jones and Posnett paper (1991a), the authors of this study express some surprise at uncovering any statistically significant cross-tax-price coefficient given the limited extent of tax-efficient giving in the UK. Finally, Knapp et al. (1995) do not find any statistically significant effect from the own-wage price of volunteering. They suggest that this may be due to the fact that this variable was not reported in their dataset but rather had to be proxied by average wage rates for particular occupational groups. Summary remarks Thus, overall, the analysis of philanthropic behaviour has reached a much higher level of development in the case of monetary giving than in the case of volunteering. Moreover, the associated empirical literature is far more extensive in the USA than in the UK. While there is a considerable body of evidence which supports the finding that US donations are tax-price-elastic, some of the more recent and sophisticated studies reach the opposite conclusion (Clotfelter, 1980; Reece and Zieschang, 1985). By contrast, the few British studies that exist either exclude the issue of price-responsiveness
132
Capturing the economic value of charities
altogether (Banks and Tanner, 1997; Pharoah and Tanner, 1997) or are subject to data limitations which make it difficult to interpret the price effects obtained (Jones and Posnett, 1991a,b).
6.3
THE DATASET
The empirical analysis presented in this chapter is based on data collected in the Individual Giving Survey, during three repeated annual cross-sections covering the years 1991–93 and comprising just over 1000 observations each (Halfpenny et al., 1994). Each of these annual surveys was based on a representative cross-section of the British population (excluding Northern Ireland), achieved by quota sampling on demographic characteristics such as sex, age and size of household. Interviews were conducted face to face in the respondents’ homes and spread evenly throughout the year to avoid seasonal biases. It is important to stress the value of being able to rely on a repeated crosssection of data as opposed to a single cross-section, as has been the case with much of the philanthropic literature (recall Tables 6.1 and 6.3). Of particular importance for the identification of price effects is the fact that there were numerous modifications to the tax system during the period 1990 to 1993. The introduction of independent taxation occurs just a few months before the beginning of the study period, while the creation of the additional 20 per cent income tax band took place right in the middle of the study period, in 1992– 93. There were also numerous modifications to the real value of tax allowances, most notably the fall in the value of the married couples’ allowance by more than 10 per cent in real terms. All of this means that the tax-price effects are not merely being identified from differences in behaviour between individuals at very different income levels, but also from variations in the tax price faced by individuals at similar points in the income distribution. During the interviews, respondents are required to recall philanthropic behaviour during the last month (or year in the case of tax-efficient methods of donation). This can be contrasted with the Family Expenditure Survey (FES) – used in earlier UK empirical studies – which requires respondents to keep a diary of casual cash donations over a two-week period. The various differences that exist between these two survey methodologies suggest that the corresponding measures of donations are likely to diverge. This is so for a number of reasons. First, the IGS definition of charitable donations is broader than that adopted in the FES, in that it includes purchases of goods and services from charities in addition to pure donations. This will tend to inflate the estimate of donations resulting from the IGS. Second, the method of collecting donations differs. In the FES respondents
Providing fiscal incentives for giving
133
are asked to keep a diary of their donations, while in the IGS respondents are asked to recall their donations during a household interview. To the extent that respondents have trouble in remembering what contributions they have made, this may create a downward bias in the IGS data. Finally, the FES collects information about donations over a two-week period while the IGS uses a monthly period. It is possible that the use of a longer recall period may lead to ‘telescoping’ effects in the IGS, whereby respondents erroneously ascribe donations from the more distant past to the survey period. This would tend to inflate the donations measure obtained from the IGS and would consequently offset the earlier problem associated with imperfect recall. Comparing summary statistics from the FES and IGS for 1993, it becomes possible to gauge the relative magnitude of these effects. The IGS records participation rates that are very much higher than the FES (77.5 per cent versus 29.1 per cent), which is probably a reflection of the longer survey period and also the wider definition of charitable gifts. However, in terms of the average weekly donations recorded, the FES gives slightly higher values than the IGS (£4.11 per week versus £3.46 per week), although in terms of the median this ranking is reversed (£1.50 per week in the IGS and £1.23 per week in the FES).
6.4
ANALYTICAL FRAMEWORK
This section introduces the three different concepts of price that are deemed relevant to philanthropic behaviour and explains how the data contained in the IGS were used to construct the corresponding variables. As noted in the introduction, the concept of price considered here includes but is not limited to the tax price, which has formed the central focus of the earlier literature. In addition, the analysis will examine to what extent gifts of time and money are affected by the efficiency price and the time price. The efficiency price refers to the rate at which charities absorb donations on ‘unproductive’ administrative expenditures, whereas the time price is related to the productivity differential between time in employment and time in volunteering. 6.4.1 Tax Price When donors give via one of the government’s tax-efficient schemes, described in Table 6.2, they effectively obtain a reduction in the price of giving. This is because the government waives the tax that would normally be levied on the donation. Thus, in order to give £1 to a charity, the donor will only need to sacrifice (1 – µ) of his after-tax income, where µ is the marginal income tax rate. This tax price of giving is defined mathematically as below.
134
Capturing the economic value of charities
Since the marginal income tax rate is always greater than or equal to zero, the tax price will always be less than or equal to one: PXi = (1 − µ i )
(6.1)
In order to ascertain the tax price faced by each survey respondent it is necessary to have data on their marginal income tax rates (µ). Unfortunately, the Individual Giving Survey does not contain this information. However, the survey does include many of the underlying variables that determine an individual’s income tax position, such as personal income, household income, age, marital status, and single-parent status. On the basis of these variables, and guided by the tax rules for the relevant period, it is possible to create a simple algorithm which allocates respondents to a particular marginal income tax band. The algorithm generates the real objective tax price faced by each respondent. However, it is legitimate to question whether people actually perceive the tax price in this way. One of the questions included in the survey asks respondents whether they are aware of the opportunity to give through one of the government’s tax-efficient schemes. Interestingly, only 70 per cent responded that they knew about the schemes. The significant minority did not effectively face a tax price of unity irrespective of their marginal income tax band. In order to ensure a closer reflection of reality, the tax price variable was adjusted accordingly. Consequently, the analysis is based on this subjective tax-price variable, as opposed to the objective one. 50 40 30 20 10 0
0.60
Figure 6.1
0.75
0.80
Percentage frequency distribution of tax price (£)
1.00
Providing fiscal incentives for giving
135
Figure 6.1 presents the percentage frequency distribution of the subjective tax-price variable which indicates that about half of the sample face a tax price of unity and thus do not perceive any fiscal incentives to give. The next largest group, representing 40 per cent of the total, face a tax price of 0.75, corresponding to the standard income tax band. Fewer than 5 per cent of the sample face the lowest tax price of 0.6, corresponding to the higher-rate income tax band. 6.4.2 Efficiency Price Not all of the money that donors give to charity will end up being spent on the good cause. The reason is that charities are not perfectly efficient at converting gifts into philanthropic services (Rose-Ackerman, 1986; Steinberg, 1986; Weisbrod and Dominguez, 1986). Inevitably, some fraction of the receipts (α) will be absorbed in administrative expenditures needed to run the organization. Hence a £1 donation will increase philanthropic expenditure by only (1 – α). This efficiency price of giving is defined mathematically as follows: 1 PEi = 1 − αi
(6.2)
The IGS does not contain information on the precise charity to which a particular respondent contributes, thus it is not possible to ascertain the objective value of α, which would (otherwise) be retrievable from charity accounts. However, the IGS does collect data on the subjective value of (1 – α), by explicitly asking respondents: ‘For every £1 you give to charity, how much do you think gets to the needy cause?’ This subjective value is – in fact – more helpful for the construction of the corresponding price variable than the objective value would have been. This is because it seems likely that many donors are uninformed as to the true value of α; thus their behaviour will in all probability be guided more by their perceptions of this price variable than by what it actually is. Figure 6.2 illustrates the percentage frequency distribution of the efficiencyprice variable. The modal value of the distribution lies in the 1.5 to 2.0 range, indicating that more than a third of respondents believe that administrative expenditures absorb between 25 per cent to 50 per cent of each pound donated. It is striking that as many as 15 per cent of respondents believe that the efficiency price is greater than 4, or in other words that administration absorbs more than 75 per cent of each pound donated. These subjective assessments indicate a considerable degree of scepticism about the efficiency of voluntary organisations. However, the reality is more
136
Capturing the economic value of charities
50 40 30 20 10 0 Figure 6.2
1–1.5
1.5–2
2–4
>4
Percentage frequency distribution of efficiency price (£)
favourable than respondents seem to imagine. During the period of the early 1990s covered by the survey, the top 500 of the UK’s fundraising charities were allocating on average no more than 15 per cent to expenditures such as administration and fundraising which do not contribute directly to the good cause. 6.4.3 Time Price The concept of price for volunteering is somewhat more complex, although many of the foundations have already been laid in Chapter 4. A person who volunteers self-evidently sacrifices, say, an hour of their own time in order to give that time to charity. The value that the volunteer places on their own time will, in all likelihood, be very different from the value that the charity places on the time they have volunteered. For reasons already discussed at length in Chapter 4, the value that the volunteer places on their own time is often approximated as their net hourly wage in employment (ϖ), while the value that the charity places on the volunteer’s time will be related to the full hourly (gross wage plus non-wage labour) costs of employing somebody else to perform the same task as the volunteer (υ). The ratio of the value of the time sacrificed by the volunteer to the value of the time gained by the charity effectively defines the relevant price of time. This is stated mathematically as follows: ϖ PTi = i υi
(6.3)
Providing fiscal incentives for giving
137
Any differences between ϖ and υ are attributable to one of two sources. First, there is the fact that υ includes taxation and non-wage labour costs whereas ϖ does not. Second, it is conceivable that time spent in volunteering may be more or less productive than time spent in employment, which would be reflected in differences between the corresponding wages. For example, if a merchant banker volunteers at a soup kitchen over the weekend, the wage earned by the merchant banker in their usual employment will be very different to the wage that the charity would have to pay to get an employee to do kitchen work. In order to quantify the time-price variable defined above, it is necessary to have a measure of the net hourly wage rate of each respondent, together with the full hourly labour cost to the charity of replacing the volunteer with a paid employee. While neither of these two variables is recorded directly in the IGS, the survey does contain adequate information from which to construct estimates of the two corresponding wages for each respondent. A detailed description of how this was done can be found in Chapter 4. Figure 6.3 summarizes the percentage frequency distribution of the price of time for those who volunteer. The first point to note is that the estimates are extremely low in absolute value. This may in part reflect the underprediction of the net hourly wage noted in Chapter 4 above. However, even if the net hourly wage rate is multiplied by a factor of three (which is undoubtedly a gross overestimate of the degree of underprediction), the average price of time among volunteers remains low in absolute value, rising from £0.16 per hour to £0.62 per hour. Moreover, the proportion of volunteers who face a
50 40 30 20
Figure 6.3
>0.20
0.16–0.20
0.11–0.15
0.05–0.10
0
<0.05
10
Percentage frequency distribution of time-value price (£)
138
Capturing the economic value of charities
time price in excess of unity rises only slightly, from 2.4 per cent to a mere 12.4 per cent. Whatever the true absolute level of the time price of volunteering, Figure 6.3 remains informative in illustrating the relative level of this price among volunteers. In particular, there is a noticeable skew in the distribution of the price of time towards the lowest values, which provides some evidence that participation in volunteering may be price-sensitive. 6.4.4 Analysis of Price Variables Before proceeding to estimate a full econometric model of philanthropic contributions, it is helpful to analyse some of the broad patterns discernible in the data. To this end, Table 6.4 presents the percentage participation rates for giving and volunteering at different levels of the three price variables developed above. What is immediately striking about Table 6.4 is the presence of a clear negative monotonic relationship between price levels and participation rates. This relationship, which applies to all three of the price variables for both Table 6.4
Monthly philanthropic participation rates against different price variables (%)
Efficiency price 1.0–1.5 1.5–2.0 2.0–4.0 >4.0 Tax price 0.60 0.75 0.80 1.00 Time price <1 >1 Overall
Giving
Volunteering
84.12 81.40 78.67 70.96
35.30 24.94 24.61 21.84
93.02 88.36 86.84 73.55
50.00 34.40 28.95 20.12
92.12 73.27
– –
77.66
26.43
Notes: since the time-price variable is observed only for volunteers, the participation rate for volunteering against the time price cannot be calculated.
Providing fiscal incentives for giving
139
giving and volunteering, provides prima facie evidence of price-responsiveness in philanthropic behaviour. It is also interesting that participation in volunteering appears to be associated with the tax price, while participation in giving appears to be associated with the price of time. This is suggestive of a cross-price effect whose negative sign would seem to indicate complementarity between giving and volunteering (Dye, 1980; Brown and Lankford, 1992). Table 6.5 presents a parallel set of results, showing how the size of the average monthly philanthropic contribution varies with the price level. There continues to be some evidence of a negative association between the two, even though this looks somewhat weaker than the evidence presented for participation rates. Table 6.5
Mean monthly donations against different price variables
Efficiency price 1.0–1.5 1.5–2.0 2.0–4.0 >4.0 Tax price 0.60 0.75 0.80 1.00 Time price <1 >1 Overall
Giving (£)
Volunteering (hours)
17.15 12.96 11.45 12.46
16.48 12.50 16.86 17.99
24.51 14.67 11.59 9.42
15.21 16.63 16.77 12.56
21.65 10.34
16.21 13.51
13.65
15.94
However, given the correlation between the tax price and the household income level, it is not possible to say to what extent these results reflect a pure price effect or simply an indirect income effect. Thus a firm conclusion on price-responsiveness must await the results of multiple regression that controls for factors such as income. Another way of analysing the data is to examine how the level of the price variable differs between those who make philanthropic contributions and those who do not. The data indicate that the tax price is somewhat lower for
140
Capturing the economic value of charities
those who make monetary donations as opposed to those who do not, at 0.86 versus 0.92. The contrast is much more striking in the case of the efficiency price, which takes an average value of 3.46 for contributors and 5.59 for noncontributors. In the case of volunteering, the data for the time price are only available for those who contribute, so that the same comparison between contributors and non-contributors cannot be made. However, it is notable that the mean time price faced by volunteers at 0.50 is substantially lower than the prices of monetary giving. This finding raises questions as to why donors do not have greater recourse to this relatively low-cost mode of philanthropy. One possible explanation is that volunteers face restrictions on the availability of time. The data provide only mixed support for this hypothesis. Participation in volunteering is indeed higher among part-time workers than full-time workers, particularly for women, where the figures are 37.5 per cent versus 27.4 per cent. However, participation is actually lower among those who are not in employment than among those who are, both for men and women. As regards the hours volunteered, there is evidence that men in fulltime employment contribute fewer hours (less than 15 hours per month) than men who are not in full-time employment (who volunteer more than 20 hours per month). However, there is no such effect for women.
6.5
EMPIRICAL RESULTS
6.5.1 Coefficient Estimates This section presents the results of estimating giving and volunteering equations based on the IGS dataset and incorporating the various price variables defined above. The objective of these equations is to estimate the extent to which donations of time and money are sensitive to the associated prices. The regression makes it possible to control for other (non-price) influences on philanthropic behaviour, so that the price effect can be isolated with greater precision. Following the earlier UK literature, two separate Heckman selectivity models are estimated for giving and volunteering (Jones and Posnett, 1991a,b; Banks and Tanner, 1997; Pharoah and Tanner, 1997). The Heckman specification has the strength of incorporating separate models for the participation decision and the decision of how much time or money to contribute. This allows for the fact that separate processes may be governing these two aspects of the philanthropic decision (Greene, 1993). Thus the demand for charitable contributions of time and money is specified as a function of prices and income and conditioned on a range of
Providing fiscal incentives for giving
141
socioeconomic characteristics such as age, birth year, education, employment status and region. The equations include both own-price and cross-price terms. However, the time price could only be included in the hours equation for volunteering given that it is only observed in the case of those who actually volunteered. The reason for including age and birth year separately is to be able to distinguish between pure age effects and cohort effects. This is one of the advantages of having repeated cross-sections of data over a series of years. Following an extensive specification search, the results presented here are based on a double log functional form (Foster, 1999c). That is to say that the dependent variable is the logarithm of monthly donations or monthly hours volunteered respectively, while the key economic variables, price and income, are also entered logarithmically. Table 6.6 presents the results of the Heckman models for giving and volunteering. The giving model reported in the left-hand side of the table is identified by a variable that captures the importance of religion to the respondent. To wit, respondents were presented with the statement ‘Religion is very important in my life’ and were asked to what extent they agreed or disagreed with it. The answers are recorded on a five-point Likert scale, ranging from –2 (strongly disagree) to +2 (strongly agree). Religious views are thought to be a suitable instrument for the identification of the selectivity term. This is because it seems plausible that religious beliefs may be correlated with altruistic preferences and hence could explain participation in philanthropic giving (Smith et al., 1995). Both the efficiency and tax-price variables are found to have a significant impact on participation in spite of the fact that only about 10 per cent of the sample give through one of the government’s tax-efficient schemes. Participation is also positively associated with higher income. Turning to the socioeconomic factors, giving was found to be significantly more likely among the highly educated and among women. There is a negative effect attributable to age, combined with a negative cohort effect. This suggests that, within any particular generation, people give less money as they grow older. However, looking across generations, older generations tend to be more generous than younger generations. Finally, the variable capturing religious beliefs is statistically significant and positively associated with participation. Turning to the contributions equation, both income and the efficiency price continue to be statistically significant, with the expected signs. The positive coefficient on birth year indicates that, although younger generations are less inclined to give to charity, the gifts that they do make tend to be larger than those of older generations. The results of the volunteering model are reported on the right-hand side of Table 6.6. The selectivity term for this model is identified by means of a
142
Correlation Variance Log-likelihood Observations
2.531 2.279 –3.123 –4.242 –10.166 0.547 0.975 0.525 0.869 –1.204 3.199 9.576
0.170 0.096 –0.286 –0.012 –0.419 0.061 0.131 0.052 0.099 –0.133 0.078 35.623 –3.618
–19.192
–3.572 –0.941
t-stat.
2.201 2.335 0.307 0.907 3.943 –2.305 –0.523 –2.534 –1.802 –1.149
–1.076 –0.366
Coeff.
Contribution
0.193 0.122 0.037 0.003 0.242 –0.325 –0.086 –0.319 –0.259 –0.163
–0.76 1.52 –2117.11 1254
–2.878 –3.527
t-stat.
–0.645 –1.087
Coeff.
Participation
Giving
Heckman selectivity models for giving and volunteering
Efficiency price Tax price Hourly wage Replacement wage Household income Years of education Sex Age Birth year Full-time employed Part-time employed North of England Scotland or Wales Southeast of England School-aged children Constant
Table 6.6
0.247 0.103 –0.107 0.006 –0.162 –0.136 0.170 –0.071 –0.193 –0.158 0.147 9.263
–1.204 –0.922
Coeff.
3.574 2.608 –1.190 2.254 –4.158 –1.210 1.349 –0.728 –1.691 –1.437 3.395 2.653
–5.243 –3.131
t-stat.
Participation
–0.19 1.10 –1291.75 1307
–5.354
–0.022 –0.758 –0.007 –0.543 –0.204 0.088 0.084 0.010 0.071 –0.443 –0.647 –0.054 –0.065 0.341
–0.895
–0.048 –1.522 –0.150 –4.869 –1.691 1.431 0.573 2.206 0.980 –2.652 –3.362 –0.366 –0.360 2.068
t-stat.
Contribution Coeff.
Volunteering
Providing fiscal incentives for giving
143
dummy variable capturing the presence of school-aged children in the household. The reasoning behind this is that many voluntary activities are organized in and around schools, suggesting that parents of school-aged children are more likely to become involved. In the participation equation, the price and income variables are statistically significant, with the expected signs. Other variables found to be positively associated with volunteering are age, education and the presence of schoolaged children in the household. The greater participation of older people may reflect the fact that many of them are retired and thus have more time available. Birth year is found to be negative and significant, indicating that younger generations are less inclined to participate in volunteering. It is perhaps surprising that neither full-employment status nor gender is found to have a significant effect on participation in volunteering. In the contribution equation, the time-price variable has been disaggregated into its two constituent components – the net hourly wage and the (inverse of) the replacement wage – so that the corresponding impacts can be separately identified. The only economic variable which proves to be significant in explaining the hours supplied by volunteers is the inverse of the replacement wage, indicating that the volunteer labour supply responds positively to the opportunity to participate in relatively high-productivity tasks. As might be expected, employment status has a significant negative effect on the number of hours volunteered, reflecting the fact that those who work have fewer hours available for other activities. The results show that older people, as well as being more likely to participate in volunteering, also tend to supply a larger number of volunteer hours. In order to gauge the impact of using the subjective tax-price variable advocated here as opposed to the objective tax-price variable used in the earlier literature, the above models were re-estimated using an objective tax price. In all cases, the use of the objective tax price substantially reduced the statistical significance of the variable but did not otherwise materially affect the pattern of results obtained. 6.5.2 Elasticities The price and income elasticities implicit in the model coefficients reported above are calculated and presented in Table 6.7. Separate elasticities are calculated for the participation decision and the level of contributions. Finally an overall elasticity estimate takes into account the combined effect of price and income variables on participation and contribution. In general terms, philanthropic behaviour is inelastic with respect to income and price, since just about all of the overall elasticities are below one in absolute value.
144
Table 6.7
Capturing the economic value of charities
Price and income elasticities for different types of philanthropic activity Time price
Giving Participation Contribution Overall Volunteering Participation Contribution Overall
Efficiency price
Tax price
–0.101 –0.342 –0.459
–0.557 –0.366 –0.969
–0.452 –0.007 –0.067
–1.135 –0.758 –0.904
Hourly wage
Replacement wage
Income
0.087 0.193 0.286
–0.007 –0.007
–0.543 –0.543
0.303 –0.204 –0.165
The own-tax-price effect for giving as well as the cross-tax-price effect for volunteering both lie in the interval –0.9 to –1.0, indicating that philanthropic behaviour is inelastic with respect to the tax price. The negative sign on the cross-tax-price of volunteering provides evidence of complementarity between these two types of philanthropic activity. The efficiency-price elasticity for giving has a substantial effect on behaviour, with an overall value of around –0.46. However, this is approximately half the size of the corresponding tax-price elasticity. In the case of volunteering, the efficiency price has a substantive effect on participation, with a value of around –0.45 but the overall impact is negligible. As regards the time price, the impact of the net hourly wage is shown to be negligible. However, the replacement wage does have a substantial effect upon the contribution of volunteer hours, given an elasticity of –0.54. It should be noted that the time-price effects are in all likelihood underestimated. This is partly because the method used to impute the hourly wage was such as to dampen the degree of variation in this variable. Furthermore, the fact that the time price could not be observed for non-volunteers means that it is not possible to take into account the impact of this price variable on the decision to participate in volunteering. The income elasticities are small. For giving, the overall income elasticity lies just below 0.30, while for volunteering, the overall income elasticity is close to zero, given that the positive elasticity for participation is more than offset by the negative elasticity for contribution. Thus, in general, income elasticities are substantially lower than price elasticities.
Providing fiscal incentives for giving
145
These results can be related back to the concepts of treasury efficiency and neutrality developed by Barrett (1991) and defined in the introduction to this chapter. The finding that the tax-price elasticity of giving generally exceeds the income elasticity indicates that an increase in the marginal tax rate will be non-neutral, generating a positive net impact on donations. The reason is that the positive incentive arising from a lower tax price is large enough to outweigh the reduction in giving that arises from a reduction in net household income. However, this same increase in the marginal tax rate will not be treasury-efficient. This is because donations are inelastic with respect to the tax price, so that the additional donations resulting from the lower tax price will not be large enough to offset the resulting loss of revenue to the Treasury. Consequently the overall supply of public goods (those provided by charities plus those provided by the state) will fall. Finally, the results obtained raise an important concern about equity. Given that the tax system is broadly progressive, whereas the income elasticity of voluntary contributions is well below unity, a switch in government policy away from direct grants to voluntary organizations and towards enhanced fiscal incentives for giving will tend to be regressive in its distributional impact.
6.6
CONCLUSIONS
This chapter has identified three concepts of price which are relevant in studying the economic aspects of philanthropic behaviour: the efficiency price, the tax price and the time price. Furthermore, it was argued that these price variables should ideally be measured in subjective rather than objective terms, taking into account the effect of donor perceptions. The data indicated that the perceived efficiency price of contributions is comparatively high, while the time price among those who undertake volunteering is comparatively low. The resulting estimates of price and income elasticities have important implications for fiscal policy towards the sector. On the one hand, the finding that tax-price elasticities are in general higher than income elasticities provides comfort to charities that fiscal incentives do indeed have a net positive effect on the overall volume of contributions. On the other hand, the fact that tax-price elasticities were found to be somewhat below unity indicates that fiscal incentives for philanthropic giving come at the expense of reducing the overall supply of public goods produced in aggregate by the public and charitable sectors. This result clearly reduces the attractiveness of such fiscal incentives from the government’s point of view and suggests that direct grants may be a more efficient means of lending public sector support to voluntary organizations.
146
Capturing the economic value of charities
The study has also revealed that other concepts of price, which fall more directly within the influence of charity fundraisers, also have an important impact on philanthropic behaviour. In particular, both giving and volunteering are significantly affected by the level of charity efficiency, suggesting that improvements in this parameter will yield dividends in terms of increased donations. Given that donors were found to underestimate substantially the efficiency with which charities convert monetary contributions into ‘good works’, it probably makes most sense for charities to begin by bringing public perceptions of efficiency closer to reality rather than by improving the actual underlying level of efficiency. Furthermore, the supply of volunteer labour is positively related to the productivity of the activities undertaken by volunteers. This suggests that it may be possible to increase the number of volunteer hours by paying more careful attention to the nature of volunteering opportunities and in particular by ensuring that these include a greater number of relatively high-value activities. An important limitation of the study summarized in this chapter was the absence of data that might provide a more direct estimate of the price of time. As a result of this, the time price had to be inferred using procedures that were not always entirely reliable. In addition, these procedures could only be applied to those who were actually volunteering, thereby precluding any estimates of the impact of the time price on the participation decision or of the cross-price effects on giving. Future surveys of volunteer behaviour should pay more careful attention to this issue.
7. Choosing fundraising methods 7.1
INTRODUCTION
The previous chapter showed that fiscal incentives in the UK do not have such a powerful effect on the disposition to give. Probably more effective than government efforts to promote charitable giving are those that charities make themselves in the form of fundraising. This chapter estimates the returns to these fundraising efforts, and explores the relative efficacy of alternative fundraising techniques. The charity fundraiser is a ubiquitous figure in the real world of philanthropic finance. Indeed, everyday experience suggests that most philanthropic gifts are made in response to some sort of request by a charity fundraiser. A number of voluntary sector statistics indicate that charitable organizations have developed a substantial fundraising apparatus. To give an idea of scale, among the top 500 fundraising charities in the UK, fundraising accounted for 8.8 per cent of total expenditure in 1996/97, amounting to nearly £370 million for the year (Pharoah and Smerdon, 1998). Indeed, between professional charity fundraisers and volunteers, fundraising is estimated to occupy the equivalent of over 150 000 full-time employees. Notwithstanding these fairly well-known features of voluntary sector organization, the charity fundraiser has featured surprisingly little in economic models of philanthropic behaviour, most of which assume that giving is a spontaneous decision of the utility-maximizing consumer. Although there have been some notable recent attempts to remedy this omission (Andreoni, 1998), very little attention has been devoted to the question of how the efforts of the charity fundraiser affect the potential donor’s disposition to make a charitable contribution. The purpose of this chapter is to incorporate the charity fundraiser into the conventional model of philanthropic behaviour, and to conduct empirical tests of the resulting theoretical predictions exploiting a rather unique data source. The results provide some considerable support for the view that the role of the fundraiser is to enhance the net private benefit – or warm glow – associated with philanthropic giving, both by reducing transaction costs and increasing social prestige. They also suggest that some fundraising methodologies are more successful than others in this respect. 147
148
Capturing the economic value of charities
The remainder of this chapter is organized as follows. Section 7.2 puts the present research in the context of the existing literature on the motivation for philanthropic giving. Section 7.3 develops a conceptual framework which incorporates the charity fundraiser into the conventional models of charitable giving and uses this framework to derive a number of testable hypotheses. Section 7.4 introduces the dataset and conducts some preliminary hypothesis tests based on simple descriptive statistics. Section 7.5 develops a statistical model of charitable giving which controls for the intensity of fundraising activity and the socioeconomic characteristics of the targeted population. The predictions of this model are used to repeat the earlier hypothesis tests in a more reliable manner. The main conclusions of the analysis are drawn out in Section 7.6. A more extensive discussion of the material can be found in Foster (1999b).
7.2
MOTIVATION
According to traditional theory, philanthropic giving is purely an expression of donors’ altruistic preferences for the supply of the associated public good (Samuelson, 1954; Olson, 1965). As noted by Andreoni (1988, 1989, 1990), the theory of pure altruism predicts that in large economies the free-riding problem becomes so acute as to drive charitable donations virtually to zero. This is because the incentive to free-ride increases as the effect of each individual’s contribution on the overall supply of the public good becomes vanishingly small. Such a conclusion is clearly contradicted by the existence of a substantial voluntary sector. Much of the recent theoretical literature on charitable donations has in large part been motivated by the need to reconcile the traditional theory with observed reality (Sugden, 1982; Steinberg, 1987). To this end, a number of extensions have been made to the traditional model, and while there remain subtle but important differences between them, what they all seem to have in common is the incorporation of some sort of private benefit which is generated as a by-product of philanthropic giving. Thus Sugden (1984) hypothesizes that individuals hold moral principles which require them to contribute their fair share of effort to the financing of public goods; that is to say, they believe it is morally wrong to free-ride as long as other people are giving. Consequently, utility functions depend not only on the aggregate supply of the public good, but also on the effort that each individual contributes in relation to the effort of his peers. In a recent empirical study Andreoni and Scholz (1998) find some limited, though not overwhelming, evidence in support of this view.
Choosing fundraising methods
149
Andreoni (1989) shows how the difficulties associated with the traditional model can be resolved by taking into account the fact that donors experience a warm glow from the act of contributing that is somehow related to the magnitude of the donation. He describes the resulting model as one of impure altruism, to reflect the fact that people care both about the supply of the public good and about the size of their own personal contribution to it. The exact nature of the warm glow is not specified in the paper, but the concept is susceptible to a wide range of interpretations. Perhaps the simplest example of a warm glow is the case where the charitable donation is tied to the purchase of a private good, such as charity Christmas cards or a ticket to a charity concert. However, the warm glow could equally be taken to be referring to a sense of moral satisfaction which attaches to the philanthropic act, and to that extent bears some relationship to Sugden’s theory. Finally, a more recent strand of the literature has postulated that particularly those donors who make relatively large-scale donations reap private benefits in the form of social prestige (Glazer and Konrad, 1996; Harbaugh, 1998). This theory is based on the observation that charities often take special care to publicize the identity of donors. Thus the authors argue that charitable giving provides a signal of wealth that could be considered as a substitute for more conventional forms of conspicuous consumption. Once again, there are parallels with the warm glow, except that in this case the amount of prestige attached to a particular donation will depend not only on the size of the donation itself but also on the methods available to the recipient charity for generating prestige. All of these theories implicitly assume that the decision to give is taken spontaneously by the utility-maximizing consumer. This overlooks a particularly striking and prevalent feature of charitable giving as it occurs in practice, namely the fact that philanthropic donations are invariably made in response to some form of direct elicitation from a charity fundraiser, for example, being targeted in a door-to-door collection or through a charity mail shot. Indeed, the voluntary sector has developed a substantial fundraising apparatus for undertaking such activities. As noted above, among the top 500 fundraising charities in the UK, fundraising accounted for 8.8 per cent of total expenditure in 1996/97, amounting to nearly £370 million for the year (Pharoah and Smerdon, 1998). Moreover, the UK Institute of Charity Fundraising Managers, which covers the majority of paid professional fundraisers active in the country, boasts a current membership of some 3000 people. However given that the bulk of fundraising effort is provided by volunteers, these figures substantially understate the true scale of fundraising activity. For example, during the early 1990s, fundraising absorbed some 15.5 per cent of time volunteered to charitable organizations (Halfpenny et al., 1994). This is approximately equivalent to a further
150
Capturing the economic value of charities
150 000 full-time employees, or about 50 times the number of professional fundraisers. A recent paper by Andreoni (1998) goes some way towards acknowledging the importance of fundraising activity by developing a model of the role of charity fundraisers in securing seed money for large-scale capital campaigns. The model explains the existence of fundraising activities as the solution to a potential coordination failure that can arise when there are increasing returns to scale in the supply of the public good. If the public good can only be provided when aggregate donations exceed a certain high-level threshold, donors will be unwilling to give at all unless they can be assured that complementary funds will be forthcoming from other sources. Charity fundraisers are able to provide this assurance by initially securing a small number of large gifts from strategic donors. This seed money subsequently serves to elicit a large number of small donations from the general public. However, Andreoni’s paper (1998) does not consider how the interaction between the fundraiser and his target influences the latter’s decision to make a contribution. This effect was found to be particularly important in Freeman’s (1997) study of volunteering in the USA. Freeman shows that the single most important factor determining volunteer status is whether or not a person has been asked to volunteer. Specifically, his results are that 89 per cent of those asked to volunteer during the course of the year did so, whereas only 29 per cent of those who were not asked spontaneously chose to participate in voluntary activities. When a dummy variable indicating whether or not a person was asked to volunteer is incorporated into a probit equation for participation in volunteering, the resulting effect is found to be much larger than that obtained from conventional socioeconomic and demographic factors, and indeed reduces the coefficients attached to these other variables. On the basis of these results, Freeman (1997) hypothesizes that volunteering is a conscience good and concludes that ‘people have a latent demand for such a good, which a request brings to the fore, even if they would prefer to free-ride on the provision of the good’. Building on these findings, the purpose of this chapter is to develop a simple conceptual model of philanthropic contributions which acknowledges the central role of the charity fundraiser. This model generates a number of hypotheses, which are subsequently subjected to empirical testing.
7.3
CONCEPTUAL FRAMEWORK
The standard model of impure altruism developed by Andreoni (1989, 1990) postulates that philanthropists derive utility from their donations in two dis-
Choosing fundraising methods
151
tinct ways. They benefit indirectly, in so far as their gift contributes to expanding the supply of the philanthropic good. Although in situations where the individual’s gift makes only a negligible contribution to the overall supply of the public good, as is the case for the vast majority of donors who make small contributions to large charities, these indirect benefits could be expected to be vanishingly small. Donors also benefit directly from their charitable gifts inasmuch as these generate warm glows. These warm glows are typically modelled as a function of the size of the donation that is made. In the present chapter, this conventional framework will be extended by postulating that the magnitude of the warm glow depends not only on the size of the donor’s charitable gift, but also on the amount of fundraising effort directed towards the individual. Whereas the size of the charitable gift is clearly a choice variable for the individual, the exposure to fundraising effort is determined exogenously by the charity fundraiser. Warm glows thus arise out of the interaction between the actions of the charity fundraiser and the reactions of the potential donor. That is not to say that warm glows would not exist at all in the absence of interventions by fundraisers, but rather that the initiatives of fundraisers serve to amplify the net private benefits of giving. There are at least two different mechanisms that explain how the actions of a charity fundraiser might be expected to magnify the warm glow experienced by the donor. The first of these is that charitable fundraisers may reduce the transaction costs of making a donation. They do this both by passing on information about charitable causes and by providing a simple procedure for giving, such as handing over a few coins, placing a cheque in an envelope or ringing a toll-free number. By reducing the private costs associated with making a donation, fundraising efforts serve to increase the net private benefit of giving. The second way in which fundraising efforts may enhance the warm glow of giving is that they often create situations where the donor’s philanthropic actions can be observed by others and thus attract some degree of social recognition or prestige. This prestige mechanism could also function in reverse. For example, while there may not be a great deal of prestige associated with placing a few coins in a plate, a person who failed to do so might well meet with social disapproval. As noted by Becker (1974), ‘apparent “charitable behaviour” can also be motivated by a desire to avoid the scorn of others’. Thus the encounter with a charity fundraiser may elicit in the potential donor a sense of shame, embarrassment or remorse about not making a charitable contribution. In this case, the act of giving provides relief from the moral or social discomfort generated by the charity fundraiser. The level of fundraising effort targeted at any given individual can be measured along two dimensions. The first is the number of times the individual is approached by a charity fundraiser, and the second is the quality of
152
Capturing the economic value of charities
the fundraising methodology used. High-quality methods are defined as those that, other things being equal, have the greatest capacity to enhance the warm glow of giving. For example, it appears likely that any fundraising method entailing direct face-to-face contact with the potential donor will be of higher quality in this sense than a fundraising method entailing only remote contact with the potential donor, such as via mail appeals, or newspaper or television advertisements. There are two reasons for thinking that this might be the case. The first is that direct fundraising methods are likely to represent lower transaction costs for donors than remote methods. This is because the former tend to involve simply handing over cash, whereas the latter tend to require some sort of formal financial transaction such as writing a cheque, which involves additional time and effort on the part of the donor. The second reason for thinking that direct fundraising methods may be more effective at generating warm glows is that they are more likely to be witnessed by third parties, and thus more effective in generating social recognition. At the very least, the charity fundraiser will be there to observe whether or not the person gives, and in some cases also how much they give. Furthermore, in many cases, direct fundraising methods exploit social situations such as the workplace, pub or church. Thus the philanthropic behaviour of potential donors is observed by a wider peer group, whether it be fellow colleagues, drinkers or worshippers. In contrast to this, remote giving tends to be more of a private decision that may not be observed by anyone else. The conclusion of all this is that a person who gives by a direct fundraising method will experience a larger warm glow, and thus a higher utility level than a person who gives by means of a remote fundraising method. This implies that, other things being equal, those approached by direct methods will be more inclined to give than those approached by remote methods. As a basis for testing these conjectures, two null hypotheses are established. According to these, the choice of fundraising method, whether direct or remote, will have no significant effect either on the probability of obtaining a charitable gift or on the size of the gift obtained. 1.
2.
Participation null hypothesis (H01): direct fundraising methods are no more successful than remote fundraising methods in eliciting charitable contributions. Alternative participation hypothesis (HA1): direct fundraising methods are more successful than remote fundraising methods in eliciting charitable contributions. Size of the gift null hypothesis (H02): there is no difference in the size of the donations generated by direct and remote fundraising methods.
Choosing fundraising methods
153
Alternative size of the gift hypothesis (HA2): donations generated by direct fundraising methods are larger than those generated by remote fundraising methods. Failure to reject both of these null hypotheses would contradict the overall theory that the choice of fundraising method affects the magnitude of the warm glow obtained. Rejection of the first hypothesis, without being able to reject the second hypothesis, would lead to the conclusion that the choice of fundraising method does matter, but that it matters only in determining the probability of participation. Finally, rejection of both null hypotheses in favour of the corresponding alternatives would indicate that the choice of fundraising method influences both the probability of participation and the size of the resulting gift.
7.4
RESULTS OF SIMPLE HYPOTHESIS TESTS
The hypotheses advanced in the previous section are tested making use of data on charitable donations contained in the Individual Giving Survey (IGS) for the period 1990–93 (Halfpenny et al., 1992, 1993, 1994). The IGS was an annual repeated cross-section survey of philanthropic behaviour in the UK funded by the Charities Aid Foundation up to 1993. It covered about 1000 individuals each year, chosen to be representative of the UK population. A structured interview conducted in the respondent’s home was used to collect information about philanthropic activities during the month before the survey. The IGS has a number of unique features that make it particularly well suited to testing the hypotheses stated above, and more generally to examining the role of fundraising in generating charitable gifts. First, the IGS distinguishes between 11 different fundraising methods commonly used for eliciting philanthropic contributions. These are street collections, door-todoor collections, church collections, sponsoring schemes, shop-counter collections, pub collections, work collections, advertisement appeals, television appeals, mail appeals and telephone appeals. Second, for each of these fundraising methods, the IGS establishes how many times the individual was approached, how many times they actually gave, and the total value of donations. Third, the IGS distinguishes between a number of different methods of obtaining charitable donations by means of selling conventional goods and services, such as raffles, charity events, charity shops, jumble sales and charity goods catalogues. However, attention here is confined to fundraising methods that aim to elicit gifts for which there is no tangible private return. One limitation is that the survey does not record how much was donated on each giving occasion, but only the total donations for each fundraising method.
154
Capturing the economic value of charities
This creates a problem when it comes to comparing methods that are used with differing intensities, in that only average gifts can be compared. To the extent that the size of the marginal contribution is likely to decline with each successive gift, the fact that one method elicits a higher average gift than another may simply be attributable to its lower frequency of use rather than to any intrinsic characteristic of that method. A second limitation of the dataset is the possibility that respondents may have had the incentive to overstate the extent of their donations in order to appear charitable to the interviewer. This would have the effect of biasing the survey data upon which this analysis is based. It is not possible to prove whether or not this effect is actually present in the data. However, even if it were, the resulting bias should only affect the absolute levels of the variables rather than their relative values across different fundraising methods, and the tests reported in this chapter are concerned solely with relativities. Table 7.1 provides an initial overview of the relationship between fundraising approaches and giving behaviour. The first point to note is the wide reach of fundraising activity in that 74.1 per cent of respondents had been approached by a charity fundraiser during the preceding month. Furthermore, fundraising approaches seem to enjoy a very high success rate: among those who were approached at least once, as many as 91.2 per cent gave money at least once. On the other hand, spontaneous unsolicited giving is very much a minority activity; only 1.9 per cent of those who were not approached by a charity fundraiser during the preceding month had chosen to make a donation. Table 7.1
Cross-tabulation of fundraising approaches and giving behaviour (% of the sample) Gave at least once?
Approached at least once?
No Yes
No
Yes
25.5 6.5
0.5 67.6
Table 7.2 provides a more detailed breakdown of donating behaviour according to which of the 11 fundraising methods was used. The table presents a number of different summary statistics for each method. The first column refers to the response that respondents made to an attitudinal question enquiring how likely they were to give when approached in each of these different ways. The results are coded on a Likert scale ranging between +2 (signifying ‘very likely to give’) and –2 (signifying ‘very unlikely
155
0.02 0.02 0.03 0.02 0.02 0.03 0.03 0.02 0.03 0.02 0.01 0.01 0.01 0.01
0.54 0.68 0.27 1.17 –0.14 –0.29 0.22
–0.61 –0.06 –1.05 –1.47
0.38 –0.80 –0.06 3.05 0.48 3.58
0.14 0.11 0.21 0.03
0.87 0.63 0.50 0.41 0.34 0.25 0.10
0.08 0.03 0.09
0.02 0.01 0.02 0.01
0.04 0.02 0.03 0.02 0.02 0.02 0.01
Std. err.
Mean
Mean Std. err.
Number of approaches per person (#)
Attitude towards fundraising method (#)
Summary statistics for different fundraising methods
Direct methods Street collection Door-to-door collection Church collection Sponsor scheme Shop-counter collection Pub collection Work collection Remote methods Appeal by advertisement Television appeal Mail appeal Telephone appeal Overall Direct methods Remote methods All methods
Table 7.2
91.55 37.54 91.18
24.84 48.39 27.21 23.26
84.32 85.57 95.41 93.73 77.78 92.22 91.04
Mean
0.43 2.00 0.60
3.50 3.18 2.60 6.52
1.12 1.01 0.92 0.83 2.12 1.47 2.02
Std. err.
Probability of making a gift (%)
1.86 8.85 1.94
9.96 7.57 9.16 16.28
0.67 1.76 3.27 2.41 0.43 0.91 3.82
Mean
0.13 0.80 0.13
2.97 0.95 1.15 7.69
0.03 0.65 0.54 0.11 0.03 0.09 1.42
Std. err.
Donation (£)
1.46 2.69 1.40
2.18 3.17 1.97 2.17
0.51 1.36 2.62 2.28 0.33 0.80 3.17
Mean
0.08 0.31 0.08
0.80 0.48 0.33 1.05
0.02 0.54 0.36 0.11 0.03 0.07 1.17
Std. err.
Revenue (£)
156
Capturing the economic value of charities
to give’). The second column reports the average number of times a respondent was approached by each method, while the third column gives the probability that those who were approached at least once by each method went on to make a donation. The average size of the donation is reported in the fourth column, and the average revenue per approach in the fifth and final column. The final row of the table, which reports the summary statistics for all fundraising methods, confirms the earlier impression that fundraising is a ubiquitous phenomenon. On average, the survey respondents were approached by a fundraiser 3.6 times during the course of the month, or about once every eight and a half days. About 91.2 per cent of those approached made a donation and the average value of such donations was £1.94. Overall this implies an average return to each fundraising approach of £1.40. Summary statistics for direct and remote fundraising methods provided in Table 7.2 testify to some striking differences in these two different fundraising methods. The direct methods are characterized by relatively high probabilities of making a gift (in excess of 90 per cent) and relatively low average cash donations (of less than £2.00). By contrast, the remote methods are characterized by relatively low probabilities of making a gift (less than 40 per cent) and relatively large average cash donations (in excess of £8.00). Interestingly, the lower disposition to make a donation when approached by a remote method is reflected in the lower scores that these methods received on the attitudinal question, where respondents were asked subjectively to evaluate their likelihood of making a donation. Thus direct methods score an average of 0.38 while remote methods score only –0.80. In order to test whether these differences are statistically significant, and the extent to which they conform to the conjectures stated above, a series of more formal hypothesis tests is conducted. The first hypothesis stated above concerns possible variations in the probability of obtaining a gift using different fundraising methods. This is tested by creating a variable (∆P), which is defined as the difference between the probability (P) of obtaining a gift from the direct (D) versus the remote (R) methods. The null hypothesis, which is stated formally as (H01), stipulates that this difference should be equal to zero, while the alternative (HA1) holds that this difference should be strictly positive. Using the difference variable, a simple t-test results in the rejection of the null hypothesis in favour of the alternative. This indicates that there is a significantly higher probability of success (the average difference is 41.9 per cent) when direct fundraising methods are used as opposed to remote methods. (H01) ∆ P = PD − PR = 0
(HA1) ∆ P = PD − PR > 0
(7.1)
Choosing fundraising methods
157
The second hypothesis stated above concerns possible variations in the average donations received from different fundraising methods. Once again, this is tested by creating a difference variable (∆C) for the average donations (C/N) resulting from the two types of fundraising methods. In this case, the null hypothesis (H02) requires that this difference be equal to zero, while the alternative hypothesis (HA2) holds that the difference is strictly positive. Once again, a t-test procedure based on the difference variable results in rejection of the null hypothesis. There are significant differences in the size of the average gift obtained from direct versus remote fundraising methods. However, curiously this difference does not run in the direction implied by the alternative hypothesis (HA2). In fact, rather, the average cash gifts obtained from direct fundraising methods are significantly smaller than those obtained from remote fundraising methods. The average difference is –£6.76.
(H02) ∆ C =
CD CR − =0 ND NR
(HA2) ∆ C =
CD CR − >0 ND NR
(7.2)
Further confirmation of this result comes from comparing the frequency distributions of the average donations obtained by the two methods. These are illustrated graphically in Figure 7.1. From the figure it is clear that the distribution of average gifts generated by direct fundraising is heavily skewed towards smaller values, with the modal point occurring in the size category of
45 40 35 30 25 20 15 10 5 0
<£1
£1–2
£2–3 Direct
Figure 7.1
£3–4
£4–5
£5–10
>£10
Remote
Percentage frequency distribution of average donations for different fundraising methods (£)
158
Capturing the economic value of charities
gifts worth less than a pound. By contrast, the distribution of average gifts from remote fundraising methods is skewed towards larger donations. The overall distribution is bimodal with a local mode occurring in the size band £2 to £3 and the overall mode occurring in the size band £5 to £10. Using the Kolmogorov–Smirnov procedure (Daniel, 1990), it can be shown that the differences between the two distributions are in fact statistically significant. Thus, overall, the results of the hypotheses tests provide strong support for the general view that the choice of fundraising method materially affects the disposition to give. While the higher success rate of direct fundraising methods is in line with prior conjectures, the fact that remote methods tend to generate larger gifts provides something of a puzzle. One possible explanation is that the larger transaction costs associated with making a remote gift via a formal financial transaction serve to discourage smaller donations. People rarely write (and institutions rarely accept) a cheque for a sum as small as 50 pence, whereas they often make cash purchases of this size. The frequent use of cash in direct giving may therefore account for the prevalence of smaller gifts by this means.
7.5
CONTROLLED HYPOTHESIS TESTS
7.5.1 Potential Biases in Previous Results The results presented in the preceding section are potentially misleading for a number of reasons. In particular, on the basis of these simple tests, it is not possible to say whether the differences in performance observed between the two types of fundraising methods are attributable to the methods themselves. They may instead be due to some other difference, which happens to be correlated with the use of one or other of the two methods. One possible reason why the earlier results may not be entirely trustworthy is that the two fundraising methods are used with differing degrees of intensity. For example, if people tend to be targeted more frequently by direct fundraising methods than by remote fundraising methods, it is not surprising that the average donation reported would be lower for direct methods than remote methods. Table 7.2 confirms that direct methods are indeed more intensively used than remote methods, with an average of 3.05 approaches per month for the former and 0.48 approaches per month for the latter. A second plausible explanation is that the direct methods have a higher success rate not because of their superior ability to generate warm glows but because their use tends to be concentrated in segments of the population that have a higher pre-existing disposition to give. For example, it may be that
Choosing fundraising methods
159
direct methods are more likely to be used on relatively high-income individuals and that high-income individuals are more likely to make charitable donations. In order to test for this effect a number of ordinary least squares (OLS) regression models were estimated relating the logarithm of the number of fundraising approaches to a range of exogenous factors including socioeconomic characteristics of the targeted respondents, regional and population density factors, and seasonal dummies. The results are reported in Table 7.3, which presents separate models for direct and remote fundraising methods. The overall explanatory power of these models is comparatively low, with adjusted R2 statistics ranging between 6 per cent and 7 per cent. However, the statistics for the F-test, which lie in the range 8 to 12, indicate that the regressions are statistically significant overall. Moreover, the presence of a number of statistically significant Table 7.3
Ordinary least squares regression models for the number of fundraising approaches Direct methods
Log (household income) Years of education Sex Age Birth year Full-time employed Part-time employed North of England Scotland or Wales Southeast of England Population density Winter season Autumn season Spring season Constant Correlation F-statistic Adjusted R2 Observations
Remote methods
Coeff.
t-stat.
Coeff.
t-stat.
0.119 0.121 –0.024 0.019 –0.022 –0.095 0.039 –0.107 0.067 –0.180 –0.051
3.878 5.694 –0.528 0.915 –1.077 –1.669 0.578 –2.047 1.095 –3.056 –2.838
0.075 0.077 0.007 –0.020 0.022 –0.020 –0.013 –0.085 –0.108 –0.020
4.047 6.056 0.026 –1.592 1.742 –0.584 –0.325 –2.763 –3.081 –0.589
0.089 0.051 0.031 0.211
2.525 1.427 0.877 0.180
–3.672
–1.946 0.267 11.20 0.07 1615
0.257 8.48 0.06 1578
160
Capturing the economic value of charities
coefficients indicates that the targeting process used by charity fundraisers is far from random. A comparison of the separate models estimated for direct and remote fundraising methods reveals some substantial differences in the magnitude of coefficients and in the pattern of significance of the explanatory variables. For example, younger generations are more likely to be approached by remote methods, while those in full-time employment are less likely to be approached by direct methods. Furthermore, there are some significant regional differences. Direct methods are less likely to be used in the southeast of England, whereas remote methods are less likely to be used in Scotland or Wales. In both cases, the point of reference is the Midlands region. A Wald test was conducted to test the null hypothesis that there are no significant differences between the types of people approached using direct versus remote fundraising methods (Greene, 1993). This hypothesis was strongly rejected, indicating that the two different types of methods tend to reach very distinct constituencies. A third possible explanation for the results reported in the previous section is that the use of these two different fundraising methods is not uniform across different types of charities. For example, if direct fundraising methods tend to be used disproportionately by more popular charities, this could provide an alternative explanation of why these techniques tend to meet with a higher rate of success. The presence of this effect is much harder to establish, because it is very difficult to observe and measure a subjective variable such as charity popularity, which may in part be related to the nature of the good cause and in part to the nature of the specific charitable organization (its age, size, reputation and so on). For all of these reasons, the present study will not attempt to control for the popularity of the cause in determining the response to different types of fundraising methods. This position draws some support from a recent study of fundraising methods (Sargeant and Kaehler, 1998), which finds that the performance of various different methods does not vary significantly according to the nature of the charity. 7.5.2 Controlling for Extraneous Factors In view of the issues identified above, the subsequent analysis of fundraising performance will control for differences in the nature of the targeted population and differences in the number of fundraising methods. On the basis of the conceptual framework developed above, this can be done by estimating a donations equation. The corresponding statistical model is used to isolate the effect of fundraising method on giving behaviour, while controlling for other influencing factors, including the number of fundraising approaches and a
Choosing fundraising methods
161
range of socioeconomic and demographic factors as well as economic variables such as price and income. The two price variables used in the giving equation are described extensively in Chapter 8 below. Briefly, the efficiency price represents the rate at which charities absorb donations in administrative expenditures, while the tax price reflects the presence of fiscal incentives for charitable giving. The quality of the fundraising methods directed towards each individual respondent (Q) is measured by an index variable defined in equation (7.3) below, which gives the ratio of direct (D) versus remote (R) approaches. A high value of this index (greater than one) indicates that the respondent was approached predominantly by means of direct fundraising methods, and vice versa. The average value of the index in the sample population was found to be 3.39, indicating the preponderant use of direct fundraising methods.
D + 1 Q= R + 1
(7.3)
Donations are modelled using a Heckman selectivity framework. The principal strength of this model is that it takes into account the fact that different processes may be at work in determining the decision to participate in charitable giving and the size of the donation made. This is achieved by estimating separate equations for the decision to participate in philanthropic giving versus the decision of how much to contribute (Greene, 1993). In the context of the Heckman selectivity model, the two hypotheses advanced above can be restated as hypotheses about the sign of the coefficients on the quality index (Q). In the case of the first hypothesis, the null (H01) requires that the coefficient on the quality index in the participation equation (P) be equal to zero, while the alternative (HA1) entails that it be strictly positive.
(H01) β QP = 0
(HA1) β QP > 0
(7.4)
In the case of the second hypothesis, the null (H02) requires that the coefficient on the quality index in the contribution equation (C) be equal to zero, while the alternative (HA2) entails that it be strictly positive.
(H02) β QC = 0
(HA2) β QC > 0
(7.5)
Table 7.4 reports the results of the corresponding Heckman selectivity model. The first point to note is that the two variables that capture the number and quality of fundraising methods are highly significant, both in the participation equation and in the donations equation. Moreover, in the participation
162
Table 7.4
Capturing the economic value of charities
Selectivity models for all philanthropic gifts with controlling for fundraising effort Participation
Log (efficiency price) Log (tax price) Log (household income) Years of education Sex Age Birth year Full-time employed Part-time employed North of England Scotland or Wales Southeast of England Importance of religion Log (approaches) Log (quality index) Constant Rho Sigma Log-likelihood Observations
Contribution
Coeff.
t-stat.
Coeff.
t-stat.
–0.824 –0.311 0.143 0.006 –0.373 –0.077 0.080 0.041 0.079 0.032 0.494 0.309 0.045 1.627 0.953 4.459
–2.077 –0.569 1.168 0.073 –2.283 –0.946 0.979 0.194 0.289 0.181 2.117 1.543 0.852 12.407 8.422 0.598
–0.983 –0.200 0.202 0.087 0.006 –0.029 0.032 0.094 0.200 –0.010 0.002 –0.010
–4.444 –0.713 3.234 2.372 0.074 –0.710 0.801 0.908 1.703 –0.106 0.016 –0.099
0.777 0.185 3.060
11.616 2.505 0.824
0.27 1.08 –1554.83 1110
equation, the coefficients for these two variables are substantially larger than for any of the others. The positive coefficient on the quality index in the participation equation indicates a clear rejection of the null hypothesis stated as (H01): controlling for all other factors, direct fundraising methods do appear to be more successful in eliciting charitable gifts. Moreover, the quality index continues to carry a positive coefficient in the contributions equation, a result that constitutes a clear rejection of (H02). When all other factors are taken into account, it turns out that the more that people are approached predominantly through direct fundraising methods, the larger the gifts that they tend to make. This result is at variance with the conclusion drawn from the analysis of the raw data. It suggests that the larger average gifts observed from remote fundraising
Choosing fundraising methods
Table 7.5
Estimated marginal effects of fundraising
Aggregate approaches Separate approaches Direct Remote
163
Participation (%)
Contribution (£)
Overall (£)
12.39 (0.47)
0.92 (0.05)
0.81 (0.05)
20.11 (0.74) 2.38 (0.22)
1.11 (0.07) 0.27 (0.03)
0.93 (0.05) 0.27 (0.04)
methods may have been a consequence of the relatively low incidence of remote fundraising approaches, or of differences in the generosity of the population that tends to be reached by means of remote methods. With a view to quantifying the size of these differentials, Table 7.5 reports the marginal effect for each fundraising approach. In overall average terms, an additional fundraising approach adds just over 12 per cent to the probability of obtaining a charitable gift, and the expected value of that gift is £0.92. Taking these two effects in combination, the expected return from an additional fundraising approach is £0.81. Separating out the marginal effects between direct and remote fundraising methods reveals that the marginal effect of a direct approach on the probability of securing a contribution is of the order of 20 per cent. The corresponding value for a remote approach is almost an order of magnitude smaller, at around 2 per cent. The value of the marginal gift is £1.11 for direct methods, and substantially smaller at £0.27 for remote methods. Thus the overall expected marginal gain of an additional approach varies between £0.93 for direct approaches and £0.27 for remote approaches. These marginal effects, which are reported at the sample averages, conceal some informative variations across the sample. In particular, from a charity perspective it is interesting to enquire how rapidly the marginal pay-off declines as the intensity of fundraising effort is increased. To this end, Figure 7.2 plots the marginal effect on the probability of participation against the number of approaches of each kind that have already been made to the corresponding individuals. A number of interesting conclusions emerge. As might be expected, these marginal effects decline with additional fundraising approaches. This is no more than to say that the second derivative of the donation function with respect to fundraising effort is negative. However, the nature of this decline differs markedly between direct and remote fundraising methods.
164
Capturing the economic value of charities Direct
Marginal effect on probability of giving (%)
50
Remote
40 30 20 10 0 0
Figure 7.2
1
2
3
4 5 6 7 No. of approaches
8
9
10
>10
Marginal effect of an approach on the probability of a gift against number of approaches
For direct methods, the probability of obtaining a gift is initially very high. It is estimated that nearly half of those who have not been approached by a fundraiser in the preceding month would make a gift if this were elicited directly. However, this disposition to give tails off dramatically as the number of approaches increases. By the second request, the probability of success is little over 10 per cent, while by the fourth request it has tailed off virtually to zero. The case of remote methods provides a strong contrast. The probability of success is initially very much lower, at around 10 per cent; however, the rate of decline is much more gradual, so that even by the tenth approach the probability of success still falls just short of 5 per cent. This difference can probably be explained as follows. Owing to the higher probability of making a gift by direct methods, people are very likely to have made at least one such gift once they have been approached a couple of times. They may consequently be satiated with respect to charitable giving for that time period, and hence would fail to respond to any subsequent requests. Given the much lower probability of an immediate positive response to a remote approach, it is quite possible that even people who have been approached half a dozen times may not have made any charitable gifts. Hence the satiation effect is far less prevalent.
Choosing fundraising methods
7.6
165
CONCLUSIONS
The current chapter began with the observation that conventional economic models of philanthropic behaviour have tended to overlook the pervasive presence of the charity fundraiser. With a view to overcoming this deficiency, fundraising effort was incorporated as a new component within the conventional analytical framework for charitable giving. The expanded model postulates that fundraising effort interacts with individual donations in the production of the warm glow or private benefit associated with philanthropic behaviour. Two ways in which fundraising effort was thought to enhance the warm glow were by reducing the transaction costs of giving and increasing the prestige associated with making a donation. It was argued that fundraising effort can be measured along two dimensions: the number of times a person is approached by a charity and the quality of the fundraising method used in that approach. Two types of method were distinguished: direct methods, which involve face-to-face contact between the fundraiser and the potential donor, and remote methods, which do not involve face-to-face contact. It was hypothesized that direct methods had the greater potential for enhancing the warm glow of charitable giving and thus that they would be more successful in eliciting philanthropic gifts. Access to a unique dataset – containing very detailed information on exposure to fundraising effort – provided the opportunity to put this hypothesis to the test. An analysis of the raw data revealed that direct fundraising approaches were indeed more likely to be successful, but that the resulting gifts were on average substantially smaller than those obtained from remote fundraising drives. However, it was recognized that these results could be biased by a number of factors. First, there was the fact that direct fundraising methods are used much more intensively than remote methods. Second, it was found that the two methods were reaching significantly different populations, which could be expected to have different underlying dispositions to give. In order to control for these confounding influences, donations were modelled as a function of both the number and quality of fundraising approaches as well as a wide range of other socioeconomic and demographic factors. On the basis of these models, direct methods were still found to be more successful in eliciting contributions, but in this controlled context they were also found to generate larger contributions. Thus the overall marginal effect of an additional fundraising approach was found to vary between £0.93 for direct methods and £0.27 for remote methods. The implication is that the earlier result derived from a failure to take into account the effect of other influences on philanthropic behaviour.
166
Capturing the economic value of charities
These results provide considerable support for the theoretical framework advanced, suggesting that fundraising expenditures are an essential input in the creation of warm glows.
8. Targeting donors 8.1
INTRODUCTION
Fundraising expenditures by charities are akin to advertising expenditures by private firms. Both represent an attempt to persuade individuals to part with their money, whether in return for consumer goods or more intangible philanthropic benefits. Both are typically undertaken in a competitive environment, where shifting demand from one supplier to another may be as much of an issue as raising demand overall. Anecdotal evidence suggests that, like advertising executives, charity managers pay great attention to where they target their persuasive efforts. Some individuals have a higher predisposition than others towards purchasing certain types of goods, or contributing to certain types of causes. Where this predisposition is correlated with observable socioeconomic and demographic characteristics, an opportunity is created to increase the returns from fundraising (or advertising activities) by targeting campaigns on particular segments of the population. The literature on charitable fundraising to date has focused on the question of whether charity managers aim to maximize gross or net revenues (Weisbrod and Dominguez, 1986). However, little attention has been given to the role of targeting in fundraising activity, and consequently a number of interesting questions remain unanswered. What factors influence the choice of targeting strategy adopted by any particular institution? How far is it desirable to take the targeting process? To what extent does targeting succeed in dissipating the competition for funds between rival organizations? This chapter extends the existing literature on fundraising by incorporating targeting as a basic decision-making variable for the charity manager. Using a dataset on environmental group membership in the UK, a number of statistical models are estimated which identify how the decision to join an environmental group correlates with a number of observable characteristics that could be used as a basis for targeting. The models are then used to simulate alternative fundraising strategies as well as to examine their potential implications for competition between charity fundraisers. Targeting is found to produce a dramatic improvement in the probability that a given fundraising approach will be successful, but this comes at the 167
168
Capturing the economic value of charities
expense of reducing the size of the total pool of potential fundraising targets. The optimum balance between these two factors is shown to depend on the size of the environmental organization, the objective function of its managers, and the basic parameters of the cost and revenue functions. Although there is some evidence of market segmentation among environmental charities, this does not appear to be sufficient to succeed in softening the competition for potential members between different types of groups. The discussion is organized as follows Section 8.2 sets out the conceptual framework for the analysis. Sections 8.3 and 8.4 provide a descriptive overview of UK environmental groups and the data on those groups that are available from the British Social Attitudes Survey (Brook et al., 1991). Section 8.5 presents the statistical models of environmental group membership, while the simulations are conducted in Section 8.6. Conclusions are provided in Section 8.7. A more extensive discussion of the material can be found in Foster (1999a).
8.2
CONCEPTUAL FRAMEWORK
An important area of research in the economics of non-profit institutions has been to provide models of charitable fundraising activity. One strand of this literature examines how the objective function of charity managers would affect their optimal fundraising strategy (Steinberg, 1986; Weisbrod and Dominguez, 1986). Analogous to advertising expenditures for private firms, fundraising effort generates costs for non-profit organizations while at the same time increasing their revenue from donations. Drawing on the theory of the firm, this literature identifies two polar cases of managerial objectives. First, there are managers who aim to maximize the net revenues from fundraising – and thus the overall level of charitable services provided. Second, there are managers who aim to maximize the total revenues from fundraising – and thus the overall size of the charitable organization (Baumol, 1962). Empirical studies of fundraising behaviour have found evidence of both strategies being adopted by organizations operating in different sectors of charitable activity (Weisbrod and Dominguez, 1986; Posnett and Sandler, 1989; Khanna et al., 1995). Indeed, both of these objectives would appear to be plausible depending on the type of charity. On the one hand, net revenue maximization is particularly appropriate for charities whose primary aim is to provide direct assistance to needy groups. These needy groups are effectively analogous to the shareholders of a private firm in that they have residual claim on the services financed by the stream of net revenues generated by the charity managers. Thus the larger the net revenues generated by the charity, the higher the volume of
Targeting donors
169
philanthropic services that can be financed. On the other hand, gross revenue maximization may make sense for charities which are primarily concerned with campaigning and whose political clout may be enhanced by the ability to demonstrate a large following of supporters. Even charities that are primarily service providers may lean towards the maximization of gross revenues if managerial objectives threaten to eclipse the interests of the needy groups, or ultimate shareholders. Clearly, gross revenue maximizers will tend to allocate a higher level of effort to fundraising than net revenue maximizers. The reason is that gross revenue maximizers care primarily about raising additional revenues without much regard to the associated costs, as long as some breakeven constraint is met. They will consequently continue to pursue fundraising initiatives even when the additional revenues secured are comparatively low in relation to the additional costs incurred. By contrast, net revenue maximizers will be more concerned to minimize the costs of fundraising, which directly reduce net revenues. Consequently, they will not want to pursue fundraising efforts beyond the point where the additional costs incurred exactly offset the additional revenues raised. Both the costs and benefits of fundraising will depend on the extent to which resources are well targeted towards those segments of the population with the highest propensity to make financial contributions. As regards costs, fundraising expenditures are likely to be proportional to the number of potential donors targeted. As regards revenues, targeting could be expected to have two contrary effects on the yield from fundraising expenditures. On the one hand, the application of more stringent targeting criteria will increase the probability that each person targeted responds positively with a donation. On the other hand, as more stringent targeting criteria are applied, the pool of potential donors will shrink, since fewer and fewer people will satisfy the corresponding conditions. For example, imagine that the targeting criterion is gross annual income. The higher the targeting threshold of gross annual income selected by the charity fundraiser, the higher could be expected to be the probability of obtaining a positive response, but at the same time the smaller would be the pool of individuals who met the criterion. This reality faces fundraisers with a basic trade-off between raising the probability of successful strike (P) and reducing the number of potential targets (N). The optimal point of balance between P and N essentially depends on the cost and revenue structure of fundraising. In particular, the higher the ratio of the marginal cost to the marginal revenue of fundraising, the more attractive it becomes to target fundraising efforts. The marginal cost of fundraising is the amount that it costs to approach an additional person (for example, via a mail shot), while the marginal revenue of fundraising is the additional amount of money that a new member could be expected to contrib-
170
Capturing the economic value of charities
ute. When the marginal cost of fundraising is large relative to the marginal revenue, the cost of failure is high and it is therefore important to focus efforts on those segments of the population that are likely to be most responsive. However, when the marginal cost of fundraising is small relative to the subscription fee, the cost of failure is low and so it makes sense to approach as many potential members as possible. Furthermore, for any given value of the marginal cost–revenue ratio, it can be shown that net revenue maximizers will find it in their interest to pursue targeting to a greater extent than gross revenue maximizers. This is because maximizing gross revenue is broadly equivalent to maximizing the number of donors, and hence it is less attractive to apply criteria that reduce the size of the potential target population. Net revenue maximizers, on the other hand, are concerned to keep down the costs of fundraising. Targeting helps them to do this by reducing the number of people that have to be approached to raise a given amount of revenue. Finally, the issue of targeted fundraising links with another important strand in the existing literature. This concerns the problem of excessive fundraising expenditures when charities must compete for donations from the same target population (Rose-Ackerman, 1982). The industrial organization literature suggests that product differentiation can be used to attenuate this kind of competition (Tirole, 1988). In the traditional models of product differentiation, consumers have heterogeneous tastes for product characteristics. By locating themselves at a particular point in characteristics space, firms can act as local monopolists, breaking up the overall market into a number of distinct and homogeneous segments of consumers (Hotelling, 1929). Successful product differentiation by non-profit organizations would similarly have the effect of reducing competition for donations by enabling charities to target their fundraising expenditures on their respective market segments, thereby raising their returns from fundraising expenditure. In order to investigate both of these issues, this chapter will focus on one particular sector of charitable activity, namely environmental groups (Richer, 1995). The probability of joining an environmental group is modelled as a function of the socioeconomic and demographic characteristics thought to affect the disposition to give. Using the models, it is possible to conduct simulations of the effect of alternative targeting criteria on the probability of obtaining an environmental group member, and on the size of the available target group. Furthermore, the models shed light on the extent to which different types of environmental groups may be competing for members from the same target populations.
Targeting donors
8.3
171
ENVIRONMENTAL GROUPS IN THE UK
The charitable environmental sector in the UK grew rapidly during the 1980s and is now substantial in size. Recent research on the scale of voluntary sector activity in the sphere of environmental conservation found that there were 4000 groups with an aggregate expenditure of £634 million (Fenyo et al., 1993). To put this figure into context, it is estimated that non-profit organizations account for between 36 per cent and 69 per cent of national spending on environmental conservation, depending on which measure of non-profit expenditure is used (Ecotec, 1993; Fenyo et al., 1993). However, shares of total expenditure probably understate the true significance of environmental groups, since to the extent that the activity of these groups focuses on campaigning it is designed to bring about an increase in environmental expenditure by other bodies. To a greater degree than other voluntary organizations, environmental groups have tended to organize themselves as clubs, using annual membership fees as an important fundraising vehicle. Fenyo et al. (1993) estimate that 20 per cent of the income of environmental organizations derives from this source. Table 8.1 provides details of membership numbers and subscription income for some of the largest of the UK environmental charities, and indicates that they raise 25–50 per cent of their income from subscriptions and assign 10– 25 per cent of their expenditure to fundraising. All of these groups experienced significant growth in their membership during the 1980s, the most notable examples being the National Trust (NT), the Ramblers’ Association (RA) and the Royal Society for the Protection of Birds (RSPB), whose members approximately doubled over the decade, and the World Wide Fund for Nature (WWF) whose membership more than tripled over the same period. Table 8.1 also illustrates how these environmental groups differ with respect to their portfolio of activities and thus provides some indication of the extent to which they are engaging in product differentiation. Groups such as the Council for the Protection of Rural England (CPRE), Friends of the Earth (FoE) and Greenpeace (GP) are largely concerned with campaigning activities aimed at promoting environmental protection, whether on a domestic or global level. The WWF also has an important campaigning function relating to the specific problem of biodiversity, but differs from the other groups in devoting a considerable portion of its budget to funding conservation projects directly. At the other end of the spectrum, the National Trust (NT) deliberately eschews campaigning activity with a view to focusing its efforts on its perceived primary role of direct environmental protection via land ownership (The National Trust, 1995). By purchasing sites, the NT’s aim is to preserve them in perpetuity while at the same time providing a recreational resource to its members. The two remaining groups, the Royal Society for the
172
Table 8.1
Capturing the economic value of charities
A comparative profile of some of the UK’s major environmental groups CPRE
FoE
Members (m) Budget (£m) Subscriptions (%) Fundraising (%)
0.05 2.1 24.8 19.5
0.20 3.5 n.a. 14.0
Campaigns Research Land management Recreational sites Project funding
✔
✔ ✔
GP
NT
RA
0.40 2.21 0.10 7.2 142.6 1.9 n.a. 29 54 18.8 9.5 12.6 ✔
✔ ✔ ✔
✔ ✔
RSPB WWF 0.89 33.9 33 18.0
0.23 21.1 25 25.0
✔ ✔ ✔ ✔
✔ ✔ ✔
Notes: data compiled from the 1994/95 Annual Reports for the respective organizations: Council for the Protection of Rural England (CPRE); Friends of the Earth (FoE); Greenpeace (GP); National Trust (NT); Ramblers’ Association (RA); Royal Society for the Protection of Birds (RSPB); World Wide Fund for Nature (WWF). The definitions may not be entirely consistent across groups due to different accounting conventions. It should be noted that the NT and the RSPB manage 239 600 and 91 000 hectares of land as well as operating 240 and 130 recreational sites, respectively.
Protection of Birds (RSPB) and the Ramblers’ Association (RA), occupy something of an intermediate position in that they both supply recreational services to their members and perform some campaigning activities. Following Hotelling’s model (1929), the positioning of environmental groups described above can be understood in terms of the two-dimensional characteristics space illustrated in Figure 8.1. The two axes of the space represent the extent to which the focus of the group is on the domestic or global environment, on the one hand, and the extent to which the group is concerned with environmental recreation or protection, on the other. Those groups located towards the top left-hand corner of the figure (global environmental protection) are supplying what is close to being a pure public good, while those located towards the bottom right-hand corner of the figure (domestic environmental recreation) are supplying what is close to being a pure club good. This suggests that the former may face a more substantial free-rider problem than the latter. Indeed, the substantive nature of the club facilities provided by the NT and the RSPB may go some way towards explaining their relative success in terms of membership numbers and budget sizes. The positioning of groups in the diagram provides some tentative indication of the extent to which these organizations may be competing with each
Targeting donors
173
Global FoE GP WWF RSPB
Domestic
CPRE
RA
NT
Protection Figure 8.1
Recreation
Positioning of some of the UK’s major environmental groups in characteristics space
other in fundraising. Thus, for example, FoE might be expected to face much more direct competition for funds with GP and the WWF than with organizations such as the NT or the CPRE, which have a very different focus of activity.
8.4
BRITISH SOCIAL ATTITUDES SURVEY
The present analysis is based on data collected as part of the British Social Attitudes Survey (BSAS), an annual repeated cross-sectional survey covering a wide range of social issues. In 1985, the BSAS started to run an occasional section focusing on attitudes towards the countryside, which covered – among other things – membership of environmental groups. The section was run for three consecutive years between 1985 and 1987, and was repeated in a more extensive form in 1990, providing a total of 6142 observations. The surveys required respondents to indicate whether or not they belonged to the National Trust (NT), the Royal Society for the Protection of Birds (RSPB), a recreation-oriented environmental group or a protection-oriented environmental group. It is important to note that the last two are self-assessed categories. Therefore, it is not inconceivable – given the composite nature of environmental group activities as recorded in Table 8.1 above – that two members of a given organization might assess themselves in one case as
174
Capturing the economic value of charities
belonging to a recreation-oriented group and in the other case as belonging to a protection-oriented group. This subjective feature of the data is not necessarily problematic, since what is of primary interest is the preference of individuals for the recreation or protection activities undertaken by such organizations. The patterns of environmental group membership arising in the data are summarized in Table 8.2. The overall proportion of the sample claiming environmental group membership is relatively static across the period at between 15 per cent and 20 per cent – rising slightly during the late 1980s only to fall back by 1990. Table 8.2
Environmental group membership patterns from BSAS (%)
NT BSAS Social Trends RSPB BSAS Social Trends Recreation-oriented group Protection-oriented group Multiple groups All groups
1985
1986
1987
1990
5.68 2.12
7.88 2.49
8.53 2.72
8.47 3.54
3.52 0.90 8.19 4.19 3.28 16.81
5.47 0.90 6.60 4.52 3.58 19.40
4.39 0.99 6.55 3.80 2.91 19.17
4.38 1.47 3.92 6.35 3.96 14.78
Notes: all BSAS figures are adjusted by the survey weighting factor; this leads to a slight reduction in the proportion of members relative to the unadjusted figures. The survey weighting factor is designed to adjust for the fact that there tend to be some differences between the number of electors listed on the register and the number of adults actually found at any particular address. Source:
Brook et al. (1991).
At the beginning of the period, recreation-oriented groups were the most popular, accounting for just over 8 per cent of the sample. However, this category exhibited a steep decline in membership numbers throughout the late 1980s, so that by 1990 it had halved its proportion of members. For the reasons cited above, this trend need not necessarily be interpreted as a genuine shift in tastes away from recreation-oriented environmental groups. It could simply reflect a change in individuals’ perceptions of the environmental organizations they have always belonged to. All the other categories of groups experience steady growth in membership proportions across the period, par-
Targeting donors
175
ticularly the NT, which tops the membership league in 1990 with well over 8 per cent of the total sample reporting an affiliation. Finally, about 3 per cent of the sample claim affiliation with multiple environmental groups. Cross-group membership patterns differ significantly between groups. The NT and the recreation-oriented environmental groups tended to have a higher proportion of exclusive membership (in excess of 60 per cent), whereas less than half of the members of the RSPB and the protection-oriented environmental groups belonged exclusively to those organizations. The BSAS is designed with a view to obtaining a representative crosssection of the UK population, with a sample frame constructed on the basis of the electoral register (Brook et al., 1991). To verify this representativeness with respect to environmental group membership, the sample proportions claiming affiliation with the NT and the RSPB are compared in Table 8.2 with the figures for the population as a whole taken from the government publication Social Trends (Office of Population Censuses and Surveys, 1992). The figures reported in Social Trends can be regarded as accurate inasmuch as they are based on the actual membership rosters of the respective environmental groups. In order to obtain the reported percentages, these membership statistics are divided by official estimates of the UK adult population. The figures show that the proportions reported in the BSAS are up to three times as large as those found in the population for the NT and up to six times as large as for the RSPB. None the less, it is important to note that the upward trends over time are broadly consistent, as are the relative membership rates for the two groups. This indicates that the BSAS data are more reliable when it comes to relative membership rates than in terms of the absolute level of membership. There are a number of possible explanations for this upward bias in the membership percentages recorded in the BSAS. The first explanation relates to the low response rates achieved by the survey. For example, in 1990, only 64 per cent of those households originally sampled actually completed the questionnaire. This overall average masks regional variation in participation rates of between 52 per cent and 74 per cent. To the extent that participation in the survey was correlated with environmental group membership, a selectivity bias problem could be expected to arise in the data. The second explanation relates to ‘yea-saying’ induced by the ordering of questions within the survey. Since group membership questions were asked directly after a series of questions on environmental attitudes, it is possible that respondents who had expressed significant concern about the state of the environment then felt embarrassed to admit that they did not actually belong to an environmental group, and were thus led to make false claims about membership.
176
Capturing the economic value of charities
Third, respondents may have answered in the affirmative if any member of their household (apart from themselves) held a subscription to an environmental group, for example, parents whose adolescent children might be more environmentally conscious than themselves. Finally, people may have responded in the affirmative if they had at any time in the past belonged to such an organization, irrespective of whether or not they were currently paid-up members. This last point highlights a significant drawback of the BSAS data, namely the absence of longitudinal information from which to ascertain a given individual’s membership history. In addition to providing data on club membership, the BSAS is a rich source of information on the socioeconomic and demographic characteristics of respondents. For the purposes of the empirical model, attention will focus on six respondent characteristics. These are household income, age, newspaper preference (whether or not the respondent is a broadsheet reader), region (whether or not the respondent resides in the south of England), years of full-time education and the presence of children in the household. These six variables fulfil two important criteria for inclusion in the present study. First, they are potentially observable (or at least could be inferred) by charity managers and might therefore plausibly be used as targeting variables for fundraising campaigns. Second, it is thought that they jointly capture the main underlying factors determining preferences for environmental preservation, namely, purchasing power (household income), access to information about the environment (newspaper preference and education), opportunities for environmental recreation (age, region and the presence of children) and tastes (all of the above).
8.5
EMPIRICAL ANALYSIS
Table 8.3 summarizes the coefficient estimates obtained from a series of binary logit models for membership of the NT, the RSPB, and recreation and protection-oriented environmental groups respectively (Maddala, 1983). The purpose of these models is to identify whether there is any statistical association between environmental group membership and the six explanatory variables described above. A time trend is also included in the model in order to capture any general shift in preferences across the period. The coefficients indicate the magnitude of the effect that any particular explanatory variable has on the probability that a person joins an environmental group. The results show that many of the explanatory variables are statistically significant in affecting the disposition to become an environmental group member, as can be seen from the fact that the corresponding t-statistics are greater than two in absolute value. The chi-squared statistics for the test of
177
Observations Chi-squared (7df) Log-likelihood
6.924 7.117 6.057 3.425 7.512 –1.978 –0.210 –0.056
t-stat.
3700 381.19 –831.42
0.725 0.035 0.953 0.467 0.376 –0.308 –0.007 –3.722
Coeff.
NT
5.011 3.849 2.349 0.877 3.344 0.380 –1.414 1.248
t-stat.
3700 92.76 –648.07
0.642 0.023 0.466 0.143 0.209 0.069 –0.059 104.084
Coeff.
RSPB
6.090 –0.665 1.352 1.966 1.202 1.415 –4.688 4.555
t-stat.
3700 105.47 –866.41
0.665 –0.003 0.239 0.267 0.066 0.199 –0.180 346.780
Coeff.
Recreation-oriented groups
Summary of coefficient estimates for logit models of group membership
Log (household income) Age Broadsheet reader South of England Years of education Child in household Year Constant
Table 8.3
4.307 1.910 3.424 2.826 4.194 –0.641 1.501 –1.666
t-stat.
3700 141.11 –668.53
0.525 0.011 0.641 0.454 0.249 –0.110 0.059 –130.024
Coeff.
Protection-oriented groups
(b) 15
12
0
Age
>60
Household income
46–60
(a) >£25 000
£20 001–£25 000
£15 001–£20 000
£10 001–£15 000
£5001–£10 000
<£5000
20
31–45
Fitted probability of group membership (%) 25
0
<30
Fitted probability of group membership (%)
178 Capturing the economic value of charities
NT RSPB Recreation Protection
15
10
5
NT RSPB Recreation Protection
9
6
3
30 25
179
NT RSPB Recreation Protection
20 15 10
(d)
Year
12
>18
15
1990
Age completed full-time education
18
(c)
1987
17
0
<17
5
Fitted probability of group membership (%)
Fitted probability of group membership (%)
Targeting donors
NT RSPB Recreation Protection
9 6
Figure 8.2
1986
0
1985
3
Fitted probability of group membership against continuous explanatory variables
180
Capturing the economic value of charities
overall significance of the regressions are reported at the foot of Table 8.3. These indicate that while all of the models have significant explanatory power, the model for NT membership performs significantly better than the rest, with a test statistic that is more than double the size of any of the others. The income variable is a highly significant determinant of group membership in all cases. In most cases, membership is also significantly associated with age, broadsheet readership, residing in the south of England and years of education. However, the members of recreation-oriented groups stand out as being substantially different from the rest. In particular, age, broadsheet readership and residing in the south of England were not found to be significantly correlated with belonging to this category of environmental groups. Furthermore, recreation-oriented groups are the only category to exhibit a statistically significant time trend, which indicates that the disposition to join such groups was declining over the latter half of the 1980s. The absence of a significant time trend in the NT and RSPB membership models stands in contrast to the growing number of members for these groups reported in Table 8.2 above. An explanation for this divergence is that the growing membership may have more to do with a greater prevalence in the population of the underlying characteristics that predispose towards membership, rather than to any positive shift in tastes. A series of Wald tests were conducted in order to test the null hypothesis that there are no significant differences between the types of people attracted to joining each of these different categories of environmental groups (Greene, 1993). This hypothesis could not be rejected in the case of the RSPB and the protection-oriented groups, suggesting that very similar types of people are attracted to joining both of these organizations. However, in all other cases, the null hypothesis was strongly rejected, indicating that members of the NT and the recreation-oriented groups are significantly different in profile from members of the RSPB and the protection-oriented groups. The models can be used to estimate the probability that any particular respondent would join any particular environmental group as a function of their observable characteristics. In order to illustrate how the probability of successfully identifying a group member varies with different levels of the explanatory variables, Figure 8.2 plots the fitted probability of membership against different levels of the four continuous explanatory variables – household income, age, years of education and the time trend. The fitted probability of membership is found to rise substantially with household income, particularly beyond the £20 000 per year threshold. Indeed, households with incomes in excess of £25 000 per year are between five and seven times more likely to join an environmental group than households with incomes below £5000 per year. The effect of income on the fitted probability of membership is particularly pronounced in the case of the NT.
Targeting donors
181
The relationship between the age variable and the probability of group membership is not so pronounced. There is some evidence that this probability peaks in late middle age (between 45 and 60), particularly in the case of the NT. An exception is the recreation-oriented category of environmental groups, which appear to have a much stronger appeal among younger adults (under 45). A positive correlation between educational attainment and the fitted probability of membership is found for all groups. Specifically, those who remain in full-time education beyond the age of 18 are three to six times more likely to belong to environmental groups than those who terminate their full-time education at the age of 16 or below. This effect is particularly striking for NT membership. Finally, the time trend looks somewhat different for each of the four categories of environmental groups. While the fitted probability of membership for the RSPB is relatively static across the period, it shows some increase for the NT and the protection-oriented groups. The recreation-oriented groups, on the other hand, show a clear downward trend in the fitted probability of membership.
8.6
SIMULATION OF FUNDRAISING STRATEGIES
The binary logit models reported above can be used to address the issues of prime interest identified in the earlier conceptual discussion. The first of these was the question of how to design an optimal targeting strategy for environmental group fundraising. The second was the question of whether market segmentation succeeds in attenuating competition for subscriptions between different types of groups. However, before turning to these matters, it is important to take note of two important limitations of the estimated model in these particular applications. First, the descriptive discussion of the BSAS dataset suggested that environmental group membership in the sample was three to six times larger than in the general population (see Table 8.2). This indicates that the absolute levels of the predicted probabilities should be calibrated, by scaling down with this inflationary factor. However, the descriptive statistics showed a much closer correspondence in the relative values of membership probabilities both across groups and over time. This suggests that greater confidence can be attached to the results that relate to the proportionate change in probabilities of membership under different fundraising strategies. Second, what is of primary interest to the fundraiser is the conditional probability of a person taking out a subscription following the receipt of publicity material from the organization. This is likely to differ, as a function
182
Capturing the economic value of charities
of individual characteristics, from the unconditional probability of group membership. Results reported in Chapter 5 indicate that the unconditional correlations between the probability of group membership and population characteristics will in all likelihood be confounding two different effects. One is the disposition of charity fundraisers to approach people with those characteristics. The other is the disposition of people with those characteristics to respond positively to an approach by a charity fundraiser. Since the BSAS does not provide information on which of the respondents had been asked to join an environmental group, it is not possible to separate out these two effects. Consequently, the subsequent analysis is necessarily based on the somewhat problematic assumption that the pattern of membership observed in the population is an adequate guide to how individuals with particular socioeconomic characteristics would respond to a fundraising approach. These two issues, both of which suggest that the absolute level of predicted probabilities from the model will overstate the true conditional probabilities of membership, should be borne in mind when interpreting the results that follow. 8.6.1 Optimal Fundraising Strategy The coefficients estimated in the binary logit models identify segments of the population with a significantly greater predisposition to join particular types of environmental groups. Charity fundraisers could potentially target their efforts using any or all of these variables. In order to shed light on the appropriate choice of targets, Table 8.4 shows the effect which various different targeting strategies have both on the probability of a subscription (P) and on the number of potential fundraising targets (N). Clearly, there are numerous ways in which the socioeconomic variables studied in the model could be used to define targeting criteria for fundraising. The targeting criteria presented here are taken to be illustrative rather than exhaustive. Part (a) of Table 8.4 examines the effect of targeting on the basis of each of the variables used in the model individually. For the dummy variables, the target group is defined as the half of the population exhibiting the highest propensity to join each corresponding type of environmental organization. For the continuous variables, there are clearly many ways in which the target group could be defined. The analysis is based on defining the target group as the segment of the population exhibiting the highest propensity to join on the basis of the banding used to construct Figure 8.2. There is no reason why targeting strategies should confine themselves to the use of a single targeting variable. The information contained in the singlevariable analysis could be combined together so as to develop a multi-variable targeting approach. In principle, there are any number of ways in which the
Targeting donors
Table 8.4
P and N against number of variables used in targeting
NT P (a) Single-variable targets No targeting 0.080 Targeting on Income 0.198 Age 0.099 Newspaper 0.195 Region 0.112 Education 0.251 Children 0.092
Recreationoriented groups
RSPB
Protectionoriented groups
N
P
N
P
N
P
N
3700
0.046
3700
0.067
3700
0.050
3700
519 870 959 1673 415 2348
0.102 0.055 0.084 0.056 0.109 0.047
519 870 959 1673 415 2348
0.140 0.068 0.089 0.082 0.115 0.058
519 870 959 1673 415 2348
0.114 0.056 0.107 0.069 0.136 0.052
519 870 959 1673 415 2348
3700
0.067
3700
0.050
3700
415 157 125 80 19 15
0.139 0.162 0.166 0.179 0.166 0.166
519 157 125 80 19 15
0.136 0.163 0.209 0.235 0.255 0.265
415 293 125 80 19 15
(b) Multi-variable targets – ‘AND’ basis No targeting 0.080 3700 0.046 Targeting on best variable 0.251 415 0.109 2 best variables 0.311 293 0.154 3 best variables 0.404 125 0.169 4 best variables 0.441 80 0.175 5 best variables 0.535 19 0.203 all 6 variables 0.566 15 0.209 Note:
183
P: the probability of a subscription. N: the number of potential fundraising targets.
information could be so combined (Foster, 1999a). For the purposes of illustration, attention here will be confined to one particular multi-variable targeting strategy. This will be referred to as the ‘AND’ strategy, and involves targeting individuals who simultaneously satisfy two or more of the single variable criteria. Accordingly, part (b) of Table 8.4 illustrates the effects of using between two and six targeting criteria simultaneously. The results for the single-variable approach indicate that substantial increases in the probability of membership (P) can be secured by targeting, although the magnitude of the benefits varies considerably with the type of group. Thus the NT can increase its probability of subscription by more than three times, while the other categories of groups do just better than doubling their probability of obtaining a subscription. These increases in the probability of success come at the expense of substantially reducing the number of potential fundraising targets. For example, restricting the target population by income or educational criteria immediately reduces the size of the potential market to some 10 per cent of the overall sample.
184
Capturing the economic value of charities
NT RSPB Recreation Protection
Marginal cost–revenue ratio
0.6 0.5 0.4 0.3 0.2 0.1 0 0
(a)
6
Gross revenue maximizer
NT RSPB Recreation Protection
0.6 Marginal cost–revenue ratio
1 2 3 4 5 Optimal number of targeting variables
0.5 0.4 0.3 0.2 0.1 0 0
(b)
1 2 3 4 5 Optimal number of targeting variables
6
Net revenue maximizer
Figure 8.3
Marginal cost–revenue ratio against optimal number of targeting variables for the ‘AND’ strategy
Targeting donors
185
These trade-offs become even more pronounced when an ‘AND’ strategy is adopted, as indicated in part (b) of Table 8.4. For example, the NT can raise its probability of obtaining a subscription from each targeted individual to over 0.5, but doing so reduces the number of potential targets to less than 1 per cent of the sample. For all groups, the marginal improvements in the probability of success show a tendency to decline in the number of target variables applied. Further to the discussion in Section 8.2 above, it is interesting to enquire how the optimal degree of targeting might vary according to the marginal cost–revenue ratio. Figure 8.3 illustrates the case of the ‘AND’ strategy, for both gross and net revenue maximization. The figures show that targeted fundraising does not become worthwhile until the marginal cost–revenue ratio exceeds 5–8 per cent for gross revenue maximizers and 4–6 per cent for net revenue maximizers. As the marginal cost–revenue ratio increases, the number of targeting variables needed to maximize the objective function rises at an increasing rate. For a given value of the marginal cost–revenue ratio, the NT requires the smallest number of targeting variables and the recreationoriented groups the largest number of targeting variables. Although it is not visually evident from inspecting the graphs, the curves for net revenue maximizers lie everywhere to the right of the corresponding curves for the gross revenue maximizers. This is an empirical reflection of the point described above, that net revenue maximizers will engage in higher levels of targeting for any particular level of the marginal cost–revenue ratio. For most of the groups, no amount of targeted fundraising will enable them to survive financially once the marginal cost–revenue ratios reach the level of around 20 per cent. The exception is the NT, which can survive even when marginal fundraising costs absorb more than half of the corresponding marginal fundraising revenues. This is due to the fact that the relationship between socioeconomic characteristics and the probability of membership is particularly strong for the NT, making targeting particularly effective. It is interesting to compare the rates of return (that is the ratio of gross revenue to fundraising cost) predicted in these simulations with the real rates of return observed in charity fundraising. A recent survey of charity managers (Sargeant and Kaehler, 1998) found that fundraising by such methods as mail shots, telemarketing and press advertisements yielded an average return in the range 175 per cent to 200 per cent. However, these returns rose to 560 per cent when fundraising efforts were targeted towards those with a history of giving. Taking a strategy of single-variable targeting and a plausible marginal cost– revenue ratio of 0.05, the logit models predict rates of return in the range 217 per cent to 501 per cent depending on the environmental group. Increasing the marginal cost–revenue ratio to 0.1 lowers the level of returns from this form of targeted fundraising to the range 109 per cent to 251 per cent.
186
Capturing the economic value of charities
Clearly strict comparisons between these two sets of values are not possible given that the Sargeant and Kaehler (1998) study does not report details of the targeting strategies and marginal cost–revenue ratios faced by the charity managers. None the less, the fact that the predictions of the binary logit models are broadly similar in size to those found in the other study provides some reassurance that the simulations are not seriously overstating the possible outcomes. In order to relate the predictions of the binary logit models to real fundraising situations, it is necessary to have the real-world values for the marginal cost and marginal revenue of fundraising. Both of these parameters are clearly choice variables for charity fundraisers in the long run. However, using information on typical current values for marginal cost and revenue, it becomes possible to comment on the degree of targeting that is likely to be optimal, conditional on this current choice. A current typical annual subscription fee for an environmental group is £20. However, it is important to note that once a person subscribes, they are likely to continue to do so over time and furthermore may make additional contributions to special appeals. Thus the marginal revenue should ideally reflect the present value of all these donations, so that the £20 can probably be regarded as a lower bound. On the other hand, it is also important to recall the costs that are incurred by environmental groups in order to provide a range of private benefits that are typically offered to members, including periodicals and access to special services. The cost of providing these private benefits would need to be netted out of the subscription fee in order to provide a clearer reflection of the pure untied revenue generated by an additional member. In this sense, the figure of £20 is likely to be an overestimate of marginal revenue. As regards the appropriate value of the marginal cost parameter, this is likely to vary substantially according to the chosen fundraising technology. In the context of a mailshot, which is a fairly traditional method, the marginal cost could be expected to be particularly low since it would only reflect the expenditure associated with postage and additional printing. Consequently, in this case it is hard to imagine that the marginal cost could be any greater than £1, as an absolute upper bound. These boundary values would give a maximum ratio for the marginal cost–revenue ratio of 0.05 and thus lie in the range where targeting is still not an attractive proposition. This illustrative calculation suggests that targeting may be of limited value in practice to environmental groups. However, before reaching this conclusion, it is important to note that the analysis presented thus far was premised on the assumption that the charitable organization would exhaustively approach everybody in the identified target group, including the entire population in the case of zero targeting. Whether or not this is a realistic assumption will
Targeting donors
187
depend on the size of the environmental organization relative to the size of the country in which it is operating. For small groups operating in large countries it would appear likely that budgetary restrictions could preclude such an exhaustive approach to fundraising. Indeed, mailshots by the RSPB – one of the UK’s largest environmental charities – have historically been of the order of 200 000–400 000, while the WWF’s largest fundraising campaign in recent years encompassed a target population of 3 million. These figures can be compared with the total number of households in the UK, which is of the order of 20 million. It may therefore be more realistic to consider the benefits from targeting in a context where the charity must maximize the revenues generated from a fixed fundraising budget. This may, in fact, be true of large groups as well as small groups in the short run. The fixed-budget setting carries a number of important implications. First, the value of population that the charity can afford to reach is essentially fixed by the size of the budget and is likely to represent only a very small fraction of the overall population. This means that the loss of market size associated with targeting may no longer be material, and in the limit could be disregarded altogether. Second, since costs are fixed by the size of the budget, gross and net revenue maximizing behaviour become equivalent and both entail maximizing the probability that a targeted individual will respond with a subscription. The combination of these factors indicates that relatively small charities operating in relatively large countries should adopt an ‘AND’ strategy, targeting to the fullest possible extent on as many variables as possible. The figures presented in Table 8.4 indicated that, by targeting on six variables, such charities could raise their gross revenues by between three- and sixfold, depending on the type of organization. This conclusion is based on the assumption that the marginal cost of fundraising does not rise with the number of targeting variables used, as it might, for example, if it became increasingly difficult to identify members of the target group. 8.6.2 Competition for Funds In addition to illuminating issues of fundraising strategy, the binary logit models can be used to identify the extent of competition for funds between these different types of environmental charities. The Wald tests performed above provide evidence that environmentalists loyal to different groups are a segmented population comprising a number of distinct types of people. However, the coefficients reported in Table 8.3 indicated that the members of different environmental groups also had numerous characteristics in common. Moreover, from examining the benefits of
188
Capturing the economic value of charities
alternative targeting strategies in Table 8.4, it becomes clear that the characteristics (high income, high educational attainment and broadsheet readership) which different types of environmental group members have in common have a much stronger impact on their propensity to subscribe than the characteristics (age, residence in the south of England and the presence of children in the household) which distinguish them from each other. The extent to which the different types of groups compete for funds will therefore depend on the way in which their chosen fundraising strategies link back to these underlying characteristics of the population. Table 8.5 indicates the percentage overlap between the target populations for each pair of groups according to the number of targeting variables used and the chosen strategy. The higher the percentage overlap, the greater the extent to which groups are competing for funds from the same segments of the population. The results indicate that there is only very limited scope for avoiding competition, and this only when a small number of targeting variables are being used. In the case of the NT and the protection-oriented groups, targeting does not provide any means of avoiding competition at all. Table 8.5
Overlap ratios for target population segments between groups for the ‘AND’ strategy (%) NT: NT: NT: RSPB: RSPB: Recreation: RSPB Recreation Protection Recreation Protection Protection
No targeting Targeting on Best variable 2 best variables 3 best variables 4 best variables 5 best variables All 6 variables
8.7
100.0
100.0
100.0
100.0
100.0
100.0
100.0 79.6 100.0 100.0 100.0 100.0
30.3 79.6 100.0 100.0 100.0 100.0
100.0 100.0 100.0 100.0 100.0 100.0
30.3 100.0 100.0 100.0 100.0 100.0
100.0 42.7 100.0 100.0 100.0 100.0
37.8 42.7 100.0 100.0 100.0 100.0
CONCLUSIONS
This chapter has extended the existing literature on charitable fundraising by examining to what extent charities can improve their performance by targeting effort towards specific segments of the population characterized by particular socioeconomic and demographic attributes. The conceptual framework suggested that targeting could affect fundraising both positively and negatively. On the one hand, it could be expected to increase the probability that a targeted individual would make a donation. On the other hand, it could
Targeting donors
189
be expected to reduce the number of individuals who could potentially be targeted. A set of logit models was estimated in order to quantify the impact of a range of observable characteristics on the probability of subscribing to the National Trust, the Royal Society for the Protection of Birds, and other recreation- or protection-oriented environmental groups. Using the models, a variety of alternative fundraising strategies were simulated. The conclusions were shown to differ depending on the size of the charitable organization. For charities that are large relative to the size of the country in which they operate, the optimal fundraising strategy was found to depend both on the charity’s objective function (net or gross revenue maximization) and on the ratio of marginal fundraising costs to the group’s subscription fee. The higher this ratio, the greater the extent to which the charity should engage in targeting. None the less, there is a limit on the extent to which targeting can overcome the handicap implied by a high marginal cost–revenue ratio. All the environmental groups considered ceased to be financially viable once this ratio exceeded a certain threshold. For plausible values of the marginal cost–revenue ratio, little if any targeting appeared to be advisable. The intuition behind this result is that the shrinking size of the market brought about as a result of targeting more than offsets the greater probability that any targeted individual will subscribe, unless the costs of fundraising are comparatively high. However, other things being equal, net revenue maximizing charities will tend to undertake a greater degree of targeting than those whose objective is to maximize gross revenues. For charities that are small relative to the size of the country in which they operate, that is, just about all UK environmental charities, the optimal strategy irrespective of the charity’s objective function and marginal cost–revenue ratio is to engage in as much targeting as possible. The intuition behind this result is that small charities are unlikely to be constrained by the shrinking size of the market, and are thus best advised to maximize the probability that those whom they approach will choose to subscribe. For such charities, targeting can increase their returns to fundraising by between three- and sixfold, although there is substantial variation in the magnitude of these gains across groups. Groups such as the National Trust, which succeed in appealing to a sector of the population with very welldefined observable characteristics, will reap the benefits in terms of the greater efficacy of their fundraising efforts. In spite of some demonstrated degree of segmentation in the market for environmental charities, the model indicated that those factors with the strongest influence on the propensity to become a group member were in fact held in common between the different categories of environmental
190
Capturing the economic value of charities
groups considered. The implication is that targeting does not go very far towards diluting the intensity of competition for members between environmental groups, as they are all obliged to recruit from a very similar pool of individuals.
PART III
Policy and Social Capital
9. On social capital 9.1
INTRODUCTION
In this chapter we argue that the phenomenon of giving is part of what has come to be called ‘social capital’, the ‘glue’ that holds society together and which plays an as yet generally unquantified role in economic and social development. Social capital concerns the relationships between individuals, between institutions (including governmental institutions) and between individuals and institutions. Social capital includes the norms governing these relationships. It has been found that different societies can have broadly equal endowments of other forms of capital, but that some perform better in terms of economic and social development. The ‘missing link’ is thought to lie in the fact that the better-performing societies have less conflict between social groups, more participatory decision-making procedures and greater trust between economic agents. Thus Putnam (1993) found that one of the factors explaining northern Italy’s better economic performance compared to southern Italy was the presence of many more voluntary associations. Putnam’s focus was on horizontal associations between people and on the rules of behaviour for those associations, rules designed to ensure that members secured the full benefit of mutual cooperation. Fukuyama (1996) identifies successful communities and companies with the presence of reciprocity and trust. The breakdown of social capital results in more crime, more violence, family breakdown and more distrust. Others, such as North (1990), have emphasized institutions generally and political institutions and the rule of law in particular. Clearly, definitions of social capital vary in scope. World Bank (1997) suggests that there are three concepts of social capital. First, the narrow concept is the one employed by Putnam. It is a set of horizontal associations between people and the norms governing those associations. Associations exist for the mutual benefit of members, and the associations in turn help sponsor economic efficiency and therefore economic development. Second, Coleman’s (1988) concept of social capital expands the narrow concept above to include vertical associations, that is, linkages between members of an association who (or which) are unequal in terms of the power they 193
194
Policy and social capital
possess. Thus hierarchical relationships would be included in this definition and the agents include firms. The third definition, building on North (1990), embraces a wider focus still, and includes government, the rule of law and political structures. This more comprehensive focus opens up the potential for more empirical indicators, for example of political and civil liberties or corruption as a negative indicator. To some extent, it would be helpful to define these components as political capital but, clearly, the distinction between political and social is very blurred. In all three cases, it is argued that these forms of social capital are generally linked in a positive way to economic development, although not necessarily so (see below). None of the three definitions stresses the giving relationship, which is our focus. But trust, cohesiveness, involvement and a wider concern for society as an entity, and of the disadvantaged in particular, are all features of the acts of giving and charity. Thus measuring the giving relationship offers one important insight into the means of quantifying social capital. It cannot be a comprehensive indicator, of course, but it is a meaningful indicator none the less. We suggest, then, that the money value of the willingness to give, as derived in Part I, is at least a candidate indicator of the value of social capital. In order to set the context we digress for a while on the existing literature on social capital.
9.2
SOCIAL CAPITAL IN THE THEORY OF SUSTAINABLE DEVELOPMENT
Social capital is one form of capital asset upon which sustainable development depends. In recent years, the capital-based theory of sustainable development has advanced to a fair degree of sophistication. The briefest sketch is given here. For a detailed assessment, see Atkinson et al. (1997). Definitions of sustainable development are not in themselves very interesting, although there is an interesting debate on how ‘development’ might be measured in terms other than per capita GNP (see, for example, Atkinson et al., 1997). What matters is what has to be done to secure it. Pearce (1999) suggests that the conditions for sustainable development are likely to be invariant with the definition since the conditions will be couched in terms of capacities and capabilities: that is, sustainable development becomes an enabling concept rather than solely a particular path of change. What determines the ability of a given set of humans to improve their wellbeing (utility) is the quantity and quality of capital assets available at the time. This notion can be traced back to the economic growth theory of the
On social capital
195
1970s. It is elegantly summarized in Solow (1992). The concept of capital has widened from the classical approach, with its focus on produced goods or man-made capital (KM in notation), to embrace the skills and knowledge embodied in humans, or human capital (KH) and natural capital (KN). Natural capital refers to traditionally defined natural resources, such as oil or gas, forests and to the stocks of assimilative capacities in the environment. Modern economic growth theory would add social capital (KS), which is the focus of concern here. To complete the discussion of the capital base for sustainable development we need to add two further factors. The first is the rate of technological change, which is a disembodied stock of knowledge and skills. If this rate of technological change is positive, then the capital stocks listed above can yield higher and higher flows of services. A declining capital stock may not therefore be of major concern provided the rate of technological progress compensates for the loss of that stock. Of course, some technological change is not ‘good’. Chlorofluorocarbons (CFCs), for example, were thought to be a major technological advance for the development of aerosols, solvents and cooling systems. But today we recognize CFCs as the major source of stratospheric ozone depletion and a risk to human and ecosystem health. The second additional concept is population change. It is possible that growth in population can improve human well-being by stimulating technological change. There is some evidence for this in historical terms, where changes in invention and innovation appear to be linked to increases in population, and there is some modern evidence for this relationship too (Boserup, 1981; English et al., 1994). But population change is more likely to reduce per capita capital stocks. As population expands, pressure is put on marginal productive land, for example, and forests are cut down or burned to make way for agriculture, urban expansion and roads. All this affects KN especially, but it is also reasonable to argue that rapid population change reduces KS by increasing the likely conflicts over natural resource availability, especially land and water. Absolute population levels and the rationale for large families do, however, need to be distinguished. There is evidence that large families exist precisely because of the need to bind the family together. In turn, the focus on the family as the social unit of concern suggests that the assurances one would get from community-wide or nation-wide social capital have never existed, or have broken down. To be more specific, large families provide both labour (subject to labour laws, if any) and social insurance in old age. Large families therefore have an economic rationale based on the costs and benefits of family size to the family itself (Dasgupta, 1995). But the rationale for large families is greatly reduced if the responsibility for, say, old-age security is transferred to the state or community through either legislation or tradition.
196
Policy and social capital
The negative effects of population change are likely to be more pronounced than any positive effect of population growth in stimulating the expansion of capital, although this remains a possibility. And if sustainable development is concerned with per capita utility, then population growth will mean that more and more output will be needed just to keep pace with additional numbers of people. Whereas technological change is likely to make capital more efficient, population is likely to depress some capital stocks and make sustainable development less feasible. The intertemporal condition for sustainable development, therefore, amounts to each generation leaving the next generation a stock of productive capacity, in the form of capital assets and technology that is capable of producing more utility or well-being per capita than that enjoyed by the current generation. Notationally, dK / dt ≥ 0,
where K = KM + KH + KN + KS
This intertemporal requirement is only part of the story: concern with the poor now is also widely regarded as an important feature of sustainable development. This provides us with one immediate link to the charitable sector since so much of this sector is focused on the underprivileged groups in society. The analysis above can be applied to this equity issue as well, for the poor cannot improve their lot without access to productive capacity. If their wellbeing is to improve, then they must secure better education as a means of improving human capital (KH), better technology, more man-made capital (KM), and more natural capital (KN). Social capital will matter as well in the sense of the need for more participation in decisions that affect their lives, more consultation and more concern from others. It is incidentally worth noting that the questionnaire approaches used in Part I are themselves consultative and participatory since it is the general public and the users who are involved in expressing their preferences directly rather than through some political process. Indeed, in the case of the homeless, the political process may not reflect preferences at all well. The equation above suggests that the intertemporal part of sustainable development can be analysed in terms of the conditions necessary for its achievement, and that those conditions can be interpreted in terms of a constant capital rule. This rule requires modification in so far as: ●
the rate of technological change results in higher capital efficiency (a higher ratio of generated services to the capital stock). In this case, the constant capital stock rule can be relaxed in favour of one that ensures at least constant capital services; and
On social capital ●
197
the rate of population growth results in less capital per capita.
Of course, population growth may occur together with exogenous technological change, and this is typically the real-world case. If they work in opposite directions, as suggested previously, then a rough rule of thumb is that sustainable development will require that technological change exceed the rate of population growth. This intuitive result bears a close resemblance to some of the 1970s theory rules for sustained economic growth, for example Stiglitz (1979). Note that the links between technological change and social capital are not clear. Technological change does not obviously augment the productivity of social capital, although it could be argued that, say, communication technology vastly improves the chances of being in touch with others where before a letter or a journey might be involved. In other respects, technology may actually distance people from each other by substituting for more traditional communal or even family gatherings. An implicit assumption in the constant capital rule is that all forms of capital are substitutable for each other. On this rule, known as the weak sustainability (WS) rule, any one form of capital can be run down provided proceeds are reinvested in other forms of capital. Weak sustainability does not imply that substitution is easy or inexpensive. We may have to surrender a great deal to obtain one extra unit of some forms of capital, which is a feature of weak sustainability that tends to be ignored by those who have criticized it. Moreover, WS requires that the running down of any form of capital is compensated by investment in some other form of capital. It is not consistent with running down capital stocks and ‘consuming’ the proceeds. As a ‘weak’ rule, then, WS is not particularly weak, and empirical tests show that it is quite easy for a country to fail a weak sustainability test (see Atkinson et al., 1997). Objections to weak sustainability tend to centre on the assumed substitutability of capital stocks (Victor et al., 1994). Indeed, it can be argued that the philosophy of sustainable development arose precisely because there were concerns about the unsustainability of forms of economic development that sacrificed the environment and social cohesion in the name of economic growth. The problem can be formalized by saying that the critics regard at least some forms of capital as having no substitutes. Those who believe in the non-substitutability of natural capital support strong sustainability (SS). Far less attention has been paid to strong sustainability based on social capital, yet opinion polls repeatedly indicate that there are widespread concerns in society about those aspects of modern life that indicate a breakdown of social capital: crime, violence, drugs and sheer inhumanity. As noted above, these are often advanced as negative indicators of social capital. SS does not imply that WS is irrelevant. What SS requires in addition to WS is that the stocks of KN or KS or KH (or perhaps all three) should not
198
Policy and social capital
decline. ‘In addition to’ is needed because a situation in which any one form of capital is preserved but other forms of capital are allowed to decline could hardly be called ‘sustainable development’ (it might be ‘survivable’ but even that seems very unlikely). SS implies WS, but WS does not imply SS.
9.3
THE NATURE OF SOCIAL CAPITAL
Because the focus on social capital is comparatively new, it is worthwhile considering in a little more detail what it might mean and whether it necessarily contributes to sustainable development. Putnam (1993) speaks of social capital as comprising certain features of social organization – norms of behaviour, networks of interactions between people and between institutions, and trust between people. Empirical studies of economic growth have shown that conventional growth accounting models (stressing labour, capital and technology) explain only a limited amount of the difference between growth rates in different economies. World Bank (1997) refers to studies of the ‘Asia Miracle’ economies (as they were before the Asian economic collapse) which suggest that institutional arrangements for cooperation and information exchange may be as, if not more, important than conventional factors. But close interpersonal and interinstitutional arrangements may not always be good for sustainable development. After all, price-fixing cartels are a form of social arrangement, as is the Mafia. This suggests that social capital may have positive and negative aspects. On the positive side it is suggested that social capital contributes to economic development because: ●
●
●
flows of information between economic agents are better and higher if there are closer social relationships. Such flows may relate to anything from price information, information on the availability of materials or labour, through to information on the credit-worthiness of individual agents; trust reduces the need to search out information in order to make a transaction; that is, transaction costs can be reduced. Trust may also result in behaviour which avoids the need to make laws and hence to intervene via government; and social links between individuals and organizations and government also reduce the need for overt public control. Governments may find it easier and more efficient to operate via established social links than to legislate. The rise of voluntary agreements as a means to control environmental problems may be a case in point. Polluters simply agree to self-regulate and, in turn, self-regulation will be all the more
On social capital
199
efficient if the polluters have social arrangements whereby they trust each other. Environmental improvement is also regarded as an integral part of modern sustainable development (Atkinson et al., 1997). Social capital contributes to environmental improvement by: ●
●
●
●
●
substituting for other forms of capital, especially man-made capital. Arrangements to share machinery, for example tractors, harvesters, mean that fewer tractors are needed; reducing the high discount rates that often imperil the environment. This happens because individual insecurity is reduced by ganging together to fight particular causes and by spreading risks among the social group; reducing external effects, that is the spillover effects of one agent’s actions on the well-being of another agent. Effectively, such behaviour is inhibited by the concern for neighbours and third parties arising from social norms of behaviour; resolving the risks arising from common property. Common property involves a whole community owning and managing a resource, a situation that has risks of environmental destruction if the resulting communal management system breaks down. The stronger the social ties, the less likely the management system is to collapse; and inhibiting antisocial behaviour that damages the environment, whether it is simply the dumping of illicit waste and litter or the perverse destruction of wildlife.
Social capital could have negative results by keeping contracts with those within the social circle, when those outside are more efficient. Examples include price-fixing, closed-contract award systems, and even the requirement that small firms institute some social welfare system to look after those in the social group, imposing costs that impair productivity. One might summarize these problems as the creation of rent by restrictive activity and through lobbying of government and others. The idea that a capital asset may have negative impacts on sustainable development perhaps distinguishes social capital from other forms of capital assets. On the other hand we have already noted the potential negative effects of some forms of technological change. Social capital therefore presents a new and challenging dimension of sustainable development. It may, as some have argued, account for the dynamism of some economies and even for lower environmental damage and better social stability than might otherwise occur. Others suggest that it could just as easily
200
Policy and social capital
contribute in a negative fashion by restricting economic interests to specific groups. And, of course, the social group could also be persuaded to destroy the environment or wider social cohesion in the name of its own group interests.
9.4
INDICATORS OF SOCIAL CAPITAL AND THE VALUE OF GIVING
The existing literature on indicators of social capital is small and several commentators have remarked on the absence of a theory to guide model building (Klitgaard and Fedderke, 1995). As noted above, it is also not very clear what it is that one is trying to measure. Most of the work on indicators takes measures of such things as civil and political liberties and then correlates these with some indicator of development, for example GNP per capita. Efforts are also made to see if there are associations between the different indicators of social capital, the aim being to see if there are key indicators that could be used instead of a proliferation of indicators. Indicators also need to be distinguished from measures: there is no suggestion that indicators of, say, civil liberties are measures of social capital. Indeed, in the sense employed in the theory of sustainable development above, the only correct measure would be a monetary value. To date, this type of measure has been only indirectly referred to in the literature. Some writers have focused more on indicators of the absence of social capital rather than its presence. This is still useful since it reminds us that what may have been taken for granted in the past is disappearing. Klitgaard and Fedderke (1995), for example, focus on indicators of social integration and disintegration, but much of their list of indicators relates to the breakdown of social (and political) cohesion. The positive indicators tend to be political and civil liberties, thus embracing the much wider definition of social capital above. The negative indicators include strikes, assassinations, ethnic separatism, suicides, riots and so on. While the data sources are weak, they speculate that: ●
what they call ‘Factor I’ indicators, relating to the stability of political institutions, and political and civil rights, are correlated with indicators such as homicides, corruption and separatist movements. The higher the instability and the lower the rights, the worse are the other social indicators: – ‘Factor II’ indicators, which relate to disorder that does not threaten wholesale destruction of regimes (riots, assassinations and the like), are correlated with inequality of land ownership, low voter turnout and low level of citizen satisfaction;
On social capital
●
●
●
201
– Factor I and Factor II indicators are not correlated; economic growth is positively correlated with social integration in the sense that countries with low social scores have low economic growth, to some extent bearing out the Putnam–Fukuyama hypothesis outlined earlier. But the authors say that they cannot conclude that social cohesiveness causes better economic development; those issues that most concern Western countries are not correlated with either Factor I or Factor II: crime, drugs, rapes, suicide, ethnic fractionalization and so on; and the problems of social integration in the developing world appear to be very different to those of the developed world.
Cross-sectional analysis of this kind is potentially interesting. Principal components analysis can be used to reduce the large list of potential indicators to a few key indicators, but, even then, the linkages between indicators remain complex and there is no suggestion of a general theory linking social capital to economic development. World Bank (1997) suggests that data on the types of institutions that exist can be collected and correlated with the success or otherwise of development projects. Measuring institutions, rather than relationships, is attractive because the relationships are, as the Bank argues, already embedded in the institution. But, even then, institutions will vary substantially in their effectiveness: the existence of an institution does not mean that it is successful. Data on membership go some way to overcoming this problem since high membership presumably acts as a surrogate indicator of institutional success. An alternative procedure, also suggested in World Bank (1997), and which indirectly aims for measurement in the monetary sense, is to adopt a ‘production function’ approach whereby the contributions of man-made and natural capital to income growth are determined. What is left over (the residual) must be due to human and social capital. If human capital contributions can be isolated, then the final residual (the residual of the residual, as it were) measures social capital. Gaarder and Hamilton (1999) test this approach by taking various social capital indicators to see if they make any difference to an explanation of the determinants of human resources per capita. For example, human resources per capita can be regressed on an educational variable, say mean years of education per capita, and a social capital variable to see if the addition of the second variable makes any difference to the explanatory power of the equation overall. If it does, then social capital is significant. Gaarder and Hamilton (1999) find that the addition of a corruption index, a civil/political rights index, and indicators of trust and civic norms (Knack and Keefer, 1997) makes no significant difference to mean years of education as the explanation for human resources per capita. What matters, they argue, is
202
Policy and social capital
human capital, not social capital. Literacy alone provides the most powerful explanation of human resource availability. We suggest that the economic value of giving provides a measure of social capital, though not a comprehensive one. In other words, the kind of aggregate measure we found in Chapter 5 offers a way of measuring KS. The attractions of such an approach are (a) that it is a monetary measure rather than an indicator, and hence is more readily built into the theory of sustainable development outlined above; (b) it captures at least part of an absolutely essential ingredient of social capital (the relationship of care and concern for others) which is neglected in the existing literature; and (c) potentially, it is capable of being standardized across countries so as to permit comparisons. While we are not primarily concerned with the issue of whether large voluntary sectors are more or less correlated with development, it is easy to think of various hypotheses linking the two. Moreover, which hypothesis is correct does affect the extent to which the size of the voluntary sector is a reasonable measure of what is at least part of social capital. Thus we might expect a large voluntary sector in economies where state provision of welfare and other services is low, as in developing economies. What the state does not, or cannot, provide, has to be provided by voluntary associations. This would suggest that social capital would be high in lowincome countries. The appropriate indicator would be the money value of giving (plus consumers’ surplus) relative to gross domestic product (GDP). Against this, state provision may be minimal and voluntarism may also be limited if the family unit is strong: social capital is more confined to the family and a measure of social capital based on the scale of extra-family giving would miss the critical component of social capital, the intra-family associations, in such a context. The more developed an economy, the higher its income per capita, and the larger the voluntary sector is – when measured in terms of willingness to pay – simply because willingness to pay is positively linked to ability to pay, that is, to income. Expressed as a proportion of GDP, however, the money value of the voluntary sector might be less or more than in the low-income case. We have no prior expectations about the income elasticity of giving, although some evidence suggests that it is greater than unity for within-nation analyses (Ribar and Wilhelm, 1995). If income elasticities are greater than unity, then one would expect the share of voluntary sector value in GDP to be higher in rich than in less rich countries. But since our willingness to pay surveys reported in Part I are the only studies of their kind, we do not have comparative data on which to base any analysis. The best that can be done is to look at charities’ expenditure as a percentage of GDP. Salamon et al. (1995) report such a statistic for eight developed and transitional countries. While voluntary expenditures do rise as a percentage of GDP as GDP rises, the UK and
On social capital
203
the USA appear to behave differently from the other six countries, with very much higher expenditure shares for income levels that are comparable to those for France and Germany (the data relate to circa 1990). The data are clearly still limited but, a priori, we see no particular reason to suppose that social capital should be higher for a richer nation, even if we accept the Putnam–Fukuyama hypothesis that social capital figures prominently in the sustainability of the development process. There may, for example, be an element of take-off, with social capital being essential for the launch into sustained development rather than being clearly present once development has taken place. There is clearly a major research agenda here for the future. Given the uniqueness of our own study and the limited nature of other data, we can only speculate here on the possible links between social capital and development.
10. Conclusions and policy implications 10.1
THE SIZE OF THE CHARITABLE SECTOR
Part I of this volume was dedicated to measuring the size of the voluntary sector. While most studies have determined size by the measurement of income and expenditure by designated charities or the amount of employment associated with them, our own approach departs substantially from those studies by seeking the willingness to pay (WTP) for charitable services. The relevant WTP is that of the general public and the users, or beneficiaries, of the charities. To this end, we conducted extensive questionnaires with the general public, using stated preference procedures, and with one group of beneficiaries, the homeless. Our approach has the following advantages. First, as demonstrated in Chapter 1, it is rooted in the theory of welfare economics. Second, it enables us to lay the foundations for an approach to determining the efficiency of different forms of social provision. Third, it highlights the potential revenues that, in theory, charities could capture but which currently they do not, that is, the excess of WTP over actual donations. Fourth, it offers far more scope for a beneficiary-oriented approach to social provision. We argue that our approach provides a more accurate measure of the size of the voluntary sector. Rather than focusing on income and expenditure, which are, effectively, measures of the cost of the sector, WTP gives some idea of the benefits of the sector to society at large. Our calculations for the UK suggest that the social value of charities is some 40 per cent higher than the costs of providing that output. This result is based on our survey of the WTP of the general public to ensure that charities continue in existence. This is not, of course, income to the charities but an upper bound on potential additional income. The challenge is how to devise strategies for capturing as much as possible of this excess of true WTP over actual donations. Because of resource constraints, we were able to test the beneficiary approach (looking at actual users of charities’ services) only for one charitable sector, the provision of accommodation and care for the homeless. But we deliberately chose what we knew would be a difficult sector on the grounds that, if it works for that sector, it is likely to work for most other sectors. Questionnaire approaches among the homeless have obvious hazards and 204
Conclusions and policy implications
205
difficulties. But the results were reassuring and they suggest that the beneficiary value of homelessness charities is more than 200 per cent; that is, the social value of homelessness charities is some three times the cost of providing those services. Questionnaire approaches have a number of drawbacks, even though the degree of sophistication now applied to their design and execution is considerable. One obvious drawback is that the answers are hypothetical and we cannot be sure that people are giving true answers; that is, they might not actually be willing to pay what they say they are willing to pay. The general public questionnaire could suffer from this problem but it is notable that we would expect an upwards bias in WTP over actual donations, and that the excess WTP is not absurdly high. The homeless survey has one other reassuring factor, namely that the responses (in this case to a WTA question) were graded according to the quality of the hostel service in the manner we would expect and that the compensation required was anchored in what the homeless would actually have had to pay for alternative accommodation. These considerations give us faith that the answers we secured are reasonable.
10.2
ON THE EFFICIENCY OF CHARITABLE PROVISION
The procedure for estimating the social value of the charitable sector opens the way for tests of the efficiency of different forms of social provision. For example, consider the alternative ways in which some form of social care is provided. It can be provided totally by charities, totally by government (local or central) or totally by the private sector. Obviously there are also variations of ownership and management, for example state provision with private management. From the standpoint of efficient care, which is a central concern of government, what matters is the ratio of output to input, that is the ratio of social benefit to social cost of provision. Our approach permits these ratios to be measured. For any form of care, i, output (Qi) can be measured by the WTP of the beneficiaries for that output (WTPi). WTP captures the perceptions of the beneficiaries, so it allows for the fact that users may prefer one form of provision over another, even if the essential features of that care are the same. ‘Quality’ is measured as well as ‘quantity’. The cost of provision (Ci) can also be measured, so that a cost-efficiency indicator (ei) becomes: ei = WTPi / Ci A broader notion of efficiency would include the public’s WTP for particular forms of provision which, as Part I showed, reflects the public’s option values and existence values.
206
The price of virtue
While we have not investigated the costs and benefits of different forms of provision (only the costs and benefits of prevailing charitable institutions), the potential is clearly there for developing efficiency measures which could help guide the future provision of the services of the voluntary sector.
10.3
CAPTURING THE SURPLUS
Part I showed that the general public is willing to pay more for the provision of charitable services than they actually pay. This surplus reflects an untapped source of revenues, although capturing all of the additional 35 per cent is clearly not feasible. Part II looked at various features of this ‘capture’ problem in more detail. Like most charitable sectors, the UK charitable sector is financed by volunteer contributions (including volunteer time), sales of goods and government grants. Including volunteer time in the analysis shows that voluntary contributions account for about 35 per cent of charitable sector financing. We have shown that these voluntary contributions could be higher still if two issues could be tackled: the free-rider problem, whereby individuals give less than they are willing to because they know others are giving too, and the transaction costs problem, whereby individuals perceive that there are costs to the act of giving and that these outweigh the benefits of giving. Free-riding and transaction costs continue to exist even in the face of altruistic behaviour by individuals. That is, what look like purely self-interested features of behaviour, for example reducing contributions because others are giving, coexist with what may be very altruistic motives for giving. The existence of free-riding provides one of the traditional arguments for government financing. Government grants remain a major source of financing for charities in the UK. However, current government grants to charities stand at about £35 per person per year, which is well below the additional £40 to £60 per year that people say they would be willing to pay to support charities. The existence of the 35 per cent surplus suggests, clearly, that charities are underfunded, inasmuch as their true social value is 35 per cent higher than the incomes they receive. The issue is where this extra 35 per cent should come from. Clearly, one possibility would be to increase government grants to the charitable sector. However, it is not the only option. Another possibility would be to sharpen fiscal incentives for giving. Alternatively, the additional money could be captured by increasing charity fundraising efforts. It is interesting to enquire, from a social perspective, which of these three approaches would be the most effective and efficient. Effectiveness can be gauged in terms of the ability of the method to have a substantial impact on the flow of resources towards the charitable sector. Efficiency can be measured in terms of the deadweight loss – or social cost – per pound of additional
Conclusions and policy implications
207
resources yielded to the charitable sector. On the basis of the research presented in Part II, it is possible to present a tentative ranking of the three approaches against these two criteria. In terms of effectiveness, the results reported in Chapters 6–8 suggest that fundraising initiatives are likely to dominate fiscal incentives as a means of channelling additional resources towards the charitable sector, since the empirical results indicate that the gains available from improving fundraising methods are much more substantial than those which could be generated by sharpening fiscal incentives. On the one hand, given the limited degree to which giving responds to changes in fiscal incentives, as shown in Chapter 6, the kinds of changes in the tax price that might be considered feasible – such as increasing the scope of tax-efficient giving or altering the tax price in the order of a few pence in the pound – could not be expected to stimulate tax-efficient giving by much more than about 10–15 per cent. On the other hand, the results reported in Chapters 7 and 8 indicate that shifts between direct and remote fundraising methods or improvements in targeting techniques can increase the return to fundraising efforts several times over. Furthermore, the finding in Chapter 6 that donations were inelastic with respect to the tax price indicates that a reduction in the tax price of giving costs more in tax relief than it generates in additional philanthropic contributions. Thus, by reducing fiscal incentives to give, and spending part of the additional tax revenue on government grants to the voluntary sector, both charities and the Exchequer could end up better off than might otherwise have been the case. Consequently, from the government’s point of view, it is more efficient to increase direct grants rather than to sharpen fiscal incentives. The discussion thus far has concluded that both fundraising and government grants are likely to dominate improved fiscal incentives as a means of channelling resources towards voluntary organizations. It remains to compare the relative merits of fundraising and government grants. In doing so, it is important to recognize that both methods of financing have their associated deadweight loss. In the case of government grants, the raising of tax revenue to finance such grants imposes a deadweight loss on the economy. The quantification of this social cost is a controversial issue. However, estimates of the average deadweight loss per unit of revenue raised for the US economy range from 2.5 per cent (Harberger, 1964) to 30 per cent (Feldstein, 1995), while estimates of the marginal deadweight loss can be as high as 200 per cent (Feldstein, 1995). In the case of charitable fundraising, the information on the cost of raising funds is relatively sparse. One recent study suggests an
208
The price of virtue
average cost equal to between 20 per cent and 60 per cent of the revenue raised (Sargeant and Kaehler, 1998); however, no estimates of the marginal cost are available. Based on this comparison, a tentative conclusion would be that the social cost of raising funds is greater for the charity fundraiser than for the Exchequer. In summary, the results of Part I suggest that – based purely on free-rider arguments – there is some case for increasing the flow of resources to the voluntary sector. The three possible methods identified for achieving this are sharpening fiscal incentives, increasing government grants and redoubling fundraising efforts. Based on the research presented in Part II, fiscal incentives would appear to be the least effective of these three options, while government grants would appear to be the most efficient. Redoubling the efforts of charity fundraisers seems to be substantially more effective than sharpening fiscal incentives, but considerably less efficient than increasing government grants.
10.4
SOCIAL CAPITAL
Finally, in Chapter 9 we speculated that our measure of the economic value of the charitable sector provided one measure of social capital. Social capital is widely discussed as one of the essential features of successful economies. It reflects the degree of trust and stability in personal relationships, and between people and institutions. While the literature has focused on a number of indicators, ranging from surveys of the extent to which people trust each other to measures of civil and political liberties, it has not, to our knowledge, raised the interesting issue of the measurement of the giving relationship. Yet our measure of the economic value of the charitable sector is just such a measure. Because ours is the only study that has taken the full economic approach, we cannot offer comparative measures of social capital by country or even over time for one country. But we would like to think we have planted the idea and that others will take it forward.
References Abrams, B.A. and M.D. Schmitz (1978), ‘The “crowding-out” effect of governmental transfers on private charitable contributions’, Public Choice, 33, 29–39. Adamowicz, V. (1995), ‘Alternative valuation techniques: A comparison and movement to a synthesis’, in K.G. Willis and J.T. Corkindale (eds), Environmental Valuation: New perspectives, Wallingford, UK: CAB International. Adamowicz, W., V. Bhardeaj and B. Macnab (1993), ‘Experiments on the difference between willingness to pay and willingness to accept’, Land Economics, 69, 416–27. Adamowicz, W., J. Louviere and M. Williams (1994), ‘Combining revealed and stated preference methods for valuing environmental amenities’, Journal of Environmental Economics and Management, 26(3), 271–92. Andreoni, J. (1988), ‘Privately provided public goods in a large economy: the limits of altruism’, Journal of Public Economics, 35, 57–73. Andreoni, J. (1989), ‘Giving with impure altruism: applications to charity and Ricardian equivalence’, Journal of Political Economy, 97, 1447–58. Andreoni, J. (1990), ‘Impure altruism and donations to public goods: a theory of warm glow giving’, The Economic Journal, 100, 464–77. Andreoni, J. (1998), ‘Toward a theory of charitable fundraising’, Journal of Political Economy, 106, 1186–213. Andreoni, J. and J.K. Scholz (1998), ‘An econometric analysis of charitable giving with interdependent preferences’, Economic Inquiry, 36, 410–28. Arellano, M. and C. Meghir (1992), ‘Female labour supply and on-the-job search: an empirical model estimated using complementary datasets’, Review of Economic Studies, 59, 464–77. Arrow, K., R. Solow, P. Portney, E. Leamer, R. Radner and H. Schuman (1993), ‘Report of the National Oceanic and Atmospheric Administration Panel on contingent valuation’, Federal Register, 58, 4602–14. Atkinson, G., F. Machado and S. Mourato (2000), ‘Balancing competing principles of environmental equity’, Environmental and Planning A (in press). Atkinson, G., R. Dubourg, K. Hamilton, M. Munasinghe, D.W. Pearce and C. Young (1997), Measuring Sustainable Development: Macroeconomics and the Environment, Cheltenham, UK and Brookfield, USA: Edward Elgar. 209
210
The price of virtue
Banks, J. and S. Tanner (1997), The State of Donation: Household Gifts to Charity, 1974–96, Commentary 62, London, UK: Institute for Fiscal Studies. Banks, J. and S. Tanner (1998), Taxing Charitable Giving, Commentary 75, London, UK: Institute for Fiscal Studies. Barrett, K. (1991), ‘Panel data estimates of charitable giving: a synthesis of techniques’, National Tax Journal, 44, 365–81. Baumol, W. (1962), ‘On the theory of the expansion of the firm’, American Economic Review, 52, 1078–87. Becker, G. (1974), ‘A theory of social interactions’, Journal of Political Economy, 82, 1063–93. Beggs, S., S. Cardell and J. Hausman (1981), ‘Assessing the potential demand for electric cars’, Journal of Econometrics, 16, 1–19. Bishop, R.C. and T.A. Heberlein (1979), ‘Measuring values of extra-market goods: are indirect measures biased?’, American Journal of Agricultural Economics, 61(5), 926–30. Boserup, E. (1981), Population and Technological Change: a Study of Long Term Trends, New York: Aldine Publishing Co. Boskin, M.J. and M. Feldstein (1977), ‘Effects of the charitable deduction on contributions by low income and middle income households: evidence from the National Survey of Philanthropy’, Review of Economics and Statistics, 59, 351–4. Brook, L., B. Taylor and G. Prior (1991), British Social Attitudes 1990 Survey: Technical Report, London, UK: Social and Community Planning Research. Brown, E. and H. Lankford (1992), ‘Gifts of money and gifts of time: estimating the effects of tax-prices and available time’, Journal of Public Economics, 47, 321–41. CAF (Charities Aid Foundation) (1999), Dimensions of the Voluntary Sector, West Malling, Kent: Charities Aid Foundation. Cameron, T. (1988), ‘A new paradigm for valuing non-market goods using referendum data: maximum likelihood estimation by censored logistic regression’, Journal of Environmental Economics and Management, 15, 355–79. Carson, R., T. Groves and M. Machina (1997), ‘Informational and Strategic Properties of Value Elicitation Methods’, paper presented at the National Science Foundation Workshop on Valuation and Environmental Policy, Arlington, VA, USA. Clotfelter, C.T. (1980), ‘Tax incentives and charitable giving: evidence from a panel of tax-payers’, Journal of Public Economics, 13, 319–40. Clotfelter, C.T. (1985), Federal Tax Policy and Charitable Giving, Chicago: The University of Chicago Press.
References
211
Coleman, J. (1988), ‘Social capital in the creation of human capital’, American Journal of Sociology, 94 (supplement), 95–120. Coursey, D., J. Hovis and W. Schulze (1987), ‘The disparity between willingness to accept and willingness to pay measures of value’, Quarterly Journal of Economics, 102, 679–90. Cummings, R., D. Brookshire and W. Schulze (eds) (1986), Valuing Environmental Goods – A State of the Arts Assessment of the Contingent Valuation Method, Totowa, NJ: Rowman and Allanheld. Daniel, W.W. (1990), Applied Non-Parametric Statistics, 2nd edn, Boston, MA: PWS–Kent Publishing Company. Dasgupta, P. (1995), ‘The population problem: theory and evidence’, Journal of Economic Literature, 33, 1875–902. Day, K.M. and R.A. Devlin (1996), ‘Volunteerism and crowding out: Canadian econometric evidence’, Canadian Journal of Economics, 29, 37–53. Department of Employment (1991, 1992, 1993), New Earnings Survey, London: Her Majesty’s Stationery Office. Dye, R.F. (1980), ‘Contributions of volunteer time: some evidence on income tax effects’, National Tax Journal, 33, 89–93. Ecotec Research and Consulting Ltd (1993), A Review of UK Environmental Expenditure: A Final Report to the Department of the Environment, Department of the Environment, London. EFTEC (Economics for the Environment Consultancy Ltd) (1997), The Social Value of the Charitable Sector, report for the Charities Aid Foundation, London. EFTEC (Economics for the Environment Consultancy Ltd) (2001), Guidance on Using Stated Preference Techniques for the Economic Valuation of NonMarket Effects, report for UK Department of the Environment, Transport and the Regions, London. English, J., M. Tiffen and M. Mortimore (1994), Land Resource Management in Machakos District, Kenya, Environment Department Paper no. 5, Washington DC: World Bank. Feldstein, M. (1995), Tax Avoidance and the Deadweight Loss of the Income Tax, Working Paper no. 5055, Cambridge, MA: National Bureau of Economic Research. Feldstein, M. and C. Clotfelter (1976), ‘Tax incentives and charitable contributions in the United States’, Journal of Public Economics, 5, 1–26. Feldstein, M. and A. Taylor (1976), ‘The income tax and charitable contributions’, Econometrica, 44, 1201–22. Fenyo, A., J. Kendall, M. Knapp, J. Pinner and 6 Perri (1993), The Voluntary Sector and the Environment: Scope, Contributions, Issues, Research and Statistics Unit Occasional Paper 5, Tonbridge, Kent: Charities Aid Foundation.
212
The price of virtue
Foster, V. (1997), ‘What value should be placed on volunteering?’, Dimensions of the Voluntary Sector, Tonbridge, Kent: Charities Aid Foundation. Foster, V. (1999a), ‘How much targeting is optimal? A simulation of fundraising strategies for environmental charities’, in On the Voluntary Financing of Public Goods, PhD Thesis, University College London. Foster, V. (1999b), ‘How are “warm glows” created? The role of fundraising technology in charitable giving’, in On the Voluntary Financing of Public Goods, PhD Thesis, University College London. Foster, V. (1999c), ‘Are British philanthropists price-sensitive? New evidence from the Individual Giving Survey’, in On the Voluntary Financing of Public Goods, PhD Thesis, University College London. Foster, V. and S. Mourato (2000), ‘Measuring the impacts of pesticide use in the UK: a contingent ranking approach’, Journal of Agricultural Economics, 51 (1), 1–21. Foster, V., I. Bateman and D. Harley (1997), ‘Real and hypothetical willingness to pay for environmental preservation: a non-experimental comparison’, Journal of Agricultural Economics, 48, 123–38. Freeman, A.M. (1993), The Measurement of Environmental and Resource Values: Theory and Methods, Washington, DC: Resources for the Future. Freeman, R.B. (1997), ‘Working for nothing: the supply of volunteer labour’, Journal of Labour Economics, 15, 140–66. Fukuyama, F. (1995), Trust: the Social Values and the Creation of Prosperity, Harmondsworth, UK: Penguin. Gaarder, M. and K. Hamilton (1999), Estimating the economic returns to human and social capital: is literacy the whole story?, mimeo, Washington, DC: World Bank. Glazer, A. and K.A. Konrad (1996), ‘A signalling explanation for charity’, American Economic Review, 86, 1019–28. Greene, W.H. (1993), Econometric Analysis, New York: Macmillan. Greene, W.H. (1997), Econometric Analysis, 3rd edn, New York: Macmillan. Halfpenny, P. and C. Pettipher (1992), Individual Giving and Volunteering in Britain: Who Gives What and Why?, 5th edn, Tonbridge, Kent: Charities Aid Foundation. Halfpenny, P. and C. Pettipher (1993), Individual Giving and Volunteering in Britain: Who Gives What and Why?, 6th edn, Tonbridge, Kent: Charities Aid Foundation. Halfpenny, P. and C. Pettipher (1994), Individual Giving and Volunteering in Britain: Who Gives What and Why?, 7th edn, Tonbridge, Kent: Charities Aid Foundation. Hanemann, W.M. (1984), ‘Welfare evaluations in contingent valuation experiments with discrete responses’, American Journal of Agricultural Economics, 66, 332–41.
References
213
Hanemann, W.M. (1991), ‘Willingness to pay and willingness to accept: how much can they differ?’, American Economic Review, 81, 635–47. Hanemann, W.M. and B. Kanninen (1996), The Statistical Analysis of Discrete-Response CV Data, Working Paper no. 798, Department of Agricultural and Resource Economics, Berkeley, CA: University of California. Hanemann, W.M., J. Loomis and B. Kanninen (1991), ‘Estimation efficiency of double-bounded dichotomous choice contingent valuation’, American Journal of Agricultural Economics, 73, 255–1263. Hanley, N., D. MacMillan, R.E. Wright, C. Bullock, I. Simpson, D. Parsisson and B. Crabtree (1998), ‘Contingent valuation versus choice experiments: estimating the benefits of Environmentally Sensitive Areas in Scotland’, Journal of Agricultural Economics, 49(1), 1–15. Harbaugh, W.T. (1998), ‘What do donations buy? A model of philanthropy based on prestige and warm glow’, Journal of Public Economics, 67, 269– 84. Harberger, A. (1964), ‘Taxation, resource allocation and welfare’, in J. Due (ed.), The Role of Direct and Indirect Taxes in the Federal Reserve System, Princeton, NJ: Princeton University Press. Heckman, J.J. (1979), ‘Sample selection bias as a specification error’, Econometrica, 47, 153–61. Hotelling, H. (1929), ‘Stability in competition’, The Economic Journal, 39, 41–57. Inland Revenue (1998), Inland Revenue Statistics, London: Her Majesty’s Stationery Office. Janes, M. and B. Roberts (1990), ‘Labour costs in 1988’, Employment Gazette, September Issue, Department of Employment, London: Her Majesty’s Stationery Office. Jencks, C. (1994), The Homeless, Cambridge, MA: Harvard University Press. Johnson, F.R. and W.H. Desvousges (1997), ‘Estimating stated preferences with rated-pair data: environmental, health, and employment effects of energy programs’, Journal of Environmental Economics and Management, 34, 79–99. Jones, A. and J. Posnett (1991a), ‘Charitable donations by UK households: evidence from the Family Expenditure Survey’, Applied Economics, 23, 343–51. Jones, A. and J. Posnett (1991b), ‘The impact of tax deductibility on charitable giving by covenant in the UK’, The Economic Journal, 101, 1117–29. Jurgen, F. (ed.) (1988), Affordable Housing and the Homeless, New York and Berlin: de Gruyter. Kahneman, D. and J. Knetsch (1992), ‘Valuing public goods: the purchase of moral satisfaction’, Journal of Environmental Economics and Management, 22, 57–70.
214
The price of virtue
Kahneman, D. and A. Tversky (1979), ‘Prospect theory: an analysis of decisions under risk’, Econometrica, 47, 263–91. Kendall, J. (1996), ‘The voluntary sector: international comparisons’, in Charities Aid Foundation, Dimensions of the Voluntary Sector: How is the Voluntary Sector Changing, London: Charities Aid Foundation. Kendall, J. and M. Knapp (1995), Voluntary Means, Social Ends, Canterbury, Kent: Personal Social Services Research Unit. Khanna, J., J. Posnett and T. Sandler (1995), ‘Charity donations in the UK: new evidence based on panel data’, Journal of Public Economics, 56, 257– 72. Kingma, B. (1989), ‘An accurate measurement of the crowd-out effect, income effect and price effect for charitable contributions’, Journal of Political Economy, 97, 1197–207. Klitgaard, R. and J. Fedderke (1995), ‘Social integration and disintegration: an explanatory analysis of cross-country data’, World Development, 23, 357–69. Knack, S and P. Keefer (1997), ‘Does social capital have an economic payoff: a cross-country investigation’, Quarterly Journal of Economics, 112(4), 1251–88. Knapp, M. (1990), Time is Money: The Costs of Volunteering in Britain Today, Paper no. 3, Voluntary Action Research, London: Volunteer Centre. Knapp, M., V. Koutsogeorgopoulou and J. Davis Smith (1995), The Economics of Volunteering: Examining Participation Patterns and Levels in the UK, Discussion Paper 1060, Personal Social Services Research Unit, Canterbury, Kent: University of Kent. Krinsky, I. and L.A. Robb (1986), ‘On approximating the statistical properties of elasticities’, Review of Economics and Statistics, 68, 715–19. Lareau, T. and D. Rae (1985), ‘Valuing willingness to pay for diesel odour reduction: an application of the contingent ranking technique’, Southern Economic Journal, 55(3), 728–42. Louviere, J. (1988), Analyzing Decision-Making: Metric Conjoint Analysis, no. 67, Quantitative Applications in the Social Sciences Series, Newbury Park, CA: Sage University Papers, Sage Publications. Maddala, G.S. (1983), Limited Dependent and Qualitative Variables in Econometrics, Econometric Society Monographs no. 3, New York: Cambridge University Press. McFadden, D. (1994), ‘Contingent valuation and social choice’, American Journal of Agricultural Economics, 76, 689–708. Menchik, P.L. and B.A. Weisbrod (1987), ‘Volunteer labour supply’, Journal of Public Economics, 32, 159–83. Mitchell, R. and R. Carson (1989), Using Surveys to Value Public Goods: The Contingent Valuation Method, Baltimore, MD: Johns Hopkins Press.
References
215
The National Trust (1995), 1994/95 Annual Report and Accounts, London: The National Trust for Places of Historic Interest or Natural Beauty. North, D. (1990), Institutions, Institutional Change and Economic Performance, Cambridge: Cambridge University Press. Office of Population Censuses and Surveys (1992), Social Trends, London: Government Statistical Service, Central Statistical Office. O’Flaherty, B. (1996), Making Room: The Economics of Homelessness, Cambridge, MA: Harvard University Press. Olson, M. (1965), The Logic of Collective Action, Boston, MA: Harvard University Press. Pearce, D.W. (1999), ‘Sustainable development’, in D.W. Pearce, Economics and Environment: Economics: Essays in Ecological Economics and Sustainable Development, Cheltenham, UK and Brookfield, USA: Edward Elgar. Pharoah, C. (1997), ‘Broad trends: sector size, scope, income and expenditure’, in Charities Aid Foundation (CAF), Dimensions of the Voluntary Sector, West Malling, Kent: Charities Aid Foundation. Pharoah, C. and M. Smerdon (eds) (1998), Dimensions of the Voluntary Sector, West Malling, Kent: Charities Aid Foundation. Pharaoh, C. and S. Tanner (1997), ‘Trends in charitable giving’, Fiscal Studies, 18, 427–44. Posnett, J. and T. Sandler (1989), ‘Demand for charity donations in private non-profit markets: the case of the UK’, Journal of Public Economics, 40, 187–200. Putnam, R.D. (1993), Making Democracy Work: Civic Traditions in Modern Italy, Princeton, NJ: Princeton University Press. Randall, A. and J. Stoll (1980), ‘Consumer surplus in commodity space’, American Economic Review, 70, 449–55. Randall, G. and S. Brown (1996), From Street to Home: An Evaluation of Phase 2 of the Rough Sleepers Initiative, Department of the Environment, London: The Stationery Office. Reece, W.S. (1979), ‘Charitable contributions: new evidence on household behaviour’, American Economic Review, 69, 142–51. Reece, W.S. and K.D. Zieschang (1985), ‘Consistent estimation of the impact of tax deductibility on the level of charitable contributions’, Econometrica, 53, 271–93. Resource Information Service (1996), London Hostels Directory 1996, London: RIS. Ribar, D. and M. Wilhem (1995), ‘Charitable contributions to international relief and development’, National Tax Journal, 47(2), 229–44. Richer, J. (1995), Green Giving: An Analysis of Contributions to Major US
216
The price of virtue
Environmental Groups, Discussion Paper 95–39, Washington, DC: Resources for the Future. Rose-Ackerman, S. (1982), ‘Charitable giving and “excessive” fundraising’, Quarterly Journal of Economics, 97, 193–212. Rose-Ackerman, S. (1986), ‘Charitable giving and excessive fundraising’, in D. Rose-Ackerman (ed.), The Economics of Non-Profit Institutions, New York: Oxford University Press. Salamon, L. (1995), Partners in Public Services: Government – Non-profit Relations in the Welfare State, London: Johns Hopkins Press Ltd. Salamon, L., H. Anheier, W. Sokolowski and associates (1995), The Emerging Sector: a Statistical Supplement, Institute for Policy Studies, Baltimore, MD: Johns Hopkins University Press. Samuelson, P. (1954), ‘The pure theory of public expenditure’, Review of Economics and Statistics, 36, 387–9. Sargeant, A. and J. Kaehler (1998), ‘Charities and fundraising costs – just what is the benchmark?’, Dimensions of the Voluntary Sector, West Malling, Kent: Charities Aid Foundation. Schiff, J. (1985), ‘Does government spending crowd out charitable contributions?’, National Tax Journal, 38, 535–46. Schwartz, R.A. (1970), ‘Personal philanthropic contributions’, Journal of Political Economy, 78, 1264–91. Shyamsundar, P. and R. Kramer (1996), ‘Tropical forest protection: an empirical analysis of the costs borne by local people’, Journal of Environmental Economics and Management, 31, 129–44. Smith, J., S. Mourato, E. Veneklaas, R. Labarta, K. Reategui and G. Sanchez (1997), Willingness to Pay for Environmental Services Among Slash-andBurn Farmers in the Peruvian Amazon: Implications for Deforestation and Global Environmental Markets, CSERGE/CIAT/ICRAF Working Paper, London: University College London. Smith, V.H., M.R. Kehoe and M.E. Cremer (1995), ‘The private provision of public goods: altruism and voluntary giving’, Journal of Public Economics, 58, 107–26. Solow, R. (1992), An Almost Practical Step Towards Sustainability, Washington, DC: Resources for the Future. Steinberg, R. (1986), ‘Should donors care about fundraising?’, in S. RoseAckerman (ed.), The Economics of Non-Profit Institutions, New York: Oxford University Press. Steinberg, R. (1987), ‘Voluntary donations and public expenditures in a federalist system’, American Economic Review, 77, 24–36. Stiglitz, J. (1979), ‘A neoclassical analysis of the economics of natural resources’, in V.K. Smith (ed.), Scarcity and Growth Reconsidered, Baltimore, MD: Johns Hopkins University Press, pp. 36–66.
References
217
Sugden, R. (1982), ‘On the economics of philanthropy’, The Economic Journal, 92, 341–50. Sugden, R. (1984), ‘Reciprocity: the supply of public goods through voluntary contributions’, The Economic Journal, 94, 772–87. Taussig, M.K. (1967), ‘Economic aspects of the personal income tax treatment of charitable contributions’, National Tax Journal, 20, 1–19. Tirole, J. (1988), The Theory of Industrial Organisation, Cambridge, MA: The MIT Press. Victor, P., H.E. Hanna and A. Kubursi (1994), How strong is weak sustainability?, paper presented at the International Symposium on Models of Sustainable Development, Paris, France. Volunteer Centre UK (1995), ‘The economic value of volunteering’, Volunteering Research Bulletin, February, London: Volunteer Centre. Volunteer Centre UK/Social and Community Planning Research (1991), On Volunteering: A Qualitative Research Study of Images, Motivations and Experiences, Berkhamstead, UK: Volunteer Centre UK. Wang, H. (1997), ‘Treatment of “don’t know” responses in contingent valuation surveys: a random valuation model’, Journal of Environmental Economics and Management, 32, 219–32. Weisbrod, B.A. and N.D. Dominguez (1986), ‘Demand for collective goods in private non-profit markets’, Journal of Public Economics, 30, 83–95. Willig, R. (1976), ‘Consumer’s surplus without apology’, American Economic Review, 66, 589–97. World Bank (1997), Expanding the Measure of Wealth: Indicators of Environmentally Sustainable Development, Washington DC: World Bank.
Index accommodation, hostels expenditure, proposed scenario 97 preferences 87–9 price of 73 additional welfare, estimating 13 age environmental group membership 178, 179, 181 fundraising methods 160 participation rates of giving 141 user benefits survey 79–80 WTP 55, 56 aggregate benefits, estimating 114–19 accounting framework 114–16 net social value 116–18 interpretation and policy implications 118–19 alcohol, expenditure, proposed scenario 97 altruistic benefits 11 donating 36 volunteering 43, 44–5 see also impure altruism AND strategy 183–5 Asia Miracle studies 198 associations, social capital 193–4 attitudes general public survey explanatory variables 54 payment ladder results 56 towards donations 33–42 towards volunteering 42–6 WTP 55 towards countryside 173–6 towards hostels 84–9, 89, 90, 99 beneficiaries 11 benefits charity’s outputs 10 economic theory definition 9 to beneficiaries 11–14
to general public 19–71 to homeless 72–100 of volunteering 101–13 see also aggregate benefits Benthamite welfare function 10 bid levels, contingent valuation 23, 47, 55–8 binary logit models, targeting donors 183–6, 187 British Social Attitudes Survey 173–6 broad voluntary sector 5 campaigning, environmental groups 171 capital-based theory, sustainable development 194–8 cash donations 30, 32 charitable sector aggregate benefits 114–19 indicators of size 5–8 measurement of output 3–4 measuring economic value 9–18 size of 204–5 social value of 205–6 charities benefits to general public survey 19–71 design 20–28 results 29–61 discussion and conclusions 61–4 statistical appendix 65–71 benefits to homeless survey 72–100 design 74–8 results 79–98 discussion and conclusions 98–100 net social value 10, 17–18, 102, 116–18 total social cost 16–17, 116 total social value 11–14, 114–15 Charities Aid Foundation 5, 8, 102 chlorofluorocarbons (CFCs) 195 choice modelling techniques 16 219
220
The price of virtue
club facilities, membership numbers 172 coefficient estimates price responsiveness, giving 140–43 targeting donors 176–81 commitment, in giving 39, 40, 41 compensation hostel closures 89–95, 99 WTA measure 13 competition environmental groups 172–3 for funds 187–8 product differentiation 170 contingent ranking general public survey 20 design features 25–7 results 58–9, 62 statistical and theoretical framework 69–71 contingent valuation 15–16 general public survey 20 questions 22–5 results 46–58, 62 statistical and theoretical framework 65 user benefits survey 73 questionnaire 74–8 results 89–95 contributions equation 142 cost-effectiveness measures 4 costs charitable services 10 fundraising 169–70 see also labour costs; production costs; replacement cost approach; total social cost; transaction costs Council for the Protection of Rural England (CPRE) 171, 173 counselling services, use of 84–5, 86 credit cards 30 cross-membership patterns 175 cross-tax-price elasticity, volunteering 129–31 culture donations 32, 33 WTP 59 debriefing questions general public survey 59–61
user benefits survey 95–8 demographic variables BSAS 176 user benefits survey 79 Diana, Princess of Wales, death, WTP 55 dichotomous choice see double-bounded dichotomous choice direct fundraising frequency distribution, donations 157–8 geographic variables 160 gift size 165 likelihood of giving 159 marginal effects 163 popular charities 160 probability of obtaining a gift 164 success rates 158–9 summary statistics 155, 156 transaction costs 152 warm glow benefit 152, 165 discrete response formats 65–6 donations general public survey attitudes towards 33–42 current 29–33 results 54, 55, 57, 62 in kind 7, 102 size of, warm glow 149 see also giving donations equation 160–62 donors opportunity costs 16 targeting 167–90 British Social Attitudes Survey 173–6 conceptual framework 168–70 conclusion 188–9 empirical analysis 176–81 environmental groups 171–3 simulation, fundraising strategies 181–8 don’t know options, survey questions 23–4 double-bounded dichotomous choice elicitation questions 22–3 results 46–9, 62 statistical and theoretical framework 65–8
Index drugs, expenditure, proposed scenario 97 economic development, social capital 198–9 economic growth 201 economic value, charities 9–18 aggregate benefits 114–19 general public survey 19–71 homeless survey 72–100 volunteering 101–13 economic variables general public survey 53, 54, 56 user benefits survey 79, 93 education environmental group membership 179, 181 general public survey, WTP 55, 56 user benefits survey 79, 80, 95 effectiveness, charitable provision 206, 207 efficiency charitable provision 205–6 measures of 4 efficiency price 135–6 donation equation 161 elasticities for giving 144 mean monthly donations 139, 140 participation rates 138, 141 effort, level of fundraising 151–2 elasticities, price-responsiveness, giving 143–5 elicitation, giving 149 embarrassment motivation 35, 36, 42 emergency night shelters 77, 78, 95 employment user benefits survey 79 volunteering 140 employment related costs, volunteering 102 endogeneity problems 125 environment donations to 32, 33 WTP 59 environmental groups, targeting donors 171–3 environmental improvement, social capital 199 ex-offenders’ projects 78 existence value, donating 36
221
expenditure measures of 8 user benefits survey breakdown of 80–81 desired compensation, hostel closure 93–5 income, proposed scenario 96–7 external benefits 11 Factor I indicators 200, 201 Factor II indicators 200, 201 families, large 195 Family Expenditure Survey (FES) 103 analysis of donations 128 definition of charitable donations 132–3 favourite charities 42 fiscal incentives, for giving 123–46 fixed-budget setting 187 food, expenditure, proposed scenario 97 foresight, in giving 37, 39, 40, 41 free-riding depression of voluntary contributions 7 government financing 206 pure altruism theory 148 tax exemptions, donations 123 frequency distributions donations, fundraising methods 157–8 efficiency price 135, 136 housing preferences 88 tax price 134, 135 time price 137–8 Friends of the Earth (FoE) 171, 173 fundraising costs 169–70 income measures 5 managerial objectives 168–9 methods, valuation 147–66 motivation 148–50 conceptual framework 150–53 hypotheses results 153–8 controlled hypotheses tests 168–4 conclusions 165–6 simulation of strategies 181–8 targeting 167–8 funds, competition for 187–8 gender, WTP 55, 56
222
The price of virtue
general public benefits survey objectives 19–20 design 20–28 results 29–61 discussion and conclusions 61–4 statistical appendix 65–71 capturing surplus (WTP) 206–8 incremental WTP 118 total social value, benefits 115 giving motivation 148–50 price-responsiveness 123–46, 207 private 6–8 process of 37–42 value of 200–203 see also donations goods sale of 6 WTP 14 government funding 4, 8 government grants estimation, net hourly wage of volunteers 105–6 free riding 206 and general public, incremental WTP 119 opportunity costs 17 Greenpeace (GP) 171, 173 gross domestic product (GDP), money value of giving 202–3 gross revenue maximization 169, 184, 185 health and medical research donations 32, 33 importance of charities 64 WTP 58, 59 Hekman models net hourly wage, volunteers 103–6 performance, fundraising methods 160–64 price-responsiveness of giving 140–43 homelessness causes 83 defined 72 duration of 83–4 hostels 72 characteristics 78 price of accommodation 73
shutdown compensatory amounts 89–95 hypothetical scenario 74–5 types 77 use of services 84–9 hourly value, volunteers’ time 103–6 hours, volunteered 30, 32 house rules, hostels 85–6 housing history, homeless 82–3 housing and homelessness general public survey contingent ranking results 59 contingent valuation results 47, 48, 49, 51, 55 donations 32 importance of 64 net social value 116, 117 user benefits survey 72–100 design 74–8 discussion and conclusions 98–100 results 79–98 housing projects 72 housing schemes 77, 78 human capital 201–2 hypotheses see null hypotheses importance, of charities 33, 63–4 impure altruism 12, 149, 150–51 income environmental group membership 178, 180 from government 8 general public survey, WTP 53–5, 56 participation rates, giving 141 private earned 6 success of direct fundraising 158–9 user benefits survey allocation of 81–2 expenditure, proposed scenario 95–6 sources of 80 income elasticities, for giving 144, 145 income measures 5–6 incremental WTP all beneficiaries 117 general public 118–19 indicators size of charitable sector 5–8 social capital 1, 200–203 indirect users, benefits to 11–13, 36
Index Individual Giving Survey (IGS) 30, 102 activity based categories 106–8 definition of charitable donations 132 efficiency price 135 hypothesis testing 153 price and income elasticities 124 tax price 134 Institute of Charity Fundraising Managers 149–50 institutions, data on 201 intensity, fundraising 158 Labour Cost Survey 108 labour costs, estimating, volunteers 106–8 land ownership 171 literature, price-responsiveness of giving 123, 125–32 low-support hostels 77, 78, 92 managerial objectives, fundraising 168–9 marginal cost, fundraising 169–70 marginal cost-revenue ratio, targeting 184, 185, 189 membership dues, private income 6 environmental groups fees 171 numbers and club facilities 172 patterns 174–6 probability of subscribing 181–2 moral principles, giving 148 moral satisfaction 12, 149 motivations for donating 33–7, 148–50 towards volunteering 42–3 multi-variable targeting strategy 183 narrow voluntary sector 5 National Trust (NT) 171, 173, 175, 185 needy groups, net revenue maximization 168–9 net hourly wage, estimating, volunteers 103–6, 129 net revenue maximization 168–9, 184, 185 net social value, of charities 10, 17–18, 102, 116–18 New Earnings Survey 106, 107
223
non-tax-efficient donations 31 null hypotheses, fundraising 152–3 controlling extraneous factors 160–64 potential biases 158–60 results 153–8 open ended questions 24–5 opportunity cost approach 15 volunteering benefits survey 102–6, 112 optimal fundraising strategy 182–7 option benefits 11 donating 36 volunteering 43 output measures, charitable sector 3–4 output-based approach, volunteering benefits survey 109–10, 112 overseas aid donations 32, 33 WTP 59 panel data 125 parametric probability models 67–8 participation equations net hourly wage estimation 103–4 price-responsiveness, giving 143 participation rates IGS records 133 price variables 138–9 socioeconomic variables 141 volunteering 140 paternalism 7, 11 payment ladder elicitation 24, 25 results 46, 49–53, 56–7, 62 statistical and theoretical framework 68–9 personal enjoyment, volunteering 42 policy implications, of benefits 118–19 political capital 194 popular charities 160 population change, capital stocks 195, 196, 197 preferences hostels 84–9 measurement of person’s 9 prestige, giving 149, 151 price responsiveness, giving 123–46, 207 literature overview 125–32 analytical framework 133–40
224
The price of virtue
dataset 132–3 empirical results 140–45 conclusions 145–6 principal components analysis 201 private benefits 12 private earned income 6 private giving 6–8 private sector equivalent 14 probability models, parametric 67–8 product differentiation, competition 170 production costs, output of charities 8 production function approach 15, 201 protection-oriented groups 174 purchases, donations through 31 quality index, fundraising methods 161–2 questionnaires 205 questions general public survey contingent ranking method 25–7 contingent valuation method 22–5 debriefing 59–61 user benefits survey 74–8 debriefing 95–8 WTA format 99 Ramblers Association (RA) 171, 172 random utility model approach 66–7, 70 rank-ordered logit model 58–9, 71 recreation-oriented groups 174, 180 religion, importance in giving 141 remote fundraising frequency distribution, donations 157, 158 geographic variables 160 likelihood of giving 159 marginal effects 163 probability of obtaining a gift 164 summary statistics 155, 156 transaction costs 152 replacement cost approach 106–9, 112 resources, opportunity costs 18 Royal Society for the Protection of Birds (RSPB) 171–2, 175 sales, income from 6 satisfaction, with hostels 85, 86 scope, of giving 37, 38, 39–40, 41 scorn, charitable behaviour 151
selectivity models 128, 131 selfish motivation in donating 35–6 volunteering 42–3, 44–5 semi-supportive projects 77, 78 services indirect benefits 18 WTP 14 sex, WTP 56 simulation, fundraising strategies 181–8 single-variable targeting strategy 183, 185 social capital 208 concepts 193–4 indicators of 200–203 measuring 3–4 nature of 198–200 sustainable development 194–8 social prestige, giving 149, 151 social services donations 32, 33 importance of 64 WTP 58, 59 social situations, fundraising 152 Social Trends 175 social value estimating 205–6 output of charities 6 see also net social value; total social value social welfare functions 9–10 society importance of charities to 33 value of charities to 5 well-being of 9 socioeconomic variables BSAS 176 general public survey comparison 29 description 54 valuation functions 56 WTP 55 participation rates of giving 141 targeting donors 176–81, 188 user benefits survey compensation, hostel closures 93 profile 79–84 specific needs projects 77, 78 spontaneous giving 42 stated preference techniques 15, 20
Index statistics, fundraising methods 154–6 subscription fees, environmental groups 186 support services, use of 84–5, 86 supportive projects 77, 78, 92 surplus, WTP, capturing 206–8 sustainable development, social capital 194–8 target groups benefits 11 opportunity costs 16 targeting, donors 167–90 tax incentives, for giving 123–46 tax price 133–5 elasticities for giving 144, 145 mean monthly donations 139 participation rates 138, 139, 141 tax relief, UK 128 tax-efficient donations 31 tax-efficient schemes, awareness of 134 technological change, capital stocks 195, 196 telescoping effects, IGS 133 time hours volunteered 30, 32 volunteers’, valuing 7, 101–13 time price 136–8 elasticities for giving 144 mean monthly donations 139 participation rates 138, 139 volunteering 140 tobacco, expenditure, proposed scenario 97 Tobit specification models 128 top-down approach, general public survey 25 results 51, 62 total social costs of charities 16–17, 116 see also opportunity cost approach total social value, of charities 11–14, 114–15 traditional hostels 77, 78, 92 transaction costs 7 fundraising methods 152 size of gifts 158 United Kingdom, studies price responsiveness 127, 128, 129
225
volunteering 131 United States price responsiveness, giving 123, 125 tax treatment of donations 127 volunteer status 150 user fees 6 users benefits of charities to 72–100 output measures 4 utility, from donations 150–51 valuation general public survey results 46–61 of volunteering 43 returns, fundraising methods 147–66 user benefits survey, results 89–95 valuation functions general public survey 53, 56–7, 69 user benefits survey 93, 94 vertical associations, social capital 193–4 Volunteer Centre UK 101 volunteering attitudes towards 42–6 benefits survey reason for valuation 101–2 opportunity cost approach 102–6 replacement cost approach 106–9 output-based approach 109–10 comparison of approaches 110–12 future valuation 112–13 employment 140 participation in 3–4 price and income elasticities 129–31, 140 volunteers benefits to 12–13 opportunity costs 17 status, US 150 total social value 115 valuing time given 7, 101–13 wage equation 103, 104–6 Wald tests environmental group membership 180, 187 fundraising methods 160 warm glow benefits fundraising 151, 152, 158–9 impure altruism 12
226
The price of virtue
magnitude of donations 149 selfish motivation 35–6, 42 welfare economics 9–10 well-being in economic theory 9 impact of charities on 102 willingness to accept (WTA) 9 in environmental economics 73–4 estimating additional welfare 13 shutdown of hostels 90–95, 99 user benefits survey 75 willingness to pay (WTP) 9 demand for goods or services 14 estimating additional welfare 13
general public survey 21–2 contingent ranking results 58–9 contingent valuation results 47–58 debriefing questions 59–61 incremental 117, 118–19 World Wide Fund for Nature (WWF) 171, 173 WTA see willingness to accept WTP see willingness to pay younger generations, fundraising methods 160 zero donations, modelling 128