LIST OF CONTRIBUTORS Christopher P. Adams
Bureau of Economics, Federal Trade Commission, Washington, D.C., USA
Jan Er...
27 downloads
732 Views
1MB 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
LIST OF CONTRIBUTORS Christopher P. Adams
Bureau of Economics, Federal Trade Commission, Washington, D.C., USA
Jan Erik Askildsen
Department of Economics, University of Bergen, Bergen, Norway
José Alberto BayoMoriones
Departamento de Gestión de Empresas, Universidad Pública de Navarra, Pamplona, Spain
Chris Brewster
South Bank University, London, UK
Chris Doucouliagos
School of Economics, Deakin University, Victoria, Australia
Tor Eriksson
Department of Economics, Aarhus School of Business, Denmark
Greetje Everaert
LICOS, Centre for Transition Economics, K.U. Leuven, Belgium
Pedro Javier GalileaSalvatierra
Departamento de Gestión de Empresas, Universidad Pública de Navarra, Pamplona, Spain
Antje Hildebrandt
Department of Economics, Humboldt University, Berlin, Germany
Norman J. Ireland
Department of Economics, University of Warwick, Coventry, UK
Patrice Laroche
Institut d’Administration des Entreprises, University of Nancy 2, Nancy, France vii
viii
LIST OF CONTRIBUTORS
Javier Merino-Díaz de Cerio
Departamento de Gestión de Empresas, Universidad Pública de Navarra, Pamplona, Spain
Andrew Pendleton
Manchester Metropolitan University Business School, Manchester, UK
Erik Poutsma
Nijmegen School of Management, University of Nijmegen, Nijmegen, Netherlands
Andrew Robinson
Leeds University Business School, The University of Leeds, UK
Agustin J. Ros
NERA, Cambridge, MA, USA
Jos van Ommeren
Department of Regional Economics, Free University, Amsterdam, Netherlands
Jaroslav Vanek
Department of Economics, Cornell University, Ithaca, NY, USA
Nicholas Wilson
Leeds University Business School, The University of Leeds, UK
Hao Zhang
Leeds University Business School, The University of Leeds, UK
FOREWORD This volume of Advances in the Economic Analysis of Participatory and Labor Managed Firms marks the first volume in this series to be produced by Elsevier. It is also the first volume of Advances for which Jan Svejnar is not a series coeditor. As the remaining series editor, I wish to express my deep debt of gratitude to Jan for the enormously valuable role he has played in helping to launch Advances and in the development of the general field of participation. These changes are being used as an opportunity to re-launch the series. One of the key changes relates to the frequency of publication. Whereas in recent years Advances has appeared erratically, henceforth the aim is to publish Advances on a regular and annual basis. Reflecting a deepening pool of talent as the field of participation has grown during the last twenty years or so, another change, as is evident in the present volume, is to make frequent use of guest editors for issues of Advances. As series editor, I welcome suggestions and proposals from readers for particular issues. Other changes concerning Advances will be modest. Advances will continue to act as a forum for high-quality original theoretical and empirical research in the broad area of participatory and labor managed organizations. The original rationale for the series was the observation that while general and specialized journals publish work in this field, many do so only occasionally. There continues to be a need for an annual periodical that presents some of the best papers in a single volume. While the focus will continue to be on economic issues, analytical studies on closely related areas are also welcome. Advances will also continue to serve as an outlet for high quality pieces that regular journals often consider to be too long. The broad area of participation and labor management has changed much since the inception of the series in 1985. The tragic disintegration of the Former Republic of Yugoslavia also meant the disappearance of the principal systemic example of self management. But the collapse of the former USSR has also triggered widespread experimentation with diverse forms of participation in many transition economies, notably many firms with large degrees of employee ownership. Amongst firms in western economies we also witness the continued growth of diverse institutional arrangements that provide for participation by ix
x
FOREWORD
employees in decision-making as well as in enterprise results. Also several important examples of worker cooperatives continue to thrive, with the Mondragon Cooperative Consortium now representing the seventh largest consortium in Spain. Against this institutional backdrop much new and innovative theoretical and empirical work in the broad field has appeared. The key aim of the Advances series continues to be to make it a broad-based periodical within which is presented both new theoretical results and fresh evidence on the performance of participatory firms and sectors. The intent is to maintain high quality and to place this periodical among other successful Elsevier series. I hope you will be informed and stimulated by this volume and that you will consider contributing to it and conveying information about Advances to other interested colleagues. Derek C. Jones Series Editor
INTRODUCTION This volume of Advances in the Economic Analysis of Participatory and Labor Managed Firms consists of ten original papers. The first five papers address the effects of institutions of governance (at the workplace and corporate levels), including new forms of workplace governance (e.g. self-directed teams), a traditional form (or trade unions) and financial participation schemes. The subsequent three papers turn to the issues of the determinants of the incidence of such institutions, followed by two theoretical contributions. The paper by Tor Eriksson introduces a new survey of participatory employment practices in Danish firms, and connects these practices to productivity gains for the firm and wage gain for workers. Like similar work elsewhere, the author matched the survey data with annual financial data over 1992–1997 for each responding firm. However, unusually, the author was able to further match the resulting data with annual employee data (such as, average age, average wage, average tenure, average schooling, and the proportion female) over 1992–1997 for each firm. As such, this is a unique employeremployee matched panel that provides data on participatory employment practices. The author finds among other things that the inability to control for time-variant worker characteristics tends to result in overestimating the effects of participatory employment practices and that the productivity and wage gains take some time to appear. José Alberto Bayo-Moriones, Pedro Javier Galilea-Salvatierra, and Javier Merino-Díaz de Cerio introduce a new telephone survey of participatory employment practices in 965 manufacturing establishments in Spain, and investigate whether these practices lead to gains for the firm and workers. An attractive and unique feature of their data is that both capitalist firms and cooperatives are included in their sample. One of their most notable findings is that the introduction of HPWPs generates positive outcomes both for capitalist firms and cooperatives, without differences in the magnitude of the impact. While the above two papers focus on new institutions of workplace governance, the next paper studies a traditional institution of workplace governance, i.e. trade unions. Chris Doucouliagos and Patrice Laroche conduct a meta-analysis of the effect of unions on productivity growth. Specifically, the authors use meta-analysis to summarize the estimated effect across 27 studies xi
xii
INTRODUCTION
and examine how this estimated effect varies over time and between the U.S. and other countries. Meta-regression analysis is used to identify factors that account for differences in the findings reported by the different studies. The paper represents yet another fine example of the applications of meta-analysis to economics by the authors. All three papers so far concentrate on non-financial aspects of governance. The next two papers tackle such financial aspects. The contribution by Agustin Ros is an empirical study of the effects of employee ownership on effort/ shirking and horizontal monitoring based on rich survey data collected by the author on an employee owned firm and six comparable private firms. Ros examines the perceptions of employees and the reported results complement studies that use objective measures. The author finds that being a member of an ESOP does not affect effort, shirking, or horizontal monitoring. However, perceptions of employee participation in the decisions of the firm are found to increase horizontal monitoring and perhaps effort. The paper by Everaert and Hildebrandt contributes to the literatures on transition economies and participatory firms by examining the determinants of the incidence of soft budget constraints (SBCs), in particular enterprise ownership structure (including different forms of private ownership). Also the paper is unusual insofar as the author is able to use impressive panel data sets for two countries, namely Bulgaria and Romania. The paper reaches some provocative findings, notably the conclusion that ownership does not matter in determining SBCs. The next three papers turn to the issues of the determinants of participation. Andrew Pendleton, Erik Poutsma, Jos Van Ommeren and Chris Brewster use a unique cross-national survey of financial participation schemes in 2,506 establishments in 14 EU countries, and try to study the determinants of the adoption of such schemes. They find among other things that country effects are important predictors of both profit sharing and share ownership schemes. Considering the relative scarcity of cross-national evidence on participatory employment practices, we hope that their study will inspire more researchers to pursue similar cross-national surveys elsewhere (e.g. Asia). Christopher Adams uses rich data on 1,153 product line workers in 162 British private sector manufacturing establishments to examine the use of group incentives (profit sharing or employee share ownership) and worker participation in decision making (specifically over the range of tasks performed). He develops principal-agent model based on workers having private information about the production process, which yields testable hypotheses. As the accuracy of a worker’s information increases, the model implies that the worker is given more decision making responsibility. The
Introduction
xiii
model also implies that the firm is more likely to use either profit sharing or employee share ownership to motivate a worker with decision making responsibilities when it is difficult to monitor why the worker chooses a particular action. The empirical results provide mixed support for the theory. The paper by Nicholas Wilson, Hao Zhang, and Andrew Robinson is an empirical study that examines hypotheses arising from a transaction cost economics (TCE) framework to explain employee share ownership. The TCE framework suggests that employee share ownership would be more common when workers have firm-specific human capital, it is difficult to monitor the use of human capital and worker effort, the firm’s external environment is uncertain, and transactions are more frequent. To test these hypotheses, the authors use unique data that they collected from 638 British firms to construct proxies for the use of firm-specific human capital, uncertainty, and frequency of transactions, which are included in logit models of employee share ownership. The authors find evidence in support for the hypotheses. The last two papers are theoretical contributions. Jan Erik Askildsen and Norman Ireland carefully develop a model of bargaining by a union and a firm over future benefits (e.g. a defined benefit pension) when workers may not receive these benefits either because the firm goes out of business before the benefit is to be paid or the worker leaves the firm before the benefit is vested. The authors investigate the characteristics of the Pareto efficient outcome as well as the bargaining outcome under different bargaining regimes and when the firm’s workforce consists of heterogeneous workers. One of Askildsen and Ireland’s notable findings is that the relative bargaining power of the firm and the union does not affect the level of benefits provided by the firm; bargaining power only affects current period wages. In his paper “Comparative Systems, Destructive Trade and World Distributive Justice,” one of the pioneers on the broad field of participatory and labor managed firms, Jaroslav Vanek, extends earlier work by presenting an analysis of the impact of international trade in today’s globalized economy. In a provocative analysis Vanek comes clearly down on the side of those who argue that the impact of free trade is primarily destructive, especially for the bulk of the world’s less privileged. Furthermore, he argues that policies that promote democratic and participatory arrangements, have a vital role to play in stalling and perhaps reversing existing tendencies. Takao Kato and Jeffrey Pliskin Guest Editors
THE EFFECTS OF NEW WORK PRACTICES: EVIDENCE FROM EMPLOYER-EMPLOYEE DATA Tor Eriksson 1. INTRODUCTION The purpose of this paper is to examine some of the effects of the much discussed new work practices, sometimes also labelled high performance work organizations, common for which are less hierarchical organizations, decentralized decision rights and jobs designed to include broader sets of tasks. Much of the discussion about these practices has been based on case studies that are typically concerned with successful experiments, whilst relatively little causal evidence exists. Outside the United States, evidence is particularly scant even as to how widely spread the new work organizations are. In fact, only recently has there been an attempt to compare internationally the prevalence of the adoption of new workplace practices; see OECD (1999). Likewise is there considerably less research evidence regarding the effects of these practices for the European countries. As labour markets and labour market institutions in most European countries differ quite a lot from those in Northern America, the consequences may be different in Europe, too. In this paper I provide some additional evidence from Denmark using a data set which is constructed by merging information from a survey investigation into firms’ pay and work practices with an employer-employee linked panel. The two main questions addressed are: Do firms that use new work practices outperform other firms in productivity? Are the productivity gains shared with The Determinants of the Incidence and the Effects of Participatory Organizations, Volume 7, pages 3–30. Copyright © 2003 by Elsevier Science Ltd. All rights of reproduction in any form reserved. ISBN: 0-7623-1000-6
3
4
TOR ERIKSSON
the firms’ employees? Of course, in order for there to be something to share, the productivity gains should be sufficiently large. Why would new work practices lead to improved firm performance? One reason is that they typically involve some worker empowerment, and in particular extended decision rights regarding which tasks to perform and how. This implies that costs of information transfer and communication are lowered as compared with the hierarchical, Taylorist work organizations. Also the costs of monitoring are reduced as a consequence of decentralisation of authority. However, as discussed by Thesmar and Thoenig (2000) and Caroli and van Reenen (2000), firms may also take on new costs when introducing the new work practices.1 These may arise because of duplication of information, higher occurrence of mistakes, and less scale economies due to multi-tasking. Hence, higher performance does not necessarily follow in all, or even in many, firms. Whether it does, and to which extent, are largely empirical questions. Why would firms be sharing the gains with their employees? One reason is that workers have the power to extract some of benefits from the higher productivity. As demonstrated by Freeman and Lazear (1995), profits are maximized by giving workers less power than which would maximize productivity. Another reason is that the new practices are likely to function best if firms commit to reward their workers for the extra effort required of them in adjusting to a new system, the future and performance of which is a associated with some uncertainty. The remainder of the paper is organized into four sections. The next section briefly summarizes previous research and Section 3 contains a data description. In the fourth section the estimation results are presented and discussed. Section 5 concludes.
2. PREVIOUS RESEARCH Applied economists have examined the new work organizations from two perspectives. The first attempts to identify factors influencing the adoption of the new, allegedly superior, work practices; see e.g. Osterman (1994) and Gittleman et al. (1998). The other has looks at whether the alternative work organizations deliver the improvements promised by their proponents. This has generated a small but growing literature that examines the effects of new work practices on corporate performance. Most of this research has been concerned with the effects of firm productivity levels; for good recent summaries of the U.S. studies, see Black and Lynch (2001) and Neumark and Cappelli (2001). The management literature has also considered some other outcome variables like financial performance (Huselid, 1995), innovation (Michie & Sheehan,
The Effects of New Work Practices
5
1999), and organizational commitment (Eaton, 2001). Another strand of the literature has looked at their impact on the well-being of the firm’s employees, and on their wages in particular; see in particular Caroli and van Reenen (2000) and Bauer and Bender (2000).2 The key issue in these studies is whether the benefits from the new work organizations accrue predominantly to the firms or if these share them with the employees. A related question is whether benefits are distributed equally among employees, or if, as argued by Snower (1999), gains are in the main going to skilled labour, which is complementary to the new work designs. If so, this may contribute to a widening of wage differentials both within and between firms; cf. Aghion et al. (1999). Most of the earlier literature was based on case studies.3 More recent studies have employed large data sets, but are in many cases based on cross-sectional data and/or on rather unrepresentative samples. A major problem with the cross-section analyses is that there is potential reversed causality. Firms that are performing well (and therefore pay higher wages than other firms) can afford to experiment with alternative workplace organizations and job designs. This problem can be circumvented by using lagged work practice variables, but does not solve the firm fixed effects problem. For the latter, panel data is needed. But even with longitudinal data, the endogeneity is not solved if changes in unobserved features of firms give rise to changes in new work practices; see Athey and Stern (1998) for a comprehensive discussion. Furthermore, a potentially serious problem with previous studies is that they only to a limited extent have been able to control for firm differences in the skill structure and changes therein. This is in particular important, if as has been suggested in the discussion, the new organizations tend to be skill-biased. Four recent investigations, which both utilize a longitudinal design and include workforce characteristics as control variables are Bauer and Bender (2000), Black and Lynch (2000), Caroli and van Reenen (2000), and Neumark and Cappelli (2001). The evidence produced so far appears to be rather mixed. Thus, for example Osterman (2000, p. 193) finds “very little evidence that . . . have delivered on the promise of ‘mutual gains’ ”, whereas Black and Lynch (2000) and Neumark and Cappelli (2001) obtain results which differ considerably.4 Black and Lynch find that firms in the manufacturing sector which have made use of NWP’s both had a better productivity record and were paying higher wages to their employees than other firms. The conclusion of Neumark and Cappelli’s analysis, which make use of the same data source as Black and Lynch, is that NWP’s “may raise productivity, although the statistical case is weak”. To the best of my knowledge, there are rather few studies of the effects of alternative work practices from outside the United States. Some notable
6
TOR ERIKSSON
exceptions are Bauer and Bender (2000) for Germany, Nabb and Whitfield (2000) for U.K., Caroli and van Reenen (2000) for the U.K. and France, BayoMoriones et al. (2001) for Spain, Leoni et al. (2001) for Italy, and Kato and Morishima (2001) for Japan. The institutional setting – influence of unions, legal or other restrictions on hiring and firing, worker co-determination, wage bargaining, etc. – may very well make a difference for the consequences of introducing new work organizations, however. Many of the new work practices imply not only jobs with a wider variety of tasks, but also that employees are allowed to a higher extent than before to take part in the decision making, that is, a decentralization of authority. Worker empowerment is, however, also associated with some increase in costs, in particular agency and co-ordination costs. Moreover, and more importantly for firms, worker empowerment also gives workers the ability to extract a larger share of value added; see Freeman and Lazear (1995). These costs can to some extent be reduced by an appropriate design of compensation.
3. THE DATA The data set used in this paper has been constructed by merging two separate data files: a survey directed at firms and containing information about the work and compensation practices, and a longitudinal employer-employee data set5 which provides the information about firm characteristics and performance as well as about the firms’ employees. The survey was administered by Statistics Denmark as a mail questionnaire survey in May and June 1999, which was sent out to 3,200 private sector firms with more than 20 employees. The firms were chosen from a random sample, stratified according to size (as measured by the number of full time employees) and industry. The survey over-sampled large and medium-sized firms: all firms with 50 employees or more were included, and 35% of firms in the 20–49 employees range. The response rate was 51%, which is relatively high for a rather long and detailed questionnaire of the type that was used. Moreover, the response rates for the size and one-digit industry cells vary only little: between 47 and 53%. Thus, the representativeness of the sample is of no major concern. The survey represents a unique source of information on Danish firms’ internal labour markets and changes therein. Besides some background information about the firm, the firm was asked about its work organisation, compensation systems, recruitment, internal training practices and how it evaluates its employees. For a brief description of the questionnaire and the main results (in Danish), see Eriksson et al. (2000). In a companion paper,
The Effects of New Work Practices
7
Eriksson (2001), I have examined the determinants of firms’ adoption of pay and work practices. Owing to the high response rate, the survey provided me with 1,605 useful observations. For the questions concerning compensation practices in the firms, the respondents were asked to distinguish between four different categories of employees: (1) top managers, (2) middle management, (3) other white collar workers, and (4) blue collar and other hourly paid workers. In connection with the questions concerning work design and practices, they were asked to differentiate between hourly paid and salaried employees. Using unique firm identification numbers within Statistics Denmark, the survey data were next supplemented with information about the firms as well as about their workforces. This information is taken from a large employeremployee linked database, which covers all private sector firms and all the employees who worked in them (in Denmark) in any year during the period 1980 to 1997.6 The panel contains detailed information about employee characteristics7 (and hence, firms’ workforces in any year) and about their labour earnings and other income. In addition, the panel has economic information about the firms, with 20 or more full-time equivalent employees, for the years 1992 to 1997. Firms’ use of work (and pay) practices can be measured along several dimensions. The measure adopted in the survey questionnaire is whether a firm has implemented one of six work designs: (i) Self-managed teams. Self-managed teams is a work organization which gives its members authority over decisions regarding how to perform tasks or even, which tasks to perform. Important aspects of team working is pooling of skills and skills development of individual workers. (ii) Job rotation. Job rotation is a system where the workers are explicitly required to rotate between different jobs. This increases the variety of tasks to be performed by the employee and is also likely to enhance the employee’s understanding of the operation. (iii) Quality circles. Groups of workers that meet regularly to solve problems concerning productivity and people and to discuss aspects of performance and quality. (iv) Total quality management (TQM); an important element is that TQM programmes, of which ISO9000 probably is the best known, include employee involvement. (v) Benchmarking. Benchmarking is a formal system of learning about practices in other firms and organisations.
8
TOR ERIKSSON
(vi) Project organisation. Groups of workers are organised in projects with defined targets, timetables, budgets and frequently considerable authority with respect to how to perform tasks. The respondents were, moreover, asked to distinguish between hourly paid workers on the one hand, and salaried employees (including managerial staff) on the other. One important piece of information I do not have regarding the implementation of the practices is the proportion of employees affected by the particular work designs.8 Nor were the respondents asked to rank the practices according to some notion of their importance. The firms were also asked a corresponding question regarding the implementation of performance related pay (PRP) practices. More precisely the firms were asked whether they had adopted one of four PRP methods – team bonus, individual bonus, stock and stock options and profit sharing – for four different categories of employees: top managers, middle-management, other white collar workers and blue collar workers; see Eriksson (2001), for details. The questionnaire survey was carried out in May–June 1999. However, the data set currently at my disposal only contains information regarding the performance of the firms during years 1992 to 1997. Thus, the analysis has to be restricted to concern what has transpired by 1997, implying that the observations for firms, which have adopted the new practices in recent years have to be discarded. In order to solve some of the causality problems, I will use observations on firm performance for years 1995–1997 and correlate them with adoption of new work practices in the years up to 1995. The data: the primary advantages of the data set used in this paper are that it contains relatively rich information about both firms and their employees, and the high response rate in the survey part. The major weakness is the lack of information about the proportion of workers affected by the new practices. Table 1 gives some information about the incidence of new work practices in Danish companies. The first two columns show the proportion of firms that use the six different work designs asked about in the questionnaire for salaried and hourly paid employees, separately. The three remaining columns give information as to when the firms were reorganizing work: in recent years, in the first half of the nineties, or earlier. Although the level of the proportions of firms which have implemented the different new work practices is relatively low compared to some other countries (see OECD (1999) for international comparisons of the use of newer workplace practices), accounting for when the practices were adopted, clearly shows that new forms of organizing work have been taking root in Danish firms in the nineties.10
The Effects of New Work Practices
9
We may notice that the most widely used of the new work organizations are self-managing teams – for both salaried and hourly paid workers – and project organization for salaried employees. These have been adopted by about a forth of the firms. Another relatively frequently implemented design, especially for the hourly paid, is job rotation. As can be seen from the table, 30–40% of the firms that have adopted the new work practices have done so during the previous three years, and a third earlier in the nineties. Two practices have a slightly different pattern of adoption: benchmarking, which very few firms have implemented (and only in Table 1.
Incidence of New Work Practices and the Timing of Their Introduction in the Firm (%).
Work practice: Teams All Foreign Domestic Job rotation All Foreign Domestic Quality circles All Foreign Domestic TQM All Foreign Domestic Bench-marking All Foreign Domestic Project organisation All Foreign Domestic
Salaried employees
Hourly paid workers
1996–1999
1990–1995
Before 1990
26.5
21.8
41.6 14.1 31.0
31.3 31.8 31.0
27.1 54.1 38.0
6.2
17.4
37.9 19.3 23.5
39.9 42.1 39.3
22.2 38.6 37.2
3.7
3.4
37.0 0 30.6
39.1 60.0 33.3
23.9 40.0 36.1
8.3
4.1
38.9 9.4 10.5
51.1 59.4 45.6
10.0 31.2 43.9
7.9
1.6
57.6 5.3 10.7
34.8 34.2 35.7
7.6 60.5 53.6
24.5
5.7
35.8 21.6 29.0
37.7 40.6 36.5
26.5 37.8 34.5
Source: Firm questionnaire; 1,605 firms.
10
TOR ERIKSSON
recent years) and total quality management, which was introduced in the early nineties and the incidence of which has remained relatively low. Foreign owned firms have a clearly higher NWP incidence than domestic owned firms.11 One possibility is that foreign firms adopted the practices earlier and the Danish firms imitated them. However, as can be seen from the table, there are no clear traces of such a diffusion process; rather it seems as the new practices have been introduced at roughly the same pace in both domestic and foreign owned firms. Approximately equal proportions of foreign and Danish owned firms that have introduced a practice did it in the early nineties or before. Table 2 gives some information about the clustering of work practices. Part A shows the proportion of firms with different numbers of work practices, for hourly paid and salaried workers, separately. The first thing to note here is of course that a small proportion of companies have implemented these practices,
Table 2.
Prevalence and Clustering of New Work Practices.
A. Prevalence Number of new work practices
Hourly paid workers
Managerial and salaried employees
70.8 6.2 17.8 4.2 0.6 0.3
53.0 27.8 12.0 4.5 1.7 1.0
None 1 2 3 4 5 or 6 B. Clustering (per cent of firms) 1. Teams
1 2 3 4 5 6 7
29.1 7.2 12.7 8.9 33.2 48.4
2. Job rotation
3. Quality circles
4. Total quality
5. Benchmarking
6. Projects
51.7
63.4 53.7
51.1 40.0 18.9
48.5 48.5 16.7 31.8
53.8 34.0 7.6 18.4 16.1
10.8 17.7 15.8 37.5 54.2
37.0 23.9 37.0 49.4
23.3 45.6 33.1
54.5 57.6
54.5
7. Performance pay (min. 2) 23.1 15.9 5.1 9.2 4.0 12.1
The Effects of New Work Practices
11
and the second is that those firms which have, have as a rule implemented more than just one practice. There is a difference between managerial and salaried employees and hourly paid workers. When a NWP is adopted for the latter, it is more much likely than for salaried employees that another new work practice is implemented as well. This may reflect stronger complementarities between work practices typical for blue-collar workers, such as TQM, quality circles and job rotation. Part B looks at patterns in adoption of practices in more detail. From the first column we may see that about a third of firms with teams working also have implemented job rotation schemes. From the second column we can also see that every second firm with job rotation programs have teams, too. Moreover, also the other work practices seem to combine quite frequently with team and job rotation schemes. Another practice that also occurs frequently together with some the others is project organisation. The last column and row in the table refer to firms that have adopted at least two performance related pay schemes (out of: team bonus, individual bonus, stock or stock options, and profit sharing). Worth noting here is that firms that adopted teams, job rotation, quality circles or benchmarking are more likely to have introduced performance pay schemes than those with TQM or projects. Note, furthermore, that of the firms that have not adopted any of the new work practices – about 45% of firms – only 6.4% report that the have adopted two performance related pay schemes.
4. ECONOMETRIC ANALYSIS In this section I will report some results of efforts to estimate the effects of new work practices on firm productivity and mean firm wages. More specifically, I estimate equations of the following form: Yit = Xit + NWPit + i + it
(1)
where Y is the log of labour productivity in firm i at time t, X is a vector of control variables, NWP is a vector of indicators for the use of work practices in the firm, and i is a firm fixed effect. I have observations on productivity and the X variables for the years 1992 to 1997 and have information from the survey about when each of the new practices were implemented in each firm, and so, this allows me to estimate (1) on a first difference form and thereby removing the heterogeneity: Yi = Xi + NWPi + i
(2)
where the difference is taken between years 1997 and 1992. Labour productivity is computed as the ratio between annual sales and the number of
12
TOR ERIKSSON
full-time equivalent employees. The vector Xit includes a set of firm and workforce characteristics. The former are: log of capital stock, age of firm, and one-digit level industry indicators.12 The workforce characteristics, which are entered in order to control for differences in the skill structure, are the annual averages of age, years of schooling and ongoing tenure as well as the share of females.13 The NWP indicators consist of a set of dummies. The first group of dummies indicates whether the firm has implemented at least N of the six practices in (1) 1995–1997, (2) 1992–1995, and (3) before 1992, where N is either 1, 2 or 3. Of course, the requirement that the firm have implemented at least one, two or three practices is inevitably arbitrary. The second group of dummies refer to the individual practices using the same division into time periods. Because they are likely to correlated with each other, these dummies are entered one by one. Thus, each coefficient represents the effect of a single practice controlling for X but not for the influence of the other practices. The fourth group is constructed by combining some of the individual practices into bundles. The idea is here to test the complementarity hypothesis, that is, that it is clusters of practices rather than piecemeal, which generate high performance. Complementarity may not, however, only be limited to work practices but may also be present with regard to compensation practices. The fifth group of dummies interacts work and compensation practice indicators with the purpose of picking up these types of complementarities. Finally, instead of dummy variables, I have also included the number of practices used by each firm in the three, above-mentioned time-periods, as explanatory variables in (1). In the estimations for the log of the mean firm wage I include the same sets of explanatory variables as for productivity with the only exception that firm size, as measured by the number of employees, is added to the X-vector. I have run both OLS on Eq. (1) and the fixed effects model in Eq. (2). I will, however, only present the latter, and will only briefly mention the main differences in the estimates obtained. Likewise, I have estimated the models both with and without the workforce characteristics, but will mainly discuss the former. The main differences in estimates will be described below. Beginning with the productivity estimates, displayed in Table 3, we may note first that there are many positive coefficients about half of which differ significantly from zero. Thus, there is some evidence of firms benefiting in terms of higher productivity from introduction of at least some of the new work practices. Although sales per worker is only a proxy for productivity, the magnitude of the effects appear non-negligible, given that the productivity growth rate in Denmark has varied between 1.5% and 2%. The results from
1996–1997
1992–1995
At least three NWPs: Before 1992
1996–1997
1992–1995
At least two NWPs: Before 1992
1996-1997
1992-1995
At least one NWP: Before 1992
Work practice variables
0.017 (0.014) 0.028 (0.011) 0.014 (0.010)
0.027 (0.013) 0.026 (0.012) 0.012 (0.011)
0.022 (0.011) 0.024 (0.014) 0.013 (0.015)
With work-force characteristics
0.021 (0.013) 0.032 (0.012) 0.015 (0.010)
0.029 (0.014) 0.025 (0.012) 0.012 (0.010)
0.026 (0.012) 0.028 (0.014) 0.014 (0.013)
Without work-force traits
0.013 (0.014) 0.020 (0.010) 0.008 (0.032)
0.023 (0.012) 0.024 (0.102) 0.011 (0.024)
0.012 (0.017) 0.022 (0.012) 0.009 (0.021)
With work-force characteristics
0.017 (0.015) 0.031 (0.015) 0.010 (0.030)
0.028 (0.016) 0.039 (0.012) 0.012 (0.024)
0.014 (0.018) 0.027 (0.013) 0.010 (0.009)
Without work-force traits
Salaried employees
Labour Productivity Equation Estimates.*
All employees
Table 3.
0.018 (0.012) 0.028 (0.013) 0.015 (0.018)
0.023 (0.012) 0.030 (0.015) 0.010 (0.006)
0.021 (0.012) 0.026 (0.013) 0.010 (0.009)
With work-force characteristics
0.022 (0.012) 0.040 (0.012) 0.018 (0.023)
0.025 (0.012) 0.036 (0.014) 0.016 (0.009)
0.025 (0.013) 0.036 (0.014) 0.015 (0.010)
Without work-force traits
Hourly paid employees
The Effects of New Work Practices 13
1996–1997
1992–1995
Quality circles Before 1992
1996–1997
1992–1995
Job rotation Before 1992
1996-1997
1992–1995
Teams Before 1992
Work practice variables
0.013 (0.195) 0.063 (0.030) 0.041 (0.030)
0.015 (0.018) 0.054 (0.023) 0.073 (0.028)
0.037 (0.016) 0.030 (0.014) –0.029 (0.032)
With work-force characteristics
0.008 (0.253) 0.070 (0.030) 0.041 (0.020)
0.012 (0.017) 0.061 (0.020) 0.072 (0.027)
0.042 (0.015) 0.041 (0.014) 0.036 (0.054)
Without work-force traits
All employees
–0.017 (0.055) 0.029 (0.028) 0.050 (0.070)
0.009 (0.020) 0.014 (0.037) –0.017 (0.025)
0.029 (0.014) 0.027 (0.013) 0.005 (0.100)
0.009 (0.051) 0.035 (0.020) 0.065 (0.041)
0.006 (0.021) 0.018 (0.033) –0.017 (0.025)
0.034 (0.017) 0.028 (0.013) 0.010 (0.008)
Without work-force traits
Salaried employees
Continued.
With work-force characteristics
Table 3.
0.019 (0.056) 0.072 (0.029) 0.029 (0.032)
0.011 (0.012) 0.057 (0.024) 0.080 (0.036)
0.041 (0.013) 0.032 (0.018) –0.030 (0.025)
With work-force characteristics
0.036 (0.038) 0.072 (0.034) 0.044 (0.021)
0.018 (0.010) 0.074 (0.026) 0.078 (0.037)
0.044 (0.013) 0.034 (0.016) –0.000 (0.025)
Without work-force traits
Hourly paid employees
14 TOR ERIKSSON
1996–1997
1992–1995
Project org. Before 1992
1996–1997
1992–1995
Bench-marking Before 1992
1996–1997
1992–1995
TQM Before 1992
Work practice variables
0.001 (0.020) –0.097 (0.124) 0.033 (0.077)
0.101 (0.190) 0.055 (0.111) 0.056 (0.025)
0.015 (0.014) 0.009 (0.012) –0.005 (0.020)
With work-force characteristics
0.023 (0.034) –0.047 (0.073) 0.035 (0.066)
0.100 (0.183) 0.060 (0.072) 0.069 (0.027)
0.022 (0.011) 0.032 (0.018) 0.012 (0.018)
Without work-force traits
All employees
–0.012 (0.030) –0.105 (0.099) 0.024 (0.044)
–0.003 (0.070) 0.012 (0.090) 0.040 (0.023)
0.017 (0.014) –0.017 (0.031) –0.014 (0.060)
0.005 (0.046) –0.050 (0.084) 0.028 (0.027)
–0.001 (0.098) 0.020 (0.124) 0.042 (0.016)
0.020 (0.010) –0.016 (0.030) –0.020 (0.032)
Without work-force traits
Salaried employees
Continued.
With work-force characteristics
Table 3.
0.002 (0.044) –0.014 (0.024) –0.012 (0.020)
0.141 (0.256) 0.060 (0.075) 0.064 (0.023)
0.026 (0.012) 0.014 (0.023) –0.003 (0.044)
With work-force characteristics
–0.005 (0.040) 0.009 (0.018) –0.036 (0.025)
0.098 (0.242) 0.066 (0.030) 0.078 (0.024)
0.036 (0.013) 0.054 (0.022) 0.004 (0.056)
Without work-force traits
Hourly paid employees
The Effects of New Work Practices 15
1996–1997
1992–1995
Number of practices: Before 1992 0.022 (0.010) 0.015 (0.007) 0.015 (0.009)
0.034 (0.011) 0.022 (0.006) 0.015 (0.009)
0.068 (0.022) 0.074 (0.030) 0.036 (0.021)
1996–1997
1992–1995
Job rotation, TQM and quality circles Before 1992 0.063 (0.021) 1992–1995 0.067 (0.030) 1996–1997 0.044 (0.022)
Without work-force traits 0.034 (0.012) 0.024 (0.015) 0.011 (0.023)
With work-force characteristics
All employees
0.029 (0.011) 0.014 (0.012) 0.009 (0.030)
Bundles: Teams + proj. org. Before 1992
Work practice variables Without work-force traits
Salaried employees
Continued.
With work-force characteristics
Table 3.
With work-force characteristics
Without work-force traits
Hourly paid employees
16 TOR ERIKSSON
Indiv. Incentives
*Indiv. Incentives
NWP 96/97
*Indiv. Incentives
NWP 92/95
NWP92* Indiv. Inc.
NWP before 92
PRP
NWP 96/97*PRP
NWP 96–97
NWP 92/95*PRP
NWP 92–95
NWP 92*PRP
Interactions: NWP before 92
Work practice variables
0.024 (0.009) 0.011 (0.030) 0.020 (0.010) 0.012 (0.004) 0.010 (0.011) 0.011 (0.005) 0.009 (0.018)
0.027 (0.011) 0.017 (0.011) 0.044 (0.022) 0.020 (0.009) 0.010 (0.030) 0.014 (0.008) 0.033 (0.020)
With work-force characteristics 0.024 (0.011) 0.024 (0.014) 0.063 (0.021) 0.018 (0.009) 0.027 (0.032) 0.025 (0.015) 0.041 (0.024)
Without work-force traits
All employees Without work-force traits
Salaried employees
Continued.
With work-force characteristics
Table 3.
With work-force characteristics
Without work-force traits
Hourly paid employees
The Effects of New Work Practices 17
0.017 (0.008) –0.006 (0.024) 0.012 (0.047) –0.010 (0.023) 0.011 (0.015) –0.002 (0.023)
With work-force characteristics
Without work-force traits
All employees With work-force characteristics
Without work-force traits
Salaried employees
Continued.
With work-force characteristics
Without work-force traits
Hourly paid employees
* Standard errors in parenthesis. Coefficients which differ from zero at 5% or 1% levels are presented with bold numbers. Control variables in the equations, see text.
*Group incentives
NWP 96/97
*Group incentives
NWP 92/95
NWP92* Group inc
NWP before 92
Work practice variables
Table 3.
18 TOR ERIKSSON
The Effects of New Work Practices
19
examining different numbers of practices adopted differ somewhat. In short what they show is that the effects on firm productivity and firms’ average wages of having adopted at least one practice are positive, significant except for salaried employees and the period 1996–1997, and in general slightly smaller in magnitude than in the at-least-two-practices case. On the other hand, the effects of the adoption of three or more practices turn out to have a similar pattern as for two practices: the estimates are positive and significant for all employees and for hourly paid workers in the 1992–1995 period. Probably due to rather few firms having adopted several practices before 1992, I found no effects for the pre-1992 period. One noteworthy feature of the results is that firms which have implemented the new practices in the two last years of the observation period (1996–1997) have gained considerably less than those which introduced them earlier. There are two ways of interpreting this. The first is that it takes time for new practices to deliver the higher performance. On the other hand, according to estimates, once the gains are there, they are not lost quickly but remain for several years. The second explanation is that late adopters differ from early adopters in that they have smaller marginal benefits from practice adoption. The early adopters and the late-comers differ in respects not captured by the control variables. Obviously, which of these interpretations is correct, is of some importance. But, with the data at hand, I am not able to discriminate between them. Another striking feature is that changing work organizations for salaried workers appear to have a clearly smaller impact on productivity than changing the hourly paid workers’ work practices. This picture is somewhat changed if the workforce skill structure variables are excluded from the controls. Then some additional, albeit not many, coefficients for salaried workers grow in magnitude and become statistically significant. Several of the coefficients for the hourly paid increase both in magnitude and significance. Consequently, controlling for skills is important. Not recognizing differences in skill levels between firms and that the new work practices are likely to be complementary with employee skills, may lead to an upward bias in the estimated magnitudes of the performance benefits. Indeed, only a few earlier studies have had information about workforce skills in their data sets. Turning next to the results concerning single practices, we may note that TQM and project organizations have little or no effect on firm productivity levels. Self-managed teams seem to have beneficial effects both when adopted for salaried and hourly paid employees, whereas job rotation and quality circles pay off in higher productivity if introduced for hourly paid workers. Benchmarking, which is a relatively rare and new practice, generates according to the estimates fairly large performance gains. It should be remembered,
20
TOR ERIKSSON
however, that this might be due to early adopters reaping higher benefits than later adopters. As for bundles of practices, I have only tried out two: one, consisting of teams and project organizations, both of which are practices more likely to be found in use among salaried employees and associated with free rider problems, another including job rotation, TQM and quality circles, a combination which is more frequent among the hourly paid. The latter attaches large and significant coefficients, whereas the first bundle carry only small coefficients. Instead of examining more bundles for which theory offers very little guidance as to which ones to focus on, I have looked at the number of practices (counting a practice used for both groups of employees as two practices). According to Table 3, the number of practices is quite important, implying that using them in tandem may generate positive productivity effects. As we saw earlier, however, only a few firms have introduced three or more practices. Finally, I interacted the dummy for adopting at least two new work practices with a dummy for whether a firm has adopted at least two of the following four new pay practices: team bonus, individual bonus, stock or sock options, and profit sharing. It should be noted that the latter dummy only describes the firm’s compensation systems at the time of the survey, and so, the dummies only pick up changes in new work practices, not in payment schemes.14 At any rate, the estimates suggest that new work organizations and new pay practices may be complementary; if a firm that has adopted new work practices introduces performance related pay schemes, this increases productivity further. Although all the interaction terms are positive, only one of them (NWP92/95*PRP) is statistically significant. Distinguishing between individual- and group-based pay schemes reveals that there is more complementarity between new work practices and compensation schemes building on individual incentives. The firm wage equations are reported in the same way as those for productivity; see Table 4. There is one crucial difference, though. In the estimations where I consider the adoption of work practices for salaried and hourly paid workers separately, I also use separate average wage measures for the two groups. Consequently, the results reported in the third and columns refer to the average wage for salaried employees and workers on hourly pay, respectively. Similar to productivity, there are a number of positive coefficients, some of which differ from zero. (To save space I have not included the estimations with at least one practice and with at least three practices, as they differ from the two-practices case in the same way as they did in the firm productivity estimations). A closer inspection of Tables 3 and 4 reveals a pattern: positive
1996–97
1992–1995
Job rotation Before 1992
1996–97
1992–1995
Teams Before 1992
1996–97
1992–1995
At least two NWPs: Before 1992
Work practice variables
–0.015 (0.030) 0.024 (0.011) 0.028 (0.013)
0.009 (0.024) –0.005 (0.075) –0.010 (0.056)
0.030 (0.014) 0.031 (0.013) 0.020 (0.025)
With individual traits
0.009 (0.010) 0.038 (0.011) 0.040 (0.015)
0.008 (0.025) –0.002 (0.061) –0.015 (0.060)
0.034 (0.013) 0.037 (0.013) 0.025 (0.017)
–0.010 (0.020) –0.015 (0.044) –0.009 (0.014)
0.010 (0.036) –0.002 (0.019) 0.090 (0.067)
0.013 (0.012) 0.017 (0.013) 0.018 (0.011)
With individual traits
0.009 (0.034) –0.011 (0.034) –0.002 (0.015)
0.009 (0.045) –0.010 (0.025) 0.080 (0.054)
0.014 (0.011) 0.016 (0.015) 0.020 (0.010)
Without individual traits
Salaried employees
Firm Mean Wage Equation Estimates*.
Without individual traits
All employees
Table 4.
0.008 (0.014) 0.036 (0.018) 0.044 (0.020)
0.003 (0.010) –0.007 (0.018) –0.030 (0.027)
0.029 (0.013) 0.033 (0.014) 0.015 (0.038)
With individual traits
0.011 (0.015) 0.038 (0.016) 0.056 (0.020)
–0.007 (0.012) 0.008 (0.021) 0.011 (0.024)
0.035 (0.012) 0.040 (0.013) 0.020 (0.012)
Without individual raits
Hourly paid employees
The Effects of New Work Practices 21
1996–1997
1992–1995
Bench-marking Before 1992
1996–1997
1992–1995
TQM Before 1992
1996–97
1992–1995
Quality circles Before 1992
Work practice variables
0.001 (0.027) 0.004 (0.043) 0.017 (0.008)
0.011 (0.005) 0.009 (0.014) –0.010 (0.017)
0.010 (0.009) 0.043 (0.020) 0.024 (0.016)
With individual traits
–0.003 (0.030) 0.005 (0.027) 0.024 (0.010)
0.015 (0.006) 0.012 (0.012) 0.005 (0.015)
0.009 (0.009) 0.042 (0.021) 0.030 (0.015)
Without individual traits
All employees
Table 4.
–0.003 (0.029) –0.002 (0.045) 0.010 (0.034)
0.010 (0.006) 0.003 (0.040) –0.008 (0.016)
–0.013 (0.029) 0.010 (0.018) 0.020 (0.048)
With individual traits
–0.001 (0.047) 0.010 (0.080) 0.011 (0.047)
0.010 (0.008) –0.007 (0.034) –0.009 (0.025)
–0.012 (0.030) –0.005 (0.021) 0.012 (0.056)
Without individual traits
Salaried employees
Continued.
–0.036 (0.072) 0.006 (0.022) 0.024 (0.012)
0.012 (0.006) 0.012 (0.007) 0.004 (0.012)
0.021 (0.011) 0.053 (0.020) 0.030 (0.013)
With individual traits
–0.027 (0.036) –0.008 (0.024) 0.025 (0.011)
0.020 (0.009) 0.018 (0.011) 0.012 (0.010)
0.025 (0.011) 0.060 (0.021) 0.046 (0.021)
Without individual raits
Hourly paid employees
22 TOR ERIKSSON
0.035 (0.015) 0.052 (0.018) 0.053 (0.014)
1996–1997
1992–1995
Job rotation, TQM and quality circles Before 1992 0.040 (0.019) 1992–1995 0.038 (0.019) 1996–97 0.036 (0.018)
–0.033 (0.047) 0.020 (0.028) –0.001 (0.042)
Without individual traits
0.030 (0.014) 0.036 (0.011) 0.021 (0.012)
–0.034 (0.056) 0.010 (0.026) 0.003 (0.019)
With individual traits
All employees
0.024 (0.013) 0.024 (0.012) 0.018 (0.013)
Bundles: Teams + proj. org. Before 1992
1996–1997
1992–1995
Project org. Before 1992
Work practice variables
Table 4.
–0.019 (0.030) 0.010 (0.007) 0.005 (0.020)
With individual traits –0.011 (0.067) –0.012 (0.010) 0.015 (0.034)
Without individual traits
Salaried employees
Continued.
0.010 (0.079) 0.012 (0.011) 0.007 (0.036)
With individual traits
–0.011 (0.066) 0.014 (0.011) 0.013 (0.026)
Without individual raits
Hourly paid employees
The Effects of New Work Practices 23
PRP
NWP96/97*PRP
NWP 96–97
NWP92/95*PRP
NWP 92–95
NWP92*PRP
Interactions: NWP before 92
1996–1997
1992–1995
Number of practices: Before 1992
Work practice variables
0.020 (0.012) 0.012 (0.006) 0.023 (0.011) 0.011 (0.007) 0.020 (0.014) 0.011 (0.005) 0.024 (0.009)
0.013 (0.006) 0.013 (0.006) 0.008 (0.007)
With individual traits
0.020 (0.012) 0.016 (0.007) 0.028 (0.012) 0.009 (0.011) 0.015 (0.014) 0.012 (0.006) 0.032 (0.007)
0.015 (0.006) 0.014 (0.005) 0.003 (0.010)
Without individual traits
All employees
Table 4.
With individual traits
Without individual traits
Salaried employees
Continued.
With individual traits
Without individual raits
Hourly paid employees
24 TOR ERIKSSON
* See note to Table 3.
Group incentives
* group incentives
NWP 96/97
*group incentives
NWP 92/95
NWP92* group inc.
NWP before 92
Individual incentives
*individ. incentives
NWP 96/97
*individ. incentives
NWP 92/95
NWP92* indiv. Inc.
NWP before 92
Work practice variables
0.013 (0.006) 0.009 (0.008) 0.012 (0.005) –0.007 (0.020) 0.008 (0.009) –0.010 (0.018) 0.016 (0.007)
0.012 (0.006) 0.010 (0.007) 0.016 (0.006) 0.008 (0.004) –0.002 (0.018) 0.003 (0.014) 0.045 (0.015)
With individual traits
Without individual traits
All employees
Table 4.
With individual traits
Without individual traits
Salaried employees
Continued.
With individual traits
Without individual raits
Hourly paid employees
The Effects of New Work Practices 25
26
TOR ERIKSSON
wage coefficients are found more often in those cases where there was a positive effect on productivity than when there was not. This indicates that some of the gains of new work practices accrue also to the employees, or that firms need to pay workers higher wages in order to reap productivity gains. The effect on wages may be positive for several reasons. One is gain sharing, another skill upgrading in adopting firms, adopters sorting their workers more rigorously, or skill complementarity. Indeed, in the estimations where the skill controls were omitted, I obtained larger and some additional significant and positive effects on firm wages. OLS estimations for the 1997 cross-section yielded even higher employee returns to new work practices. This is consistent with that unobservables, particularly those related to productivity, are also increasing the likelihood of NWP adoption. Another pattern clearly borne out by the estimations is that introducing new work organizations for salaried workers do not appear to pay off in terms of higher wages. This is not very surprising in view of the findings that changing salaried workers’ work practices did little good for firm productivity. As for the individual practices, firms that have implemented job rotation, quality circles and benchmarking (in recent years) pay higher wages to their employees. These are, however, relatively rarely adopted practices, save job rotation among workers on hourly pay. Firms that have several practices also pay higher wages, indicating that productivity gains from bundling practices are shared with employees. Interacting with performance related pay-practices does not increase the main effects much, rather new pay schemes seem to be equally important for differences in firms’ average wages. The same is true also when I distinguish between individual (individual bonus, stock and stock option schemes) and group incentives (team bonus and profit sharing schemes).
5. CONCLUDING REMARKS The key findings emerging from this exercise can be summarised as follows: There is some, although not very strong, evidence of the new work practices being productivity enhancing. The evidence also indicates that the average wage is higher in firms that have adopted the new work organizations. Thus, some part of the gains accrues also to the employees involved. Furthermore, there are also some signs of complementaries both between different work practices and between new work and pay practices. The analysis also reveals that not controlling for the skill structure of firms’ workforces yields considerably higher estimates of the returns in productivity and wages to innovative work practices. Hence, in addition to controlling for firm heterogeneity, recognising the skill complementarity of new work
The Effects of New Work Practices
27
practices is important in assessing their effects on firm performance. On the other hand, controlling for the quality of the workforce is also important as not doing it tends to overstate the increase in wage costs associated with adopting new work practices. In future work I hope to extend the analysis by using data on wages for years prior to 1992 (that is, 1980–1991) and on wages and firm performance from the years closer to the survey (1998–1999). Other potentially interesting extensions concern other consequences for firms and employees than those focussed on here. One is workforce turnover in firms, another is wage dispersion between and within firms, and a third is worker satisfaction. The first two pieces of information is in the current data set, but the last one would require merging on information from yet another data source.
NOTES 1. For further analyses of the costs of adopting new work organizations, see also Pil and MacDuffee (1996), Ichniowski and Shaw (1995), Jones and Kato (1995) and Kato and Morishima (2001). 2. But also other outcomes have been studied; Osterman (2000) examines the effects on the firms’ lay-off rates, Godard (2001) looks at employee satisfaction, and Askenazy et al. (2001) consider worker absenteeism. 3. Although case studies can be very informative because of the level of detail of the information collected, they are likely to be unrepresentative as success stories are much more likely to be chosen as cases. 4. Another study using U.S. data, which does not find a significant effect on productivity at the level of the firm is Freeman et al. (2000). 5. See www.cls.dk Pay and performance. 6. The important feature of the panel is the link between firms and employees, which is consistent over time. The data originate from two separate registers maintained by Statistics Denmark: the integrated database for labour market research (IDA) and the business statistics database (BSD). 7. Worker characteristics at the person-year level include gender, age, ongoing tenure, and level and years of education. 8. If seems plausible to assume that the higher the number of practices used, the larger is the proportion of workers in the firm involved in some of the new work practices. Thus, the number of practices implemented can serve as a proxy for coverage. 9. The questionnaire also contained an open alternative. As rather few firms used this option and as no additional work practice was implemented by more than one per cent of the sample firms, I restrict the analysis the six practices listed in the table. 10. It should be noted, however, that the information provided by the survey refers to the situation in only point in time (1999) and does not tell us anything about whether and to what extent firms have implemented these schemes but then stopped using them.
28
TOR ERIKSSON
11. In fact, in logit analyses of factors influencing the adoption of new practices, reported in Eriksson (2001) foreign ownership, show that foreign ownership is one of the key variables in explaining differences in firms’ implementation of the new practices. 12. In the OLS estimates of Eq. (1), I also experimented with including indicators for foreign ownership, competition coming mainly from abroad and the demand conditions in the firm’s product market. The two last indicators are constructed from the survey. Both foreign ownership and foreign competition turned out to have a positive significant effect on firm productivity. Omitting them from the equation had only tiny effects on the estimates to the work practice variables. However, the estimates to the workforce characteristics changed more. Thus, the estimated effects of skills are reduced, but remain significant. 13. Note that entering workforce characteristics may also pick up other influences. Freeman and Lazear (1995) argue workers with a more long-term employment relationship are more able to extract value added. Thus, firms with employees with longer tenure may be less willing to empower their workers, but given they have, the may have to pay them more. Another workforce characteristic emphasized by Freeman and Lazear is the unionisation rate that may reflect the opportunities to receive outside help in extraction of higher shares of profit. Unfortunately, the data do not contain adequate unionisation measures. 14. As in the questionnaire 41% of the responding firms report that they have “implemented new payment systems within the last three years”, due caution in interpreting the results should clearly be taken as many of the firms may not have had performance pay schemes in the early nineties or before.
ACKNOWLEDGMENTS This is a substantially revised version of a paper presented at the International Conference on Organisational Design, Management Styles and Firm Performance at the University of Bergamo in June 2001. I am grateful to Kjell Salvanes and a referee for many useful suggestions to improve the paper. Andreas Sørensen and Jens Therkelsen provided excellent research assistance. Financial support from Danish Social Science Research Council is gratefully acknowledged.
REFERENCES Aghion, P., Caroli, E., & Garcia-Peñalosa, C. (1999). Inequality and Economic Growth: the Perspective of the New Growth Theories. Journal of Economic Literature, 37, 1615–1660. Askenazy, P., Caroli, E., & Marcus, V. (2001). New Organizational Practices and Working Conditions: Evidence from France in the 1990s. Mimeo. Athey, S., & Stern, S. (1998). An Empirical Framework for Testing Theories about Complementarity in Organizational Design. NBER working paper No. 6600.
The Effects of New Work Practices
29
Bailey, T., Berg, P., & Sandy, C. (2001). The Effect of High Performance Work Practices on Employment Earnings in Steel, Apparel, and Medical Electronics and Imaging Industries. Industrial and Labor Relations Review, 54, 525–543. Bauer, T., & Bender, S. (2000). Organizational Change and Wages: Evidence from Matched Employer-Employee Data. Manuscript, IZA, Bonn. Bayo-Moriones, J., Galilea-Salvatierra, P., & Merino-Diaz de Cerio, J. (2001). Participation, Cooperatives and Performance: An Analysis of Spanish Manufacturing Firms. Mimeo. Black, S., & Lynch, L. (2000). What’s Driving the New Economy: The Benefits of Workplace Innovation. NBER working paper 7479. Black, S., & Lynch, L. (2001). How to Compete: the Impact of workplace Practices and Information Technology on Productivity. Review of Economics and Statistics, 83, 434–445. Caroli, E., & van Reenen, J. (2001). Skill Biased Organizational Change? Evidence from a Panel of British and French Establishments. Quarterly Journal of Economics, 116, 1449–1492. Conyon, M., & Freeman, R. (2000). Shared Modes of Compensation and Firm Performance: U.K. Evidence. Forthcoming in: R. Blundell, D. Card & R. Freeman (Eds), Seeking a Premier League Economy. University of Chicago Press. Cristini, A., Gaj, A., Labory, S., & Leoni, R. (2002). New Work Practices in Italy. Adoption and Performance Effects. Mimeo, University of Bergamo. Eaton, S. (2001). If You Can Use Them: Flexibility Policies, Organizational Commitment, and Perceived Productivity. JFK School of Government, Harvard University, research working paper 01-009. Eriksson, T. (2001). How Common Are the New Compensation and Work Organization Practices and Who Adopts Them? Aarhus School of Business, Department of Economics working paper 01-8. Eriksson, T., Poulsen, C., & Westergård-Nielsen, N. (2000). Har nye arbejdsorganisationer og aflønningsformer fundet plads i danske virksomheder? CLS rapport 00-01. Freeman, R., Kleiner, M., & Ostroff, C. (2000). The Anatomy of Employee Involvement and Its Effects on Firms and Workers. NBER working paper 8050. Freeman, R., & Lazear, E. (1995). An Economic Analysis of Works Councils. In: J. Rogers & W. Streeck (Eds), Works Councils: Consultation, Representation, and Cooperation in Industrial Relations (pp. 27–50). University of Chicago Press. Gittleman, M., Horrigan, M., & Joyce, M. (1998). Flexible Workplace Practices: Evidence from a Nationally Representative Survey. Industrial and Labor Relations Review, 52, 99–115. Godard, J. (2001). High Performance and the Transformation of Work? The Implications of Alternative Work Practices for the Experience and Outcomes of Work. Industrial and Labor Relations Review, 54, 776–805. Handel, M., & Gittleman, M. (1999). Is There a Wage Payoff to Innovative Work Practices? Jerome Levy Economics Institute working paper 288. Huselid, M. (1995). The Impact of Human Resource Management Practices on Turnover, Productivity, and Corporate Performance. Academy of Management Journal, 38, 635–672. Ichniowski, C., & Shaw, K. (1995). Old Dogs and New Tricks: Determinants of the Adoption of Productivity-Enhancing Work Practices. Brookings Papers on Economic Activity: MicroEconomics, 1–55. Jones, D., & Kato, T. (1995). The Productivity Effects of Employee Stock Ownership Plans and Bonuses: Evidence from Japanese Panel Data. American Economic Review, 85, 391–414. Kato, T., & Morishima, M. (2001). The Productivity Effects of Participatory Employment Practices: Evidence from New Japanese Panel Data. Forthcoming in: Industrial Relations.
30
TOR ERIKSSON
McNabb, R., & Whitfield, K. (2000). New Work Practices, Compensation Systems and Performance in U.K. Workplaces. Manuscript, Cardiff Business School. Michie, J., & Sheehan, M. (1999). HRM Practices, R&D Expenditure and Innovative Investment: Evidence from the U.K.’s 1990 Workplace Industrial Relations Survey. Industrial and Corporate Change, 8, 211–234. Neumark, D., & Cappelli, P. (2001). Do High Performance Work Practices Improve Establishment-level Outcomes? Industrial and Labor Relations Review, 54. OECD (1999). Employment Outlook. Paris: OECD. Osterman, P. (1994). How Common is Workplace Transformation and Who Adopts It? Industrial and Labor Relations Review, 47, 173–188. Osterman, P. (2000). Work Reorganization in an Era of Restructuring: Trends in Diffusion and Effects on Employee Welfare. Industrial and Labor Relations Review, 53, 179–196. Pil, F., & MacDuffie, J. (1996). The Adoption of High-Involvement Work Practices. Industrial Relations, 35, 423–455. Snower, D. (1999). Causes of Changing Earnings Inequality. IZA discussion paper 29. Thesmar, D., & Thoenig, M. (2000). Creative Destruction and Firm Organization Choice: A New Look into the Growth-Inequality Relationship. Quarterly Journal of Economics, 115, 1201–1237.
PARTICIPATION, COOPERATIVES AND PERFORMANCE: AN ANALYSIS OF SPANISH MANUFACTURING FIRMS José Alberto Bayo-Moriones, Pedro Javier GalileaSalvatierra and Javier Merino-Díaz de Cerio ABSTRACT This paper attempts to analyze the effects that shopfloor participation has on firm performance and examine if cooperatives show better results than capitalist firms. Moreover, it tries to study if the impact of participation on performance is different in cooperatives and capitalist firms. To fulfill these objectives information about 965 Spanish manufacturing plants with at least fifty employees is used. The results indicate that there are not significant differences in the performance of cooperatives and the rest of companies. Our findings also show that the introduction of participatory practices generates positive outcomes both for capitalist firms and cooperatives, without differences in the magnitude of the impact.
1. INTRODUCTION The last decade has witnessed a marked increase in companies adopting new organizational practices aimed at promoting shopfloor participation (Osterman, 2000). These efforts to achieve greater employee involvement have been launched in two directions. Firstly, there has been an attempt to create a series The Determinants of the Incidence and the Effects of Participatory Organizations, Volume 7, pages 31–56. Copyright © 2003 by Elsevier Science Ltd. All rights of reproduction in any form reserved. ISBN: 0-7623-1000-6
31
32
J. A. BAYO-MORIONES ET AL.
of communication channels to permit the two-way flow of information between managers and employees. Meanwhile, there is a growing tendency to allow workers more participation and decision making rights in their jobs, in a departure from the traditional organizational structure in which the worker’s tasks were strictly defined and employees were merely required to follow the supervisor’s instructions to the letter. The growing diffusion of these practices, aimed at promoting shopfloor participation in firms, has been accompanied by keen academic interest in researching the various issues that this involves, particular attention being paid to whether or not these practices actually lead to an improvement in the firm’s performance and competitive edge (e.g. Batt & Appelbaum, 1995; Fernie & Metcalf, 1995; Black & Lynch, 1997; Ichniowski et al., 1997; Cappelli & Neumark, 1999; Ichniowski & Shaw, 1999; Eriksson, 2001; Kato & Morishima, 2002). Another example of employee participation, though of a very different nature, is to be found in producer cooperatives. Here, unlike the situation in typical capitalist firms, the workers are the owners and therefore hold all the control and return rights. The fact that these two rights are vested in the workers creates the need for an incentive framework completely different from that usually found in capitalist firms, and this will presumably have its effect on worker behavior and firm performance. The study of the effects of employees’ whole ownership rights on firm performance has given rise to a wide range of theoretical and empirical literature (Porter & Scully, 1987; Bonin et al., 1993). The results of both shopfloor participation and worker co-operatives have received enormous attention in the literature. Less interest, however, has been directed towards the analysis of possible differences between cooperatives and capitalist firms, in terms of the effects of shopfloor participation programs on workers’ morale and organizational performance. Cooperative firms, in which workers own return rights and elect the board of directors, present an employment relationship scenario that clearly differs from that found in conventional firms. This could explain different levels of effectiveness in the adoption of practices aimed at promoting more direct participation of workers in their jobs. This chapter takes data from an initial sample of 965 Spanish manufacturing plants, to study the relationship between employee involvement, cooperatives and performance. The aim is to find the answers to three questions. First of all, an assessment will be made of the results obtained when shopfloor participatory schemes are used. Next, cooperatives will be analyzed to see if they achieve better performance measures than other types of companies.
Participation, Cooperatives and Performance
33
Finally, this investigation will attempt to ascertain whether the effectiveness of employee involvement practices varies according to whether they are introduced in a cooperative or a capitalist firm. Shopfloor participation in both these kinds of companies will be analyzed in order to detect any possible differences in outcome. There are several features of this paper that need to be underlined. First, it must be emphasized that the range of establishments analyzed means that they do not belong to a narrow sector of industry with limited relevance to the economy of the country as a whole. This, together with the size of the sample used, allows a higher degree of generalization of the results and conclusions achieved. Secondly, to the knowledge of the authors, this is the first paper to take such a large sample to study the relationship between cooperatives, participation and performance in Spain.1 This enlarges the geographical area covered by the empirical evidence surrounding this issue. Thirdly, in analyzing the effects of participation, both measures of plant performance, in its various management aspects, and measures of the situation and behavior of workers will be considered. Fourthly, nine different practices are considered in order to assess the degree of effort being aimed at developing participatory arrangements in the firm. This number of practices is large enough to provide a more reliable picture of the importance attached by each firm to promoting employee participation. This chapter is organized as follows. The next section contains a brief review of the economic literature dealing with employee participation, cooperatives and performance. Following that, there will be a description of the details of the process of collection of the data used in the empirical analysis, then a definition of the variables employed and an explanation of the estimation methodology. The empirical results will then be presented, after which the paper will end with a discussion of the main conclusions that have been reached.
2. THEORETICAL FRAMEWORK Employee Participation There are several reasons for believing that shopfloor participation can produce positive outcomes both for employees and the firm. One of the usual reasons put forward is that employees gain access to productivity enhancing information. It has long been recognized that information is not evenly spread throughout organizations and that employees possess private information from which management could benefit. Employees engaged in routine, day-to-day, tasks are usually in a better position to detect inefficiencies in operations that
34
J. A. BAYO-MORIONES ET AL.
may affect productivity (Nalbantian, 1987). The information derived from such activity is potentially very valuable to the firm as an input to production. By improving the use of information and knowledge in organizations and turning decision making rights over to those who possess this first-hand knowledge, it is possible to speed up action and increase efficiency. From the point of view of motivation, it must be underlined that participation has a significant effect on job characteristics. In so far as participatory schemes alter such aspects of the job as variety and autonomy (Hackman & Oldham, 1980), they can be expected to bring about an improvement in the intrinsic motivation of workers. This would reduce the disutility of effort and would, as a consequence, improve job satisfaction, morale and performance. These positive effects of participation on employee satisfaction and performance would be accompanied by a better performance by the firm as a whole, as well as a reduction in absenteeism and labor conflicts. Also with respect to raising motivation, it is argued that employee involvement is likely to promote greater trust between managers and employees and greater commitment and identification of workers with the firm (Fernie & Metcalf, 1995), thus aligning individual objectives with those of the organization as a whole (Levine & Tyson, 1990). In spite of all the above-mentioned advantages, however, employee participation also presents some drawbacks. In so much as employee participation involves the handing over of decision rights from owners to workers, it gives rise to an agency relationship. This relationship has some unavoidable costs, due to the imperfection of motivational devices (Levine & Tyson, 1990). Employees may use the greater autonomy available to them to their own ends, by shirking in their job. The latest institutional economics literature suggests that these agency costs increase with the number of people taking part in the decision making process (Williamson, 1975; Jensen & Meckling, 1979). Although participation can improve intrinsic motivation among employees (Batt & Appelbaum, 1995; Berg, 1999), it can also create a negative effect, since it may prove more demanding and stressful and, therefore, increase disutility in the job (Ben-Ner & Jones, 1995). It is also necessary to take into account the costs involved in implementing and running schemes aimed at providing mechanisms for employee participation (Ichniowski & Shaw, 1995; Jones & Kato, 1995; Pil & MacDuffie, 1996; Kato & Morishima, 2002). It is immediately obvious, therefore, that participation is an extremely complex phenomenon, making it very difficult to establish an unambiguous hypothesis as to the sign and magnitude of its effects on firm performance and whether they will be statistically significant. It would seem that the size and
Participation, Cooperatives and Performance
35
sign of the relationship can not be determined from a theoretical perspective and that they ultimately pose an empirical question (Jones & Pliskin, 1991). A large number of empirical papers have analyzed the effects of participation2 without, of course, always reaching the same conclusions. Nevertheless, it can be stated that the effects of shopfloor participation have generally been found to be favorable, especially if the performance measures analyzed have to do with worker motivation. Cooperatives and Their Efficiency Producer cooperatives are labor-managed firms where the workers are the owners and participate on a democratic basis in the ultimate authority over the decisions of the company, including the right to delegate some or all decisions to managerial organs.3 Cooperatives must be considered from the ownership perspective. There are two aspects involved in the ownership of an asset. The first of these has to do with the rights to control the asset, which implies the right to make decisions concerning its use; all decisions are determined by majority vote among worker-members, either directly or through an elected manager or board of directors. The second aspect relates to the return rights, that is, the rights to net earnings from the asset. The combination of these two rights is seen as a powerful incentive to promoting efficiency in the use of the firm’s assets and in any investment decisions that must be made. Cooperatives are an example of this combination, but their distinctive feature is that control and return rights are held by the workers. Most of the economic literature sustains that the reason for there being so few cooperatives in market economies is that they are inefficient (Porter & Scully, 1987). The traditional picture of cooperatives is one of small firms unable to survive in the long term in a capitalist environment because of a variety of problems, mainly of a managerial and financial nature. However, little theoretical consensus has been reached, and few attempts have been made to solve the problem through systematic empirical research. There are some disadvantages associated with cooperatives. Their inefficiency – compared to other forms of business organization, mainly public corporations – arises from various factors. Employees generally have limited wealth and hence may have trouble in financing labor-managed firms by themselves, while adverse selection and moral hazard problems limit the ability of co-operatives to raise funds in the credit markets. As employees, cooperative members already have much at stake in the success of their firm; as owners, they also have to invest financially in the firm, which leaves them even further exposed to firm-specific risk. As a consequence, cooperatives will tend to take
36
J. A. BAYO-MORIONES ET AL.
very conservative decisions and will try to diversify more than other firms might, with all the negative consequences that this has on profitability. The existing numbers of producer cooperatives can be partly explained by the substantial tax subsidies to which they are eligible.4 However, they are also sometimes established for other reasons, since they may enjoy substantial advantages over conventional firms owned by capital suppliers. Cooperatives are formed for a variety of reasons; some of them idealistic, such as the desire to establish a workplace where everyone can participate (so that no individual assumes the role of boss). A brief examination will now be made of the main advantages of cooperatives from the economic point of view. The fact that a firm is a cooperative does not necessarily mean that it is managed by the workers. These have the decision making rights and elect the cooperative board and managers. The managers need not necessarily be worker-owners; they may be outside professionals. The recruitment of professional staff usually compromises internal democracy and, as a consequence, problems may arise with involvement among worker-owners. In other words, agency problems between management and principals also exist in cooperatives. However, it is often argued that this problem is less acute in worker-owned firms than in capitalist firms. Workers’ wealth is more concentrated in a single firm than is the case with capitalist investors, who usually diversify their investments and have no special need to be informed about the management of the firm. Workers, on the other hand, have both the opportunity and the need to stay informed about the effectiveness of management. Closely related to the agency costs between ownership and management, are the agency costs between ownership and workers. When team production is present, cooperatives must face the incentive problem of free-riding by team members. Teamwork means that it is difficult to observe and verify individual effort (Alchian & Demsetz, 1972). As a consequence, an incentive problem arises, since any team member can stint on effort without facing a proportionate reduction in income. However, when workers are the owners of the firm, and are therefore assigned the return rights, the problem is alleviated and some of these costs are internalized, with the result that productivity is enhanced. In producer cooperatives, where employees have control and return rights, productivity and performance are improved as a result of fewer labormanagement disputes and increased incentive towards effort and cooperation among employees, whose income and wealth is tied to company performance. There is general agreement among researchers in organizational behavior that, when workers are owners, individual motivation is clearly enhanced (Poole & Jenkins, 1990; Pendleton et al., 1998).
Participation, Cooperatives and Performance
37
In addition to the incentive explanation just mentioned, there are other reasons why cooperatives may achieve a superior performance. Mutual monitoring is encouraged in cooperatives because they are owned by the workers. It is often argued that worker-owners are better than managers at monitoring workers. Mutual monitoring may be more efficient than a hierarchical supervision mechanism (Williamson, 1975). Co-workers are often better able than outsiders to devise cheaper or more accurate ways of observing effort (Putterman, 1984). Empirical proof that this mutual monitoring takes place is that worker-owned firms are often cited as characteristically using fewer supervisors than capitalist firms (Bonin et al., 1993). Another advantage of cooperatives is to be found in specific human capital investments that would not be possible in conventionally owned firms (Williamson, 1985). Moreover, in cooperatives the usual gaps in the spread of information between management and workers do not occur. These information gaps are frequently the cause of disputes in the regular bargaining process over employment conditions and it often happens that strikes and lock-outs can be traced back to them (Hansmann, 1990). Participation and Cooperatives Some recent contributions to the literature suggest that the benefits to be gained from participation depend on other aspects of the employment relationship and on the context in which it takes place (Levine & Tyson, 1990). Undoubtedly, the characteristics of the relationship between employee and firm in a producer cooperative differ substantially from those found in a conventionally owned company. In cooperatives, employees have the ultimate right of control over the firm, while also owning the return rights. Therefore, the introduction of shopfloor participatory arrangements may produce a very different outcome in firms where workers have full ownership rights. Although initially it seems reasonable to expect that there will be greater diffusion of shopfloor participatory practices among cooperatives, this is may not necessarily be the case. These are two aspects that, although potentially linked, are different in nature. Whereas the distinctive feature of cooperatives is employee ownership, with the political and return rights that this entails, shopfloor participation has to do with the day-to-day life of the worker at his job. It is quite possible in cooperatives for individuals as owners to have control and return rights, while, as workers, they perform their duties in narrowly defined jobs with no autonomy. As already mentioned, one of the problems generated by the handing over of decision rights, which comes about as a result of participation, is that they may
38
J. A. BAYO-MORIONES ET AL.
be used by workers for their own benefit and, as a result, work against the interests of the firm. This negative consequence of participation will be less acute in the case of cooperatives. Although the incentive to free-ride does not entirely disappear, it is not as strong. Whereas cooperative workers, as owners of the firm, will suffer an indirect loss if they fail to behave as they ought, this is not the case among workers in conventional capitalist firms. Since inappropriate behavior reduces the firm’s profits and value, the return on the individual’s share in the ownership of the firm will also be reduced. Moreover, fellow workers, being fellow owners, will be keener to devote some time to monitoring tasks. If a worker is found by his colleagues to be guilty of shirking, social sanctions may be applied. In cooperatives, workers are subject to peer pressure, which makes it more costly for them to abuse their assigned decision rights. As far as empirical evidence is concerned, we must point to the conclusions reached in the review made by Doucouliagos (1995). This article shows that there is a positive correlation between some forms of worker participation and productivity. The most interesting point for the purposes of the present discussion is that this correlation is stronger in the case of labor-managed firms than among participatory capitalist firms.
3. METHODOLOGY The Process of Obtaining Data The information used in this paper was obtained in the context of a wider research project focused on the analysis of the new systems of production and work organization in Spain. The Spanish manufacturing industry constitutes the scope of our study. The concept of manufacturing industry is clearly defined in the National Classification of Economic Activity (NACE), which includes all the manufacturing industries (from code 15 to code 37) with the exception of oil refining and the treatment of nuclear fuel (code 23). Plant was chosen as the unit of analysis, because it is at this level where the policies analyzed in the paper are determined, and therefore where problems arise and where the results must be analyzed. Further, attention focuses on the plants that employ fifty or more workers. This limit has been used in other studies relating to the research area (see Osterman, 1994) and it serves to cover a wide spectrum of the population employed in Spanish manufacturing firms, what is more it simplifies the field work. With these criteria, manufacturing
Participation, Cooperatives and Performance
39
plants with fifty or more employees, the population consists of 6,013 establishments. 965 plants were surveyed, after the adoption of a stratified sampling procedure based on industry and size. Primary data was obtained from a pre-tested questionnaire specifically designed to carry out the research project. The questions concerning human resource management refer to blue-collar workers. In order to get the information, a personal face to face interview was conducted to a manager of the plant. In most cases this manager was either the plant manager or the operations or the human resource manager. Since they work in the plant and not in central headquarters far from the shopfloor, we consider that, given the content of the questionnaire, they can provide accurate information on the practices analyzed in this article. After an initial process that required contacts with 3,246 plants to arrange interviews, 965 of them agreed to participate (a response rate of 29.7%). Describing the Dependent Variables The performance measures, the dependent variables, to be used are of two types. On the one hand, we have a series of operational performance measures, that is, results more closely related to the productive activity of the organizations. On the other hand, we also possess some results that relate more directly to human resources management. No analysis will be made of financial results, since these data are not available. It is true to say, however, that participatory practices will have a more direct impact on results in the operational and human resource management areas. This is because financial results are also subject to the effects of other variables outside the strict realm of industrial management. As a measure of direct performance outcomes, we will take the degree of improvement over the three previous years on five indicators, linked to the basic objectives of any productive system, and therefore, of any firm (Corbett & Van Wassenhove, 1993). These indicators are: the percentage of productive hours over the total number of hours of direct labor (EFFICIENCY); the percentage of faulty finished products (QFINISHED); the percentage of defective products in process (QPROCESS); the percentage of meeting agreed deadlines (PUNCTUALITY), and the time that elapses between receiving the materials and delivering the product to the customer (LEADTIME). OPERATIONAL captures in how many of these indicators there have been improvements in comparison with three years before. Therefore OPERATIONAL is an ordinal variable that takes integer values between zero and five and is based in the subjective judgment of the respondents.
40
J. A. BAYO-MORIONES ET AL.
Results of a subjective nature are often used in research on organizations. The different objective manufacturing performances, which are measured in absolute terms, depend to a great extent on the technology and type of process found at the plant. It therefore becomes difficult to establish comparisons when the data is obtained from a group of heterogeneous plants. Some studies (Venkatraman & Ramanujan, 1986; Powell, 1995) have shown that there is a strong link between objective performance measures and the same performance measures evaluated subjectively. Moreover, Machin and Stewart (1990) and McNabb and Withfield (1998) point out that there is no consensus about the most appropriate performance measure and that all objective measures have shortcomings. Three measures of results in the area of human resource management are used: the rate of absenteeism in the plant (ABSENTEEISM), an indicator of worker satisfaction measured on a scale of 0 to 10 (SATISFACTION) and a binary variable (STRIKE) indicating whether there were hours lost to strikes over the last year (value 1) or not (value 0). Describing the Independent Variables Core Independent Variables The first of our independent variables tries to capture whether the firm is or not a producer cooperative. COOPERATIVE is a binary variable that indicates whether the plant belongs to a cooperative firm (equals 1) or to a capitalist firm (equals 0).5 In order to measure the extent to which firms encourage employee participation, we have created an index (PARTICIPATION) which indicates how many of the nine involvement practices taken into account are applied in the factory. Therefore, PARTICIPATION takes integer values between 0 and 9. Most of the practices considered appear in other studies about employee participation and high-performance work practices. We have also included some practices that, although not previously considered, are deemed to be indicative of participatory management. The nine practices taken into account are the following ones: • The autonomy of workers in planning and organizing tasks (Arthur, 1992; Geary, 1999). • The participation of workers in designing the jobs for which they are employed.
Participation, Cooperatives and Performance
41
• Their collaboration in the training of new workers. • The installation of autonomous work teams (Osterman, 1994; MacDuffie, 1995). The work team is responsible for a whole product (or part of a product) and makes decisions about task assignments and work methods. • The organization of improvement groups (Osterman, 1994; Gittleman et al., 1998). They are a structured type of employee participation in which groups of employees from a particular work area meet regularly to identify and suggest improvements to work-related problems. • The use of suggestion systems (MacDuffie, 1995; Roche, 1999). They are programs that elicit individual employee suggestions on improving work or the work environment. • The consultation of workers via surveys (Huselid, 1995; Roche, 1999). Employee attitude surveys are conducted in order to know the employees’ opinion. • The celebration of open-days. The firm organizes visits to the firm, usually together with social gatherings, for the family and friends of the employees. • Meetings to inform the workers (Wood & Albanese, 1995). Regular meetings of one or several of the managers of the plant with the workers to give them information about the situation of the company, future plans, etc. In Table 1, a description of the diffusion of the participatory practices in cooperatives and capitalist firms is displayed. Only for three of the nine practices considered (meetings, autonomy in planning and collaboration in job
Table 1.
Adoption of Participatory Practices in Cooperatives and Capitalist Firms (% of Plants in Each Group).
Suggestion systems Improvement groups Surveys to employees Meetings Open says Autonomous work teams Autonomy in planning Collaboration in training Collaboration in job design *** p < 0.01, ** p < 0.05, * p < 0.10
Cooperatives
Capitalist firms
Total
2
54.17 38.78 14.29 79.59 22.22 44.68 36.17 85.10 60.00
56.74 39.03 22.57 57.27 20.00 43.26 25.58 78.75 43.18
56.63 39.03 22.22 58.67 20.38 43.72 26.02 78.67 43.98
0.124 0.001 1.849 9.533*** 0.013 0.036 2.590* 1.089 4.922**
42
J. A. BAYO-MORIONES ET AL.
design) there are statistically significant differences in the implementation; in the three cases the incidence is greater in cooperative firms. Control Variables AUTOMATION aims to capture the degree of flexible automation in the plant. The questionnaire enquired after four technical features that we felt to be directly related to the degree of automation: namely, robots or programmable automatons, automatic materials storage and retrieval systems (AS/RSs), computer integrated manufacturing (CIM) and computer networks for the processing of the plant’s production data. By applying factor analysis to these four variables, a single factor is obtained with an eigenvalue greater than one, which accounts for just over 47% of the variance. The factor loadings on these variables are greater than 0.44, while Cronbach’s alpha is 0.6278. AUTOMATION is defined as the average of the four variables mentioned. Plant size (SIZE) is measured in terms of the number of workers employed at the plant (the logarithm of this number, for better fit). To account for the influence of the age of the plant the variable AGE was included. AGE is measured as the logarithm of the number of years the plant has been working. The level of competition being faced by the firm is captured by the variable COMPETITION. This assesses on a scale of one to five, the evolution of competition levels over the last three years in the sector in which the plant operates. A score of one on this scale indicates a large decrease, whereas a score of five represents a large increase. Finally, AVERUNION and LOWUNION are two dummy variables that capture the influence that unions have on the workers of the plant.6 LOWUNION equals one if that influence is low, whereas AVERUNION takes value one if the influence is medium. In Table 2 differences in the level of adoption of participatory practices by degree of union influence on workers can be observed. We have found statistically significant differences for none of the practices analyzed. This indicates that for the Spanish case union influence neither helps nor hinders the implementation of these practices. Table 3 displays the mean, the standard deviation and the correlations of the variables in the study. The correlation matrix immediately provides the first clues as to how the variables interrelate. The most relevant for the purposes of this study are as follows: • COOPERATIVE shows no statistically significant relationship with the remaining variables, except with LOWUNION and LEADTIME. Union
Participation, Cooperatives and Performance
Table 2.
43
Adoption of Participatory Practices by Level of Union Influence (% of Plants in Each Group).
Suggestion systems Improvement groups Surveys to employees Meetings Open days Autonomous work teams Autonomy in planning Collaboration in training Collaboration in job design
Low
Average
High
Total
2
54.06 35.80 19.76 56.41 18.38 40.53 27.79 80.90 43.84
59.71 41.37 24.45 59.86 19.85 48.16 24.09 77.06 43.73
57.78 42.47 24.29 61.57 25.12 44.29 25.00 76.39 44.60
56.63 39.03 22.22 58.67 20.38 43.72 26.02 78.67 43.98
2.355 3.626 2.822 1.816 3.950 3.936 1.334 2.353 0.044
*** p < 0.01, ** p < 0.05, * p < 0.10
•
• •
• •
influence on employees is lower in cooperative firms than in capitalist firms. PARTICIPATION is positively related to AUTOMATION and SIZE and negatively to AGE, thereby indicating that participation tends to be more widely practiced in larger, more fully automated and newly established plants. PARTICIPATION does not show any significant correlation with the two union influence variables. PARTICIPATION has a positive and statistically significant relationship with most of the performance variables considered; i.e. with those related to operational performance and with the degree of job satisfaction among employees. COMPETITION shows no significant linkage with any of the variables. Operational performance displays a different pattern from that of human resource management performance measures (the only significant link at 0.01 level is between LEADTIME and SATISFACTION).
This is the initial picture that emerges from the correlation matrix. Relationships between variables will undergo more thorough examination after the multivariate analysis that is applied at a later stage of the study. Estimation Methodology Nine models are employed to estimate the influence of the explanatory variables on the nine dependent variables (nine performance measures) used.
3.87 0.06 4.10 4.93 3.26 3.54 0.30 0.46 0.65 0.69 0.63 0.64 0.73 3.34 4.12 6.86 0.10
1
2.11 0.22 0.05 2.38 0.32*** 0.89 0.14*** 0.80 –0.10** 0.89 –0.02 0.46 0.03 0.50 –0.06 0.48 0.13*** 0.46 0.11*** 0.48 0.13*** 0.48 0.14*** 0.45 0.21*** 1.70 0.21*** 2.67 0.01 1.50 0.21*** 0.30 0.01
*** p < 0.01, ** p < 0.05, * p < 0.10
1. PARTICIPATION 2. COOPERATIVE 3. AUTOMATION 4. SIZE 5. AGE 6. COMPETITION 7. AVERUNION 8. LOWUNION 9. EFFICIENCY 10. PUNCTUALITY 11. QFINISHED 12. QPROCESS 13. LEADTIME 14. OPERATIONAL 15. ABSENTEEISM 16. SATISFACTION 17. STRIKE
Mean s.d.
3
4
6
7
8
0.02 0.04 –0.01 –0.12*** 0.04 –0.61*** –0.02 0 0 –0.07* –0.04 0.04 –0.03 –0.03 0 0.05 –0.01 –0.05 0.037 0.02 0.02 –0.05 –0.03 0.05 0.02 –0.08** –0.02 0.03 0.02 –0.08** 0.07* –0.01 0.08*** –0.20*** –0.07** 0.04 0.01 0.03 0.03 –0.02 0.01 –0.1***
5
10
0.47*** 0.38*** 0.45*** 0.39*** 0.47*** 0.22*** 0.27*** 0.70*** 0.75*** 0.01 –0.05 0.06* –0.03 0.04 0.03
9
12
0.77*** 0.25*** 0.29*** 0.80*** 0.82*** 0 –0.05 0.01 0.02 0.05 0.04
11
Descriptive Statistics and Correlation Matrix.
–0.01 0.04 0.31*** –0.04 0 0.09*** 0.02 0.01 –0.03 –0.04 0 0.03 0.11*** –0.11*** –0.24*** –0.04 0.12*** 0.11*** 0 0.11*** 0.09*** –0.02 0.14*** 0.10*** –0.03 0.17*** 0.14*** 0.06* 0.23*** 0.13*** –0.03 0.21*** 0.16*** 0.02 0.02 0.17*** 0.03 0.17*** –0.03 –0.04 0.04 0.19***
2
Table 3. 14
15
16
0.55*** 0.02 –0.03 0.10*** 0.05 –0.10*** 0.05 0.07* 0.14*** –0.10***
13
44 J. A. BAYO-MORIONES ET AL.
Participation, Cooperatives and Performance
45
(i) For each of the five individual measures of operational results a logit model will be estimated (the dependent variables are binary). The generic equation is: RESULT7 = 0 + 1PARTICIPATION + 2COOPERATIVE + 3SIZE + 4AGE + 5AUTOMATION + 6COMPETITION + 7AVERUNION + 8LOWUNION + 9PARTICIPATION COOPERATIVE
(ii) The next equation is referred to the variable defined as the addition of the five previous individual indicators of operational results. OPERATIONAL = 0 + 1PARTICIPATION + 2COOPERATIVE + 3SIZE + 4AGE + 5AUTOMATION + 6COMPETITION + 7AVERUNION + 8LOWUNION + 9PARTICIPATION COOPERATIVE
To estimate this, an ordered logit model is selected, which is distinguished by the fact that the dependent variable is an ordered qualitative categorical variable. The decision to choose this model as opposed to discriminant analysis or a standard regression model, is based on the greater robustness of this type of model in withstanding the violation of the assumption of normality (Maddala, 1983). (iii)
ABSENTEEISM = 0 + 1PARTICIPATION + 2COOPERATIVE + 3SIZE + 4AGE + 5AUTOMATION + 6COMPETITION + 7AVERUNION + 8LOWUNION + 9PARTICIPATION COOPERATIVE
In this case a multiple regression model is used for the estimation. (iv)
SATISFACTION = 0 + 1PARTICIPATION + 2COOPERATIVE + 3SIZE + 4AGE + 5AUTOMATION + 6COMPETITION + 7AVERUNION + 8LOWUNION + 9PARTICIPATION COOPERATIVE
As in the case of ABSENTEEISM, a multiple regression model is used.
46
(v)
J. A. BAYO-MORIONES ET AL.
STRIKE = 0 + 1PARTICIPATION + 2COOPERATIVE + 3SIZE + 4AGE + 5AUTOMATION + 6COMPETITION + 7AVERUNION + 8LOWUNION + 9PARTICIPATION COOPERATIVE In this case, since the dependent variable is binary, a logit model is estimated.
4. RESULTS Results from previous models are shown in Tables 4, 5, 6, and 7. For each of the dependent variables three estimations were made; the first includes only the control variables (AUTOMATION, SIZE, AGE, COMPETITION, AVERUNION and LOWUNION); the second incorporates the independent core variables (PARTICIPATION and COOPERATIVE) and, finally, in the third estimation, these are further extended to include an interaction variable (PARTICIPATION COOPERATIVE). Tables 4 and 5 summarize the results of the estimation intended to explain improvement in individual measures of operational performance and in the global indicator. The first detail to emerge is the significant link between improvement and the extent to which participation is practiced (p < 0.10 in all the cases). This appears to suggest that the effort made by Spanish manufacturing firms to increase involvement and acceptance of responsibility among their employees is rewarded with an improvement in performance in all production related activities. Other notable features are the negligible influence of whether or not the firm is a cooperative on performance improvements and the fact that the impact of participation on operational performance is no stronger in cooperatives than in any other type of firm (PARTICIPATION COOPERATIVE is not significant). As far as the remaining variables are concerned, it is worth mentioning that AUTOMATION shows a positive and statistically significant relationship with the majority of operational results, a feature that is common to all the models. Table 6 summarizes the results obtained when the model is applied to absenteeism. Here it emerges that PARTICIPATION, COOPERATIVE and the interactive term all fail to be statistically significant. The rate of absenteeism remains unaffected by the adoption of participatory practices on the part of the firm and also fails to show any link with the fact that the employees may be owners of the firm. It is worth underlining the fact that absenteeism increases with the size of the firm and is significantly lower in firms where the union influence is weaker.
s.d. 0.710 0.037 0.107 0.089 0.107 0.231 0.219
19.353*** –448.443 0.037
–0.212 0.101*** 0.129 0.059 –0.049 –0.193 –0.366*
Coef. 0.725 0.039 0.108 0.089 0.109 0.233 0.221 0.042 0.360
26.159*** –445.041 0.050
–0.548 0.071** 0.121 0.065 –0.022 –0.188 –0.355 0.103** –0.369
EFFICIENCY Coef. s.d. 0.734 0.039 0.108 0.089 0.110 0.233 0.221 0.043 0.156 0.216
s.d.
27.536*** –444.350 0.052
–0.419 0.071* 0.117 0.057 –0.035 –0.184 –0.363 0.092** –1.423 0.245
Coef.
s.d. 0.712 0.037 0.105 0.089 0.107 0.230 0.220
19.041** –451.259 0.036
–0.643 0.115*** 0.057 0.105 0.089 –0.090 –0.267
Coef. 0.731 0.038 0.106 0.089 0.109 0.232 0.223 0.042 0.380
28.052*** –446.750 0.053
–1.063 0.081** 0.046 0.110 0.128 –0.088 –0.259 0.124*** –0.209
QFINISHED Coef. s.d. 0.740 0.038 0.106 0.089 0.110 0.232 0.223 0.043 0.971 0.216
s.d.
28.358*** –446.600 0.054
–1.000 0.081** 0.043 0.107 0.122 –0.087 –0.266 0.119*** –0.696 0.117
Coef. 0.037 0.037 0.109 0.091 0.109 0.232 0.219
s.d.
29.993*** –441.416 0.057
–1.656*** 0.136*** 0.188* 0.039 0.188* 0.235 –0.004
Coef.
0.746 0.039 0.110 0.091 0.111 0.234 0.221 0.043 0.397
37.712*** –437.557 0.072
–2.048** 0.103*** 0.172 0.045 0.228** 0.239 –0.004 0.117*** 0.047
QPROCESS Coef. s.d.
0.755 0.039 0.110 0.091 0.111 0.234 0.222 0.043 0.987 0.226
s.d.
37.968*** –437.478 0.072
–1.989*** 0.103*** 0.169 0.042 0.222** 0.239 –0.010 0.113*** –0.405 0.113
Coef.
Results of Logit Models for EFFICIENCY (n = 710), QFINISHED (n = 709) and QPROCESS (n = 706).
*** p < 0.01, ** p < 0.05, * p < 0.10
Chi-Squared LogL Pseudo-R2
CONSTANT AUTOMATION SIZE COMPETITION AGE AVERUNION LOWUNION PARTICIPATION COOPERATIVE COOPERATIVE PARTICIPATION
Table 4.
Participation, Cooperatives and Performance 47
s.d. 0.709 0.037 0.106 0.089 0.109 0.231 0.222
13.786** –450.257 0.026
0.189 0.091** 0.086 0.068 –0.063 –0.294 –0.287
Coef. 0.722 0.039 0.106 0.090 0.110 0.231 0.223 0.042 0.380
18.153** –448.074 0.034
–0.074 0.067* 0.073 0.070 –0.036 –0.298 –0.296 0.087** –0.008
PUNCTUALITY Coef. s.d. 0.731 0.039 0.106 0.090 0.111 0.232 0.224 0.043 1.005 0.233
s.d.
20.060** –447.120 0.038
0.080 0.067* 0.068 0.061 –0.053 –0.293 –0.306 0.076* –1.250 0.306
Coef. 0.752 0.041 0.117 0.093 0.112 0.243 0.226
s.d.
52.249*** –423.435 0.096
–0.757 0.228*** 0.155 0.130 –0.095 0.082 –0.168
Coef. 0.769 0.043 0.117 0.095 0.115 0.246 0.229 0.045 0.426
66.123*** –416.498 0.120
–1.220 0.188*** 0.125 0.129 –0.041 0.073 –0.200 0.158*** 0.372
LEADTIME Coef. s.d. s.d. 0.779 0.043 0.117 0.095 0.116 0.246 0.230 0.046 1.047 0.266
67.643*** –415.738 0.123
–1.070 0.187*** 0.121 0.119 –0.056 0.079 –0.209 0.148*** –0.771 0.308
Coef.
s.d. 0.371 0.020 0.055 0.039 0.056 0.122 0.116
43.012*** –1061.825 0.065
0.532 0.093*** 0.086 0.031 0.008 –0.029 –0.157
Coef.
0.379 0.020 0.549 0.055 0.056 0.122 0.117 0.022 0.210
60.626*** –1053.018 0.091
0.213 0.066*** 0.080 0.041 0.035 –0.028 –0.147 0.091*** –0.157
OPERATIONAL Coef. s.d.
0.384 0.020 0.055 0.038 0.056 0.121 0.118 0.022 0.551 0.125
s.d.
61.971*** –1052.345 0.093
0.280 0.066*** 0.077 0.038 0.028 –0.027 –0.154 0.086*** –0.685 0.127
Coef.
Results of Logit Models for PUNCTUALITY (n = 734) and LEADTIME (n = 759) and Ordered Logit Models for OPERATIONAL (n = 660).
*** p < 0.01, ** p < 0.05, * p < 0.10
Chi-Squared LogL Pseudo-R2
CONSTANT AUTOMATION SIZE COMPETITION AGE AVERUNION LOWUNION PARTICIPATION COOPERATIVE COOPERATIVE PARTICIPATION
Table 5.
48 J. A. BAYO-MORIONES ET AL.
0.864 0.045 0.127 0.111 0.128 0.277 0.264
s.d.
8.881*** 0.073
*** p < 0.01, ** p < 0.05, * p < 0.10
F (Fisher) R2
2.532*** –0.057 0.429*** –0.077 0.190 –0.279 –1.134***
Coef. 0.881 0.048 0.127 0.111 0.129 0.278 0.265 0.051 0.467
6.643*** 0.073
2.520*** –0.058 0.428*** –0.077 0.191 –0.280 –1.136*** 0.003 0.052
ABSENTEEISM Coef. s.d. s.d. 0.890 0.048 0.127 0.112 0.130 0.278 0.265 0.052 1.216 0.258
5.896*** 0.073
2.519*** –0.058 0.428*** –0.076 0.191 –0.280 –1.136*** 0.004 0.065 –0.003
Coef.
4.675 0.036
6.634*** 0.109*** –0.114* 0.081 –0.094 0.170 0.223
Coef. 0.461∫ 0.024 0.068 0.060 0.069 0.149 0.142
s.d. 0.463 0.025 0.068 0.059 0.069 0.147 0.140 0.027 0.249
6.341*** 0.064
6.233*** 0.073*** –0.133* 0.085 –0.053 0.168 0.217 0.125*** 0.082
SATISFACTION Coef. s.d.
0.468 0.025 0.068 0.059 0.069 0.147 0.140 0.027 0.633 0.136
s.d.
5.768*** 0.065
6.157*** 0.074*** –0.131* 0.090 –0.045 0.166 0.222 0.131*** 0.712 –0.148
Coef.
Results of Multiple Regression Model for ABSENTEEISM (n = 686) and SATISFACTION (n = 754).
CONSTANT AUTOMATION SIZE COMPETITION AGE AVERUNION LOWUNION PARTICIPATION COOPERATIVE COOPERATIVE PARTICIPATION
Table 6.
Participation, Cooperatives and Performance 49
50
J. A. BAYO-MORIONES ET AL.
Neither the degree of automation nor the level of competition nor the age of the plant prove to be related to rates of absenteeism. Moreover, Table 6 summarizes the results obtained on the dependent variable SATISFACTION. These show that PARTICIPATION is associated in a positive and statistically significant manner (p < 0.01) with the degree of job satisfaction among workers. In other words, in firms where participatory practices are more widely applied, workers show a higher level of satisfaction with their jobs. On the other hand, neither COOPERATIVE on its own nor multiplied by PARTICIPATION appears to play a significant role in this model, which shows that job satisfaction among workers does not increase when they own the firm nor is the impact of participation on job satisfaction any greater among cooperative workers. With respect to the impact of the remaining variables on SATISFACTION, it can be seen that both AUTOMATION (p < 0.01) and SIZE (p < 0.10) prove to be significant, albeit in opposite directions. Thus, in firms using more complex technology, there is a higher level of job satisfaction, which seems to suggest that workers feel happier as their jobs become more challenging, and less dreary and monotonous. However, the larger the firm, the less satisfied its workers appear to be with their jobs. This is possibly due to the fact that it is usually harder for workers to become identified with organizations of large dimensions. As a final point, it emerges that neither the age of the plant nor the level of competition to which it is subjected nor union influence are in any way associated with the degree of job satisfaction shown by workers. Table 7 summarizes the results obtained on STRIKE, the last of the variables considered in the model. Here three of the explanatory variables included, SIZE, LOWUNION and AVERUNION prove to be statistically significant. These results indicate that the larger the firm, the more likely it is that hours will be lost due to strikes. Furthermore, in those plants where union influence is lower, the probability of there being strikes, as expected, is lower. The other control variables are not significant. Here the dependent variable is unaffected by either the extent of the participatory practices or whether or not it is a cooperative; the interaction of the two variables also fails to have any impact.
5. CONCLUSIONS A broad literature has analyzed the advantages and disadvantages of employee participation. There is also a vast literature investigating the efficiency of the
Participation, Cooperatives and Performance
51
Table 7. Results of Logit Model for STRIKE (n = 749).
CONSTANT AUTOMATION SIZE COMPETITION AGE AVERUNION LOWUNION PARTICIPATION COOPERATIVE COOPERATIVE PARTICIPATION Chi-squared LogL Pseudo-R2
Coef.
s.d.
Coef.
s.d.
Coef.
s.d.
–3.609*** –0.041 0.598*** –0.153 –0.112 –0.556* –0.905***
0.967 0.057 0.137 0.134 0.155 0.300 0.312
–3.484*** –0.033 0.608*** –0.151 –0.134 –0.544* –0.859*** –1.295 –0.032
0.999 0.060 0.138 0.135 0.157 0.300 0.315 1.033 0.064
–3.400*** –0.032 0.608*** –0.155 –0.146 –0.537* –0.875*** –0.041 –5.945 0.910
1.004 0.592 0.138 0.135 0.157 0.301 0.316 0.064 5.283 0.874
40.065*** –232.340 0.106
42.670*** –231.038 0.113
44.220*** –230.263 0.117
*** p < 0.01, ** p < 0.05, * p < 0.10
cooperative business form, focusing mainly in their results compared to the capitalist conventional firm. Using information referred to an initial sample of 965 Spanish manufacturing plants, our paper looks at the relationship among shopfloor participation, cooperatives and performance. More specifically, we have analyzed the effect of this type of participation on performance; if cooperatives have results significantly different from the conventional capitalist firms, and if participation practices have different effects on firm performance depending on whether the firm is a cooperative or a capitalist firm. Performance measures considered in this paper are of two types. Firstly, we analyze results referred to the different aspects of the operational performance of the plant. Secondly, we examine variables related to motivation and workers behavior. From the empirical estimations made, various results emerge. Firstly, we have found that the adoption of participation practices in the plant leads to a significant improvement in her operational results. Secondly, we must underline that cooperatives have no better or worse results than capitalist firms; being a cooperative per se does not make any difference. This conclusion is applied to all the performance measures considered. Thirdly, we have found
52
J. A. BAYO-MORIONES ET AL.
that the effects of participatory arrangements do not differ between cooperatives and capitalist firms. Our results show that employee participation has positive outcomes in both types of firms. Our results have several implications. The non-economic literature highlights the impact of participation in worker motivation. They argue that participation can improve workers intrinsic motivation and, as a consequence, it brings about a better individual and firm performance. However, our results discard this explanation. We do not observe that the adoption of participation practices is undoubted fully associated with better worker behavior, and therefore a lower absenteeism rate or a smaller incidence of strikes. This makes us think that participation creates positive results for the firm mainly related to a better use of information. Since participation implies an information transfer among the different hierarchical levels of the firm, decisions are taken more efficiently and better operational results are reached. Some other implications can be obtained from the comparison of cooperatives and capitalist firm results. A possible explanation for the nonexistence of significant differences in the case of cooperatives could be the balance of the advantages – mainly in terms of lower monitoring costs – and the disadvantages, that is, the non specialization of ownership and control and, as a consequence, some financing problems and more conservative decisions made by worker-owners. In addition to that, incentives coming from the cooperative ownership structure are diluted as the number of worker-owners increase, since the internalization of effort is reduced because of the free-rider effect and the lower incentives to monitor colleagues. Our sample considers cooperatives with more than fifty workers, so it excludes small cooperatives where the advantages mentioned are presumably stronger. It is also usually argued that one of the advantages of cooperatives is the reduction, or even elimination, of the usual asymmetries of information between management and workers and, as a consequence, the number of conflicts as the case of strikes. Our study shows no differences in the incidence of strikes between cooperatives and capitalist firms. There are several limitations in our paper that should be noted. One is related to the characteristics of the performance measures used. It is true that subjective indicators allow us to analyze a larger number of aspects and questions and, at the same time, it makes possible to compare firms from very different manufacturing industries. However, we must recognize that an analysis based on objective data would have been complementary to the analyses made in this paper. A second limitation comes from the crosssectional character of the information used, which logically can create some doubts about the unambiguity of causal relationships.
Participation, Cooperatives and Performance
53
NOTES 1. The situation of cooperatives in Spain has not received great attention, except the case of Mondragon (see, for example, Bradley & Gelb, 1982; Whyte, 1991, 1999; Cheney, 1999). 2. Examples of reviews of this literature are Miller and Monge (1986), Cotton et al. (1988), Levine and Tyson (1990), Wagner (1994) and Doucouliagos (1995). 3. This assumption does not necessarily mean egalitarianism in the distribution of workers’ income, or direct-democracy in decision-making, or anti-specialization within the workforce. It is this concept of cooperatives, with these specific features, that is most typically used in the literature (see Putterman, 1984). 4. There is also an indirect way of avoiding taxes in the co-operatives. Worker-owned firms can reduce, or avoid, corporate income tax by raising wages and, simultaneously, reducing profits. 5. One specific note must be made at this point. There are two legal forms for cooperatives in Spain. The classical one, usually named ‘cooperativa’, where decisions are taken on a one-member, one-vote basis. There is a second legal form for cooperatives named ‘sociedad laboral’ (labor corporation); where membership is possible for capital and labor owners, members have different grades of control and return rights, but with the legal requirement that workers have the majority of control rights. Some of this second type of cooperatives are sometimes created as a consequence of employee buyouts of financially troubled facilities that might otherwise shut down. In spite of these differences, we will use the term cooperatives to refer to both types since economic problems are the same. 6. In Spain, unlike other countries like Britain or the U.S., it does not have much sense to distinguish between unionized and non-unionized plants. Legislation makes that in a huge majority of the firms with size similar to those of our sample, there is a representative union body for workers. Due to this fact, we have considered more appropriate to divide the sample in three groups according to the level of union influence assessed by the respondent. 7. RESULT = EFFICIENCY, QFINISHED, QPROCESS, PUNCTUALITY, LEADTIME. Therefore, five models will be estimated.
ACKNOWLEDGMENTS The authors would like to thank Fundación BBVA and Spanish Ministry of Education (PB 98-0550) for funding provided and José Enrique Galdón for helpful comments.
REFERENCES Alchian, A. A., & Demsetz, H. (1972). Production, information costs, and economic organization. American Economic Review, 62(5), 777–795. Arthur, J. B. (1992). The link between business strategy and industrial relations systems in American steel minimills. Industrial and Labor Relations Review, 45(3), 488–506.
54
J. A. BAYO-MORIONES ET AL.
Batt, R., & Appelbaum, E. (1995). Worker participation in diverse settings: Does the form affect the outcome, and if so, who benefits? British Journal of Industrial Relations, 33(3), 353–378. Ben-Ner, A., & Jones, D. C. (1995). Employee participation, ownership, and productivity: A theoretical framework. Industrial Relations, 34(4), 532–554. Berg, P. (1999). The effects of High Performance Work Practices on job satisfaction in the U.S. steel industry. Relations Industrielles, 54(1), 111–153. Black, S., & Lynch, L. (1997). How to compete: The impact of workplace practices and information technology on productivity. NBER Working Paper Series no. 6120, Cambridge, Massachusetts. Bonin, J., Jones, D. C., & Putterman, L. (1993). Theoretical and empirical studies of producer cooperatives: will ever the twin meet? Journal of Economic Literature, 31(3), 1290–1320. Bradley, K., & Gelb, A. (1982). The replication and sustainability of the Mondragon experiment. British Journal of Industrial Relations, 20(1), 20–33. Cappelli, P., & Neumark, D. (1999). Do ‘high performance’ work practices improve establishment-level outcomes? NBER Working Paper Series no. 7374, Cambridge, Massachusetts. Cheney, G. (1999). Values at work: Employee participation meets market pressure at Mondragon. London: ILR Press. Corbett, C., & van Wassenhove, L. (1993). Trade-offs? What trade-offs? Competence and competitiveness in manufacturing strategy. California Management Review, 35(4) 107–122. Cotton, J. L., Vollrath, D., Frogatt, K., Lengnick-Hall, M., & Jennings, K. (1988). Employee participation: Diverse forms and different outcomes. Academy of Management Review, 13(1), 8–22. Doucouliagos, C. (1995). Worker participation and productivity in labor-managed and participatory capitalist firms: A meta-analysis. Industrial and Labor Relations Review, 49(1), 58–77. Eriksson, T. (2001). The effects of new work practices: Evidence from employer-employee data. Paper presented at the International Conference on Organizational Design, Management Styles and Firm Performance, University of Bergamo, June. Fernie, S., & Metcalf, D. (1995). Participation, contingent pay, representation and workplace performance: Evidence from Great Britain. British Journal of Industrial Relations, 33(3), 379–415. Geary, J. F. (1999). The new workplace: Change at work in Ireland. International Journal of Human Resource Management, 10(5), 870–890. Gittleman, M., Horrigan, M., & Joyce, M. (1998). ‘Flexible’ workplace practices: Evidence from a nationally representative survey. Industrial and Labor Relations Survey, 52(1), 99–115. Hackman, J., & Oldham, G. R. (1980). Work redesign. Reading, Massachusetts: Addison-Wesley. Hansmann, H. (1990). The viability of worker ownership: an economic perspective on the political structure of the firm. In: M. Aoki, B. Gustafsson & O. Williamson (Eds), The Firm as a Nexus of Treaties (pp. 162–184). London: Sage Publications. Huselid, M. A. (1995). The impact of human resource management practices on turnover, productivity, and corporate financial performance. Academy of Management Journal, 38(3), 635–672. Ichniowski, C., & Shaw, K. (1995). Old dogs and new tricks: determinants of the adoption of productivity-enhancing work practices. Brookings Papers on Economic Activity, special number, pp. 1–65.
Participation, Cooperatives and Performance
55
Ichniowski, C., & Shaw, K. (1999). The effects of human resource management systems on economic performance: An international comparison of U.S. plants and Japanese plants. Management Science, 45(5), 704–721. Ichniowski, C., Shaw, K., & Prennushi, G. (1997). The effects of human resource management practices on productivity: a study of steel finishing lines. American Economic Review, 87(3), 291–313. Jensen, M., & Meckling, W. (1979). Rights and production functions: an application to labormanaged firms and co-determination. Journal of Business, 52(4), 469–506. Jones, D., & Kato, T. (1995). The productivity effects of employee stock ownership plans and bonuses: Evidence from Japanese panel data. American Economic Review, 35(3), 423–455. Jones, D. C., & Pliskin, J. (1991). The effects of worker participation, employee ownership, and profit-sharing on economic performance: A partial review. In: R. Russell & V. Rus (Eds), International Handbook of Participation in Organizations, Vol. II: Ownership and Participation (pp. 43–63). Oxford: Oxford University Press. Kato, T., & Morishima, M. (2002). The productivity effects of participatory employment practices: Evidence from new Japanese panel data. Forthcoming in Industrial Relations. Levine, D. L., & Tyson, L. D. (1990). Participation, productivity and the firm’s environment. In: A. S. Binder (Ed.), Paying for Productivity (pp. 183–243). Washington, D.C.: Brookings Institution. MacDuffie, J. P. (1995). Human resource bundles and manufacturing performance: organizational logic and flexible production systems in the world auto industry. Industrial and Labor Relations Review, 48(2), 197–221. Machin, S., & Stewart, M. (1990). Unions and the financial performance of British private sector establishments. Journal of Applied Econometrics, 5, 327–350. Maddala, G. S. (1983). Limited Dependent and Qualitative Variables in Econometrics. New York: Cambridge University Press. McNabb, R., & Whitfield, R. (1998). The impact of financial participation and employee participation in financial performance. Scottish Journal of Political Economy, 45(2), 171–187. Miller, K. I., & Monge, P. R. (1986). Participation, satisfaction, and productivity: A meta-analytic review. Academy of Management Journal, 29(4), 727–753. Nalbantian, H. R. (1987). Incentive compensation in perspective. In: H. R. Nalbantian (Ed.), Incentives, Cooperation and Risk-Sharing: Economic and Psychological Perspectives on Employment Contracts (pp. 3–43). Totowa, New Jersey: Rowman and Littlefield. Osterman, P. (1994): How common is workplace transformation and who adopts it? Industrial and Labor Relations Review, 47(2), 173–188. Osterman, P. (2000). Work reorganization in an era of restructuring: Trends in diffusion and effects on employee welfare. Industrial and Labor Relations Review, 53(2), 179–196. Pendleton, A., Wilson, N., & Wright, M. (1998). The perception and effects of share ownership: Empirical evidence from employee buy-outs. British Journal of Industrial Relations, 36(1), 99–123. Pil, F. K., & MacDuffie, J. P. (1996). The adoption of high-involvement work practices. Industrial Relations, 35(3), 423–455. Poole, M., & Jenkins, G. (1990). The impact of economic democracy. London: Routledge. Porter, P., & Scully, G. (1987). Economic efficiency in cooperatives. Journal of Law & Economics, 30(October), 489–512.
56
J. A. BAYO-MORIONES ET AL.
Powell, T. C. (1995). Total Quality Management as competitive advantage: A review and empirical study. Strategic Management Journal, 16(1), 15–37. Putterman, L. (1984). On some recent explanations of why capital hires labor. Economic Inquiry, 22(2), 171–187. Roche, W. K. (1999). In search of commitment-oriented human resource management practices and the conditions that sustain them. Journal of Management Studies, 36(5), 653–678. Venkatraman, N., & Ramanujan, V. (1986). Measurement of business performance in strategy research: A comparison of approaches. Academy of Management Review, 11(4), 801–814. Wagner, J. A. (1994). Participation’s effect on performance and satisfaction: A reconsideration on research evidence. Academy of Management Review, 19(2), 312–330. Whyte, W. F. (1990). Learning from Mondragon. In: R. Russell & V. Rus (Eds), Ownership and Participation. International Handbook of Participation in Organization Series, Vol. 2 (pp. 83–102). Oxford: Oxford University Press. Whyte, W. F. (1999). The Mondragon cooperatives in 1976 and 1998. Industrial and Labor Relations Review, 52(3), 478–481. Williamson, O (1975). Market and Hierarchies: Analysis and Antitrust Implications. New York: Free Press. Williamson, O. (1985). The Economic Institutions of Capitalism. New York: Free Press. Wood, S., & Albanese, M. T. (1995). Can we speak of a high commitment management on the shop floor? Journal of Management Studies, 32(2), 215–247.
UNIONS AND PRODUCTIVITY GROWTH: A META-ANALYTIC REVIEW Chris Doucouliagos and Patrice Laroche ABSTRACT The impact of unions on productivity growth has received extensive attention from researchers in industrial relations and economics. Despite a voluminous literature, controversy continues regarding the effect of unions on productivity growth. In this paper, meta-analysis and metaregression analysis is used to quantify the association between unions and productivity growth and to accomplish a quantitative assessment of the empirical literature. The results indicate that the overall association between unions and productivity growth is negative, especially for the U.S. The search for moderator variables revealed that most of the variation in the published results is artificial and can be attributed to specification differences.
1. INTRODUCTION Productivity and productivity growth are important engines for economic development and economic growth. Accordingly, productivity and productivity growth have received extensive attention from researchers, as well as from the policy, business and union communities. Particular attention has been devoted The Determinants of the Incidence and the Effects of Participatory Organizations, Volume 7, pages 57–82. Copyright © 2003 by Elsevier Science Ltd. All rights of reproduction in any form reserved. ISBN: 0-7623-1000-6
57
58
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
to what factors serve as drivers or brakes on productivity growth. A large number of factors have been considered in the empirical literature, including the role of research and development, factor utilization, human capital formation and education levels, energy prices, product market conditions, deregulation and privatization, business cycles, new technologies, innovative workplace practices and industrial relations. The focus of this paper is on the association between unions and productivity growth. This association is explored through a synthesis of the available empirical literature, using meta-analysis and meta-regression analysis. Metaanalysis is a statistical procedure that can be used to quantify associations drawn from an existing body of literature (Wolf, 1986; Hunter & Schmidt, 1990). Meta-analysis is used to identify and quantify patterns, to draw inferences from the diversity of results and to generalize from the results derived from the numerous singular studies (Rosenthal, 1984). In this paper, meta-regression analysis is used to examine the extent to which individual estimates of the association between unions and productivity growth reported in the empirical literature are influenced by the methodological features of the studies themselves, and to detect possible regularities in the association between unions and productivity growth. Meta-analysis of this issue is important for a number of reasons. Despite a voluminous literature, controversy continues regarding the impact of unions on productivity, as well as on other aspects of business such as employment, profitability and investment. Meta-analysis can assist in resolving at least some of the controversy. The available empirical evidence on unions and productivity growth has produced mixed results. It is thus instructive to explore the extent to which the differences in the empirical literature are due to the way the studies have been constructed, rather than differences in the underlying relationship between unions and productivity growth. Moreover, there continues to be significant interest on the links between industrial relations practices and human resources management and economic performance. While unions have declined in importance, they continue to be an important force in many industries and often are an integral part of high performance work systems and participatory governance systems, such as employee participation, profit sharing and employee share ownership. Eaton and Voos (1992, p. 189) consider workplace innovations as a form of productivity bargaining that is: “a union agreement to revise established work practices or work rules in exchange for a better economic settlement”. According to them (1992, p. 201): “unions clearly have the potential for improving the implementation and functioning of participative programs”.
Unions and Productivity Growth
59
2. THEORETICAL CONSIDERATIONS There is now a sizeable theoretical literature exploring both the hypothesized costs and benefits of trade unions on both productivity levels, as well as productivity growth (see for example Addison, 1982, 1985; Addison & Barnett, 1982; Freeman & Medoff, 1984; Kuhn, 1985; Hirsch & Addison, 1986; Turnbull, 1991; Belman, 1992). Freeman (1976) and Freeman and Medoff (1984) argue that unions can raise productivity. They speak of the two faces of unionism: the monopoly face emphasized by orthodox economics as well as the collective voice/institutional response face. Benefits from unions are said to arise from “shock effects” as unions may induce managers to alter methods of production and adopt more efficient personnel policies (Slichter et al., 1960). Unions may also be able to reduce labor turnover and hence help retain the benefits of firm-specific training, as well as providing an incentive for firms to provide more training (see Addison & Barnett, 1982; Freeman, 1976). Unions can improve worker morale and motivation, and unions can improve communications between workers and management (Dworkin & Ahlburg, 1985). These benefits are offset, at least in part, by a number of costs arising from unions: unions raise wages (Lewis, 1963); unions favor restrictive work practices (McKersie & Klein, 1983); unions instigate strike activity (Flaherty, 1987); unions repress R&D spending and investments (Hirsch & Link, 1987); and unions can lead to inefficient personnel practices (Lawler & Mohrman, 1987). For instance, the role of unions in supporting or resisting organizational change at the workplace has been the subject of much debate (Verma & McKersie, 1987; Eaton & Voos, 1992; Kizilos & Reshef, 1997). Some scholars have seen unions as inhibiting the introduction of alternative forms of control and ownership (Guest, 1995), while others see unions as playing a positive role in the successful adoption of such innovative practices (Eaton & Voos, 1992; Kochan & Osterman, 1994).1 Many of the hypothesized costs and benefits associated with unions have implications for productivity growth. For example, the collective voice/ institutional response view of unions implies not only that unions will be associated with higher levels of productivity, but also with higher productivity growth. One channel arises through higher union wage rates inducing the substitution of capital for labor. This can affect total factor productivity growth if, as a response to rising labor costs, firms adopt new vintages of capital which have higher productivity or if they invest in research and development and this leads to technological change. It is possible that the same factors that lead to a rise in productivity levels also reduce productivity growth. For example, Hirsch and Link (1984) note that unions can increase productivity levels by
60
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
providing an effective voice and reducing X-inefficiency, while also decreasing long-run profit expectations, restricting management flexibility and discouraging R&D and thereby restricting productivity growth. As Betts et al. (2001) point out, union wage increases act as a tax on labor which may increase or decrease investment, while union rent-seeking behavior can be a disincentive to investment. Clearly, the net impact of unions on productivity growth is an empirical matter. The focus of this paper is on productivity growth, as this is often more important than productivity levels.2 For example, firms, industries and regions that experience faster rates of productivity growth will catch up and perhaps ultimately surpass the productivity levels of their slower growing rivals. A number of studies have explored the link between unions and productivity growth. These are presented in the following section. While most have found a negative association, many of these do not produce statistically significant results, implying the possibility that unions have no impact on productivity growth at all. To complicate matters further, a number of studies established a positive association between unions and productivity growth. Most of these failed to establish statistically significant results, and hence also imply no impact on productivity growth. Given the differing theoretical arguments and at times differing empirical results, it is instructive to undertake a quantitative evaluation of the available evidence.
3. META-ANALYSIS OF UNIONS AND PRODUCTIVITY GROWTH Traditional literature reviews are conducted in a qualitative and narrative manner, and rely mainly on a subjective assessment of the hypothesis under examination. As a response to this, meta-analysis was developed to facilitate a quantitative research synthesis. There is an extensive reference literature on meta-analysis. See, for example, Cook et al. (1992), Hedges and Olkin (1985), Hunter and Schmidt (1990), Mullen (1989), and Wolf (1986). Meta-analysis commences with the identification of the relevant empirical literature. For this paper, an extensive computer search was conducted for studies written in English exploring union-productivity growth effects, from which we have chosen a total of 29 useable studies.3 All of the studies included in the meta-analysis provide direct measures of the association between unions and productivity growth, with productivity growth as the dependent variable and unionism as part of a set of explanatory variables. All of these studies used multiple regression analysis to explore the association between unions and productivity growth. While the studies differ in many respects, they do share
Unions and Productivity Growth
61
some common characteristics. For example, all use time-series data with at least two years of data and most adopt a production function framework. The main differences between the studies are the measurement of the dependent and independent variables, the type of control variables included and, obviously, the data used. However, it is these differences to which meta-regression analysis can be applied (see Section 4 below). A number of studies were excluded, leaving 29 studies for the meta-analysis. The majority (26) of these studies offer estimates of the union-productivity growth effect, while Williams and Moomaw (1989), Haskel (1991) and Denny (1997) offer estimates only of the change in union-productivity growth effect. Since the focus of the paper is on productivity growth, we do not include any studies exploring the association between unions and productivity levels.4 Second, we excluded those studies that used subjective and perception based measures of productivity growth (e.g. Holzer, 1990; Cooke, 1994). Also excluded are studies using probit models (e.g. Addison et al., 2000), as they are not comparable with the rest of the literature. A third and small group of excluded studies is those that included unionization in a regression, but did not report the relevant regression output (Brown & Medoff, 1978; Wadhwani, 1990; Morishima, 1991; Layard & Nickell, 1989; Tachibanaki & Noda, 2000), or which reported only part of the necessary results (Freeman & Medoff, 1984). The fourth group of excluded studies was those that are not published. Results presented in unpublished material, such as manuscripts and working papers, can change by the time they reach their published form, and hence many of these are less reliable than those found in published material. Finally, some studies purport to be investigating productivity growth but do so only by inference. For example, Oulton (1990) presents some results using productivity growth as the dependent variable, but this is in the form of the acceleration in productivity growth and not productivity growth itself and, hence, this study is not comparable to the other studies. In the more extensive panel data analysis Oulton (1990) uses change in employment as the dependent variable and the coefficients on the explanatory variables are then used to infer impact on productivity growth. Such studies are effectively indistinguishable from employment function studies. Hence, they are not included in the metaanalysis. Any review of a body of evidence runs into the problem of suppressed findings. For example, authors may have included unions in a productivity growth regression, found that this variable was not statistically significant and did not pursue the issue of unionization any further, and also did not report the statistically insignificant findings. Moreover, journals may prefer not to publish statistically insignificant results. However, as can be seen from Table 1,
62
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
this is not an area where insignificant findings are suppressed. Indeed, a significant number of the published studies found no relationship between unions and productivity growth. Nevertheless, if there are other insignificant results in file-drawers, then the omission of these results can impact negatively on the results of meta-analysis, just as they would in any review of the literature. There is little that can be done about this problem and it plagues any empirical literature. While the number of studies is small, it is the available pool of regression based studies from which conclusions can be drawn. It should be noted, however, that these 29 studies represent a total of 25,965 observations across several countries and spanning a number of decades. The next step in the meta-analysis is the calculation of effect sizes from each study. The partial correlation coefficients are the measure of the association between unionization and productivity growth used in this meta-analysis. Partial correlations measure the strength, and importantly the direction, of association between unionization and productivity while holding other factors constant.5 Obviously, the specification used will influence what other factors are held constant and hence influence the magnitude of the partial correlation coefficient. Regression coefficients and elasticities are similarly affected. The drawback with regression coefficients is that they are not standardized, and hence cannot be used in a meta-analytic review. Elasticities can be used. However, not all of the studies provide sufficient information from which elasticities can be calculated, limiting the pool of studies that can be used in the meta-analysis. From the 29 studies, we extract 32 partial correlations. Three studies, Allen (1988), Mitchell et al. (1990) and Bronars et al. (1994) presented separate estimates for different industries. There are two ways of handling this. The coefficients relating to different samples can be averaged using a weighted average procedure and then the weighted average can be used with the sample size being the sum of all the samples. The alternative way is to treat each subsample as a different observation. In terms of meta-analysis, the two procedures lead to an identical overall weighted average effect size. That is, it makes no difference to the summary measures if we disaggregate sub-samples or group them.6 For meta-regression analysis, it is preferable to use the sub-samples, because of degree of freedom considerations and also because the different samples relate to different industries. For the sake of brevity, the sub-samples are not presented in Tables 1 and 2. However, the sub-samples were used in the meta-analysis and meta-regression analysis. One possible criticism of meta-analysis is that statistics like partial correlation coefficients depend on the sample used, the specification chosen
Unions and Productivity Growth
63
and the control variables introduced. Hence, the question is whether a quantitative review of the literature using meta-analysis is useful. The answer is a definite yes. The whole point of research is to sequentially increase knowledge and understanding in an area under investigation. This necessitates an evaluation of several different studies, as we would rarely rely on a single study no matter how excellent it may be. There is a very high demand for literature reviews, as they synthesize an existing empirical literature. Qualitative literature reviews have to rely on the same set of empirical results as meta-analysis. So, any criticism of the primary data used in meta-analysis affects traditional literature reviews to the same degree. The important difference is that the data and specification differences can be modeled in metaregression analysis in order to test and quantify the sensitivity of the different specifications on the estimated association between unions and productivity growth. It is because there are differences between studies that meta-analysis has been developed. The partial correlation coefficient was calculated from each of the published studies.7 Where studies report more than one set of results, we used only the results deemed by the study’s author as the preferred result. Where more than one preferred result is presented by an author, and the results do not relate to different samples, we calculate the average. As noted earlier, some studies provide estimates for more than one industry. That is they use different samples. These can be treated as separate estimates (although they are presented as averages in the tables). The 29 studies are presented in alphabetical order in Tables 1 and 2, together with the country to which the data relate, the sample size used in each study (N), the t-statistic (in some cases the average t-statistic), the partial correlation coefficient (r) and the associated union-productivity growth effect. The studies reporting the association between unionization and productivity growth are presented in Table 1. This is the main group of studies. A smaller set of studies explored the association between changes in unionization levels and productivity growth. These are presented in Table 2. Unionization and changes in unionization are different variables, designed to capture different effects, with potentially different magnitudes and even different directional affects on productivity growth. Hence, it is not valid to group the two sets of studies together. Ideally, studies should explore both the role of the initial level of unionization (or the long term effect of unionization on productivity growth), as well as changes in unionization (or the impact effect on productivity growth). The later is important if the degree of unionization is changing over time. It can be seen from Table 1 that there is a wide range of results, with both positive and negative associations between unionization and productivity
64
Table 1.
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
Empirical Studies Exploring the Association Between Unionization and Productivity Growth (Number of Studies = 26).
Author Allen (1988) Bartel (1994) Bronars et al. (1994) Clark (1984) Clark & Griliches (1984) Davies & Caves (1987) Gregg et al. (1993) Hirsch (1991) Hirsch & Link (1984) Kaufman (1992) Kendrick (1973) Kendrick & Grossman (1980) Kruse (1993) Link (1981) Link (1982) Maki (1983) Mansfield (1980) Mitchell et al. (1990) Nickell et al. (1992) Pantuosco et al. (2001) Phipps & Sheen (1994) Sveikauskas & Sveikauskas (1982) Terleckyj (1980) Terleckyj (1984) Wilson (1995) Zagler (2000)
Country USA USA USA USA USA U.K./USA U.K. USA USA USA USA USA USA USA USA Canada USA USA U.K. USA Australia USA USA USA USA Austria
Sample Size t-statistic 155 155 510 4,681 4,146 61 1,772 4,258 19 81 21 20 5,652 51 97 53 20 886 1,464 816 812 138 20 192 30 20
–1.23 –0.14 –1.76 –0.81 + 1.00 –1.66 + 0.98 –3.11 –2.75 + 1.55 –2.56 –4.21 –0.52 –1.83 –1.94 –2.04 –5.53 + 2.01 –1.80 –3.61 + 8.62 –0.43 –1.843 –2.60 –1.595 + 1.30
r
Productivity Growth Differential
–0.134# –0.012# –0.120* –0.012# + 0.016# –0.226* + 0.024# –0.048*** –0.578** + 0.179# –0.516** –0.772*** –0.007# –0.263* –0.197* –0.281** –0.805*** + 0.099** –0.046* –0.126*** + 0.285*** –0.038# –0.423* –0.469** –0.304# + 0.267#
+ 0.0%m –0.2% –1.8% –0.7% + 0.4%e n.a. + 0.0% n.a. –2.9% + 6.1% –1.8% –1.7%e 0% –5.5%e –4.9%e –1.3% –2.8% + 1.6% n.a. n.a. + 1.2% n.a. –1.5% n.a. –0.1% n.a.
Source: calculated from the primary studies; *, **, *** Statistically significant at the 10, 5, and 1% levels, respectively; # not statistically significant. Superscript e denotes insufficient information to calculate the productivity growth differential from the primary study, but estimates made using supplementary information from other sources. m relates only to the manufacturing sub-sample. n.a. insufficient information to calculate the productivity growth differential.
growth established in the literature. The partial correlations range from –0.81 to + 0.29. Most of the studies (76%) found a negative association. A significant number (40%) of studies found that the association was not statistically significantly different from zero. Only two studies found a positive and statistically significant association between unions and productivity growth.
Unions and Productivity Growth
65
In addition to statistical significance, there is the issue of economic significance. The last column in Tables 1 and 2 presents estimates of the productivity growth differential. These are evaluated at the mean of the sample, using the regression coefficients and primary data.8 For Table 1, the productivity growth effect is evaluated by using data on the level of unionization. In some cases the authors do not present enough information to calculate the productivity growth differential. In other cases, we have estimated the differential from limited data available, or using supplementary information (such as information on unionization presented in Kendrick (1973) and Freeman and Medoff (1979)). The productivity differential presented in column 6 of Table 2 measures the change in productivity growth as a result of changes in unionization. For example, using the sum of the regression coefficients on the unionization variables in Denny (1997) and the decline in union density, gives a positive increase in productivity growth of over 3% as a result of the decline in unionization, for that sample. Note that a negative partial correlation coefficient in Table 2 is associated with a positive impact on productivity growth when the degree of unionization falls. 3.1. Meta-Analysis Results The third step in meta-analysis is the calculation of summary statistics relating to effect sizes. The effect size is a measure of the association between two Table 2. Empirical Studies Exploring the Association Between Change in Unionization and Productivity Growth (number of studies = 9).
Author
Country
Sample Size
Allen (1988) Denny (1997) Gregg et al. (1993) Haskel (1991) Hirsch & Link (1984) Kendrick & Grossman (1980) Maki (1983) Williams & Moomaw (1989) Wilson (1995)
USA U.K. U.K. U.K. USA USA Canada USA USA
155 702 1,772 324 19 20 53 44 30
t-statistic
r
Productivity Growth Differential
–1.55 –2.01 –1.17 –0.24 –2.50 + 4.49 + 2.49 –2.83 + 0.345
–0.178# –0.076** –0.028# –0.014# –0.535** + 0.792*** + 0.341** –0.422*** + 0.070#
+ 0.1%m + 3.4% + 2.6% + 0.6% + 0.4% n.a. + 0.2% n.a. 0.0%
Source: calculated from the primary studies; *, **, *** Statistically significant at the 10, 5, and 1% levels, respectively; # not statistically significant. m relates only to the manufacturing sub-sample. n.a. insufficient information to calculate the productivity growth differential.
66
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
variables. There are several different effect size measures. However, they are all transformations of each other and all aim to represent the results from the empirical literature in a common manner. As noted earlier, in this paper we have chosen the partial correlation as the measure of association. Correlation coefficients are restricted to values between + 1.00 and –1.00. Hence, reported correlations will tend to have non-normal sampling distributions. Accordingly, many researchers prefer to convert correlation coefficients into their associated z scores using the Fisher (1928) z-transformation (see Cooper, 1998 for details).9 For this dataset, it makes no real difference to the reported results and conclusions if partial correlation coefficients themselves are used or if the z-transformation is applied. While Table 1 lists the partial correlation coefficients, we follow the literature and use the z-transformation for the metaanalysis reported in Table 3. The key statistics of interest calculated in meta-analysis are the mean, the weighted mean, a measure of homogeneity of research results and confidence intervals constructed around the mean. These statistics are calculated in order to draw conclusions from the available pool of studies. The summary statistics are presented in Table 3. The results of meta-analysis performed on all of the studies are presented in row 2. The unweighted average and median partial correlation are both negative, –0.19 and –0.08, respectively. The weighted mean is calculated using the sample size derived from each study as weights.10 Table 3.
Meta-Analysis of Partial Correlations of Unions and Productivity Growth. Unweighted Mean r
Median r
Weighted Mean r
Heterogeneity Test
All Studies ( = 26) N = 25,965, X = 895 R = –0.81 to + 0.29
–0.189 (–0.57 to + 0.19)
–0.078
–0.010 (–0.02 to + 0.01)
205.3***
U.S. studies ( = 21) N = 21,844, X = 911 R = –0.81 to + 0.18
–0.239 (–0.66 to + 0.18)
–0.162
–0.021 (–0.04 to –0.01)
119.2***
Change in Unionization ( = 9) N = 3 119, X = 347 R = –0.54 to + 0.79
+ 0.018 (–0.75 to + 0.79)
–0.028
–0.040 (–0.08 to 0.00)
45.48***
N = total sample size. X = average sample size. R = range of partial correlation coefficients. Figures in brackets are 95% confidence intervals. *** denotes heterogeneity is statistically significant at the 10% level, Chi-square test.
Unions and Productivity Growth
67
This statistic is calculated because the studies do not use a common sample size. The weighted mean has a value close to zero. The mean is interpreted as the central tendency of the findings of this group of studies. The 95% confidence intervals are presented in brackets and these incorporate the variance associated with the estimated average partial correlations. These can be used to test the hypothesis that the unionproductivity growth effect is zero, positive or negative. The 95% confidence interval around the weighted mean includes zero. That is, taking all of the available published evidence, the conclusion is that the central tendency of the published results falls around zero or is a very small negative association between unions and productivity growth. This is a valid inference if there is an invariant underlying association between unions and productivity growth. However, the actual relationship between unions and productivity growth is likely to be influenced by many factors, such as industry, firm and plant specific industrial relations practices and their interaction with particular production processes. Since there are differences in the studies, it is instructive to explore whether there is a sub-group of studies that is relatively homogenous. A number of subgroups were compiled but for the sake of brevity only the meta-analysis for those studies using U.S. data is presented.11 The meta-analysis for this group is presented in the third row of Table 3. The meta-analysis of this sub-sample is informative. Once again, there is a negative association between unions and productivity growth. However, the weighted mean is now a larger negative number than for the full sample, but is still close to zero. The main difference is that it is now statistically significant. That is, the 95% confidence interval does not include 0. We can conclude that unions and productivity growth are negatively associated in the U.S., although the effect is very small. Note that this conclusion is different to those drawn by traditional qualitative literature reviews, where it is concluded that the evidence supports neither a negative nor a positive relationship (e.g. Wilson, 1995). From the studies which explored changes in unionization we conclude that there is a negative union association with productivity growth (row 4, Table 3). The weighted mean is negative and statistically significantly different from zero. Although this is a small group of studies, it does present solid evidence that when the degree of unionization is falling, productivity growth is increasing. Also presented in Table 3 is a Chi-square test for heterogeneity (see column 5).12 This is a test of the hypothesis that all the partial correlations are drawn from a group of studies that is homogenous. This test uses the partial
68
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
correlations and the associated sample sizes to compare the variation in the partial correlations with the variation expected if the only source of variation in partial correlations was due to sampling error. In all cases, the test results indicate heterogeneity. That is, the observed variance in partial correlations is statistically significantly different from the variation that would result from sampling error alone. We conclude that the partial correlations reported in the literature do not differ simply because of chance or sampling error. Rather, the partial correlations are drawn from studies where the union-productivity growth effect is moderated in some fashion. Hence, it is necessary to search for factors that lead to heterogeneity between studies (see Section 4 below). With respect to the association between unions and productivity levels, there is some, albeit weak, evidence of the union-productivity effect varying over time (see for example Connerton et al., 1983; Davies & Caves, 1987). This is a pattern that has been observed in some of the productivity growth studies (e.g. Gregg et al., 1993). Accordingly, it is useful to explore whether a similar pattern can be seen with respect to productivity growth. Table 4 reports two different ways of categorizing the empirical literature in order to explore differences in the association over time, for the studies listed in Table 1. For example, the studies can be divided into two rough halves, those published pre1985 (12 studies) and those published from-1985 (14 studies). The unweighted average partial correlation for the first set of studies is –0.36 and this becomes –0.02 for the later set of studies. The weighted average partial correlations are similar across the two periods. Publication date is not necessarily associated with the time over which unions impact on productivity growth. In Table 4, the studies have also been separated according to the time period of the data used, exploring differences in the studies between those using data from-1985 and pre-1985. This is a more meaningful distinction than publication date. It can again be seen clearly that
Table 4. Sub-Group
Published from 1985 Published pre-1985 Used from 1985 data Used pre-1985 data
Unions and Productivity Growth, Time Effects. Number of Studies
Unweighted Partial Correlation
Weighted Partial Correlation
14 12 9 17
–0.02 –0.36 + 0.08 –0.33
–0.01 –0.01 + 0.02 –0.03
Unions and Productivity Growth
69
the association between unions and productivity growth appears to have changed over time. Both the unweighted and weighted averages for studies using data pre-1985 data are negative and become positive for studies using from-1985 data. Importantly, the differences are statistically significant. The 95% confidence interval for the from 1985 data is –0.004 to + 0.038, while the 95% confidence interval for the pre-1985 data is –0.044 to –0.011. These intervals do not overlap. Evidently, the empirical literature from 1985 is finding a substantially smaller association between unions and productivity growth than the pre-1985 literature. Since there is no evidence to indicate that the nature of research from 1985 is different, it appears that the actual association between unions and productivity growth maybe changing over time. Another way of exploring this issue is through cumulative meta-analysis. This involves performing a sequence of meta-analysis (or recursive metaanalysis), commencing with one study and then performing a new meta-analysis each time adding a new study. The results from this can be seen graphically in Figs 1 and 2, where the weighted average partial correlations (with the z-transformations) and confidence intervals around the averages are constructed cumulatively, commencing with the first published study. It can be seen that as each new study is added, the distribution shifts towards the zero line. Classifying studies either by publication date or by the years included generates a noticeable shift in the measured association between unions and productivity growth. We can only speculate about the underlying causes of such changes in the union-productivity effect. There is no evidence to indicate that it is the use of better data sets and different estimation techniques that has led to these
Fig. 1.
Effect Sizes Arranged by Publication Date.
70
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
Fig. 2.
Effect Sizes Arranged by Year.
changes. The alternative explanation is that there is no universal underlying and invariant union-productivity growth effect. Hence, the estimated effects can change as the underlying effect changes. Fundamental shifts in this effect could be due to the changing role of unions over time, changing attitudes of unions and the weakening of unions. The meta-analysis presented in Table 3 is exploratory and does not separate the differences across studies. For example, the group of U.S. studies shares the use of U.S. data, but differs in many other respects. These differences can be explored using meta-regression analysis.
4. META-REGRESSION ANALYSIS The fourth step in meta-analysis is moderator analysis. This involves the identification of the sources of variation between published effect sizes. In this paper, we use meta-regression analysis to identify possible moderator variables. Meta-regression analysis is used to explore the relationship between the estimated union-productivity growth effect derived from each of the published studies and a set of independent variables that are hypothesized to be associated with differences in the estimated union-productivity effects. Metaregression analysis can be used to identify moderator variables, and to explore the impact of specification differences on the estimated union-productivity growth effect. Meta-regression analysis enables a quantitative assessment of the impact of differences in research design, methodology, data and estimation on reported
Unions and Productivity Growth
71
study outcomes. Meta-regression analysis offers a rich framework through which an existing body of empirical literature can be reviewed. The meta-regression equation is simply a regression where the dependent variable is the association of interest and the explanatory variables are a set of study characteristics. The meta-regression equation can be represented as follows: Y = + 1X1 + . . . + kXk + 1K1 + . . . + nKn + u
(1)
where Y is the partial correlation coefficient derived from the ith study, is the constant, the s and s are coefficients to be estimated, Xs are dummy variables representing characteristics associated with the ith study, Ks are the mean values of quantifiable variables, such as the sample size and the span of data, and u is the disturbance term, with usual Gaussian error properties (see Stanley & Garrell, 1998). There are a large number of potential moderator variables. Unfortunately, many of the factors that theory identifies as important cannot be investigated. The meta-regression analysis is restricted to data drawn from the studies themselves. Thus, for example, there are not enough observations to explore the impact of closed shop arrangements, multi-unionism and capital-intensity. Likewise, there are not enough observations to explore variables such as the goodness of fit (as measured by R-squared) of the regression, the number of industries to which the data relate, and the industry classification code (2 digits, 3 digits, etc). The meta-regression analysis presented here relates to those studies that have explored the association between unions and productivity growth. As noted earlier, there are in total 29 studies from which we can extract 32 partial correlations (counting different industries as separate estimates). The metaregression analysis involves only those studies listed in Table 1, and involves 26 studies from which we can extract 29 estimates. With only 29 observations relating to unions and productivity growth it is obviously necessary to limit the number of potential moderating variables. We consider the influence of thirteen potential explanatory variables, which form the base or general regression model. The descriptive statistics for these variables are presented in Appendix A. The variables are defined as follows:
72
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
SIZE:
the sample size used in each study. This controls for differences in sample size, testing if larger samples produce different results. INFLUEN: A dummy variable designed to capture the existence of crossauthor effects, whereby one author publishing in the same area influences the research of one author. Takes the value of 1 if the author(s) acknowledge in the study the comments/suggestions/ assistance of another author included in the pool of studies, 0 = otherwise. MANUF: dummy variable, with 1 = data relate exclusively to manufacturing industry, 0 = otherwise. USA: dummy variable, with 1 = U.S. data was used, 0 = otherwise. R&D: dummy variable, with 1 = if R&D was included as a control variable, 0 = otherwise. DENSITY: dummy variable, with 1 = trade union density or union coverage was used as a measure of union presence, and a 0 if a dummy variable was used for the presence of a union. FIRM: dummy variable, with 1 = firm level/establishment data used, 0 = otherwise, as when industry level data are used. TFP: dummy variable, with 1 = if dependent variable was total factor productivity, 0 = otherwise. FROM85: dummy variable, with 1 = if the study includes data from 1985, 0 = otherwise.13 DU: dummy variable, with 1 = if the study included the change in unionization as a control variable in addition to union presence/ union density levels, 0 = otherwise. SPAN: the number of years over which productivity growth is measured. YEAR: the year of publication. CROSS: dummy variable, with 1 = if the study used cross-sectional data, 0 = otherwise. These are studies that use only two years of data to calculate productivity growth. Some of these require explanation. Union density and union coverage are measures of union presence that are preferable to using a dummy variable, although often problems arise in finding reliable measures of union density. R&D is an important determinant of productivity growth and hence should be included as a control variable in productivity growth studies. A number of the studies defined the dependent variable as growth in multi-factor productivity (output divided by the contributions of both labor and capital), and we wish to test whether this group produces different results.
Unions and Productivity Growth
73
Obviously, in order to calculate productivity growth time series data are needed. However, some studies used panel data, while others used just a time series. Other studies used panel data, but calculated the dependent variable as a percentage change over the time period studied, or calculated the average annual rate of change in productivity (partial or total). These are effectively cross-sectional studies. A priori, the sign on many of the explanatory variables used in the meta-regression analysis is not clear. The meta-regression analysis results are presented in column 2 of Table 5. This is our starting meta-regression model. The dependent variable in all the models is the partial correlation drawn from each of the studies, without any adjustment made for the sample size. The actual partial correlation coefficients are used and not the Fisher z-transformations.14 The coefficients on the explanatory variables together with the associated t-statistics are presented. Also presented in Table 5 are three diagnostic tests conducted on each of the MRA models.15 These are the
Table 5.
Meta-Regression Analysis, Unions and Productivity Growth.
Variable
General model
CONSTANT DENSITY R&D FIRM USA YEAR SIZE/1000 TFP MANUF INFLUEN FROM85 DU SPAN CROSS R2/Adjusted R2 F-Statistic Sample Size Diagnostic Tests
–30.54 (–1.30)# –0.17 (–0.91) 0.08 (0.46) –0.01 (–0.06) –0.16 (–1.03)# 0.02 (1.30)# 0.11 (1.77)* 0.18 (1.18)# 0.10 (0.86) –0.44 (–2.99)** –0.06 (–0.35) –0.17 (–1.63)# –0.01 (–0.88) 0.04 (0.19) 0.80/0.62 4.56*** 29 J: 0.46 W: 0.95 R: 0.26
Reduced model –19.35 (–1.27)# –0.13 (–1.55)# 0.21 (1.83)* – –0.130 (–1.49)# 0.01 (1.28)# 0.09 (3.44)*** – – –0.43 (–3.62)*** – –0.21 (–2.91)*** – – 0.76/0.68 9.45*** 29 J: 0.50 W: 0.93 R: 0.17
t-statistics in parenthesis. *, **, *** denotes coefficient is statistically significant at the 10, 5 and 1% levels, respectively. # denotes t-statistic is greater than 1.
74
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
Jarque-Bera test for normality of the error terms, White’s heteroscedasticity test, and Ramsey’s RESET test (a test for general mispecification). These are denoted as J, W and R, respectively. For the sake of brevity, only the associated prob-values are reported. Heteroscedasticity in particular can be a problem in meta-regression analysis. The estimated models appear to be free from heteroscedasticity and misspecification, although it should be noted that these tests are technically large sample tests, so the results should be interpreted with some caution. Stanley and Jarrell (1998, p. 961) argue that meta-regression analysis may also be vulnerable to autocorrelation, because: “applied research often exhibits trends in methodology.” Accordingly, we also conduct a Breusch-Godfrey test for autocorrelation.16 The next step involved sequentially eliminating statistically insignificant variables from the base model in order to derive a reduced model (see Hendry, 1995), where all the remaining variables have a t-statistic greater than 1. This model is presented in column 3, Table 5. The R-squared indicates that this model explains 76% of the variation in published union-productivity growth effects, leaving 24% of the variation (1–R-squared) in published unionproductivity growth effects due to factors other than those controlled for in the reduced model. This suggests that most, although not all, of the variation in the effect of unions on productivity growth is the product of measurement, data, and specification differences, rather than differences in the underlying unionproductivity growth effect. This is indeed one of the main benefits of conducting meta-analysis and meta-regression analysis. A body of empirical literature with conflicting results can be explored and differences in these results analyzed. Four of the variables have coefficients that are statistically significant while three do not but have coefficients that are consistent with our prior expectations. The negative sign on DU suggests that studies that include a measure of the change in unionization tend to produce higher negative estimates of the union-productivity growth effect. A similar result emerges from DENSITY, although this variable is not statistically significant. The positive coefficient on R&D indicates that studies that control for the effects of R&D find smaller negative union-productivity growth effects, once other study characteristics are controlled for. Since density measures of union presence are preferable, and it is desirable to control for R&D and to allow for both the level of unionization and changes in unionization, these results indicate that failing to include these variables will impact on the estimates of the union-productivity growth effects. The negative coefficient on INFLUEN indicates that studies conducted by authors who received advice/comments/suggestions from other researchers
Unions and Productivity Growth
75
who have also published in this area, tend to find larger negative unionproductivity growth effects. This does not mean that these studies are biased, as these studies may be more accurate. In contrast to the meta-analysis presented earlier, the meta-regression analysis indicates that, once other study characteristics are controlled for, there is no evidence to suggest that the union-productivity growth effect has changed over time – the coefficient on FROM85 is not statistically significant. Surprisingly, neither was SPAN. There is no evidence to suggest that using a longer time period over which union-productivity growth effects is an important factor in explaining differences in published results. The coefficient on YEAR suggests that the more recent studies will produce a lower negative effect, but this too is not statistically significant. The positive coefficient on YEAR is however consistent with the earlier discussion about the apparent change in the partial correlations coefficients reported in the studies over time. Sample size also has a positive coefficient. As sample size rises, ceteris paribus, lower negative associations are established. The FIRM, CROSS and MANUF variables do not appear to be important in explaining differences in the reported union-productivity growth effects. However, the USA variable has a negative coefficient that is consistent with the earlier analysis of differences in the union-productivity growth effect across countries. The meta-regression analysis provides solid evidence of a significant negative association between unions and productivity growth. Estimates of the average union-productivity growth effect can by derived by using the coefficients reported in column 3 of Table 5. For example, if a study was published in 1987, uses the median sample, does not measure union presence using a dummy, and controls for changes in union density and R&D, the estimated partial correlation between unions and productivity growth is –0.02. If U.S. data is used the partial correlation coefficients become –0.07 and –0.15, if the average and median sample sizes are used, respectively.
5. CONCLUDING REMARKS It is often the case that the theoretical literature cannot resolve theoretical debates. The question then is can the theoretical debates be resolved by the empirical literature. In this paper, meta-analysis and meta-regression analysis was applied to the empirical literature exploring the association between unions and productivity growth. The main conclusion that can be drawn from this body of literature is that taking all the studies together and for all time periods, the overall association
76
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
between unions and productivity growth is negative. In particular, the metaanalysis results indicate that for the U.S., the association between initial levels of unionization and productivity growth is negative. Changes in the levels of unionization also have a negative association with productivity growth. There is some evidence that the union-productivity growth relationship has been changing over time and becoming less negative from 1985 onwards, although this is not as yet conclusive. The search for moderator variables revealed that most of the variation in results published in the literature is artificial, and can be attributed to differences in the way the studies are conducted. In particular, differences in sample size, the way the dependent and independent variables are measured and the type of control variables included all influence the reported association between unions and productivity growth. Some of the differences are however due to real economic factors, as is the case with U.S. data. Meta-regression analysis enables a quantitative assessment of the degree to which data and specification differences impact on published results. One of the main benefits of meta-analysis is that it offers assistance on the conduct of future research. The results from the meta-analysis presented in this paper indicate strongly the importance of paying particular attention to the way variables are defined as this does systematically influence estimates of the relationships under investigation. There are several remaining issues. One neglected issue in this area concerns the possibility of selectivity. For example, unions may choose to unionize more productive firms (see Chezum & Garen, 1998). The issue of causality has not received the attention it deserves in this literature. It would also be informative to undertake meta-analysis on other areas that unions are said to impact upon, such as productivity levels, investment and employment. Investment and employment affects are particularly important, as they are two of the channels through which unions affect productivity growth. This line of research will enable comparisons to be made of the impact of unions on these areas and assist in reaching conclusions about the overall affect of unions. Moreover, a meta-analysis of other determinants of productivity growth, such as impact of R&D or alternative forms of stakeholder ownership and control, will assist in identifying the relative contributions of unions to productivity growth.
NOTES 1. According to Kizilos and Reshef (1997, pp. 643–645): “the union’s perception of the innovation as a threat or opportunity may be closely related to the strength of the union”. They suggest that: “unions resist HRM innovations that are perceived as efforts to increase management control and influence at the expense of the union”. In contrast,
Unions and Productivity Growth
77
they go on to argue that: “when unions feel institutionally secure, they will be more likely to support HRM innovation”. 2. Most of the studies have been conducted within the context of a production function, where output is some function of a measure of technological change in disembodied form and capital and labor inputs. The normal form of this function in this literature is that of the Cobb-Douglas specification. Researchers can investigate the impact of unions on the level of productivity – the scale factor or they can look at the rate of change in productivity. The former involves an exploration of a change in the intercept but no change in the slope of an estimated production function. The later allows for changes in the slope but not the intercept. Rarely, do researchers allow for both level and growth effects (intercept and slope). 3. A number of databases were searched, including EconLit, Proquest/ABI Inform and EBSCO. 4. Where studies reported results for both productivity growth and productivity levels, only the former set of results was used. 5. Partial correlations are preferred to simple correlations, as the later can misrepresent the direction and magnitude of association, since they do not control for any other factors. Importantly, simple correlations cannot be derived from the published regression results. 6. For example, the partial correlation coefficient for clerical workers reported in Mitchell et al. (1990) is + 0.068 with a sample size of 491 and the partial correlation coefficient for production workers is + 0.138 with a sample size of 395. These can be combined into a weighed mean of + 0.099 with a sample size of 886. It makes no difference to any subsequent weighted mean calculations if the two sub-samples are used separately, or if the overall weighted mean is used. Note however that the unweighted statistics will be affected. 7. The partial correlation coefficients are not actually reported in any of the 29 studies. However, partial correlations can be calculated from reported regression analysis output. The partial correlation coefficients are calculated by using the tstatistics associated with the measure of trade union presence. The formula used to calculate partial correlations is: t2/(t2 + df ), where t is the t-statistic and df is degrees of freedom. This will always produce a positive number, so it is necessary to convert it to a negative number if the regression coefficient is negative (see Greene, 2000, Ch. 6). 8. Since means differ from study to study, the productivity effects are not strictly comparable. However, they are indicative of the response of productivity growth to unionization. 9. The z-transformation is simply z = 0.51 ln [(1 + r)/(1 r)], where r is the partial correlation coefficient. 10. This means that a study which uses a time series of one firm for 50 years is assigned the same weight as a study that uses two years of data for 25 firms. It is not possible to use a different set of weights, such as employment, as the majority of the studies do not report such data. 11. For example, meta-analysis was conducted only on those studies using total factor productivity as the measure of productivity growth, those that included R&D as a control variable, etc. The results from these are not informative but are available from the authors.
78
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
12. A technical appendix is available from the authors detailing the formulas (weighted mean, confidence intervals and the heterogeneity test) used in the metaanalysis. All the meta-analysis calculations were made using MetaWin version 2.2. 13. 1980 was also used as a cut-off data. This made no difference to the results. 14. The Fisher z-transformation was used to calculate the confidence intervals presented in Table 3. The meta-regression analysis can be performed using the actual partial correlation coefficients. 15. Eviews 3.1 was used for all of the meta-regression analysis, as well as the diagnostic tests. 16. This was tested by ordering the observations by time (publication date), and reestimating the model. The regression coefficients are of course identical to those presented in Table 5. The Breusch-Godfrey test for serial correlation has a prob-value of 0.86. We conclude that there is no evidence of autocorrelation in the estimated metaregression model.
ACKNOWLEDGMENTS This paper benefitted from the comments made by an anonymous referee, as well as those made by Phillip Hone. We apologize for any errors and take sole responsibility for them.
REFERENCES Addison, J. (1982). Are unions good for productivity? Journal of Labor Research, III(2), Spring, 125–138. Addison, J. (1985). What do unions really do? A review article. Journal of Labor Research, VI(2), Spring, 127–146. Addison, J., & Barnett, A. (1982). The impact of unions on productivity. British Journal of Industrial Relations, 20, 145–162. Addison, J., Siebert, S., Wagner, J., & Wei, X. (2000). Worker participation and firm performance, Evidence from Germany and Britain. British Journal of Industrial Relations, 38(1), 7–48. Allen, S. G. (1988). Productivity Levels & Productivity Change Under Unionism. Industrial Relations, 27(1, Winter), 94–113. Bartel, A. P. (1994). Productivity Gains from the Implementation of Employee Training Programs. Industrial Relations, 33(4, October), 411–425. Betts, J., Odgers, C., & Wilson, M. (2001). The Effects of Unions on Research and Development: an empirical analysis using multi-year data. Canadian Journal of Economics, 34(3), 785–806. Belman, D. (1992). Unions, quality of labor relations, and firm performance. In: L. Mishel & P. B. Voos (Eds), Unions and Economic Competitiveness (pp. 47–107). New York: Economic Policy Institute, M.E. Sharpe. Bronars, S., Deere, D., & Tracy, J. (1994). The effect of unions on firm behavior: an empirical analysis using firm-level data. Industrial Relations, 33(4, October), 426–451. Brown C., & Medoff, J. L. (1978). Trade unions in the production process. Journal of Political Economy, 86(3, June), 355–378. Chezum, B., & Garen, J. E. (1998). Are union productivity effects overestimated? Evidence from coal mining. Applied Economics, 30, 913–918.
Unions and Productivity Growth
79
Clark, K. (1984). Unionization and firm performance: the impact on profits, growth and productivity. American Economic Review, 74, December, 893–919. Clark, K., & Griliches, Z. (1984). Productivity growth and R&D at the business level: Results from the PIMS data base. In: Z. Griliches (Ed.), R&D, Patents, and Productivity (pp. 393–416). Chicago, IL: University of Chicago Press. Connerton, M., Freeman, R., & Medoff, J. (1983). Industrial relations and productivity: A study of the U.S. bituminous Coal Industry. Mimeo, Harvard University. Conte, M. A., & Svejnar, J. (1988). Productivity effects of worker participation in management, profit-sharing, worker ownership of assets and unionization in U.S firms. International Journal of Industrial Organization, 6, 139–151. Conte, M. A., & Svejnar, J. (1988). The effects of worker participation in management, profits, and ownership of assets on enterprise performance. In: K. Abraham & R. B. McKersie (Eds), New Developments in the Labor Market (pp. 59–84). Cambridge MA: The MIT Press. Cook, T., Cooper, H., Cordray, D., Hartmann, H., Hedges, L., Light, R., Louis, T., & Mosteller, F. (1992). Meta-Analysis for Explanation: a casebook. New York, The Russel Sage Foundation. Cooke, W. (1994). Employee Participation Programs, Group-Based Incentives, and Company Performance: A Union-Nonunion Comparison. Industrial and Labor Relations Review, 47(4), 594–609. Cooper, H. (1998), Synthesizing Research: A guide for literature reviews (3rd ed.). Thousand Oaks: Sage. Davies, S., & Caves, R. (1987). Britain’s Productivity Gap. Cambridge University Press. Denny, K. (1997). Productivity and trade unions in British manufacturing industry 1973–1985. Applied Economics, 29(10, October), 1403–1409. Dworkin, J. B., & Ahlburg, D. A. (1985). Unions and productivity: A review of the research. In: D. B. Lipsky (Ed.), Advances in Industrial and Labor Relations (pp. 51–68). A Research Annual Series, Vol. 2. Greenwich, Conn. and London: JAI Press. Eaton, A. E., & Voos, P. B. (1992). Unions and contemporary innovations in work organization, compensation, and employee participation. In: L. Mishel & P. B. Voos (Eds), Unions and Economic Competitiveness (pp. 173–211). Armonk, NY: M. E. Sharpe. Fisher, R. A. (1928), Statistical Methods For Research Workers (2nd ed.). London: Oliver and Boyd. Flaherty, S. (1987). Strike activity, worker militancy, and productivity change in manufacturing, 1961–1981. Industrial and Labor Relations Review, 40(4, July), 585–600. Freeman, R. (1976). Individual mobility and union voice in the labor market. American Economic Review, 66, May, 361–368. Freeman, R., & Medoff, J. (1979). New Estimates of Private Sector Unionism in the United States. Industrial and Labor Relations Review, 32, 143–147. Freeman, R., & Medoff, J. (1984). What Do Unions Do? New York: Basic Books. Gregg, P., Machin, S., & Stewart, D. (1993). Signals and cycles: Productivity growth and change in union status in British companies, 1984–89. Economic Journal, 103, 894–907. Greene, W. (2000). Econometric Analysis (4th ed.). London: Prentice Hall International. Guest, D. (1995). Human resource management, trade unions and industrial relations. In: J. Storey (Ed.), New Perspectives on Human Resource Management (pp. 41–55). London: Routledge. Haskel, J. (1991). Imperfect Competition, Work Practices and Productivity Growth. Oxford Bulletin of Economics and Statistics, 5(3), 265–279.
80
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
Hedges, L. V., & Olkin, I. (1985). Statistical Methods for Meta-Analysis. London: Academic Press. Hendry, D. (1995). Dynamic Econometrics. Oxford: Oxford University Press. Hirsch, B. (1991). Labor Unions and the Economic Performance of Firms. Michigan: W. E. Upjohn Institute for Employment Research, Kalamazoo. Hirsch, B., & Addison, J. (1986). The Economic Analysis of Unions: new approaches and evidence. London, Allen & Unwin. Hirsch, B., & Link, N. A. (1984). Unions, Productivity, and Productivity Growth. Journal of Labor Research, 5(1), 29–37. Hirsch, B., & Link, N. A. (1987). Labor Unions Effects on Innovative Activity. Journal of Labor Research, 8, 323–332. Holzer, H. (1990). The determinants of employee productivity and earnings. Industrial Relations, 29(3, Fall), 403–422. Hunter, J. E., & Schmidt, F. L. (1990). Methods of Meta-Analysis: correcting error and bias in research findings. Sage Publications. Kaufman, R. T. (1992). The effect of improshare on productivity. Industrial and Labor Relations Review, 45(2, January), 311–323. Kendrick, J. (1973). Postwar Productivity Trends in the U.S. 1948–1969, NBER, no. 98. Kendrick, J., & Grossman, E. (1980). Productivity in the United States. Baltimore and London: The Johns Hopkins University Press. Kizilos, M., & Reshef, Y. (1997). The effect of workplace unionization on worker responses to HRM innovation. Journal of Labor Research, XVIII(4, Fall), 641–656. Kochan, T., & Osterman, P. (1994). The mutual gain enterprise. Boston, Mass.: Harvard Business School Press. Kruse, D. (1993). Profit-Sharing: Does it make a difference? Upjohn Institute. Kuhn, P. (1985). Union productivity effects and economic efficiency. Journal of Labor Research, VI(3, Summer), 229–247. Layard, R., & Nickell, S. (1989). The Thatcher Miracle? American Economic Review, May, 215–219. Lawler, E. E., & Mohrman, S. (1987). Unions and the new management. Academy of Management Executive, 1, November, 293–300. Lewis, H. G. (1963). Unionism and Relative Wages in the United States: an empirical inquiry. Chicago: University of Chicago Press. Link, A. (1981). Basic research and productivity increase in manufacturing. American Economic Review, 71, December, 1111–1112. Link, A. (1982). Productivity growth, environmental regulations and the composition of R&D. The Bell Journal of Economics, Autumn, 548–554. Maki, D. R. (1983). Trade unions and productivity: Conventional estimates. Relations Industrielles, 38(2), 211–229. Mansfield, E. (1980). Basic research and productivity increase in manufacturing. American Economic Review, 70, December, 863–873. McKersie, R. B., & Klein, J. A. (1983). Productivity: The industrial relations connection. National Productivity Review, 3(1), 26–35. Mitchell, D., Lewin, D., & Lawler, E. (1990). Alternative Pay Systems, Firm Performance, and Productivity. In: A. Blinder (Ed.), Paying for Productivity: A Look at the Evidence (pp. 15–88). Washington D.C.: Brookings Institution. Morishima, M. (1991). Information sharing and firm performance in Japan. Industrial Relations, 30, 37–61.
Unions and Productivity Growth
81
Mullen, B. (1989). Advanced Basic Meta-analysis. Hillsdale NJ: LEA. Nickell, S., Wadhwani, S., & Wall, M. (1992). Productivity growth in U.K. companies: 1975–86. European Economic Review, 36, 1055–1091. Oulton, N. (1990). Labour Productivity in U.K. manufacturing in the 1970s and in the 1980s. National Institute Economic Review, May, No. 132, 71–93. Pantuosco, L., Parker, D., & Stone, G. (2001). The Effect of Unions on Labor Markets and Economic Growth: An Analysis of State Data. Journal of Labor Research, 22(1), 195–205. Phipps, A., & Sheen, J. (1994). Unionisation, industrial relations and labour productivity growth in Australia: a pooled time-series/Cross-section analysis of TFP growth. Labour of Economics and Productivity, 6, 54–77. Rosenthal, R. (1979). The ‘file drawer problem’ and tolerance for null results. Psychological Bulletin, 86, 638–641. Rosenthal, R. (1984). Meta-Analytic Procedures for Social Research. Beverly Hills: Sage Publications. Slichter, S., Healy, J., & Livernash, R. (1960). The Impact of Collective Bargaining on Management. Washington, D.C.: The Brookings Institution. Stanley, T. D., & Jarell, S. (1998). Gender Wage Discrimination Bias? A Meta-Regression Analysis. The Journal of Human Resources, 33, 947–973. Sveikauskas, C., & Sveikauskas, L. (1982). Industry characteristics and productivity growth. Southern Economic Journal, 48, January, 769–774. Tachibanaki, T., & Noda, T. (2000). The Economic Effects of Trade Unions in Japan. London: MacMillan. Terleckyj, N. (1980). What do R&D numbers tell us about technological change? American Economic Association, May, 55–61. Terleckyj, N. (1984). Comment on R&D and productivity Growth at industry level. In: Z. Griliches (Ed.), R&D, Patents, and Productivity (pp. 496–501). The University of Chicago Press. Turnbull, P. (1991). Trade unions and productivity: opening the Harvard “Black Boxes”. Journal of Labor Research, XII(2, Spring), 135–150. Verma, A., & McKersie, R. B. (1987). Employee involvement: The implications of noninvolvement by unions. Industrial and Labor Relations Review, 40, July, 160–173. Wadhwani, S. (1990). The effect of unions on productivity growth, investment and employment: a report of some recent work. British Journal of Industrial Relations, 28(3, November), 371–383. Williams, M., & Moomaw, R. L. (1989). Capital and Labour Efficiencies: A Regional Analysis. Urban Studies, 26, 573–585. Wilson, K. (1995), The Impact of Unions on United States Economy-wide Productivity. New York and London: Garland. Wolf, F. M. (1986). Meta-analysis: Quantitative Methods for Research Synthesis. Sage University Paper no. 59. Zagler, M. (2000). On the Sources of High Growth Experience in Austria. International Advances in Economic Research, 6(2), 210–220.
82
CHRIS DOUCOULIAGOS AND PATRICE LAROCHE
APPENDIX A: DESCRIPTIVE STATISTICS FOR VARIABLES USED IN META-REGRESSION ANALYSIS Variable
Mean
Standard Deviation
SIZE INFLUEN MANUF USA R&D DENSITY FIRM TFP FROM85 DU SPAN YEAR CROSS
898 0.43 0.57 0.82 0.39 0.82 0.46 0.50 0.32 0.25 13.11 1987 0.643
1646 0.50 0.50 0.39 0.50 0.39 0.51 0.51 0.47 0.44 9.99 6.26 0.48
DO ESOPS MOTIVATE EMPLOYEES? WORKER EFFORT, MONITORING AND PARTICIPATION IN EMPLOYEE-OWNED STOCK OWNERSHIP PLANS Agustin J. Ros ABSTRACT This study uses data from a 90% employee owned ESOP and six comparable private firms in the same region to investigate the effects of employee ownership and participation on effort, shirking and horizontal monitoring. Participation turns out to be a strong determinant of effort and horizontal monitoring. It was found that, on average, employees who perceive their participation level to be higher will, on average, exert greater effort, have an incentive to horizontally monitor and engage in horizontal monitoring. It is also found that participation does not have to be coupled with employee ownership to elicit motivational effects. Firms without employee ownership may achieve efficiency gains from increasing workers’ perception of their level of participation in the firm.
The Determinants of the Incidence and the Effects of Participatory Organizations, Volume 7, pages 83–103. Copyright © 2003 by Elsevier Science Ltd. All rights of reproduction in any form reserved. ISBN: 0-7623-1000-6
83
84
AGUSTIN J. ROS
1. INTRODUCTION Employee stock ownership plans (ESOPs) are employee benefit plans operating through a trust that accepts tax-deductible contributions from the company in order to accumulate company stock. The shares are then allocated to accounts for individual participants thus allowing employees the opportunity to accumulate an equity stake in the company. Since their inception in the U.S. in the early 1970s, ESOPs have experienced significant growth both in the total number of plans and the number of plan participants covered in such plans. In 1975 there were approximately 1,600 ESOPs covering a quarter of a million employees; five years later the figures were approximately 4,000 and three million employees, respectively (NCEO, 2000). While the growth of ESOPs has slowed somewhat in the 1990s, by 2000, there were approximately 11,500 ESOPs covering roughly 8.5 million employees. There are two main reasons why companies decide to implement ESOPs. First, the contributions that companies make to the ESOP are tax deductible and according to the NCEO, the ESOP is the most tax-advantaged mechanism for companies to share ownership with employees.1 Second, motivated by poor productivity performance from the early 1970s to the mid 1990s, firms began experimenting with innovative workplace programs in the hopes of improving productivity. Quality of worklife programs, employee involvement teams, gainsharing and employee ownership were being viewed as ways to align the interests of labor and capital, elicit greater worker effort, reduce monitoring costs and improve overall company performance. This paper is primarily devoted to examining the latter objective of ESOPS, namely, the ability of ESOPs to generate economic efficiency gains. The theoretical and empirical literature on the impact of worker ownership is voluminous.2 For purposes of this paper I focus on the area of the debate in the theoretical literature that suggests that employee-owned firms are incompatible with optimal economic performance because of employee behavior and motivation. Specifically, I investigate the validity of the proposition that employee-owned firms may suffer from the classic “free-rider” problem where the individual employee bears the cost of increased worker effort but all workers benefit from increased output, (Alchian & Demsetz, 1972; Williamson, 1975, 1980). Under these conditions, economic efficiency requires that some agent be given the rights to be the residual claimant in order to arrive at the optimal level of monitoring. In this paper I test for evidence of the aforementioned criticism by comparing worker behavior and motivation in an ESOP and in comparable privately-owned firms. I attempt to answer the following question: given the
Do ESOPs Motivate Employees?
85
data set used, does employee ownership improve firm performance by making employees more productive and motivated, and more interested in the overall performance of the firm, or is it incompatible with efficient economic outcomes?
2. THE EMPLOYEE-OWNED FIRM The data for this study were collected in 1993–1994. The ESOP is a manufacturing firm producing pneumatic air cylinders and is located in northern Illinois.3 At the time of the study, the ESOP was 90% employee owned and was employing approximately 300 employees.4 The questionnaires were distributed to all non-managerial personnel and were returned by 144 employees for a response rate of 48%. Of the 144 returned responses, 28 were from managerial or professional personnel, thus reducing the sample from the ESOP to 116.5 As will be shown below, on average the employees in the ESOP perceive their participation level to be higher than employees in the privately owned firms. There is an ESOP committee representing all three shifts which acts as a liaison between employees and management and as a source of information about the ESOP, benefits, corporate finances and other issues important to employees. And a non-management employee selected by the workers sits on the board of directors. Team production is the norm at the firm and employees are split into work teams on the basis of common functions. Team meetings are common and the minutes from all meetings are posted so that employees from different teams can make suggestions.
3. THE PRIVATE FIRMS Approximately fifty private firms in northern Illinois with the same four digit SIC code were contacted and asked to participate in the study. Six private firms accepted, with two accepting only on condition of being able to see the results for their employees.6 The six firms which participated in the study are privately owned, ranging in size from 30 employees to over 200. Table 1 below presents some descriptive information on all the firms that participated in the study, including the employee owned firm. “Resp” stands for respondents. Surveys were distributed to non-managerial and non-professional employees only. Emp stands for the total number of employees in the firm (including professional and managerial); as a result the survey response rate is likely biased down because the total number of non-managerial employees is less than indicated by emp.
86
AGUSTIN J. ROS
Table 1. Company Characteristics. Company
Emp.
# Resp.
% Resp
Size*
Benefit Plan/Bonus
ESOP** Firm 1** Firm 2** Firm 3** Firm 4 Firm 5*** Firm 6***
300 200 44 30 55 200 150
116 68 11 4 4 9 22
39% 34% 25% 13% 7% 45% 73%
141 145 140 17 85 267 166
Cashed based profit sharing 401(k) Profit sharing plan Profit sharing plan, Bonus 401(k) Profit sharing plan and year end bonus Gainsharing
* Plant size in square ft (000). ** A lottery was conducted in these firms to increase response rate. *** For these firms, management distributed to a select group of workers, for firm 5, 20 questionnaires were given to management (for distribution) and for firm 6, 30 were given.
As mentioned in the table, for Firms 5 and 6, management did not permit the distribution to all employees but rather agreed to distribute to a few employees. Management indicated that they would distribute the questionnaires randomly, however, there was no way of being able to confirm whether this was actually done. However, I ran the regressions described below, dropping the observations containing Firms 5 and 6 and found that the conclusions regarding ESOP and participation described below were not altered. Some of the private firms had more extensive arrangements for employee participation than others. For example, Firm 1, which accounted for 57% of the private firms respondents, had monthly meetings with the President to discuss any matter that might be of interest. The meetings were structured in such a way so that each employee met at least once a year with the President. In addition, some employees were members of the firm’s Safety Inspection Committee which has influence on safety, production and technology issues. Employees in firm 6, which accounted for 19% of the private firms respondents and which had a gainsharing program, had team meetings (3 teams) and shift meetings. Some employees were members of the Gainsharing Committee which decides on targets and rewards. In addition, there was a “What’s on your mind” form which was sent to the Vice-President. Finally, Firm 5, which accounted for 8% of the respondents, had a Better Way Committee made up of management and labor. In addition, a manufacturing meeting was held once a week and focused on production problems, problem solving or any other matter that may be of interest to the worker. The objective difference in the degree of participation between the ESOP and the private firms is difficult to measure. One would have to say that with
Do ESOPs Motivate Employees?
87
the exception of Firm 6, which has a gainsharing program, participation is somewhat higher in the ESOP than in the private firms. However, individuals differ in their perception of identical realities. An employee in the ESOP, where there is a higher objective level of participation, may feel that their level of participation is less than an employee in Firm 5 where there is a lower objective level of participation. Arguably, it is the perception which will motivate the employee and for this reason I created a variable measuring employees’ perception of their level of participation. I examined the within firm variability of participation in the six private firms, compared to the ESOP. The mean and standard deviation, respectively for the firms were as follows: ESOP (1.75 & 0.75); Firm 1 (1.36 & 0.80); Firm 2 (1.58 & 0.66); Firm 3 (1.6 & 0.89); Firm 4 (1.4 & 1.14); Firm 5 (1.7 & 1.05); Firm 6 (1.7 & 0.86). Based on this, with the exception of Firms 4 and 5, which account for only 13 of the 118 private responses, there is not much within firm variability for the participation variable; perhaps a bit more than the ESOP. What this seems to indicate is that a subjective measure may be an effective alternative to measurements on actual arrangements. An important observation that was left out of the analysis was union status of the private firms. It would have been preferable to obtain this information directly from the firm. Union membership may significantly impact work arrangements and it would have been good to know what kind of impacts the union had. However, the data in this study measure individual perceptions about things like participation, effort, horizontal monitoring, etc. While union status clearly impacts the work environment of the employee, the survey is able to directly pick up how the worker perceives her environment, which is arguably the important motivator of worker behavior.
4. DATA ANALYSIS 4.A. Dependent Variables The dependent variables used in this study measure effort, shirking, the incentive to horizontally monitor and whether employees have horizontally monitored. Appendix 1 at the end of the paper presents the questions and statements used in the survey. I used several survey responses to develop two dependent variables measuring worker effort used in the econometric analysis below. The first, question 23, earw (effort in addition to what is required . . .) was used by Bilby (1988) and appears in The Michigan Quality of Employment Survey (1977) and Time Use Longitudinal Panel Study 1975–1981. The four ordered responses ranged from none to a lot. The second dependent variable
88
AGUSTIN J. ROS
measuring worker effort is eff. Eff was used in Time Use Longitudinal Panel Study 1975–1981. This variable was created from the responses to question 9 and was used, in altered form, by Stafford and Duncan in Ehrenberg (1980). This second variable proxying effort was based on responses to the question asking respondents to indicate how much effort they put into a typical hour of work (ew1) and typical hour of watching T.V (etv). A dependent variable was then created by the formula: eff = (ew1 etv)/10.
(1) 7
The measure for T.V. was used as a control mechanism. The dependent variable measuring shirking, shk, was used by Hammermesh (1990) and Stafford and Duncan in Ehrenberg (1980). After asking how much time is spent at lunch and other regularly scheduled breaks, the survey asks question 8, “Thinking about the rest of your time at work, about how much time do you usually spend on things not related to the work that you do – like talking to friends, doing personal business or just relaxing.” This response was then converted into a weekly measure.8 Finally, I created two dependent variables that examine horizontal monitoring by employees. A dependent variable measuring incentive to horizontally monitor was created from question 17, “If I saw a co-worker slack off, I would complain.” (whm). The four ordered responses ranged from strongly disagree to strongly agree, which I coded as 0 to 3 with 3 indicating strong agreement and 0 indicating strong disagreement. The last dependent variable attempts to determine if employees have horizontally monitored by asking respondents, question 24, to answer yes or no to the following, “Have you ever done or said anything to a co-worker when you saw that worker slack off”? (hm). 4.B. Independent Variables A dummy variable (ESOP) was created measuring the effects of being a member of an ESOP and number of independent variables controlling for personal characteristics were added. This includes months worked (mw), age, sex (female = 1, male = 0), number of children (c), education in years (ed) and married and spouse works (msw). Additional independent variables were added which are hypothesized to affect our dependent variables, including employee perceived level of participation (p), employee perception of whether she is being treated fairly (tf ), whether the employee receives a bonus (bon) and the employee’s wage level (w). Finally, three additional variables were added which control for team production and compensation. The first variable controls for team production and measures employee perception of own effort
Do ESOPs Motivate Employees?
89
being affected by colleagues effort (oeace). The second variable controls for team compensation and measures employee perception of own pay being affected by colleagues effort (opace). The last variable controls for individual compensation and measures employee perception of own pay being affected by own effort (opaoe). The variable used for participation is different from those used in previous studies. Instead of creating an objective measure for participation I created a variable that measures workers’ perception of their participation level. This variable should in theory be correlated with some objective “true” measure of participation. However, as mentioned above, designing an objective measure of participation that can be applied across firms is difficult. Regardless of the objective measure of participation, workers who are employed in a work environment where there is a high degree of participation but who perceive their participation level to be low may, all else equal, be less motivated than those who perceived their level of participation to be higher. Workers in the same firm with the same objective measure of participation may perceive their level of participation differently and differ in the manner in which this impacts their behavior. In addition, participation levels may not be the same for all employees. Some employees may work in a team or an area where there is marginally less participation. This variable, and the variables measuring team compensation, production and individual compensation, attempt to control for these possibilities.9 4.C. The Model The dependent variables measuring would horizontally monitor (whm) and effort in addition to what is required at work (earw) were ordered responses ranging from strongly disagree (0) to strongly agree (3). Therefore, maximum likelihood ordered probit models were estimated for these dependent variables. The dependent variables measuring effort at work (eff ) and shirking/week (shk) were continuous variables but were truncated at 0. Therefore, Tobit models were estimated. Finally, the dependent variable measuring if the employee had ever horizontally monitored was dichotomous in nature and a univariate probit model was estimated. The dependent variable is assumed to depend on: a dummy variable ESOP (where ESOP = 1 if the employee is employed in the ESOP and 0 otherwise), a vector of personal characteristics variables (such as tenure, age, sex) and a vector of variables which may affect the dependent variables (such as wage, if bonus is received, team compensation, team production, individuals perception of the degree of their participation and treatment by the firm).
90
AGUSTIN J. ROS
4.D. Results Table 2 below presents the sample means of the five dependent variables separately for the ESOP firm and for the six private firms. The data indicate that there is more effort and less shirking at the ESOP as well as more horizontal monitoring. Of course, our econometric analysis below will examine whether there is a statistically significant difference between the ESOP and the private firms. Table 3 below contains descriptive statistics by type of organization. It presents some interesting insights. The perceived level of participation on the part of the employees seems to be significantly higher in the ESOP than in the 6 private firms. There is nothing controversial about this finding; employee owned firms tend to be more participatory than conventional private firms. In addition, employees’ perception of how they are being treated by the firm seems to be significantly higher in the ESOP as well. This is of particular interest because I consider this variable as a proxy for Akerlof’s “partial gift exchange,” theory where workers respond positively when the firm treats them fairly.10 Other interesting findings from Table 3 involve the variables controlling for team production and group compensation. The means for these variables, (oeace, opace), were almost identical. This is explained by the fact that the companies studied had the same SIC code and were producing nearly the same products and were of very similar plant size. The production process and possibly the technology used were very similar for all firms. Thus there is strong evidence to suggest that any difference in employee behavior and motivation is not due to the production process or the technology used. Other important differences seem to include the average age of the workforce, months worked (tenure) and absences. On average, the employees in the ESOP are younger, have less tenure and have more absences than their counterparts in the
Table 2. Means Comparison for Dependent Variables. Variable
ESOP
Private
EFF (effort at work) SHK (shirking/week) WHM (would hor. mon.) EARW (effort in add. . .) HM (have hor. mon.)
0.54 86.2 1.8 2.5 0.72
0.50 89.6 1.5 2.4 0.54
Do ESOPs Motivate Employees?
Table 3.
Descriptive Statistics by Organization.
Variable Mean mw (months worked) hw (hours worked/week) ab (absences during year) lb (lunch brk./day, min) ob (other brk./day, min) shk (shirking/wk., min) ew1 (effort at work) etv (effort watch. tv) eff (ew1–etv)/10 jrwh (job req. hard work) ijc (imp. of job content) ihj (imp. of having job) hcw (help co-workers) oeace (own effort affected by co-worker’s effort; team production) opace (own pay affected by co-worker’s effort; team compensation) opaoe (own pay affected by own effort; ind. comp.) whm (would horizontally monitor) p (participate in decs.) bon (bonus received) tf (am treated fairly) erw (effort required) earw (effort in addition to what is required) hm (have horizontally monitored) w (wage) ESOP (part of ESOP plan) Shares (shares in account) vested (% of shares vested) oj (have other job) age sex (female = 1, male = 0) c (# of children) ed (education in years) msw (married and spouse works = 1 otherwise = 0)
91
ESOP SD
n
Mean
68.1 43.1 3.9 23.2 3.8 86.2 8.6 3.2 0.54 2.2 1.3 1.5 2.5
54.8 5.6 9.2 8.2 10.8 88.1 1.5 2.4 0.29 0.70 0.81 0.75 0.52
116 116 114 116 116 116 116 116 116 116 116 116 116
107.2 42.3 3.2 28.7 10.9 89.6 8.5 3.5 0.50 2.2 1.2 1.6 2.5
1.7
0.96
115
1.5
1.0
2.0
Private SD
n
95.5 5.1 6.3 11.8 7.5 135.8 1.4 2.6 0.28 0.60 0.73 0.78 0.52
124 124 120 122 124 118 123 123 123 124 122 123 124
1.6
0.94
124
116
1.5
0.94
124
0.88
116
1.9
0.89
124
1.8 1.7 0.97 2.0 2.8
0.75 0.74 0.16 0.65 0.45
116 115 116 115 116
1.5 1.2 0.25 1.7 2.7
0.84 0.84 0.44 0.80 0.58
124 123 123 122 123
2.5
0.55
116
2.4
0.72
123
0.72 359.2 0.89 888.7 63.2 0.41 34.6 0.29 1.7 12.6
0.45 98.9 0.32 682.8 35.0 0.49 10.0 0.46 1.6 1.2
115 96 115 47 65 110 114 114 109 113
0.54 426.9
0.50 148.0
123 110
0.24 37 0.14 1.8 12.0
0.43 10.3 0.35 1.7 1.1
121 122 123 119 123
0.50
0.50
115
0.40
0.49
124
92
AGUSTIN J. ROS
six private firms. The former is important because of the positive theoretical relationship between motivational effects of employee benefit plans and age. For young workers who are not planning out their retirement days, belonging to an ESOP may not be as important to a middle aged worker with a family. On the margin, this may motivate some individuals more than others. Table 4 below contains results when effort and shirking are used as the dependent variables. As can be seen, there is no evidence that there are significant differences between employees in the ESOP and the private firms in worker effort while very weak evidence of less shirking in the ESOP. The most important variables explaining earw are the perceived level of participation by the employee and the sex of the employee. Workers who perceive their participation level to be higher will, on average, state they exert greater effort. As mentioned above, this is consistent with the many studies that also found benefits from greater participation. This result indicates that based on this data set, worker ownership per se does not lead to efficiency gains.11 In addition, female workers, on average, state they exert greater effort than males. This finding supports Bielby’s (1988) finding that women allocate more effort to work than do men. It also contradicts Becker’s theoretical proposition that women exert less effort on the job than do men because of women’s effort exertion in household activities. Therefore, an interaction variable, sex*c, (c = number of children) was created where the variable takes on the value 0 for males and the number of children reported by females. Higher values of sex*c indicates, all else equal, a higher likelihood that the female has greater household responsibilities. The coefficient for sex*c was negative indicating that the effort that females exert on the job declines with the number of children, however it was not statistically significant. When using eff as the dependent variable measuring worker effort on the job, one finds that there is no evidence that employees in the ESOP exert greater effort than employees in the six private firms. Evidence is again found that females exert greater effort at work than do males.12 Surprisingly, on average, workers who received a bonus exerted less effort on the job than workers who did not. This contradicts much of the literature predicting a positive relationship between bonuses and employee behavior.13 However, according to Blinder (1990), group compensation schemes, including bonuses, work better when there is participation. Therefore, I created an interaction variable, bon*p, which measures whether higher levels of participation for those who receive a bonus effects effort on the job. The addition of this independent variable made the bon coefficient positive but statistically insignificant. The bon*p variable was positively correlated with effort on the job, however, it was not statistically significant. The only other significant variable is the married and spouse works
Do ESOPs Motivate Employees?
93
Table 4. Determinants of Work Effort and Shirking, earw (Effort in Addition to What is Required) Eff ((Effort at Work-Effort Watching T.V./10)) and Shirking (time Spent on Things Not Related to Work). Variable
esop mw age ed sex msw c oj p tf bon w oeace opace opaoe constants a1 a2 a3 Chi-Sq. (15) Log Like. N
EARW (Ordered Probit)
EFF (Tobit)
SHK (Tobit)
0.30 (0.85) –0.001 (–0.55) 0.02 (1.51) –0.11 (–1.13) 0.55** (2.00) –0.07 (–0.29) 0.04 (0.61) –0.02 (–0.09) 0.26* (1.94) –0.11 (–0.70) –0.35 (–1.08) 0.001 (0.54) –0.07 (–0.67) 0.06 (0.53) 0.18 (1.48)
0.06 (0.90) –0.0004 (–1.16) –0.001 (–0.02) 0.004 (0.84) 0.13** (2.49) 0.09** (1.97) 0.002 (0.16) 0.06 (1.46) 0.03 (1.12) 0.02 (0.57) –0.12 (–1.83) 0.0002 (1.01) –0.02 (–0.80) 0.02 (0.88) –0.04 (–1.43)
–54.98 (–1.58) 0.25 (1.41) –1.81 (–1.10) –2.10 (–0.23) –6.34 (–0.23) 18.75 (0.79) 4.55 (0.58) 43.31* (1.81) –3.29 (–0.23) –8.04 (–0.46) 86.14*** (2.57) 0.10 (0.91) 12.42 (1.10) –17.78 (–1.49) 1.28 (0.10)
2.73 (2.27) 0.93 (2.01) 0.18 (1.48) 25.42 146.27 186
0.33 (1.35)
22.21 –46.31 186
78.05 (0.63)
24.88* –965.01 182
T-statistics are presented in parentheses. * Statistically significant at the 10% level; ** the 5% level; *** the 2% level.
94
AGUSTIN J. ROS
dummy variable which is positively associated with worker effort on the job. Those employees, who are married and whose spouse also works-thus making it more likely that the individual has greater household responsibilities-on average, exert greater effort on the job. This is a surprising result and contrary to the expected. When worker shirking, (shk) is used as the dependent variable, there is no evidence indicating that there are significant differences in the amount of shirking by employees in the ESOP and the six private firms. The coefficients for the ESOP dummy are negative, indicating a negative relationship between ESOP employees and shirking, however it is not statistically significant. The other important variable explaining worker shirking was whether employees received a bonus and whether they had another job. On average, employees who had another job were more likely to shirk than those who do not. One can think of effort as being a variable with a maximum fixed at a certain level. On any given day, an individual worker can not surpass this maximum amount and must allocate their effort according to the activities of the day. Having another job increases the likelihood that adjustment in effort must be made in all other activities including work. In addition, on average, workers who received a bonus shirked more than workers who did not. I included an interaction variable, bon*p to measure whether higher levels of participation for those who receive a bonus affects shirking. The variable was negatively correlated with shirking, however, it was not statistically significant. The addition of this variable had no effect on the bon coefficient. Finally, whether an employee perceives himself or herself to be being treated fairly does not seem to have an impact on employee behavior and motivation. This is counter to the Akerlof partial gift exchange hypothesis mentioned above. The last two dependent variables deal with horizontal monitoring. Table 5 below presents determinants of whether employees would horizontally monitor, (whm) and whether employees have engaged in horizontal monitoring (hm). The results indicated that horizontal monitoring is not significantly correlated with whether an employee works in the ESOP or one of the six private firms. When WHM is used as the dependent variable, the variables that are significant are employees’ perceived level of participation, whether they feel that their pay is influenced by the work of their co-workers and age of the employee. Employees who perceive their level of participation to be high and their pay to be influenced by the work of their co-worker are more likely, on average, to state they would horizontally monitor their co-workers.
Do ESOPs Motivate Employees?
95
Table 5. Determinants of Horizontal Monitoring, whm (would Complain if I Saw Co-worker Slacking Off) hm (Said Something to Worker Slacking Off). Variable
esop mw age ed sex msw c oj p tf bon w oeace opace opaoe constants a1 a2 a3 Chi-sq. (15) Log Lik. N
WHM (ordered probit) –0.075 (–0.26) 0.001 (0.59) –0.03** (–2.12) –0.08 (–0.85) 0.24 (1.00) 0.25 (1.20) 0.06 (1.14) 0.20 (0.96) 0.49**** (4.26) –0.16 (–1.00) 0.20 (0.77) 0.001 (0.77) 0.07 (0.82) 0.30**** (3.36) –0.07 (–0.57) 2.01 (1.64) 1.48 (8.73) 2.96 (14.85) 47.99 –199.83 187
HM (probit) 0.06 (0.16) –0.16 (–0.88) –0.31** (–2.24) 0.09 (1.02) –0.33 (–1.19) 0.37 (1.52) 0.17** (2.13) –0.03 (–0.13) 0.40**** (2.66) –0.16 (–0.89) 0.43 (1.31) 0.003*** (2.39) –0.09 (–0.80) 0.10 (0.78) 0.25 (0.18) –1.63 (–1.3)
44.98**** –100.17 186
T-statistics are presented in parentheses. * Statistically significant at the 10% level; ** the 5% level; *** the 2% level; **** the 1% level.
96
AGUSTIN J. ROS
When “have horizontally monitored their co-workers,” (hm) is used as the dependent variable, employees’ perceived level of participation, wage, age of the employee and number of children are the most important factors affecting horizontal monitoring (hm). Employees’ perceived level of participation is, once again, positively correlated with hm. Workers who perceive their level of participation to be high are, on average, more likely to have horizontally monitored. In addition, age is significant and negative in both equations. A possible explanation for this is that there is a lag, at times quite long, between horizontal monitoring efforts and reward, in the sense of greater productivity and higher pay. Older workers may be less likely to complain because the discounted future marginal benefits (i.e. future pay increases) are significantly less than the present current cost of complaining. While this is the case for all workers, older workers are more likely to realize this especially as they near retirement age and current monitoring effort may never be recouped. 4.E. Summary of Results Table 6 below summarizes the results for the five equations. Below each column are listed the variables which were statistically significant in determining the dependent variables. As can be seen, employees’ perceived level of participation turned out to be significant in three out of the five estimated equations. Higher levels of perceived participation on the part of the employees leads to, on average, greater effort, a higher incentive to horizontally monitor and actually engaging in horizontal monitoring. Females, on average, exert greater effort on the job than do males. Even when the interaction variable sex*c was placed in the equation, there was still evidence that females exert greater effort than males and that household responsibilities do not adversely affect effort on the job. Those employees who received a bonus were, on average, more likely to exert less effort on the job and shirk more frequently than those employees who
Table 6. Summary of Results. EARW sex ( + ) 5% p ( + ) 10%
EFF sex ( + ) 5% msw ( + ) 5% bon ( ) 10%
SHK bon ( + ) 2% oj ( + ) 10%
WHM age ( ) 5% p ( + ) 1% opace ( + ) 1%
HM age ( ) 5% c ( + ) 5% p ( + ) 1% w ( + ) 2%
Do ESOPs Motivate Employees?
97
did not receive a bonus. Bonuses or team compensation may work better when there is participation. Therefore, an interaction variable, bon*p was created. The results indicated that those individuals who received a bonus and who perceived their level of participation to be higher, exerted greater effort on the job and shirked less. However, none of the coefficients was statistically significant. Other variables which turned out to be significant in explaining the dependent variable are age, employees’ perception that their pay is affected by their co-workers’ effort (opace), married and spouse work (msw) and number of children (c). Older workers, on average, are less likely to have an incentive to horizontally monitor or to engage in horizontal monitoring. On average, increased perceptions on the part of the employee that their pay is affected by the work of their co-worker leads to a greater incentive to horizontally monitor. Msw (married and spouse works) is positively associated with greater effort and employees with greater number of children are more likely to horizontally monitor their co-workers.
5. PARTICIPATION EFFECTS According to the data, there is strong evidence that participation leads to efficiency gains. However, is it sufficient to implement participation programs without increasing employee ownership or is employee ownership a necessary condition for obtaining efficiency gains? An interaction variable was created to investigate this proposition. The variable ESOP*P was created and run in the five equations where all the controls were used. The variable takes on the value 0 to 3 for employees in the ESOP and 0 for all other employees. This variable measures the effects of employees’ perceived level of participation in the ESOP on the dependent variables. The results suggest that participation does not have to be coupled with employee ownership to elicit efficiency gains. None of the five ESOP*P coefficients was significantly correlated with cooperative employee behaviori.e. more effort, less shirking, an incentive to horizontally monitor and having engaged in horizontal monitoring. In addition, Wald tests do not reject the joint hypothesis that the coefficients on participation and ESOP are both equal to zero for any of the five models that omit the ESOP*P interaction term. Moreover, I examined the effects of participation on private firms by estimating the models over the subsample that excluded workers in the ESOP. The effects of participation on the dependent variables for the private firms were the same as for the entire sample. That is, employees in the private firms who perceive their participation to be higher, on average, exert greater effort,
98
AGUSTIN J. ROS
have an incentive to horizontally monitor and have engaged in horizontal monitoring. In addition, I estimated the dependent variables for the entire sample without participation. The exclusion of participation did not make the ESOP coefficient significant in any of the regression estimates. These results indicate that employees’ perceived level of participation, irrespective of the level of employee ownership, is associated with efficiency gains. Firms need not increase employee ownership to capture the benefits of participation on worker behavior and incentives. Of course, increasing participation in firms entails managerial, worker and administrative costs. These costs should be taken into account before such changes should occur.
6. CONCLUSIONS This study used data from a 90% employee owned ESOP manufacturing firm in northern Illinois and six similar private firms in the same industry and region to investigate the effects of employee ownership on effort, shirking and horizontal monitoring. It was found that, on average, employees at the ESOP do not exert greater effort, shirk less, have a greater incentive to horizontally monitor or engage in more horizontal monitoring than the employees in the six private firms. Participation turns out to be a strong determinant of effort and horizontal monitoring. Employees who perceive their participation level to be higher will, on average, exert greater effort, have an incentive to horizontally monitor and engage in horizontal monitoring. This result provides additional evidence that increasing levels of participation leads to efficiency gains.14 It is also found that participation does not have to be coupled with employee ownership to elicit motivational effects. Firms without employee ownership may achieve efficiency gains from increasing workers’ perception of their level of participation in the firm. Other important results indicate that, on average females exert greater effort on the job than males and older workers have less incentive to horizontally monitor and engage in less horizontal monitoring. In addition, workers’ perception that their pay is affected by their co-workers effort, (team compensation), leads to an increased incentive to horizontally monitor. Higher wages and having more children leads to, on average, more horizontal monitoring. These results support prior findings that increased levels of participation lead to efficiency gains. Unlike previous studies, which did not examine participation in private firms, this study finds that increasing workers’ perception that they participate in the decisions of the firm will lead to efficiency gains, even when it is not coupled with employee ownership. A new finding of this study
Do ESOPs Motivate Employees?
99
is that increased levels of participation elicits strong horizontal monitoring incentives and leads to more horizontal monitoring on the part of the employees. Thus, firms that increase employees’ perception of participation may increase worker effort and induce horizontal monitoring on the part of the employees, which may reduce the firm’s monitoring costs.
NOTES 1. Companies can also use ESOPs to buy the shares of an owner in a closely held company or borrow money. This is not to say that other plans, such as pension plans, are not tax deductible as they are. 2. See Bonin, Jones and Putterman (1993) for a thorough review of the theoretical and empirical literature on Labor Managed Firms, (LMFs). See also Ros (2001) for a discussion of more recent studies. 3. The ESOP program in existence is a typical one. Employees are eligible for participation after six months of service. Allocation of shares is based on relative pay, although there is a small adjustment for years of service. In addition to the ESOP program, the company has a cash based profit sharing program. Annual operating targets are set and if they are exceeded, the surplus is distributed, based on the same percentage of salary for each employee. 4. At the time of this writing the company was still operating as an ESOP. 5. The high response rate was due to two factors. The first is the type of organization. Being a part of an ESOP, members may feel more interested in sharing their views. Secondly, a lottery system was implemented for all respondents who returned the questionnaire. The winner received $100 in cash. 6. The firms which declined had many reasons, including timing, logistical constraints and uncertainty of researcher’s motives and reaction by the workforce. 7. It would be highly subjective if the values obtained from asking respondents how much effort they put into a typical hour at work were used without attempting to control for the difference in respondents. Some respondents tend to bias their responses upward while others bias them downward. The use of etv acts like an anchor to reduce the subjectivity associated with such an approach. 8. The statement, “my job requires that I work hard” was also used as a dependent variable but neither ESOP nor participation was found to be significant explanatory variables. 9. Participation in this study is an ordinal variable based on 4 categories derived from a question meant to determine whether employees view themselves as participating in the decisions of the firm. In response to a question the employee can select from a range of four from strongly disagree to strongly agree. However, I treat them as a continuous variable in the econometric analysis. In order to ascertain whether this causes any problems, I have also created dummy variables for the responses and used them in the equations. The results were not different when the dummy participation variables were used. 10. See Akerlof (1982), (1984). 11. In the last part of this section, I will perform group tests to determine whether the participation and the ESOP variables, taken together, affect the dependent variables.
100
AGUSTIN J. ROS
12. Once again, the equation was estimated with sex*c as an additional independent variable. Sex became positive and statistically insignificant and sex*c was positively and significantly correlated with eff. Contrary to Becker’s theory, females who have greater children-and thus are more likely to have greater household responsibilities-are, on average, more likely to exert greater effort than females who have fewer children and males. 13. Based on responses it is not clear what the bonus was actually based on, i.e. the productivity of the employee or group. While there was a question in the survey asking the employee to state what was the basis of the bonus (i.e. individual or group-based) many of the questions were left blank. The option most selected was that the bonus was based on the productivity of the firm. 14. See Ros (2001) for a survey of studies that examined the impact of participation. In contrast to this present study which uses subjective measures of participation, many of the previous studies attempt to construct more objective measures.
REFERENCES Akerlof, G. A. (1984). Gift Exchange and Efficiency-Wage Theory: Four Views. American Economic Review, May, 79–83. Akerlof, G. A. (1982). Labor Contracts as Partial Gift Exchange. Quarterly Journal of Economics, November. Alchian, A., & Demsetz, H. (1972). Production, Information Costs, and Economic Organization. The American Economic Review, 62, 777–795. Bielby, D. D., & Bielby, W. (1988). She Works Hard for the Money: Household Responsibilities and the Allocation of Work Effort. American Journal of Sociology, 5, 1031–1059. Blinder, A. S. (1990). Paying for Productivity. The Brookings Institute. Bonin, J. P., Jones, D. C., & Putterman, L. (1993). Theoretical and Empirical Studies of Producer Cooperatives: Will Ever the Twain Meet? Journal of Economic Literature, September. Ehrenberg, R. G. (1980). Research in Labor Economics, Vol. 3. JAI Press. Hamermesh, D. S. (1990). Shirking or Productive Schmoozing: Wages and the Allocation of Time at Work. Industrial and Labor Relations Review, 43, S121–S133. Manly, B. F. J. (1986). Multivariate Statistical Methods: A Primer. Chapman and Hall. NCEO (2000). National Center for Employee Ownership, Statistical Profile of Employee Ownership. November. Ros, A. (2001). Profits for All? The Costs and Benefits of Employee Ownership. NOVA Science Publishers. Williamson, O. (1980). The Organization of Work: A Comparative Institutional Assessment. Journal of Economic Behavior and Organizations, 1, 5–38. Williamson, O. (1975). Markets and Hierarchies: Analysis and Antitrust Implications. Free Press.
Do ESOPs Motivate Employees?
101
APPENDIX 1 The following is an ANONYMOUS survey which will be used for research only. Please take 5–10 minutes to answer the following questions. When completed place in envelope and seal. Thank you for your cooperation. 1. What is your job title? 2. For how many years or months have you worked for this company? (mw) years months 3. On average how many hours per week do you work for this company? (hw) 4. How many days of work did you miss in the last year? (ab) 5. How many paid sick days are you allowed each year? 6. During a usual work day, about how much time do you spend for a meal break? (lb) 7. Apart from meal breaks, about how much time do you spend on regular coffee or scheduled breaks? (ob) *8. Thinking about the rest of your time at work, about how much time do you usually spend on things not related to the work that you do – -like talking to friends, doing personal business or just relaxing. (nwb) *9. Now I would like you to think of a 10 point scale for the amount of energy and effort you put into the following activities, with ten representing all your energy and effort and zero representing hardly any at all. Five would be about half way in between. A typical hour at work (ew1) A typical hour watching T. V. (etv) For the following statements please circle the most appropriate choice. 10. My job requires that I work hard. (jrwh) strongly agree agree disagree strongly disagree 11. What I do at work is more important to me than the money I earn. (ijc) Strongly agree agree disagree strongly disagree. 12. Having a job is more important to me than the wage I earn. (ihj) Strongly agree agree disagree strongly disagree. 13. I help my co-workers when they need it. (hcw) Strongly agree agree disagree strongly disagree.
102
14.
AGUSTIN J. ROS
My effort at work is affected by the effort of my co-workers. (oeace) Strongly agree
15.
agree
disagree
strongly disagree.
agree
disagree
strongly disagree.
agree
disagree
strongly disagree.
agree
disagree
strongly disagree.
Do you receive a bonus. (bon) yes
21.
strongly disagree.
I am treated fairly by the company. Strongly agree
20.
disagree
I participate in the decisions of the firm. (p) Strongly agree
19.
agree
If I saw a co-worker slacking off I would complain. (whm) Strongly agree
18.
strongly disagree.
My effort at work affects my salary. (opaoe) Strongly agree
*17.
disagree
The work of my co-workers affect my pay (opace) Strongly agree
16.
agree
no
If yes what determines this bonus. my productivity my productivity and the productivity of my co-workers profitability of the firm
22.
Altogether how much effort, either physical or mental, does your job require? (eaw) none
*23.
some
a lot
How much effort do you put into your work beyond what your job requires? (earw) none
*24.
only a little
only a little
some
a lot
Have you ever said or done anything to a co-worker when you saw that worker slack off? (hm) yes no
25.
On average how much do you earn per week from this job? (w)
26.
Do you participate in the company’s Employee Stock Ownership Plan (ESOP)? yes no
Do ESOPs Motivate Employees?
103
27.
If yes, how many shares are allocated to your account?
28.
What% of these shares are vested to you? That is, how many shares do you actually own?
29.
Do you have any extra job or other way of making money in addition to this job? (oj) yes no.
30.
Age
31.
Sex
32.
Marital Status
33.
If married, does your spouse works? (msw)
34.
Do you have children?
35.
If so, how many? (c)
36.
What is the highest grade or year of school you have completed? grade
Single
* Used as dependent variables.
Married
yes
shares
Divorced yes
no
no
ON THE CAUSES OF SOFT BUDGET CONSTRAINTS: FIRM-LEVEL EVIDENCE FROM BULGARIA AND ROMANIA Greetje Everaert and Antje Hildebrandt ABSTRACT Since Kornai (1980), the adverse effects of soft budget constraints have been well-documented in the literature. More recently, several theoretical explanations for the presence of soft budget constraints have been put forward. The purpose of this paper is to empirically test these theories on the causes of soft budget constraints. We therefore use a panel data set, consisting of company account data for Bulgarian and Romanian manufacturing firms, covering the period 1995–1999. Our results suggest that the probability of finding soft budget constraints importantly depends on the degree of competition within the sector and on the ownership structure of the firm. We further find that sociopolitical concerns about employment increase the probability of soft budget constraints, but only when firms are loss making. Thus, our empirical results largely confirm the hypotheses that competition, privatization, and firm size matter in explaining soft budget constraints, as is suggested in the theoretical models on the causes of soft budget constraints.
The Determinants of the Incidence and the Effects of Participatory Organizations, Volume 7, pages 105–137. Copyright © 2003 by Elsevier Science Ltd. All rights of reproduction in any form reserved. ISBN: 0-7623-1000-6
105
106
GREETJE EVERAERT AND ANTJE HILDEBRANDT
1. INTRODUCTION The transition towards market-based economies in Central and Eastern Europe offers a natural experiment for testing the widely held view in economics that competition and private ownership contribute to improved economic performance. Indeed, there is now a vast body of empirical literature examining firm performance in terms of ownership structure and degree of competition. The consensus conclusion is that private companies perform better than state companies, though not necessarily in the case of private firms that were previously state-owned or featured insider-privatization (Boardman & Vining, 1989; Konings, 1997; Blanchard, 1997; Frydman et al., 1999), and that firms in more competitive industries also tend to perform better (Nickell, 1996; Konings, 1997; Brown & Earle, 2000). Our paper takes up the question whether product-market competition and ownership structure can explain the prevalence of soft budget constraints. The concept of soft budget constraints (SBCs) was introduced by Kornai (1980) and refers to a situation where loss-making firms are bailed out or refinanced. In Kornai’s (1980) view, bail-outs of loss-making firms reflect a paternalistic government attitude. The government wants to preserve employment and the survival of firms even when they incur losses. Closely related to the paternalistic explanation of SBCs are political economy models of SBCs. Shleifer and Vishny (1994) argue that the political influence of entrepreneurs automatically gives rise to bribes and subsidies. Alternatively, Dewatripont and Maskin (1995) explain the existence of SBCs as the outcome of a commitment problem in the presence of sunk costs. In contrast to Kornai’s (1980) assumption that the paternalistic state attitude is an exogenous given, Dewatripont and Maskin (1995) provide an endogenous explanation for SBCs. In their model, institutional conditions such as state ownership, centralization of credit and a lack of competition increase the probability of SBCs. Along similar lines, Segal (1998) argues that a lack of competition promotes SBCs. The purpose of this paper is to empirically test for the various causes of SBCs suggested in the theoretical literature. Do privately-owned firms and firms in more competitive industries suffer less from SBCs? To what extent can sociopolitical motives such as employment account for the incidence of SBCs? To date, the empirical work on SBCs has remained modest. Schaffer (1998) assesses the importance of the different forms of SBCs, arguing that tax arrears are the main channel through which the government continues to support its firms. Other documentary evidence on SBCs comes from the EBRD (1999) and the World Bank (1999), stressing the importance of non-collected bills from state utility suppliers. Clifton and Khan (1993) discuss inter-enterprise arrears
On the Causes of Soft Budget Constraints
107
in Romania. The bulk of empirical work on SBCs, however, focuses on the effects of SBCs in explaining firm or macro-economic performance in Central and Eastern European countries.1 This paper, on the other hand, tries to explain the prevalence of SBCs themselves. In this respect, our work bears some resemblance to that of Li and Liang (1998), who test SBC theories in a Chinese context. In attempting to explain the presence of SBCs, we use an unbalanced panel data set consisting of company account data for Bulgarian and Romanian manufacturing firms covering the period 1995–1999. Bulgaria and Romania are typically slow reformers (EBRD, 2000). This makes them more likely to be prone to SBCs and makes them more suitable for testing theories on SBCs. Moreover, both countries experienced arrears crises. In Bulgaria, a run-up of soft bank credit and non-performing loans formed the basis of the 1996 banking crisis. In Romania, firms repeatedly stopped paying each other, gambling on a collective bail-out (Clifton & Khan, 1993; Berglöf & Roland, 1998; Perotti, 1998). These experiences illustrate that the main source and manifestation of SBCs were different in each country. Our SBC variable will be accommodated to capture both sources of SBCs. More specifically, we use a logit approach where a dummy variable, indicating the presence of SBCs, acts as the dependent variable. The latter variable is based on a measure for bank-related SBCs as suggested by Schaffer (1998). In addition, we propose an alternative SBC measure capturing inter-enterprise arrears. Our results suggest that the probability of finding SBCs importantly depends on the degree of competition within the sector, as well as on the ownership structure of the firm. Ownership structure in Bulgaria, however, has no additional explanatory power once firms are loss-making. We further find that sociopolitical concerns about employment increase the probability of SBCs, but only when firms are loss-making. Thus, our results confirm, as suggested by the theoretical models on the causes of SBCs, the hypotheses that competition, privatization, and firm size matter in explaining SBCs. A number of papers have explicitly addressed the issue of the adverse effects of SBCs, both on theoretical and on empirical grounds. In Kornai’s (1980) framework SBCs are the causes of permanent shortages in socialist economies. Moreover, SBCs are believed to hamper innovation (Qian & Xu, 1998), the restructuring of firms and an efficient resource allocation (Kornai, 1980; Dewatripont & Roland, 1996), output (Schaffer, 1989), economic growth (Huang & Xu, 1999) and free trade (Everaert & Vandenbussche, 2001). Consequently, the hardening of SBCs is at the heart of the reform process in Central and Eastern Europe and therefore of primary concern to policy-makers. Our conclusions should draw the attention of policy-makers to the importance
108
GREETJE EVERAERT AND ANTJE HILDEBRANDT
of raising competitive pressure in transition economies and seeing through their privatization programs to reduce the prevalence of SBCs. The paper is structured as follows. In Section 2 we sketch out the theoretical framework for testing SBC theories and formulate the hypotheses to test. Section 3 describes the variables and data set we use and gives some descriptive statistics. Empirical results are presented in Section 4. A discussion of the results can be found in Section 5. A final section comments and concludes.
2. THEORETICAL BACKGROUND In this Section, we outline a theoretical framework for testing SBC theories. As the relevant theoretical literature has recently been summarized by Maskin and Xu (2001), we concentrate instead on major themes that can be distilled from this literature. More specifically, we group the theoretical predictions from this literature into three main categories and translate these into hypotheses that will be tested in Section 4 of the paper. First, we expect firms with higher levels of employment to be more likely to benefit from SBCs. Sociopolitical motives such as preservation of jobs or maximizing employment and output usually support the idea that in socialist countries firms tend to be bailed out rather than allowed to go bankrupt. This reflects managerial incentives under socialism2 and the paternalistic attitude of the state in these countries (Kornai, 1980). A second hypothesis is that decentralization helps establish harder budget constraints. This is clearly illustrated in the model of Dewatripont and Maskin (1995), who give an endogenous explanation of SBCs. In their model, bad investment decisions are made because of creditors’ lack of information about the quality of the project and due to a lack of commitment on the creditors’ side not to refinance bad projects once an irreversible investment has been made. Consequently, solving the asymmetric information problem and the commitment problem not to bail-out bad projects are at the heart of hardening SBCs in such models. These problems are remedied with decentralization, i.e. making the market more transparent3 and more competitive. More specifically, Dewatripont and Maskin (1995) argue that, if refinancing takes place in a competitive credit market with a large number of small creditors – as compared to when refinancing is done by the same monopolist creditor – the likelihood of a bail-out is smaller, given the fact that small creditors face liquidity constraints which makes ex-post renegotiation of credit more difficult. Anticipating these difficulties, bad investment decisions are avoided and the bail-out issue becomes superfluous. In other words, competition among creditors hardens the budget constraints.
On the Causes of Soft Budget Constraints
109
Dewatripont and Roland (1996) consider the case where a single creditor needs to finance a multitude of projects. Now, competition on the producers’ side decreases the probability of a bail-out such that fewer bad investments are made. Again, budget constraints are harder under competitive pressure. A similar effect of competition through trade linkages is illustrated by Berglöf and Roland (1998). When strong one-to-one relations between suppliers and buyers exist, liquidation of one loss-making firm (i.e. when it is not bailed out) will be very costly, given the negative spillover effects to the firm’s trading partners. Consequently, increased competition weakens these negative spillover effects, makes the liquidation option more credible and hardens budget constraints. Competition also works through competition between old and new projects, as in Berglöf and Roland (1998), or via competition for funds among regional governments as explained by Qian and Roland (1998) for China. In a different setting, Segal (1998) likewise demonstrates the importance of competition in the hardening of budget constraints. When the market is serviced by a sufficient number of firms, other firms can make up for the output loss following the liquidation of a defaulting firm. This makes the social cost of liquidation smaller and the liquidation itself is a more feasible option. Notably, nearly all authors suggest several ways in which competition might contribute to harden budget constraints. In all cases, the driving mechanism is that the likelihood of a future bail-out decreases under competition. The Dewatripont and Maskin model (1995) also illustrates the effects of privatization. Privatization can harden budget constraints in two different ways. Firstly, bad projects are more easily restructured into good projects after privatization, since private investors usually have better access to capital to buy new equipment and since they can more easily oppose or appease opposition from workers to restructuring (Blanchard, 1997), rendering bail-out superfluous. Secondly, privatization changes the conditions under which a bail-out of a bad project occurs. This can be seen as follows. When the creditor is a private profit maximizing agent, instead of a welfare maximizing government, his objective function is less comprehensive. In particular, a profit maximizing creditor will discard the payoff to the entrepreneur in his own utility function whereas in case a welfare maximizing government makes credit decisions, it will include both terms in its objective function.4 Therefore, the payoff from refinancing a bad investment project will less likely exceed the value of liquidation, making the latter option more attractive. Hence, private ownership reduces bad investments and hardens budget constraints. Along similar lines, one could argue that profit maximization incentives and incentives to restructure are even stronger for foreign-owned firms in Central and Eastern
110
GREETJE EVERAERT AND ANTJE HILDEBRANDT
Europe or that insider privatization will be less effective in disciplining firms as compared to domestic private ownership (Blanchard, 1997).5 In Section 4, we will test whether indeed firm size, competition and privatization can explain the incidence of soft budget constraints among firms, as predicted in the theoretical literature. We however first turn to a description of the data set in Section 3.
3. DATA DESCRIPTION To test the aforementioned theories on SBCs, we use an unbalanced panel data set of company data from Bulgarian and Romanian manufacturing firms,6 covering the period 1995–1999. Bulgaria and Romania could be considered laggards in transition when compared to more advanced CEECs such as Poland or Hungary. This can be seen e.g. from enterprise restructuring and banking reform indices in Table 1.7 Moreover, privatization in Bulgaria and Romania only took off in the second half of the 1990s, so these countries are more likely to be characterized by SBCs. Finally, the choice of Bulgaria and Romania allows us to compare results from testing SBC theories in both a small, open economy and in a larger, relatively closed economy. The firm-level data are taken from AMADEUS CD-ROMs, distributed by the Bureau Van Dijk, Belgium. Data are restricted to large and medium sized firms, i.e. firms that are either characterized by an employment level exceeding 100 or which have total assets and total sales exceeding U.S.$12 million.8 Our sample consists of 1536 Bulgarian and 2293 Romanian firms. Firm-level employment figures are used to test the first hypothesis that firm size increases the likelihood of SBCs. Descriptive statistics are presented in Table 2. Both countries experienced a decline over time in average firm size. However, initial conditions at the start of transition were rather different, with Romanian firms being much larger than Bulgarian enterprises. This legacy is reflected in the data. Table 1.
Reform Indices in Central and Eastern European Countries.
Enterprise reform index Banking reform index Small-scale privatization Large-scale privatization
Bulgaria
Czech Rep.
Hungary
Poland
Romania
2.3 2.7 3.3 3.0
3.0 3.3 4.3 4.0
3.3 4.0 4.3 4.0
3.0 3.3 4.3 3.3
2.0 2.7 3.7 2.7
Source: EBRD (1999), Transition Report.
On the Causes of Soft Budget Constraints
Table 2.
111
Size of Firms: Number of Employees.
Year
Median
Bulgaria Mean
St. dev.
Median
Romania Mean
St. dev.
1995 1996 1997 1998 1999
224 197 184 174 154
428 390 362 341 307
843 792 739 699 617
439 383 340 310 272
893 797 745 636 564
1,755 1,591 1,493 1,249 1,085
Source: Own calculations using AMADEUS Data.
To test the second hypothesis that competition makes SBCs less likely to occur, we need data on the degree of competition. For that reason we include Herfindahl indices and import penetration on the 3-digit NACE level.9 The Herfindahl index is calculated as the sum of squares of market shares of all firms in the relevant industry and ranges between 0 and 100%. A high value for the Herfindahl index corresponds to high industry concentration in the sector. The import penetration ratio is measured as total imports over the sum of total sales and imports in the sector, and also ranges between 0 and 100%. Sectors that face fierce competition from abroad will consequently feature a high import penetration ratio. Thus, these indices reflect the overall degree of domestic and foreign competitive pressure within the sector. Data were obtained from the Statistical Offices of Bulgaria and Romania for the period 1994–1998, such that we use lagged values of these variables in our regressions. Descriptive statistics on Herfindahl indices and import penetration ratios are presented in Tables 3 and 4. The overall drop in industry Table 3.
Concentration of Manufacturing Firms in Bulgaria and Romania: Herfindahl Indices.
Year
Median
Bulgaria Mean
St. dev.
Median
Romania Mean
St. dev.
1994 1995 1996 1997 1998
20.56 16.37 16.45 16.52 14.78
30.71 27.07 27.60 25.28 20.22
27.66 26.13 28.51 24.96 18.76
27.41 22.57 27.86 23.36 21.74
61.06 57.90 56.46 53.07 50.92
27.97 27.77 28.52 28.29 28.91
Sources: Statistical Offices of Bulgaria and Romania, own calculations.
112
Table 4.
GREETJE EVERAERT AND ANTJE HILDEBRANDT
Import Penetration of Manufacturing Industries in Bulgaria and Romania.
Year
Median
Bulgaria Mean
St. dev.
Median
Romania Mean
St. dev.
1994 1995 1996 1997 1998
32.40 33.82 31.17 34.38 39.60
37.81 37.77 39.12 40.37 42.85
27.60 27.96 27.65 28.87 29.76
14.06 14.76 19.48 21.25 26.93
25.80 30.05 32.43 30.49 36.66
28.97 24.54 26.61 24.63 28.96
Sources: Statistical Offices of Bulgaria and Romania, own calculations.
concentration and the general upward trend in import penetration reflect the process of reform in both countries. The fact that import penetration is on average higher in Bulgaria is consistent with our claim that the Bulgarian economy is more open and Romania more closed. The difference between Bulgaria and Romania is even more pronounced for the concentration index in Table 3. Many huge conglomerates in Bulgaria were split up following the Demonopolisation Act of 1992. This led to an important reduction in concentration ratios (Djankov & Hoekman, 2000). However, firms were often split up in complementary parts, such that lowered levels of concentration did not necessarily reflect an increase in product-market competition. For this reason and following Nickell (1996), we use first differences of the indices in the econometric analysis as they better reflect actual changes in competition patterns. However, the sector-level changes in Romania for the concentration ratios are very small, so in that case we use the levels for both variables. The AMADEUS data also allowed us to trace down the ownership structure of the firms for the years 1997–1999. The ownership structure for the preceding years was unavailable and therefore we assume it to be the same as the ownership structure for 1997 in our regressions.10 Ownership information, given in AMADEUS, includes the name and nationality of the owner and his direct ownership share. We could hence distinguish various ownership categories – state, municipalities,11 foreign investors, private investors and insider-owned companies or cooperatives – and we could construct dummies for the various owners involved,12 dummies for full, majority or minority ownership13 or for various forms of mixed ownership. Table 5 shows the percentage of firms in our sample in which the listed owner held majority stakes.
On the Causes of Soft Budget Constraints
Table 5.
113
Ownership Structure: Percentage of Firms in Which the Named Owner Has the Majority.
Majorities
Country
1997
1998
1999
State-owned
Bulgaria Romania Bulgaria Romania Bulgaria Romania Bulgaria Romania Bulgaria Romania Bulgaria Romania
26.1 37.7 19.3 41.2 5.2 10.4 6.5 2.6 2.4 n.a. 40.5 8.1
16.2 37.9 30.2 40.6 8.4 10.4 5.3 2.7 2.4 n.a. 37.5 8.4
15.9 18.6 33.6 50.6 8.3 19.4 5.2 2.7 2.2 n.a. 34.8 8.7
Private-owned Foreign-owned Insiders/cooperatives Municipalities Not majority owned
Source: own calculations using AMADEUS Data.
As can be seen from Table 5, state-ownership is still important in Bulgaria and Romania, even though it is on the downward trend, whereas private and foreign ownership are of increasing importance in our sample.14 There are also some insider-privatized firms.15 This information on ownership will allow us to test whether our third hypothesis (privatization hardens budget discipline) holds. Obviously, ownership information is sometimes incomplete, and the indices we use on competitive forces serve only as a proxy. Our data do not e.g. allow us to distinguish between de novo firms, thereby leaving the theory of Berglöf and Roland (1998) on competition between new and old firms outside the scope of our analysis. Neither can we assess the extent of competition on the creditor’s side (Dewatripont & Roland, 1996) nor the strength of trade linkages between buyers and suppliers (Berglöf & Roland, 1998). Competition in foreign export markets may be an additional source of competitive and disciplining pressure, but this information is lacking in our data set. Nonetheless, we believe that our data on ownership are sufficiently representative and that the indices on concentration and import penetration give a good overall indication of domestic and foreign competitive pressure and can be used as such in the econometric analysis. The dependent variable in our analysis is a measure for SBCs. SBCs can take various forms: tax arrears, inter-enterprise arrears, non-payment of bills from state utility suppliers, and soft bank credit. Our data allow us to identify both
114
GREETJE EVERAERT AND ANTJE HILDEBRANDT
SBCs that are due to inter-enterprise arrears and SBCs that originate from the banking sector. For the latter, we use a measure of SBCs based on Schaffer (1998). A firm is said to have net-bank financing SBCs (NSBCs) when it receives net bank financing (NBF) despite negative profitability. NBF is defined as the net increase in outstanding debts16 over total assets and multiplied by one hundred: NBFt =
debtt debtt 1 * 100 total assetst
Our measure for NSBC thus takes the value 1 when NBFt is positive and operating profit in year t is negative; it is 0 otherwise.17,18 Consequently, this measure reflects which firms “undeservedly” obtain extra credit. In Figs 1 and 2, NSBC firms are located in the upper-left side of the diagram. SBCs also appear in the form of inter-enterprise arrears19 (Clifton & Khan, 1993; Perotti, 1998). Therefore, we attempt to identify firms that benefit from unacceptably generous credit margins, reflecting their inability to pay. Along these lines, we now define a variable for credit-related SBCs (CSBCs) that takes on the value 1 when the firm is loss-making in year t and benefits from
Fig. 1.
Net Bank Financing, Bulgaria.
On the Causes of Soft Budget Constraints
Fig. 2.
115
Net Bank Financing, Romania.
a credit period that is larger than the average credit period firms with positive profits in year t enjoy. The variable is 0 otherwise. Table 6 clearly illustrates that firms with negative operating profits benefit from more credit days on average and that this credit period displays greater variation as reflected by the standard deviation. Thus, most of the credit to lossmaking firms can be considered as involuntary as on average, profit-making firms are given less time to pay. This motivates the choice of our CSBC measure. Firms with inter-enterprise arrears, under our definition, are displayed at the upper-left in Figs 3 and 4. Looking at Table 7, we see that NSBCs have been more important in Bulgaria, whereas the relative importance of CSBCs has been greater in Romania. The column BSBC indicates the percentage of SBC firms, having either NSBCs or CSBCs, or both. The data on SBCs are well in line with Schaffer (1998) and Dmitrov (1999), who argue that inter-enterprise arrears were unimportant for Bulgaria,20 unlike in Romania, where enterprises engaged in collusive arrears anticipating a general government bail-out. In Bulgaria, on the other hand, a weak banking sector made NSBCs widespread. In 1996, when the Bulgarian banking sector experienced a banking crisis, NSBCs dropped to
116
GREETJE EVERAERT AND ANTJE HILDEBRANDT
Table 6.
Credit Period in Days. Bulgaria
Year 1995 1996 1997 1998 1999
Romania
Operating profit/loss
Mean
St. dev.
Mean
St. dev.
Positive Negative Positive Negative Positive Negative Positive Negative Positive Negative
26 48 43 53 31 50 24 45 31 54
48 95 79 83 47 69 33 71 48 94
45 79 42 105 41 96 45 99 45 85
54 116 50 154 48 134 55 122 51 104
Source: own calculations using AMADEUS Data.
Fig. 3.
Credit Period, Bulgaria.
On the Causes of Soft Budget Constraints
Table 7.
117
Firms with Soft Budget Constraints (in Percent of Firms in the Whole Sample).
Year
NSBC
1995 1996 1997 1998 1999
21 2 13 31 30
Bulgaria CSBC BSBC 13 8 10 18 20
21 9 17 33 36
IBSBC
NSBC
21 9 12 28 32
4 3 2 9 5
Romania CSBC BSBC 6 5 6 11 12
8 6 7 14 13
IBSBC 7 5 6 13 13
Notes: NSBC refers to net-bank-finance related SBCs, CSBC refers to inter-enterprise related SBCs and BSBC refers to SBCs of any kind, as explained in the text. IBSBC refers to investmentcorrected SBCs. Source: Own calculations using AMADEUS Data.
Fig. 4.
Credit Period, Romania.
118
GREETJE EVERAERT AND ANTJE HILDEBRANDT
2% as bank financing dried up.21 In the empirical part, we will use the variable BSBC, indicating the presence of SBCs of any type. Given the substantial number of firms that are both classified as having NSBCs and CSBCs, we are confident that we identified genuine SBC firms and that we can use this BSBCs measure in the econometric analysis as such. We further considered a refined measure of SBCs, “investment-corrected” BSBCs (abbreviated IBSBCs), where we correct for the possibility that firms are engaged in deep restructuring22 and need the financial resources to do so. Apart from using internal resources, obtaining extra bank financing is the most straightforward option given the fact that equity markets are virtually nonexistent in these countries. Hence, firms we previously classified as NSBC firms, might be in the process of reorganizing their business in the prospect of becoming profit-making in the future. We checked whether NSBC firms, which by definition have NBF > 0, also exhibited a positive real increase in tangible fixed assets.23 In this case, we did not assume that the firm was subject to SBCs. Our refined variable for net-bank-finance related SBCs, INSBC, thus has fewer cases of SBCs, compared to the old NSBC measure. The variable for CSBCs remains unchanged. The variable IBSBC is constructed analogously, representing firms that have either INSBCs or CSBCs, or both.
4. EMPIRICAL RESULTS As our dependent variable is discrete, we run regressions for explaining SBCs using a random effects logit approach.24 We run regressions for the entire sample and for the sample, restricted to the firms with negative operating profit to distinguish between the performance aspect and the issue of softness in our dependent variable.25 All regressions include year dummies to capture yearspecific effects. Our reference regression features the following form: BSBCi,t = + 1heri,t 1 + 2impi,t 1 + 3empl i,t + 4statei,t + 5 foreigni,t
T
+ 6municipi,t + 7coopinsideri,t +
yeart + ui + i,t,
t=2
where i,t is the white-noise disturbance and ui is the disturbance term accompanying the random effects term. In Tables 8 and 9 we report the results for Bulgaria and Romania respectively when the entire sample was considered, and when the sample was restricted to loss-making firms only. To stress the impact of privatization, we add information on ownership status in columns (2) and (4–5) for both countries and both sample cases.26
On the Causes of Soft Budget Constraints
119
Regression (1) in Tables 8 and 9 only takes into account competition variables and the level of employment within the firm. Competition ratios are included as percentages in the regressions. For both countries, the Herfindahl index is positive, indicating that more concentration within the sector increases
Table 8.
Logit Results for Soft Budget Constraints in Bulgarian Firms. Estimation Method: Logit Model with Random-Effects Sample Period: 1996–1999 Dependent Variable: Soft Budget Constraints
Variables Constant Herfindahl Import Employment
Unrestricted sample (1) (2) –1.0311** (–10.61) 0.0120 (1.38) –0.0063 (–1.07) –0.0001 (–0.51)
State Insiders/cooperatives Foreign Municipalities Year 961 Year 972 Year 99 Wald-chi2 Prob > chi2 No. of obs. No. of groups
–1.9941** (–13.15) –1.1297** (–8.85) 0.2080* (1.92) 296.45 0.00 4,657 1,401
–1.1742** (–10.11) 0.0216** (2.06) –0.0006 (–0.09) –0.0001 (–0.16) 0.6221** (3.95) 0.2442 (0.93) –0.0370 (–0.16) 0.3776 (0.96) –2.2116** (–12.26) –1.2988** (–8.73) 0.3510** (2.93) 277.30 0.00 3,780 1,225
Notes: * significant at 10% level, ** significant at 5% level. 1 Referring to year 98 in regression (3). 2 Referring to year98 in regression (4).
Restricted sample (3) (4) –0.8368** (–5.20) 0.0342** (2.86) –0.0151* (–1.64) 0.0007** (2.86)
2.5671** (10.84) 1.9855** (8.55) 1.9145** (9.64) 144.67 0.00 1,557 855
–1.200** (–5.26) 0.0292* (1.92) –0.0087 (–0.76) 0.0007** (2.71) –0.1671 (–0.91) –0.2479 (–0.83) 0.4580 (1.26) 0.2406 (0.48) 2.1319** (7.59) 0.4735* (1.80) –0.1201 (–0.52) 116.50 0.00 1,283 738
120
GREETJE EVERAERT AND ANTJE HILDEBRANDT
the likelihood of SBCs. This coefficient is significant for Romania but not for Bulgaria. The coefficient for import penetration has the expected sign, i.e. increased import competition is associated with tougher budget discipline. Its Table 9.
Logit Results for Soft Budget Constraints in Romanian Firms. Estimation Method: Logit Model with Random-Effects Sample Period: 1995–1999 Dependent Variable: Soft Budget Constraints
Variables Constant Herfindahl Import Employment
Unrestricted sample (1) (2) –3.7568** (–20.73) 0.0082** (2.83) –0.0012 (–1.45) –0.0001* (–1.92)
State Foreign Insiders
–4.2896** (–19.44) 0.0082** (2.58) –0.0018* (–1.91) –0.0001 (–0.91) 0.9281** (5.69) 1.2868** (6.57) –2.5603** (–2.32)
(3) 0.7057** (2.57) 0.0121** (2.85) –0.0006 (–0.49) 0.0001 (1.10)
Restricted sample (4) 0.1849 (0.57) 0.0170** (3.49) –0.0002 (–0.17) 0.0002** (2.13) 0.2505 (1.08) 1.3818** (4.48) 1.1111 (0.57)
Purely state Purely foreign Year 96 Year 97 Year 98 Year 99 Wald-chi2 Prob > chi2 No. of obs. No. of groups
–0.2157 (–1.31) –0.1380 (–0.85) 0.9000** (6.19) 0.5807** (3.86) 105.43 0.00 8,440 1,877
–0.2706 (–1.48) –0.2806 (–1.55) 0.7807** (4.74) 0.6903** (4.07) 140.47 0.00 6,647 1,609
–0.3678 (–1.19) –0.3176 (–1.04) 0.3476 (1.25) –1.054** (–3.84) 57.75 0.00 1,278 715
Notes: * significant at 10% level, ** significant at 5% level.
–0.5457 (–1.53) –0.7591** (–2.19) –0.0483 (–0.15) –1.3138** (–4.21) 65.11 0.00 983 550
(5) 0.5479* (1.74) 0.0147** (2.98) –0.0003 (–0.21) 0.0003** (2.22)
2.8107** (2.03) 0.4037 (0.90) –0.4356 (–1.22) –0.7510** (–2.14) –0.0969** (–0.30) –1.2845** (–4.09) 53.43 0.00 983 550
On the Causes of Soft Budget Constraints
121
significance, however, is low.27 The coefficient for the level of employment has a negative sign, and, contrary to Kornai’s (1980) hypothesis, is even significant for Romania. This can be explained by the significant number of large firms with positive profit in our sample. As will be discussed below, however, this result is not robust for Bulgaria, nor for Romania in the restricted sample.28 The big negative coefficient on the 1996 year dummy for Bulgaria reflects the drop in SBCs due to the banking crisis. In both countries, however, SBCs increase over time. When the ownership structure is added to the regressions, as in columns (2) of Tables 8 and 9, the sign and significance of the variables remain largely unchanged. Now, the Herfindahl index becomes significant for Bulgaria, as does import penetration for Romania. The results on ownership refer to the case when dummies for majority ownership are included. We also experimented with dummies for full and minority ownership, or dummies for the presence of any ownership category. However, the results continue to hold in those regressions. Compared to the case of private ownership, which is our benchmark case,29 the presence of state ownership increases the incidence of SBCs significantly in both countries. For Bulgaria, all other ownership variables also bear the expected sign: foreign participation decreases the probability of SBCs, whereas insider-owned firms and municipalities increase this probability, even though these coefficients are not significant.30 The case of Romania is somewhat more puzzling: insider and foreign ownership both have an unexpected sign and are statistically significant. Insider-owned firms in Romania mostly report positive profits in our sample. Hence, they bear a negative coefficient. The involvement of the state in most foreign majorityowned companies is responsible for the negative coefficient on foreign ownership, as will become clear below. The remaining columns in Tables 8 and 9 report analogous results when the sample, restricted to loss-making firms, is considered. Here, we test whether the variables we selected to explain the incidence of SBCs, can distinguish between firms with SBCs and firms with hard budget constraints, even when all firms in the sample are loss-making. As can be seen in columns (3) and (4) of Tables 8 and 9, the results for the competition variables are now strengthened: the coefficient on the Herfindahl index is always positive and statistically significant and import penetration is always negative.31 Moreover, the coefficient on employment now consistently bears the hypothesized positive sign, most of the times being significant.32 Thus, firm size does matter in explaining SBCs, but only when firms are loss-making. The story is somewhat the opposite for ownership variables. Ownership information does not contribute towards explaining SBCs, once firms are loss-making, especially not
122
GREETJE EVERAERT AND ANTJE HILDEBRANDT
so in Bulgaria. For Romania, the results are more in line with the results from the unrestricted sample when purely state-owned and purely foreign-owned companies are considered. The former have a positive significant effect on SBCs, the latter make the coefficient and significance level for foreign ownership considerably smaller, and thus in line with theoretical predictions. However, purely state-owned or purely foreign-owned firms make up only a very small part of the majority-owned firms and the conclusions are thus not representative for the majority-owned sample as a whole. The influence of the Romanian state in case of mixed ownership thus reaches farther than its quantitative share would suggest. Using dummies for purely state-owned or purely foreign-owned did not change the results for Bulgaria. Ownership structure in Bulgaria is thus more indicative of the performance of the firms, as the incentives to restructure, explained in Section 2, would suggest. To test whether different owners have different sociopolitical objectives, we included an interaction term for majority state-ownership and employment. In none of the specifications, this term appeared significant. Big firms are thus equally perceived as “too big too fail”, irrespective of whether they are privately owned or owned by the state. The results, in line with the theoretical literature on SBCs, thus make a case for accepting that privatization, competition and firm size matter in explaining the prevalence of SBCs in Bulgaria and Romania. State-ownership is usually associated with a higher incidence of SBCs. We do not consistently find, however, that foreign participation gives rise to fewer SBCs. Foreign investors might, e.g. enjoy more freedom to concentrate on strategic considerations – such as attracting qualified workers, establishing their firm reputation and brand name and capturing market share – instead of being profit maximizing in the short run. Moreover, they usually have access to foreign equity markets such that they do not need to rely on bank financing. Finally, the significance of the variables that explain SBCs in the restricted sample ensures that we are not taking up a performance effect, but that our SBC indicators can – within the sample of possibly SBC firms, i.e. firms with negative operating profit – distinguish between SBC firms and hard budget constraint firms.
5. DISCUSSION OF THE RESULTS The results presented in Section 4 are fairly robust, since they hold across countries considered, across sub-samples and irrespective of the measurement of the variables. We further performed robustness checks for ownership structure, since we assumed that ownership prior to 1997 was the same as in 1997. This
On the Causes of Soft Budget Constraints
123
assumption is particularly strong for Bulgaria, since mass-privatization occurred in 1996. However, we could identify the state-ownership share (100%) for the years prior to 1997 for those firms (Centre for Mass Privatisation, 1996). The results remain unchanged. The same holds when using the sub-samples 1997–1999 and 1998–1999 for which the evolution of ownership structure is complete. The results are equally robust with respect to the specification of the ownership dummies included, i.e. the results continue to hold when dummies on minority ownership are included or for dummies simply indicating the presence of an ownership category irrespective of the quantitative importance of the ownership share. We further experimented with taking logarithms, squares of logarithms and with dummies for firms employing as many as 400 people in Bulgaria and 1,000 people in Romania. These figures are close to the average employment levels in both countries. Again, the results are robust. Our results go through when our investment-corrected variable for SBCs acts as the dependent variable.33 The same holds when we constructed an SBC measure that circumvents the endogeneity that is possibly present in our measure for SBCs. Firms can run positive operating profits precisely because they benefit from SBCs. However, they are not selected via the original procedure. The alternative way to proceed was to compare last year’s profit with the current year’s NBF measure or with current year’s credit period, but with no major change in the results. As far as our competition variables are concerned, the results for Romania also go through when first differences of the indices are used instead of levels. As an alternative measure for concentration, we directly computed market shares from the AMADEUS data and added this variable to our regression (Nickell, 1996). The results for the regressions are included in the Appendix in Tables 12–15. The conclusions we set out earlier for the Herfindahl and import penetration indices and for ownership information continue to hold unchanged. Market shares, however, turn out to be negative and significant, implying that higher market shares lower the probability of finding SBCs. Including squares of market shares, however, reveals that the relation between market shares and SBCs is non-linear. Smaller firms that fiercely compete for market share have little market power to price themselves to positive profits. Larger firms, however, can use their market power to earn positive profits. Firms with dominant market shares, however, are typically loss-making and qualify for SBCs. That also explains why in the restricted sample, market share always bears a positive sign, and why the significance of squared market shares in the restricted sample disappears.34
124
GREETJE EVERAERT AND ANTJE HILDEBRANDT
Therefore, we believe that the results convincingly argue that increasing competitive pressure and continuing through with the privatization process has primary importance for policy-makers in Central and Eastern Europe. In line with previous papers, our results suggest that private ownership can contribute to better performance, making firms less prone to suffer operating losses. However, if this is not complemented by competitive pressure, the merits of privatization are limited. In particular, the results from the restricted regressions suggest strong effects from competition on the hardening of SBCs. Continued efforts to increase competitive pressure are particularly warranted since time dummies suggest that SBCs are nevertheless on the rise in both countries. Possible problems with our results are twofold. First, our data set is biased towards large and medium-sized enterprises and does not allow us to distinguish between de novo private firms and privatized firms. Neither can we follow up the firms which are involved in mergers or split-ups. Second, our results possibly suffer from endogeneity problem. Firms can operate under SBCs because they occupy a strategic position within the sector. However, continued subsidies can prevent the sector from becoming more competitive. We have resolved this problem using lagged variables for Herfindahl and import penetration indices, and by carrying out regressions that restrict the sample. A more explicit dynamic approach or a two-stage procedure could be an interesting extension. Finally, reported profits from firms are often unrealistically squeezed towards zero (see Figures 1–4) and this might influence our SBC variable. This reflects, among others, the reluctance of firms to report their losses or to pay high taxes on profits (Blanchard, 1997).
6. CONCLUSION The aim of the paper was to shed new light on the causes of SBCs by assessing their potential to empirically explain the incidence of SBCs in Bulgaria and Romania. We did so by using a panel data set for Bulgarian and Romanian manufacturing firms that covered the period 1995–1999. We used a random effects logit approach to try to explain the causes of SBCs, stemming from soft bank credit or stemming from inter-enterprise arrears. We used Schaffer’s (1998) measure for net-bank-financing SBCs and proposed a new measure, based on the credit period a firm enjoys. Our results suggest that the incidence of finding SBCs can be explained by the ownership structure of the firm and competitive pressure within the sector. Firm size, as proxied by the level of employment, is another determinant of SBCs, but only when firms are lossmaking or after correcting for market shares. The opposite effect holds with respect to the ownership structure in Bulgaria: for loss-making firms, the
On the Causes of Soft Budget Constraints
125
ownership structure has no additional explanatory value in explaining SBCs. Ownership, thus, is a strong indicator for performance but not for distinguishing between soft and hard budget constraint firms, once they begin to perform poorly. Comparing our results with the theoretical predictions of the various models we consider, we can confirm the hypotheses that competition, privatization and firm size matter in explaining the incidence of SBCs. Finally, our results once again stress the importance of increasing competitive pressure and of continuing privatization reforms in transition countries. The hardening of budget constraints continues to be a top priority.
NOTES 1. For example, Earle and Estrin (1998), Dobrinsky, Dochev and Nikolov (1997), Konings and Vandenbussche (2000), Majumdar (1998), Bertero and Rondi (2000), Raiser (1993, 1994). 2. Note that the expectation of a future government bail-out will precisely distort incentives and make firms more prone to run losses. Hence, the bail-out expectation can become a self-fulfilling prophesy. 3. We do not explicitly address the asymmetric information problem in this paper. Theoretical work on solving asymmetric information effects of SBCs through screening is discussed by Bai and Wang (1998). Models were effort is monitored are suggested by Qian and Roland (1998). 4. The extent to which the pay-off of the entrepreneur reflects his concerns about employment in the firm, Dewatripont and Maskin (1995) predict that bail-outs will less likely occur to preserve employment when the credit market is privatized. Thus, sociopolitical concerns might be different under different types of ownership. This links it back with our first hypothesis. However, if the entrepreneur only cares about his personal prestige or his private benefits related to the operation of his firm, the employment issue can be treated separately from the privatization hypothesis. To allow for an explicit test on this view, the preferred specification in the econometric analysis is to treat the issues of ownership and firm size separately and to test for interaction effects (Section 4). 5. From a theoretical perspective, insider privatization could be compatible with hard budget constraint incentives and incentives to restructure. First, insiders do have incentives to generate profits and to buy new capital equipment. However, retained profits are likely to be insufficient and outside finance limited or non-existent in transition countries. Secondly, workers may accept restructuring even if it comes at the expense of their job loss provided they are sufficiently compensated as shareholders. In transition economies, this is unlikely to be the case. Moreover, if future profits are distributed as higher wages, no gain is obtained as shareholders. The empirical literature for transition economies also consistently reports inferior performance of insider-owned firms, compared to domestic or foreign outsider-owned firms (Blanchard, 1997; Frydman et al., 1999; Roland, 2000). 6. Manufacturing was the principal activity of these firms throughout the sample period.
126
GREETJE EVERAERT AND ANTJE HILDEBRANDT
7. A higher value for the indices stands for further progress towards market-based economy practices. The highest possible index value in both categories is 4.3. 8. AMADEUS is a commercial and Pan-European database, available on CDROM, created and distributed by the Bureau Van Dijk. The Bureau Van Dijk is a listed company on the Euronext stock exchange. The data are collected by local information providers (Creditreform Bulgaria OOD and the Romanian Chamber of Industry and Commerce) and the Bureau Van Dijk makes them consistent across countries. Small firms are excluded. 9. We match the Herfindahl index and import penetration according to the principal activity reported by the firm. 10. The assumption does not change our results (see Section 5). 11. Ownership for municipalities was only available for Bulgaria. 12. The dummy reports whether a certain owner category has a stake in the firm, irrespective of the importance of its ownership share. 13. A firm is said to be majority-owned when one ownership category has a stake of more than 50%. For minority ownership, an owner needs to own at least 33%. 14. Since there were only a few mass-privatized firms in Romania, we considered them together with privately owned firms in the regressions. 15. Insider-owned companies are often quite similar to cooperatives in our sample. Therefore, we consider these two categories together in the regressions for Bulgaria. 16. We include both short-term liabilities and long-term debt in our measure of debt. Unlike Schaffer (1998), we do not subtract interest paid as the non-payment of interest on existing debt already shows the presence of bank arrears. Moreover, data on interest paid were of such poor quality that it would have significantly reduced the number of observations in our sample. Note that short-term liabilities might also include items unrelated to bank finance such as wage arrears, and debts to tax administration or to state utility suppliers. Hence, our NSBC measure should not be narrowly interpreted. 17. We use “operating” profit and loss, since we want to identify the firms that are economically nonviable, irrespective of their financial structure. 18. Here, we differ from Schaffer (1998) by considering the value of operating profit and loss in period t instead of in period t 1. 19. Non-payment of bills from state utility suppliers is often reflected as a very high number of credit days, which we capture here. 20. This holds prior to 1997 at least. 21. The drop of NSBCs in Bulgaria in 1996 is not due to the number of firms having negative operating profit. 22. We refer here to strategic restructuring as opposed to defensive restructuring (Blanchard, 1997). 23. Data on tangible fixed assets were equally available on a firm-level basis in AMADEUS. 24. Including a fixed effects term could arguably better capture firm-specific characteristics that are constant over time. The estimation of fixed effects proceeds via conditional maximum likelihood and is based only on observations where the dependent variable changes status. This would lead to a substantial loss of information, especially since firms that repeatedly operate under SBCs resp. hard budget constraints are among the most informative in our sample, i.e. essentially, our analysis is a cross-sectional one. We ran tests to check the accuracy of the estimation by increasing the number of quadrature points used in the approximation. However, all coefficients were sufficiently
On the Causes of Soft Budget Constraints
127
stable within conventional levels. Only the coefficient on employment was relatively unstable at 8 quadrature points in the unrestricted sample for both countries. Increasing the number of quadrature points added to the stability of the coefficients, but without change in their size. Results reported refer to the case when 12 quadrature points are used. 25. One can argue, given our definition of SBCs, that SBCs could be a proxy for poorly performing firms. The explanatory variables in our regressions would thus explain the performance, instead of the identity, of firms with negative operating profit, which operate under SBCs. 26. The number of observations in our sample slightly drops because Herfindahl and import penetration indices were unavailable for some sectors and because ownership information for some firms was unavailable. 27. Note, however, that for Romania, the coefficient on import penetration is almost significant at a 10% level. 28. Notice that the stability of the coefficient of employment in the unrestricted sample was unstable when estimated with 8 quadrature points. 29. The benchmark case also includes firms where no majority owner is present or where this information is incomplete. Given the fact that non-majority owned firms in Bulgaria make up a large part of the sample (Table 5), we also performed regressions where the case of diffuse ownership acts as our benchmark to separate it from the case of private majority ownership, which is then separately included in the regressions. However, the results continue to hold unchanged. 30. The coefficient for municipalities ownership here is smaller than that of state ownership (and in line with the Tiebout competition hypothesis) as incorporated in the SBC literature by Qian and Roland (1998). 31. In both countries, the size of the effect, calculated at the median and at the average is greater for concentration index than for the import penetration ratio. 32. The effect of firms size on SBCs, calculated at the median and average is greater for Bulgaria (0.0001) than for Romania (0.00004). 33. See Tables 10 and 11 of the Appendix. 34. The fact that employment in the unrestricted sample earlier had a negative sign might be because it previously took up a market share effect. After correcting for market share, employment indeed becomes positive significant in Bulgaria in the entire sample. For Romania in the entire sample the sign becomes positive in the specification of column (2).
ACKNOWLEDGMENTS The first author gratefully acknowledges financial support from the Fund for Scientific Research (FWO) under research grant G.0267.01. The second author gratefully acknowledges financial support from the European Union Marie Curie Training Site Fellowships under the program “Improving Human Research Potential and the Socio-Economic Knowledge Base”. Both authors would like to thank LICOS for research support. We would equally like to thank Joep Konings, Hylke Vandenbussche, Alexander Repkine, Bas Van Aarle, Todor Gradev, Charles B. Blankart, participants of the Public Choice meeting
128
GREETJE EVERAERT AND ANTJE HILDEBRANDT
in Innsbruck in June 2001, participants of the summer school of the European University Institute in Florence in September 2001, participants of the EUNIP Conference held in Vienna in December 2001, participants of the conference on Institutional and Organizational Dynamics in Post-Socialist Transformation held in Amiens in January 2002, participants of the BOFIT seminar in Helsinki in March 2002, participants of the Irish Economic Annual Conference held in Mullingar in April 2002 as well as one anonymous referee. The authors are solely responsible for the views expressed in this paper and these do not represent the opinions of the Community. The usual disclaimer applies.
REFERENCES Bai, C.-E., & Wang, Y. (1998). Bureaucratic Control and the Soft Budget Constraint. Journal of Comparative Economics, 26(1), 41–61. Berglöf, E., & Roland, G. (1997). Soft Budget Constraints and Credit Crunches in Financial Transition. European Economic Review, 41, 807–817. Berglöf, E., & Roland, G. (1998). Soft Budget Constraints and Banking in Transition Economies. Journal of Comparative Economics, 26(1), 18–40. Bertero, E., & Rondi, L. (2000). Financial Pressure and the Behaviour of Public Enterprises under Soft Budget Constraints: Evidence from Panel Data, Journal of Public Economics, 75, 73–98. Blanchard, O. (1997). The Economics of Post-Communist Transition, Clarendon Lectures in Economics. New York: Oxford University Press. Boardman, A. E., & Vining, A. R. (1989). Ownership and Performance in Competitive Environments: A Comparison of the Performance of Private, Mixed, and State-Owned Enterprises. Journal of Law and Economics, 32, 1–33. Brown, J. D., & Earle, J. S. (2000). Competition and Firm Performance: Lessons from Russia. CEPR Discussion Paper 2444. Centre for Mass Privatisation Bulgaria (1996). Massovata Privatizatsija v Bulgaria (Mass Privatisation in Bulgaria). Clifton, E. V., & Khan, M. S. (1993). Interenterprise Arrears in Transforming Economies: The Case of Romania. IMF Staff papers, 40(3), 680–696. Dewatripont, M., & Maskin, E. S. (1995). Credit and Efficiency in Centralized and Decentralized Economies. Review of Economic Studies, 62, 541–555. Dewatripont, M., & Roland, G. (1996). Transition as a Process of Large-Scale Institutional Change. Economics of Transition, 4(1), 1–30. Djankov, S., & Hoekman, B. (2000). Market Discipline and Corporate Efficiency: Evidence from Bulgaria. Canadian Journal of Economics, 33(1), 190–206. Dmitrov, L. (1999). Budget Constraints in Bulgarian Enterprises: 1996–97. Centre of Economic Reform and Transformation Discussion Paper, 99/15. Dobrinsky, M., Dochev, N., & Nikolov, B. (1997). Debt Workout and Enterprise Performance in Bulgaria: Lessons from Transition Experience and Crisis. Centre of Economic Reform and Transformation Discussion Paper, 97/15. Earle, J. S., & Estrin, S. (1998). Privatization, Competition, and Budget Constraints: Disciplining Enterprises in Russia. SITE Working Paper 128.
On the Causes of Soft Budget Constraints
129
European Bank for Reconstruction and Development (1999). Transition Report. London: European Bank for Reconstruction and Development. European Bank for Reconstruction and Development (2000). Transition Report. London: European Bank for Reconstruction and Development. Everaert, G. M. M., & Vandenbussche, H. (2001). Does Protection Harden Budget Constraints? LICOS Discussion Paper, 98/2001. Belgium: Catholic University of Leuven. Frydman, R., Gray, Ch., & Hessel, M. (1999). When does Privatization Work? The Impact of Private Ownership on Corporate Performance in the Transition Economies. Quarterly Journal of Economics, 4, 1153–1191. Greene, W. H. (1997). Econometric Analysis (3rd ed.). New Jersey: Prentice-Hall. Huang, H., & Xu, C. (1999). Institutions, Innovations and Growth, American Economic Review, 89(2), 438–443. Konings, J. (1997). Competition and Firm Performance in Transition Economies: Evidence from Firm Level Surveys in Slovenia, Hungary and Romania. CEPR Discussion Paper 1770. Konings, J., & Vandenbussche, H. (2000). The Adjustment of Financial Ratios in the Presence of Soft Budget Constraints: Evidence from Bulgaria. Preliminary Working Paper Catholic University of Leuven, September. Kornai, J. (1980). The Economics of Shortage. Amsterdam, North Holland. Li, D. D., & Liang, M. (1998). Causes of Soft Budget Constraints: Evidence on three Explanations. Journal of Comparative Economics, 26(1), 104–116. Majumdar, S. K. (1998). Slack in the State-Owned Enterprise: An Evaluation of the Impact of Soft-Budget Constraints. International Journal of Industrial Organization, 16, 377–394. Maskin, E., & Xu, C. (2001). Soft Budget Constraint Theories: From Centralisation to the Market. Economics of Transition, 9(1), 1–27. Nickell, St. J. (1996). Competition and Corporate Performance. Journal of Political Economy, 104(4), 724–746. Perotti, E. C. (1998). Inertial Credit and Opportunistic Arrears in Transition. European Economic Review, 42, 1703–1725. Qian, Y., & Roland, G. (1998). Federalism and the Soft Budget Constraint. American Economic Review, 88(5), 251–278. Qian, Y., & Xu, C. (1998). Innovation and Bureaucracy under Soft and Hard Budget Constraints. Review of Economic Studies, 65(1), 151–164. Raiser, M. (1993). Searching for the Hole in the Public Pocket: The Institutional Legacy of Soft Budget Constraints and the Polish Transformation Process. Economic Systems, 17(4), 251–278. Raiser, M. (1994). The No-Exit Economy: Soft Budget Constraints and the Fate of Economic Reforms in Developing Countries. World Development, 22(12), 1951–1967. Roland, G. (2000). Transition and Economics: Politics, Markets and Firms. Cambridge, Mass.: MIT Press. Schaffer, M. (1989). The Credible-Commitment Problem in Center-Enterprise Relationship. Journal of Comparative Economics, 13, 359–382. Schaffer, M. (1998). Do Firms in Transition have Soft Budget Constraints? A Reconsideration of Concepts and Evidence. Journal of Comparative Economics, 26, 80–103. Segal, I. R. (1998). Monopoly and Soft Budget Constraints. Rand Journal of Economics, 29(3), 596–609. Shleifer, A., & Vishny, R. (1994). Politicians and Firms. Quarterly Journal of Economics, 109, 995–1025.
130
GREETJE EVERAERT AND ANTJE HILDEBRANDT
Tiebout, Ch. M. (1956). A Pure Theory of Local Expenditure. Journal of Political Economy, 64(5), 416–424. World Bank (1999). Dismantling Russia’s Nonpayment System: Creating Conditions for Growth. World Bank Report on the Russian Federation, September.
On the Causes of Soft Budget Constraints
131
APPENDIX Table 10.
Logit Results for Soft Budget Constraints in Bulgarian Firms with Investment-Corrected Measure. Estimation Method: Logit Model with Random-Effects Sample Period: 1996–1999 Dependent Variable: Soft Budget Constraints
Variables Constant Herfindahl Import Employment
Unrestricted sample (1) (2) –1.3385** (–12.98) 0.0095 (1.05) –0.0101 (–1.62) 0.0001 (0.30)
State Insiders/cooperatives Foreign Municipalities Year 96 Year 971 Year 99 Wald-chi2 Prob > chi2 No. of obs. No. of groups
–1.7392** (–11.34) –1.4057** (–9.90) 0.2444** (2.17) 272.93 0.00 4,657 1,401
–1.4144** (–11.50) 0.0157 (1.46) –0.0060 (–0.85) 0.0001 (0.42) 0.6112** (3.78) –0.1482 (–0.52) –0.0466 (–0.19) 0.1798 (0.44) –1.9937** (–10.94) –1.6349** (–9.82) 0.3762** (3.04) 264.38 0.00 3,780 1,225
Notes: * significant at 10% level, ** significant at 5% level. 1 referring to year 98 in regressions (3) and (4).
Restricted sample (3) (4) –0.1235 (–0.75) 0.0268** (2.24) –0.0222** (–2.48) 0.0008** (3.45)
–0.8579** (–4.06) 1.0851** (5.30) 0.7394** (4.04) 110.68 0.00 1,557 855
–0.0352 (–0.17) 0.0154 (1.04) –0.0217** (–2.02) 0.0007** (2.89) –0.1066 (–0.59) –0.7509** (–2.53) 0.2223 (0.68) –0.2161 (–0.47) –0.9330** (–3.73) 1.1352** (4.79) 0.7943** (3.72) 95.39 0.00 1,283 738
132
Table 11.
GREETJE EVERAERT AND ANTJE HILDEBRANDT
Logit Results for Soft Budget Constraints in Romanian Firms with Investment-Corrected Measure for SBC. Estimation Method: Logit Model with Random-Effects Sample Period: 1995–1999 Dependent Variable: Soft Budget Constraints
Variables Constant Herfindahl Import Employment
Unrestricted sample (1) (2) –3.8533** (–20.82) 0.0075** (2.60) –0.0011 (–1.33) –0.0001 (–1.23)
State Foreign Insiders Year 96 Year 97 Year 98 Year 99 Wald-chi2 Prob > chi2 No. of obs. No. of groups
–0.4589** (–2.62) –0.1066 (–0.65) 0.8286** (5.56) 0.8698** (5.77) 126.96 0.00 8,440 1,887
–4.4456** (–19.44) 0.0083** (2.59) –0.0019** (–1.99) –0.0001 (–0.85) 0.9934** (6.00) 1.2292** (6.19) –2.5620** (–2.31) –0.4825** (–2.47) –0.2400 (–1.29) 0.8259** (4.88) 1.0074** (5.86) 168.80 0.00 6,647 1,609
Notes: * significant at 10% level, ** significant at 5% level.
Restricted sample (3) (4) 0.4607* (1.73) 0.0090** (2.19) –0.0004 (–0.37) 0.0002* (1.75)
–0.8945** (–2.92) –0.2291 (–0.77) 0.1032 (0.38) –0.4256 (–1.61) 28.28 0.00 1,278 715
–0.1192 (–0.37) 0.0162** (3.28) –0.0007 (–0.49) 0.0002** (2.13) 0.3584 (1.52) 1.3463** (4.30) 0.7496 (0.37) –0.9728** (–2.75) –0.6060* (–1.77) 0.0782 (0.25) –0.6042** (–2.00) 47.53 0.00 983 550
On the Causes of Soft Budget Constraints
Table 12.
133
Logit Results for Soft Budget Constraints in Bulgarian Firms with Investment-Corrected Measure for SBC and Market Shares. Estimation Method: Logit Model with Random-Effects Sample Period: 1996–1999 Dependent Variable: Soft Budget Constraints
Variables Constant Herfindahl Import Employment
Unrestricted sample (1) (2) –3.0367** (–19.95) 0.0063 (0.69) –0.0101 (–1.61) 0.0002* (1.94)
State Insiders/cooperatives Foreign Municipalities Market shares Year 97 Year 981 Year 99 Wald-chi2 Prob > chi2 No. of obs. No. of groups
–0.0170** (–3.53) 0.3377** (2.10) 1.7495** (11.31) 2.0042** (13.62) 278.70 0.00 4,650 1,401
–1.3396** (–10.74) 0.0112 (1.03) –0.0063 (–0.89) 0.0002** (2.14) 0.5824** (3.58) –0.2526 (–0.879) 0.0608 (0.25) 0.0974 (0.23) –0.0191** (–3.52) –1.6324** (–9.75) –1.9834** (–10.84) 0.3786** (3.03) 268.55 0.00 3,777 1,225
Notes: * significant at 10% level, ** significant at 5% level. 1 Referring to year 96 in regression (2).
Restricted sample (3) (4) –0.9949** (–5.91) 0.0290** (2.38) –0.0223** (–2.49) 0.0007** (2.70)
0.0101 (1.54) 0.8551** (4.05) 1.9377** (9.05) 1.5913** (8.33) 112.07 0.00 1,553 852
–0.9624** (–4.64) 0.0165 (1.09) –0.0214** (–2.00) 0.0006** (2.53) –0.1026 (–0.57) –0.7373** (–2.48) 0.2097 (0.63) –0.2063 (–0.45) 0.0035 (0.50) 0.9201** (3.68) 2.051** (8.22) 1.7150** (7.74) 94.86 0.00 1,281 736
134
GREETJE EVERAERT AND ANTJE HILDEBRANDT
Table 13. Logit Results for Soft Budget Constraints in Bulgarian Firms with Investment-Corrected Measure for SBC and Market Shares (Levels and Squared). Estimation Method: Logit Model with Random-Effects Sample Period: 1996–1999 Dependent Variable: Soft Budget Constraints Variables Constant Herfindahl Import Employment
Unrestricted sample (1) (2) –2.8559** (–18.61) 0.0030 (0.33) –0.0116* (–1.84) 0.0002* (1.76)
State Insiders/cooperatives Foreign Municipalities Market shares Squared market shares Year 971 Year 98 Year 99 Wald-chi2 Prob > chi2 No. of obs. No. of groups
–0.0651** (–5.84) 0.0007** (4.97) 0.3219** (1.99) 1.7504** (11.30) 1.9851** (13.48) 291.04 0.00 4,650 1,401
–3.0830** (–15.88) 0.0086 (0.80) –0.0079 (–1.11) 0.0002* (1.69) 0.5347** (3.27) –0.4035 (–1.39) 0.1728 (0.70) –0.0393 (–0.09) –0.0693** (–5.63) 0.0007** (4.72) 0.3295* (1.71) 1.9644** (10.73) 2.3179** (13.13) 277.09 0.00 3,777 1,225
Notes: * significant at 10% level, ** significant at 5% level. 1 Referring to year 96 in regression (4).
Restricted sample (3) (4) –0.9926** (–5.87) 0.0289** (2.35) –0.0222** (–2.49) 0.0007** (2.58)
0.0078 (0.48) 0.0001 (0.15) 0.8553** (4.05) 1.9384** (9.05) 1.5922** (8.33) 111.94 0.00 1,553 852
–0.0240 (–0.12) 0.0157 (1.02) –0.0217** (–2.02) 0.0007** (2.58) –0.1060 (–0.59) –0.7614** (–2.54) 0.2434 (0.72) –0.2239 (–0.49) –0.0095 (0.53) 0.0002 (0.78) –0.9172** (–3.66) 1.1332** (4.78) 0.7971** (3.73) 94.73 0.00 1,281 736
On the Causes of Soft Budget Constraints
Table 14.
135
Logit Results for Soft Budget Constraints in Romanian Firms with Investment-Corrected Measure for SBC and Market Shares. Estimation Method: Logit Model with Random-Effects Sample Period: 1995–1999 Dependent Variable: Soft Budget Constraints
Variables Constant Herfindahl Import Employment
Unrestricted sample (1) (2) –3.8544** (–20.53) 0.0072** (2.48) –0.0011 (–1.29) –0.0001 (–0.50)
State Foreign Insiders Market shares Year 96 Year 97 Year 98 Year 99 Wald-chi2 Prob > chi2 No. of obs. No. of groups
–0.0303** (–2.55) –0.4430** (–2.50) –0.0503 (–0.30) 0.8658** (5.71) 0.9146** (5.98) 127.28 0.00 8,253 1,831
–4.4495** (–19.30) 0.0086** (2.63) –0.0019* (–1.93) 0.0001 (0.12) 0.9492** (5.69) 1.2251** (6.16) –2.4746** (–2.23) –0.0512** (–2.82) –0.4526** (–2.29) –0.1747 (–0.94) 0.8971** (5.21) 1.0689** (6.14) 169.69 0.00 6,563 1,586
Notes: * significant at 10% level, ** significant at 5% level.
Restricted sample (3) (4) 0.5689** (2.11) 0.0077* (1.90) –0.0007 (–0.60) 0.0001 (1.34)
0.0602* (1.71) –0.9519** (–3.08) –0.2990 (–0.99) 0.0321 (0.12) –0.4828* (–1.81) 30.63 0.00 1,234 692
–0.0724 (–0.22) 0.0159** (3.20) –0.0008 (–0.59) 0.0002* (1.76) 0.3715 (1.56) 1.3368** (4.26) 0.7746 (0.39) 0.0650* (1.65) –1.0518** (–2.94) –0.6911** (–1.99) 0.0141 (0.04) –0.6670** (–2.17) 49.90 0.00 971 541
136
GREETJE EVERAERT AND ANTJE HILDEBRANDT
Table 15. Logit Results for Soft Budget Constraints in Romanian Firms with Investment-Corrected Measure for SBC and Market Share (Squared and Levels). Estimation Method: Logit Model with Random-Effects Sample Period: 1995–1999 Dependent Variable: Soft Budget Constraints Variables Constant Herfindahl Import Employment
Unrestricted sample (1) (2) –3.8544** (–20.53) 0.0072** (2.48) –0.0011 (–1.29) –0.0001 (–0.50)
State Foreign Insiders Market shares Squared market shares Year 96 Year 97 Year 98 Year 99 Wald-chi2 Prob > chi2 No. of obs. No. of groups
–0.0304** (–2.55) 0.0001** (2.34) –0.4430** (–2.45) –0.0503 (–0.30) 0.8659** (5.72) 0.9146 (5.98) 127.30 0.00 8,253 1,831
–4.4496** (–19.30) 0.0086** (2.63) –0.0019* (–1.93) 0.0001 (0.12) 0.992** (5.69) 1.225** (6.16) –2.4746** (–2.23) –0.0512** (–2.82) 0.0001** (2.36) –0.4526** (–2.29) –0.1747 (–0.93) 0.8971** (5.21) 1.0690** (6.14) 169.70 0.00 6,563 1,586
Notes: * significant at 10% level, ** significant at 5% level.
Restricted sample (3) (4) 0.5702** (2.12) 0.0077* (1.89) –0.0007 (–0.61) 0.0001 (1.30)
0.0677 (1.37) –0.0002 (–0.24) –0.9524** (–3.08) –0.3018 (–1.00) 0.0284 (0.10) –0.4849* (–1.82) 30.99 0.00 1,234 692
–0.0710 (–0.21) 0.0158** (3.17) –0.0008 (–0.59) 0.0002* (1.71) 0.3716 (1.56) 1.3369** (4.26) 0.7771 (0.39) 0.0750 (0.90) –0.0004 (–0.14) –1.0511** (–2.93) –0.6949** (–1.99) 0.0090 (0.03) –0.6698** (–2.17) 49.88 0.00 971 541
On the Causes of Soft Budget Constraints
Table 16.
137
Distribution of Firms by Industries in the Sample.
Nace-code for manufacturing industries 15: food products/beverages 16: tobacco products 17: textiles 18: wearing apparel, dressing, dyeing 19: tanning, dressing of leather 20: wood/products of wood, cork 21: pulp, paper, paper products 22: publishing, printing 23: coke, refined petroleum products 24: chemicals, chemical products 25: rubber, plastic products 26: other non-metallic and mineral products 27: basic metals 28: fabricated metal products 29: machinery and equipment 30: office machinery, computers 31: electrical machinery, apparatus 32: radio, television 33: medical, precision instruments 34: motor vehicles, trailers, semi-trailers 35: other transport equipment 36: furniture 37: recycling Total Source: Own calculations using AMADEUS Data.
Bulgaria Firms In % 296 26 154 186 51 32 26 34 4 71 35 79 48 92 152 6 66 27 20 21 23 82 5 1,536
19.27 1.69 10.03 12.11 3.32 2.08 1.69 2.21 0.26 4.62 2.28 5.14 3.12 5.99 9.90 0.39 4.30 1.76 1.30 1.37 1.50 5.34 0.33 100
Romania Firms In % 463 1 227 245 113 111 30 91 11 79 66 122 63 149 168 11 47 21 26 51 38 136 24 2,293
20.19 0.04 9.80 10.68 4.93 4.84 1.31 3.97 0.48 3.44 2.88 5.32 2.75 6.50 7.11 0.48 2.05 0.91 1.13 2.22 1.65 5.93 1.05 100
THE INCIDENCE AND DETERMINANTS OF EMPLOYEE SHARE OWNERSHIP AND PROFIT SHARING IN EUROPE Andrew Pendleton, Erik Poutsma, Jos Van Ommeren and Chris Brewster ABSTRACT This paper uses a substantial international database to provide the widest and the most detailed analysis to date of financial participation across Europe. It explores the antecedents of broad-based share ownership and profit sharing schemes. It is found that country effects are important predictors of both profit sharing and share ownership schemes. Share ownership schemes are also associated with company size, stock market listing and some measures of HRM ‘sophistication’. Employee participation and representation have weak relationships with financial participation.
1. INTRODUCTION This paper examines the characteristics of business organizations in Europe with broad-based employee share ownership or profit sharing schemes. The issue we attempt to shed light on is whether country-level factors (legislation, The Determinants of the Incidence and the Effects of Participatory Organizations, Volume 7, pages 141–172. Copyright © 2003 by Elsevier Science Ltd. All rights of reproduction in any form reserved. ISBN: 0-7623-1000-6
141
142
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
tax regimes etc) or organizational factors identified in the financial participation literature (e.g. monitoring costs) are the more important reasons for the use of broad-based financial participation. So far, the importance of these two sets of factors have been dealt with in largely separate literatures, and it has not been possible to evaluate their relative merits. The comparative literature, drawing mainly on institutional information, has emphasized the importance of legislation and tax concessions in explaining national variations in the use of financial participation (Uvalic, 1991; Vaughan-Whitehead, 1995; Poutsma, 2001). By contrast, empirical studies conducted in particular countries (mainly the U.S., Japan, France, Germany, and the U.K.) have sought explanations in variations in company characteristics, as proxies for underlying economic processes and relationships such as the costs of monitoring worker performance and the need to enlist co-operation (see, for example McNabb & Whitfield, 1998; Kruse, 1996; Sesil, Kroumouva, Blasi & Kruse, 2002). We attempt to draw on the insights of both sets of literatures in this paper, and to evaluate which are the more important factors in explaining patterns of financial participation. We therefore pose several questions in the paper. To what extent can the use of financial participation by organizations in Europe be attributed to nationallevel influences? Alternatively, do certain characteristics of firms (e.g. industrial relations and human resource management practices) influence the use of financial participation, having controlled for country effects. We group these characteristics into three main types. The first relates to the activities and workforce characteristics of firms: for instance, does the composition of the workforce have any relationship with the propensity to have profit sharing and/ or employee share ownership schemes, on the assumption that certain types of occupations are typically more difficult to monitor than others. Do firms with greater growth opportunities have higher use of financial participation, on the basis that uncertainty about the merits of future decisions raises monitoring costs (as the Financial Economics literature suggests?). Employee participation and representational arrangements are the second set of characteristics of interest. Do firms with profit sharing and/or employee share ownership communicate more with their employees? Are financial participation organizations more or less likely to have representational forms of participation based on trade unions? Do firms with profit sharing and employee share ownership have higher or lower than average levels of union membership? Finally, we examine the role of factors associated with ‘high performance’ aspects of human resource management such as training. Here, the starting point is the recent literature that emphasizes the role of ‘bundles’ of HR practices (for example, Ichinowski, Shaw & Prennushi, 1997) and that suggesting that firms
Employee Share Ownership and Profit Sharing in Europe
143
that are rich in human capital will tend to use financial participation (e.g. Blair, 1995). The context for the investigation is the considerable policy interest in Europe in the use of financial participation. The Council of Ministers of the European Union passed a Council Recommendation in 1992 calling on Member States to promote the use of financial participation by employees. This initiative grew out of the Pepper Report – Promotion of Employee Participation in Profits and Enterprise Results – which summarized the incidence, characteristics, and effects of profit sharing and employee share ownership in the countries of the Union (Uvalic, 1991). During the 1990s there was a growing interest in and use of financial participation in virtually all of the Member States. New legislation was passed in countries such as Germany, Ireland and the Netherlands. Active debates have been taking place recently about the merits of employee share ownership in particular in Belgium, France, and Germany. Recently, the governments of Spain and Italy have called on the Social Partners to implement financial participation schemes (see Poutsma, 2001). The European Commission has just published a Communication advocating further promotion of financial participation (CEC, 2002). To address the research questions, we utilize the European component of the CRANET data-set (see Brewster, Tregaskis, Hegewisch & Mayne, 1996). This comprises data on fourteen of the Member States of the European Union (all except Luxembourg). This is the widest geographical scope of any empirical study yet conducted, with the Mediterranean, Scandinavian, Benelux, and northern European regions of the EU all included. The paper is based on research originally commissioned by the European Foundation for the Improvement of Living and Working Conditions1 to determine the incidence of financial participation across the European Member States. In the paper we provide background information on the incidence of profit sharing and employee share ownership, and then address the research questions listed above. Overall, we find that country level effects are more important than most company characteristics. The most important company level factors are those relating to basic corporate structure, such as size and stock market listing.
2. KEY ISSUES 2.1. Country Differences There is increasing interest in the comparative incidence of profit sharing and employee share ownership (see, for instance, the studies of profit sharing in France, Germany, Italy, and the U.K. conducted by the IPSE team,2 and the
144
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
studies reported in a special issue of Economic and Industrial Democracy in 19993). In general, these studies show that there are differences in cultural attitudes, regulatory and fiscal regimes that create differences between countries in the use and incidence of financial participation. For instance, the Pepper Report – Promotion of Employee Participation in Profits and Enterprise Results – (Uvalic, 1991) found a wide divergence between Member States (results that have been replicated by, for example Vaughan-Whitehead, 1995; Poutsma, 2001) The general consensus in the literature so far is that variations in statutory identity and fiscal concessions have a major impact on crossnational variations in the incidence of financial participation. Thus, profit sharing is especially prevalent in France because of the statutory requirement (supported by fiscal and social security concessions for employees and firms) for certain firms (those with 50 employees or above) to operate Participation. Although not fully developed in the literature so far, national differences in corporate organization and ownership also appear likely to influence the incidence of share ownership schemes. For instance, share ownership schemes seem likely to be facilitated by the extensive use of stock market listing in countries such as the U.K., and obstructed by the pyramidal structures of corporate ownership in countries such as Italy. Given the known variations between nations in institutional support for financial participation, we predict substantial differences between countries in the incidence of the various forms of financial participation. These differences in usage would broadly correspond to known differences in legal, fiscal, and other institutional features of the financial participation environment. Variations in the environmental supports for financial participation seem likely to have a major influence on managerial evaluations of the costs and benefits of financial participation as a potential solution to firm-level problems and issues, such as the difficulties of securing workforce co-operation. For instance, the costs of using share ownership schemes (set-up costs, administrative costs, and share dilution) may be off-set by a reduction in monitoring costs and tax deductions (accruing either directly to the firm or to employees). Without tax breaks, the net costs of using financial participation may be sufficient to dissuade from using a scheme. 2.2. The Costs of Monitoring Worker Behavior There is now quite a substantial literature on the characteristics of firms with financial participation (see Cheadle, 1989; del Boca & Cupiaolo, 1999; Jones & Kato, 1993; Kato & Morishima, forthcoming; Kruse, 1996; Pendleton, 1997; Poole, 1989; Poole & Whitfield, 1994; Poutsma & Huijgen, 1999; Festing,
Employee Share Ownership and Profit Sharing in Europe
145
Groening, Kabst & Weber, 1999). This literature draws on a range of theoretical approaches within Industrial Relations and Economics but a common concern is to extrapolate features that are associated with the presence of financial participation and which may be judged to predispose firms to introduce financial participation. Running throughout most of this literature is a set of predictions drawn from principal-agent/optimal contracting theory. This body of theory suggests that the difficulties of monitoring the behavior of employees with potentially divergent interests may inhibit management’s capacity to organize work and production most effectively. These problems can be ameliorated if employees are provided with an incentive to share information with other employees and with those controlling the firm, and to devote additional effort to their work. Remuneration based on collective performance provides an incentive for all employees to do this. The essence of financial participation, therefore, is inducement of performance-enhancing changes in worker behavior. In the financial participation literature a range of factors are thought to contribute to monitoring costs, and these are commonly taken as proxies for agency costs. For instance, it is widely thought that information asymmetries and monitoring become more problematic as firm size increases. Therefore size may be an important predictor of the adoption and use of financial participation, and indeed many studies find this to be the case (especially in relation to share ownership schemes) (see e.g. Festing et al., 1999; Conyon et al., 2001). However, the problem with interpreting this evidence is that size of firm is likely to be inversely related to the incentive effects of financial participation, because of free-rider effects. Thus, the effects of size are likely to be ambiguous. A strong positive relation with size is less commonly found with profit sharing than share ownership, possibly because profit sharing is more targeted at providing direct incentives than share ownership (and hence the free-rider limitations of its use in large firm settings are more relevant). Irrespective of the influence of size, information asymmetries and monitoring problems could be more intense in work situations where individual output and performance is hard to measure because of the complexity and interdependence of work tasks (Alchian & Demsetz, 1972; Ben-Ner, Burns, Dow & Putterman, 2000; Kruse, 1996; Sesil et al., 2002). These features may be observed in advanced manufacturing contexts, some service activities, and in creative occupations (Fakhfakh & Pérotin, 1993). In these cases, individual incentive payment systems, such as piecework, are not readily applicable. There are a range of measures which may be used to capture these processes: the proportions of various groups of staff, the presence of automated technology, the complexity/interdependence of work tasks, and the proportion
146
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
of highly-educated staff. These features might be proxied by the proportion of manual workers (on the basis that manual work tends to be more readily measurable than much white collar work), the proportion of graduates in the workforce (on the grounds that more highly qualified employees tend to perform relatively complex work tasks), and the proportion of professionals (on the basis that professional work is especially difficult to monitor, especially by non-professionals). 2.3. Employee Participation and Representation Employee participation characteristics are a further area of interest. Here we are concerned with both direct, often individualistic, forms of communication and with collective, representational forms of participation. We anticipate that financial participation and direct communication/participation will tend to coexist. There are two possible characterizations of this relationship. The first is that both are introduced for similar reasons, and hence are conceptually distinct manifestations of the same (unobserved) phenomenon (e.g. high quality or trusting management). Poutsma and Huijgen (1999) point out that both tend to be introduced by management, tend to confer limited decision rights, and are often aimed at enhancing flexibility, quality, and individual performance, so as to improve organizational competitiveness. The second aspect of the relationship is a symbiotic one, whereby each type of participation stimulates use of the other (Levine & Tyson, 1990). It has been noted by several authors that shifts in organizations towards more co-operation, interaction, and responsibility rather than strongly specialized routine tasks, lead to a higher use of financial participation schemes (Wächter & Koch, 1993, p. 304; Becker, 1993; FitzRoy & Kraft, 1987, p. 34). Financial participation and direct employee participation tend to reinforce each other (Kruse, 1993; Poole & Jenkins, 1990). Financial participation may be perceived as the reward for becoming more involved in the firm and hence may be seen as a pay-off for direct participation (Levine & Tyson, 1990), whilst involvement in a financial participation scheme may stimulate demand for greater communication and involvement in work decisions (Kato & Morishima, 2001). Relationships between financial participation and representative forms of participation are likely to be complex, because representative participation is a complex phenomenon. Representative ‘systems’ are often multi-faceted (e.g. the triple system of representation in France) and differ considerably between Member States. Focusing on trade union membership, there are long-standing arguments that financial participation is antithetical to representative participation. There are well-documented instances of the use of financial participation
Employee Share Ownership and Profit Sharing in Europe
147
to discourage employee participation in unions from several countries over the last one hundred and fifty years (see Ramsay, 1977). The argument goes that workers who participate in profits and ownership will come to perceive themselves as an integral part of the firm, rather than mere factors of production. They will thus come to identify less with workers’ organizations. This is especially acute in the case of employee share ownership because this blurs the fundamental distinction between capital and labor upon which unionization is based. Gregg and Machin (1988), using the U.K.’s Workplace Industrial Relations Survey data, found unionized establishments more likely to have profit-sharing or share schemes, but that stronger trade unions were associated with an absence of such schemes. Conyon et al. (2001, p. 80) found that the ‘presence of trade unions is generally positively associated with profit sharing and equity-based pay’, and Poole (1988) found that profit sharing is more likely in firms with a ‘consultative’ style of industrial relations. As the evidence cited here suggests, much of our evidence on this topic comes from the U.K. where legally-mandated requirements for representative participation have generally been less than elsewhere in Europe. Where institutional and statutory support for employee representation and participation is greater, the relationship with financial participation might be different. We might find either weak associations because participation and representation are widely dispersed throughout economies, or even negative relationships because managers view financial participation as a means of weakening those constraints on their flexibility that derive from legal or customary supports for representation and participation. 2.4. Human Resource Management Practices The recent HRM literature provides grounds for expecting that firms with financial participation also have a raft of other human resource management policies. An important strand of this literature views financial participation schemes as part of a high-performance work system, composed of several, interacting HRM instruments (Becker & Huselid, 1998). This literature identifies ‘bundles’ of human resource measures which, by interacting positively with each other, lead to performance outcomes which are greater than would be achieved by the sum of each measure independently (Ichinowski et al., 1997). As with direct participation, there is a potential reciprocity between financial participation and other human resource management measures. If employees are to accept a range of performance-enhancing managerial initiatives, such as performance appraisal, for example, it is arguable that they should receive a pay-off from any improvements in
148
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
performance that might result. Equally, if employees are to share in the performance of the firm, it is arguable that they should actively contribute to performance outcomes. We therefore expect to find firms with financial participation also having a range of other human resource management features, such as higher than average training expenditure, comprehensive performance appraisal systems etc. The relationship between financial participation and training is of particular interest as recent writing has argued that employees should have equal rights to owners and investors because employees also make firm-specific investments and incur potential opportunity costs from their investment (see Blair, 1995, 1999). They bear risk in so far as managements or owners may opportunistically capture the benefits resulting from employee investments. The corollary of this is that employees should receive a dividend on their investment in much the same way as private investors. Financial participation schemes signal to employees that the benefits of training investments will be shared with employees. Equally, from the firm’s point of view, the use of financial participation helps to protect investments made by the firm in employee training. It raises the costs to employees of shirking (i.e. dismissal may lead to lower remuneration elsewhere) and of leaving the firm (except where financial participation is in use in similar firms). In so far as financial participation schemes frequently have deferral periods, it binds the employee to the firm in the medium term. The possibility of a positive relationship between performance appraisal and financial participation might be seen as a hybrid conceptually. The potential use of appraisal for information-sharing raises considerations similar to those mentioned earlier in relation to direct employee participation. The possibility of a pay-off for exchange of performance-enhancing information may provide a stimulus for employees to engage positively with performance appraisal systems. More diffusely, it may help to create a climate of trust and openness that reinforces the claimed commitment-enhancing effects of financial participation. At the same time, the capacity of appraisal to support training and development programmes may add to the potential relationship between training and financial participation outlined above.
3. DATA AND APPROACH The approach in the paper is to assess the relationship between variables representing the four potential sets of influences outlined above and the presence of broad-based share ownership and profit sharing. The primary question we address is whether contextual or organizational characteristics are
Employee Share Ownership and Profit Sharing in Europe
149
the main influence on national variations in the use of financial participation schemes by firms. The data source used for this purpose is the Cranet Survey on International Strategic Human Resource Management. This is a major survey of company human resource management policies and practices conducted in Europe and further afield, co-ordinated at the Cranfield School of Management, Cranfield University in the U.K.4 Altogether, over thirty countries are included in the Cranet survey. Fourteen of the current Member States of the European Union are included in the survey: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, Spain, Sweden, and the United Kingdom. The latest round of data collection was in 1999–2000.5 The unit of response in each country is ‘organization’: this can mean the firm in its entirety or a relatively self-contained unit within a company, such as a subsidiary. This is potentially a damaging limitation of the dataset as we cannot be sure that we are comparing like with like. We deal with this in several ways. We incorporate information on total company size for all cases (by number of employees) so that there is consistency of treatment in this respect. We also create a dummy to record the presence of organizations that are part of larger companies, and make explicit any influences that may be due to this variation in organizational type within the dataset. We base our analysis on private sector organizations with 200 or more employees (n = 2506). The sampling frames used in each country were designed to produce stratified representative samples and do so in the main for all the countries involved. The results gained from individual countries are consistent with those generated by other studies. However, due to slightly different sampling procedures in each country, the analysis cannot claim to provide a fully reliable representative picture of the European Union. In all but two cases the survey was conducted using a postal questionnaire. In the other two, the survey was completed using face-to-face interviews with respondents. In most cases the response rates range from 12–20%. The exceptions were Italy, where the response rate was just under 10%, and Sweden, where the response rate was in excess of 60%. In each instance, the respondent was the person with responsibility for human resource management in the business organization. The topics covered in the survey include the organization of the human resource management function, human resources strategies and policies, staffing practices, flexible working practices, employee appraisal, training and development, compensation and benefits, employee relations and communications, and organizational details (size, workforce composition etc). Translations of questions were slightly amended between national questionnaires to capture
150
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
nuances in meaning between languages. The questions were blind-translated twice by different translators (from English into another language, and then back again). Since the survey was not designed specifically to investigate financial participation, it is unlikely that any completion of the survey will be biased towards firms with schemes. However, as the survey excludes small firms, the incidence of financial participation found in the results will be higher than in the economy as a whole (given a known tendency for larger firms to use financial participation). A further potential source of bias is that the emphasis on human resource management strategies and practices in the questionnaire may lead to a disproportionate response from business organizations with relatively sophisticated and well-developed human resource management systems and practices. On the basis that financial participation tends to be associated with firms with well-developed HRM, this would also lead to an upwards bias in the incidence of financial participation. The distribution of observations between the fourteen countries is shown in Table 1. For each country the cases were weighted according to Eurostat information on the proportion of European business organizations with 250 employees plus in each country.6
Table 1. Country
Austria Belgium Denmark Finland France Germany Greece Ireland Italy Netherlands Portugal Spain Sweden United Kingdom Total
The Distribution of Cases by Country.
Number of cases (unweighted)
Percentage of total (unweighted)
Percentage of total (weighted)
131 141 151 131 264 415 95 96 60 56 96 184 156 530
5 6 6 5 11 17 4 4 2 2 4 7 6 21
3 4 1 1 19 23 1 1 11 4 3 9 2 18
2,506
100
100
Employee Share Ownership and Profit Sharing in Europe
151
Two forms of financial participation are identified: profit sharing and share ownership schemes. The wording of the relevant questions on share ownership diverged somewhat between countries to reflect the predominant form of employee share ownership scheme in each country. Thus, in the U.K. the share ownership question asked about share options as the SAYE share option scheme was the main form of employee share ownership scheme at the time. In other countries, Finland for instance, the question referred to both share acquisition and share option schemes. In Germany the question asked about employee share ownership in general. Framing appropriate questions in comparative surveys always poses methodological dilemmas as similar wordings may tap different phenomenon in different national contexts (Whitfield, Delbridge & Brown, 1998). In the CRANET case, it was decided to phrase questions slightly differently where appropriate so that the primary forms of share ownership could be clearly picked-out. For each type of financial participation the occupational coverage of the scheme forms part of the question. Thus, respondents indicate which categories of employees the schemes apply to: management staff, professional and technical staff, clerical and administrative staff, and manual workers. Thus, it is possible to identify which categories of the workforce the schemes cover as well as whether there is a scheme present. This is an advance on many previous studies, in which the financial participation variable usually records just whether there is a scheme (of whatever coverage) present. In conjunction with a further question, which asks for information on the occupational composition of the workforce, we are able to calculate the proportion of the workforce that is covered by financial participation schemes (assuming each workforce category has universal eligibility). We could use this information in two ways. The first is to use the eligibility coverage as a continuous variable and undertake OLS or Tobit regressions.7 The problem with this approach is that the distribution is highly bi-polar. There are a substantial group of cases with eligibility in the 10–15% area, and a substantial group with 70% plus eligibility, with very few cases in between. This corresponds to the distinction between executive bonus/share schemes and all-employee schemes, and the determinants of the two types of scheme may be fundamentally different.8 We therefore adopted a second approach. We separated the two sub-groups for each type of scheme at the 50% coverage point: schemes with over 50% or more coverage classified as broadly-based, and those with less than 50% as narrowly-based.9 This cutting point is not that arbitrary given the distribution mentioned above, and there were only very few organizations that had coverage near the point of division. Alternative specifications of the point of division (20% and 80%) had hardly any effect on
152
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
the numbers in each category. Whilst we report the overall incidence of both narrow-based and broad-based schemes by country, we restrict our attention in the determinants analysis to the broad-based schemes. We create dichotomous variables equal to 1 where a broad-based scheme is present and 0 where there is not, and uses these as the dependent variables in a series of logistic regressions.
4. DESCRIPTIVE RESULTS Prior to presenting the results of the multi-variate analysis we provide information and commentary on the overall incidence and distribution of financial participation throughout Europe. We also include information here on narrow-based schemes. A comparison of these results with those of the earlier EPOC study reported in Poutsma (2001) can be found in Appendix 1. Most (63%) profit sharing schemes cover in excess of 80% of the workforce, whilst around 50% of organizations with share schemes operate broad-based schemes. There are very few instances of either type of financial participation that cover occupational groups other than managers but which nevertheless do not cover all employees (defined as 80% or more coverage).10 This is particularly so in the case of profit sharing where very few schemes cover more than 20% but less than 80% of employees. There are very few share ownership schemes with more than 40% but less than 80% coverage. The wider coverage of some selective share ownership schemes is mainly accounted for by the inclusion of technical and professional groups in financial participation. Tables 3 and 4 present the use of schemes per country and the proportion of these schemes that are broadly based. On the whole, the higher the incidence of financial participation schemes in a country, the higher the proportion that tend to be broadly-based.11
Table 2.
Incidence of Financial Participation Schemes Across Europe.
Type of scheme
Share ownership scheme (%)
Profit sharing scheme (%)
No scheme Narrow-based scheme Broad-based scheme
68.5 15.1 16.4
54.6 8.5 36.9
Total
100
100
Weighted as indicated in the text.
Employee Share Ownership and Profit Sharing in Europe
153
Table 3.
Proportion of Business Units in Each Country With Share Ownership Schemes 1999/2000. Percentage of business units in each country with share ownership schemes Country
Narrow-based schemes
Broad-based schemes
Combined Total
Proportion of schemes that are broad-based
Austria Belgium Denmark Finland France Germany Greece Ireland Italy Netherlands Portugal Spain Sweden United Kingdom
5 18 6 15 18 10 16 18 13 24 3 14 14 15
4 11 15 15 23 10 7 16 2 21 2 5 12 30
9 29 21 30 41 20 23 34 15 45 5 19 26 45
41 38 70 50 56 50 23 47 13 47 40 26 46 67
Total (weighted)
15
16
31
52
Notes: Business units with share schemes have either narrow-based or broad-based share ownership. Business units with both types of scheme are classed as broad-based.
It can be seen that there are wide variations between European countries in the overall incidence of share ownership schemes. France, the Netherlands, and the United Kingdom have the highest incidence, whilst Southern European countries tend to have the lowest. It is interesting to note, however, that there is less variation around the European norm for narrow-based schemes. National differences are more marked when the focus is broad-based schemes. The Southern Mediterranean countries clearly have a very low incidence of share schemes compared with the mean for Europe as a whole. These findings suggest the impact of legislation, fiscal concessions and other regulatory influences is uneven between the two types of scheme. The variations in the incidence of broad-based schemes correlate broadly with known differences in the extent of government supports for share ownership (see Poutsma, 2001). By contrast, management-only schemes appear to be driven by factors other than legislation. In most countries the incidence of narrow-based schemes is higher than broad-based despite the lack of fiscal supports for selective schemes. The incidence of profit sharing, like that of share ownership, varies quite widely between European countries. It ranges from a high of 87% in France
154
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
Table 4.
Proportion of Business Units in Each Country With Profit Sharing Schemes 1999/2000. Percentage of business units in each country with profit sharing schemes Country
Narrow-based schemes
Broad-based schemes
Total
Proportion of schemes that are broad-based
Austria Belgium Denmark Finland France Germany Greece Ireland Italy Netherlands Portugal Spain Sweden United Kingdom
41 10 7 8 3 53 12 5 7 4 8 12 8 7
25 12 8 27 84 18 7 24 8 55 17 13 19 30
66 22 15 35 87 71 19 29 15 59 25 25 27 37
38 55 52 77 97 25 39 83 53 93 67 52 70 81
Total (weighted)
9
36
45
63
Note: Business units with financial participation schemes have either narrow-based or broad-based schemes. Business units with both types of scheme are classed as broad-based.
(where a form of profit sharing (Participation) is usually compulsory for firms with more than 50 employees) to 15% in Denmark and Italy. Like share ownership, there is less absolute variation in the incidence of narrow-based profit sharing, with the exception of Austria and Germany.12 Unlike share ownership schemes, most profit sharing schemes cover all or most employees (except Austria and Germany).
5. DETERMINANTS OF FINANCIAL PARTICIPATION The frequency distributions presented so far indicate that there is considerable variation in the incidence of broad-based financial participation in Europe. As in earlier studies (e.g. Uvalic, 1991), the extent of government support for financial participation seems to correlate broadly with the level of financial participation in each country. In the remainder of the paper, we attempt to assess more precisely the role of country-level factors in explaining the overall incidence of financial participation in Europe relative to the firm-level factors
Employee Share Ownership and Profit Sharing in Europe
155
commonly identified in the literature. To do this, we mount a series of logistic regressions where the dependent variable is the presence of either a broadbased profit sharing or employee share ownership scheme. To measure country effects, we take two approaches. The first is to enter dummy variables for each country (with the U.K. as the base) on the basis that this variable will capture a cocktail of features of the legal and fiscal regime in each case. Where significant effects are observed, this indicates that these features have an impact on the distribution of financial participation across Europe even when other firm-level determinants of financial participation are taken into account, and will tend to confirm the findings of comparative work that national regime differences are an important determinant of the use of financial participation. The second approach is to enter measures for specific characteristics of national regimes. Two are used here. One, we create a dummy for tax supports for financial participation. This is a judgement-based variable, whereby those countries viewed by the authors to have substantial tax supports for profit sharing (income tax/social security charge exemptions/reductions for the profit share) or share schemes (share gains treated as capital gains rather than income) in 1995 (four years before the observation) were coded as 1, with others coded as 0 (see Poutsma, 2001).13 Two, we provide a measure of the extent of stock market development in each country, as shown by the valuation of domestic listed firms as a proportion of GDP in 1997 (source: Federation of European Stock Exchanges). We predict this will have a positive influence on the presence of share schemes but not profit sharing. We use various variables to test for monitoring cost explanations of financial participation. We include a measure of size (log of employees in entire company) though there are ambivalent expectations about its effects. It may proxy for monitoring difficulties increasing with organizational size but conversely may proxy for free-rider effects. In the absence of data on work tasks in the CRANET survey, we use various measures of workforce composition on the grounds that the work tasks often associated with certain types of staff may be more or less complex and difficult to monitor. We use the percentage of manual workers, professional workers, and graduate employees in the organization’s workforce, though we recognize that the occupation-work task relationship may be over-simplified by the use of these proxies. Furthermore, the latter two measures are by no means unambiguous, for they could equally be used to measure human capital-type explanations for financial participation. Finally, we include a measure of the rate of innovation (firms who judge themselves to be in the top 10% of innovators in their industry are coded as 1) and a measure of market growth (growth is coded as 1, 0
156
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
otherwise). These variables are based on arguments in the recent Financial Economics literature that firms with good growth opportunities will be more likely to use share schemes because there is greater scope for employee discretion where opportunities are greater (Yermack, 1995).14 To measure the importance of communication, participation, and representation factors we make use of a range of measures. We include two measures of information disclosure: one is a dummy for the release of information on business strategy to manual or clerical employees, the other is a dummy for the release of information on company finances to these employees. We also include a variable for the presence of a works council or joint consultative committee. Finally, we have a measure of trade union density based on the rank fractions procedure. In CRANET union density is organized as a categorical question and we are therefore unable to use a simple continuous measure of union density. One alternative is to create dummies corresponding to all but one of the CRANET categories. The problem with this approach, however, is that national differences in the level and structure of union membership will be controlled for by the country variables described above. We do not wish to do this, as we wish to keep these variables distinct from other variables. The benefit of the rank fractions procedure is that controls for the level and structure of union membership are in effect built into the union membership variable itself. The measures to test human resource management/human capital arguments are as follows. We measure the general sophistication of HRM with a variable that records whether there is a written HRM strategy. We also measure whether there is a person responsible for HRM on the company board, on the grounds that board presence may lead to more sophisticated HRM. We include a measure for the proportion of the workforce who have been on a formal training activity within the last year, and a dummy variable for the use of performance appraisal for manual or clerical workers (alternatively, appraisal might be seen as an alternative solution to monitoring problems rather than as a complement to financial participation). Finally, we include several organizational and performance variables. We record whether the organization is foreign-owned. We also include a dummy for subsidiary status to control for the differences (mentioned earlier) in organizational type contained within the data-set. There is a measure for productivity (the top 10% are coded as 1) on the grounds that use of financial participation has been widely observed to be associated with higher levels of performance in the literature (see e.g. Kruse & Weitzman, 1990). Finally, we include a measure for adverse cash-flow on the basis that firms with cash-flow
Employee Share Ownership and Profit Sharing in Europe
157
difficulties may use share schemes as a substitute for cash wages. Whilst we hypothesize that this variable has a positive association with share schemes, we expect a negative relationship with profit sharing for obvious reasons. Details of each variable are summarized in Table 5. 5.1. Country Influences It is clear from the results that country specific influences are very important in explaining patterns of broad-based financial participation. In the case of share ownership every country dummy, except those for the Netherlands and Portugal, displays a significant negative sign relative to the U.K. This undoubtedly reflects the well-developed state of legislative and fiscal support for share ownership schemes in the U.K. The picture is slightly different in the case of profit sharing. France and the Netherlands show significant positive signs, reflecting well-developed legislation in support of profit sharing in those countries. Sweden, Spain, Denmark, Italy, Greece and Belgium have significant negative signs. Overall, the significance of country dummies indicates the importance of country specific factors relative to organizational factors in explaining national differences in the use of financial participation. The problem with using country dummies is that they combine together all possible influences on the use of financial participation emanating from national regime characteristics. It is not possible to discern what the primary country level influences actually are. To deal with this problem we replaced the country dummies with two measures of potential influences on the use of financial participation. The first is a measure of tax concessions available for share ownership or profit sharing. The second is a measure of the relative importance of national equity markets, as indicated by the market value of domestic listed companies as a proportion of GDP. As can be seen in Table 6, both measures are significantly associated with the presence of share ownership schemes, suggesting that both tax and stock market factors influence the incidence of share schemes. The highly significant association between stock market listing and share schemes further indicates the importance of equity market factors in influencing patterns of employee share ownership. Tax factors are also very important in explaining the incidence of profit sharing. However, there is a negative association between stock market size and profit sharing (and the market listing variable is insignificant). This implies that in some instances profit sharing may be an alternative to share ownership schemes when stock market factors are less facilitative of share ownership schemes.
Proportion of professional/technical employees in the workforce Measure of adverse cashflow (where revenue insufficient to cover costs) High relative rate of productivity (in top 10% of sector for productivity) Information released to non-managerial employees on business strategy Information released to non-managerial employees on company finances Presence of a works council Total employment in the company
Listed on a stock market
CASHFLOW
PRODUCTIVITY
INFORMATION ON STRATEGY
INFORMATION ON FINANCE
WORKS COUNCIL
SIZE
LISTED SUBSIDIARY
Proportion of manual employees in the workforce
PROFESSIONAL/ TECHNICAL (%)
MANUAL (%)
0,1 0,1
Natural log
0,1
0,1
0,1
0,1
0,1
Percentage
Percentage
Percentage
Proportion of the workforce who are graduates
GRADUATES (%)
Percentage
0,1
0,1
Proportion of the workforce participating in a training activity in the last year
TRAINING 0,1
Use of performance appraisal for non-managerial staff
APPRAISAL
0,1 0,1
High relative rate of innovation (in top 10% of sector for innovation)
Presence of a written HRM strategy
WRITTEN
INNOVATION
Presence of a director with responsibility for HRM
HRM DIRECTOR
Variable type
GROWTH OPPORTUNITIES Measure of market growth
Variable content
Independent Variables Used in the Analysis.
Variable name
Table 5.
0.42 0.44
7.97 (21,495)
0.75
0.66
0.39
0.32
0.09
22.68
48.58
20.94
0.272
0.595
47.16
0.73
0.48
0.58
Mean
158 A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
Union density from 26–50% Union density from 51–75% Union density from 76–100% Country dummy (U.K. as the benchmark) As above As above As above As above As above As above As above As above As above As above As above Share schemes taxed under capital gains tax regime Profit shares attract income tax/social security concessions Market valuation of domestic listed firms as a proportion of GDP in 1997
VERY HIGH DENSITY
FRANCE
GERMANY
SWEDEN
SPAIN
DENMARK
NETHERLANDS
ITALY
PORTUGAL
FINLAND
GREECE
AUSTRIA
BELGIUM
TAX (ESO)
TAX (PS)
EQUITY MARKETS
Union density from 10–25%
LOW DENSITY
HIGH DENSITY
Measure of domestic ownership
FOREIGN
Continued.
MEDIUM DENSITY
Variable content
Variable name
Table 5.
Percentage
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
0,1
Variable type
84.23
0.39
0.54
0.06
0.05
0.04
0.05
0.04
0.02
0.02
0.06
0.73
0.06
0.17
0.11
0.17
0.13
0.21
0.11
0.32
Mean
Employee Share Ownership and Profit Sharing in Europe 159
HRM DIRECTOR WRITTEN APPRAISAL TRAINING GROWTH OPPORTUNITIES INNOVATION GRADUATES (%) MANUAL (%) PROFESSIONAL/TECHNICAL (%) CASHFLOW PRODUCTIVITY INFORMATION ON STRATEGY INFORMATION ON FINANCE WORKS COUNCIL LOW DENSITY MEDIUM DENSITY HIGH DENSITY VERY HIGH DENSITY SIZE LISTED
Variable
Table 6.
–311* (0.152) 0.149 (0.149) 0.088 (0.184) 0.768** (0.265) 0.158 (0.140) 0.068 (0.151) 0.663 (0.409) –0.424 (0.503) 0.833 (0.603) –0.304 (0.289) 0.114 (0.148) 0.139 (0.148) 0.127 (0.171) 0.118 (0.194) 0.031 (0.234) –0.235 (0.248) 0.117 (0.234) –0.032 (0.249) 0.160*** (0.039) 1.181*** (0.143)
B 0.732 1.160 1.092 2.156 1.171 1.070 1.940 0.654 2.301 0.738 0.442 1.149 1.136 1.125 1.032 0.791 1.124 0.986 1.173 3.259
Odds
Logistic regression
–0.269 (0.146) 0.144 (0.482) 0.100 (0.180) 0.750** (0.259) 0.135 (0.139) 0.080 (0.150 0.646 (0.403) –0.633 (0.482) 0.816 (0.581) –0.293 (0.288) 0.109 (0.146) 0.163 0.147) 0.180 (0.166) 0.209 (0.174) –0.023 (0.230) –0.294 (0.240) 0.166 (0.221) 0.057 (0.225) 0.163*** (0.039) 1.145*** (0.142)
B
Characteristics of Organisations With Share Ownership Schemes.
0.764 1.154 1.105 2.116 1.145 1.084 1.908 0.531 2.262 0.746 1.115 1.177 1.197 1.233 0.977 0.746 1.180 1.059 1.177 3.143
Odds
160 A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
2235 409.062*** 1496.848 85.9
0.008
–4.436*** (0.622)
Odds 2.235 0.269 0.504 0.346 0.309 0.136 0.508 0.724 0.045 0.374 0.001 0.367 0.266 0.146 0.381
B
Continued.
0.804*** (0.155) –1.313*** (0.181) –0.685** (0.253) –1.062*** (0.244) –1.175*** (0.329) –1.994*** (0.391) –0.676* (0.297) –0.323 (0.389) –3.096** (1.040) –0.984 (0.391) –6.661 (6.896) –1.1001** (0.340) –1.323** (0.472) –1.923*** (0.496) –0.965** (0.337) –
Note: Sector dummies included (all insignificant).
N Model chi–square 2 log–likelihood Percentage correct predictions
SUBSIDIARY FOREIGN FRANCE GERMANY SWEDEN SPAIN DENMARK NETHERLANDS ITALY IRELAND Portugal FINLAND GREECE AUSTRIA BELGIUM TAX (ESO) EQUITY MARKETS Constant
Variable
Table 6.
2235 381.671*** 1524.238 86.2
– – – – – 0.549** (0.183 0.006*** (0.001) –5.363*** (0.545)
0.837*** (0.153) –1.360 (0.174) – – – – – – –
B
1.732 1.006 0.005
2.309 0.257
Odds
Employee Share Ownership and Profit Sharing in Europe 161
HRM DIRECTOR WRITTEN STRATEGY APPRAISAL TRAINING GROWTH OPPORTUNITY INNOVATION GRADUATES (%) MANUAL (%) PROFESSIONAL/TECHNICAL (%) CASHFLOW PRODUCTIVITY INFORMATION ON STRATEGY INFORMATION ON FINANCE WORKS COUNCIL LOW DENSITY MEDIUM DENSITY HIGH DENSITY VERY HIGH DENSITY SIZE LISTED
Variable
Table 7.
0.039 (0.124) 0.098 (0.124) 0.132 (0.142) 0.135 (0.221) –0.073 (0.117) 0.077 (0.130) –0.176 (0.375) –1.194** (0.432) –0.319 (0.541) –0.807*** (0.240) 0.181 (0.124) 0.164 (0.125) 0.548*** (0.142) 0.074 (0.165) –0.008 (0.190) –0.403* (0.200) –0.289 (0.199) 0.017 (0.203) 0.030 (0.033) 0.254 (0.121)
B 1.039 1.103 1.141 1.145 0.930 1.080 0.838 0.303 0.727 0.446 1.198 1.178 1.729 1.077 0.992 0.668 0.749 1.017 1.031 1.289
Odds 0.076 (0.117) 0.057 (0.118) 0.223 (0.135) 0.078 (0.210) –0.074 (0.112) 0.050 (0.126) –0.061 (0.352) –1.171** (0.389) –0.422 (0.500) –0.572* (0.223) 0.076 (0.120) 0.113 (0.119) 0.421*** (0.129) 0.169 (0.144) –0.136 (0.177) –0.543** (0.187) –0.603*** (0.187) –0.552*** (0.171) 0.056 (0.032) 0.177 (0.114)
B
Characteristics of Organizations With Profit Sharing.
1.079 1.058 1.250 1.081 0.928 1.051 0.941 0.310 0.656 0.564 1.078 1.120 1.523 1.184 0.873 0.581 0.547 0.576 1.057 1.193
Odds
162 A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
N Model chi–square 2 log likelihood Percentage correct predictions
SUBSIDIARY FOREIGN FRANCE GERMANY SWEDEN SPAIN DENMARK NETHERLANDS ITALY IRELAND PORTUGAL FINLAND GREECE AUSTRIA BELGIUM TAX (PS) EQUITY MARKETS Constant
Variable
0.322
–1.132*(0.453) 2235 565.778*** 2044.467 79.5
1.340 0.523 10.304 0.623 0.437 0.373 0.221 3.271 0.260 1.200 0.740 0.733 0.237 0.901 0.254
Odds
Continued.
0.293* (0.134) –0.648*** (0.151) 2.333*** (0.237) –0.473* (0.196) –0.828** (0.278) –0.987*** (0.291) –1.507*** (0.350) 1.185*** (0.323) –1.346** (0.553) 0.182 (0.295 ~ ) –0.302 (0.380) –0.311 (0.274) –1.441** (0.459) –0.104 (0.263) –1.372*** (0.341) –
B
Table 7.
7.282 0.991 0.244
1.985*** (0.150) –0.009*** (0.001) –1.410*** (0.424) 2235 448.310*** 2161.935 78.7
1.416 0.510
Odds
0.348** (0.128) –0.673*** (0.144)
B
Employee Share Ownership and Profit Sharing in Europe 163
164
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
5.2. Influences Arising from Monitoring Costs There is little support for the argument that financial participation is used to resolve monitoring difficulties. The suggestion, derived from predictions in the Financial Economics literature, that firms with relatively high growth opportunities will use financial participation is not borne out as the relevant variable is insignificant in each estimation. Furthermore, innovative firms are not more likely to use either profit sharing or share ownership. The various measures of workforce composition are also insignificant (though signs are in the expected direction) in most instances. However, profit sharing organizations have a significantly lower proportion of manual workers, as predicted by monitoring arguments. The measure of company size is significantly and positively related to the presence of share ownership schemes in both estimations. This is entirely consistent with findings from elsewhere in the literature but has ambiguous implications for employee monitoring. Whilst monitoring costs may be proportionally higher in larger organizations, free rider effects are likely to counteract these. Given the strength of the associations shown (employee share ownership is clearly a large-firm phenomenon), it seems likely that administrative economies of scale may be important in explaining the distribution of share schemes. The results are less clear-cut for profit sharing: coefficients are positive but insignificant suggesting that the balance between monitoring costs and free rider effects are more finely balanced. 5.3. Employee Participation and Representation In contrast to predictions from the literature, participation and communication variables are not significantly related to the use of financial participation in most cases. The presence of a works council or joint consultative committee is not significantly related to either profit sharing or employee share ownership. This may reflect the greater statutory provision for consultative institutions in Europe than in the U.S. Disclosure of information on finances and business strategy is not significantly associated with the presence of share ownership. However, disclosure of financial information is significantly related to the use of profit sharing. This is consistent with the suggestion that employees will require more information about company finances when some of their remuneration is linked to the financial performance of the firm. However, this does not occur in the case of share ownership schemes as the stock market provides the relevant information instead.
Employee Share Ownership and Profit Sharing in Europe
165
As far as the dummies for various levels of trade union membership are concerned, attention should focus on the estimations with country dummies included as these control for national differences in union density levels. It is noticeable that most of the dummies are negatively significant in the profit sharing results without country dummies – this may well reflect a ‘France effect’ (very high incidence of profit sharing, very low level of union membership). When country dummies are included, the trade union dummies are insignificant in the share ownership model. The dummy for 26–50% density is significant (negatively) for profit sharing. Overall, it is reasonable to conclude that levels of trade union membership do not have a strong or clear relationship with the use of financial participation. 5.4. Human Resource Management Various measures are used to test the proposition that organizations with more sophisticated HRM will be more likely to use financial participation. With one important exception, the results do not provide much support for this proposition. The presence of a formal, written HRM strategy does not predict the use of financial participation, whilst the presence of a HRM director is either insignificant (profit sharing) or negatively related to the use of employee share schemes. The use of appraisal for manual or clerical workers is not related to financial participation. However, the measure of training – the proportion of workers receiving formal training in the last year – is significantly related to the presence of employee share schemes. This supports a human capital rationale for financial participation. Share schemes provide a pay-off for employees for investments in training, whilst binding them to the firm in the medium term (because of the deferred character of share-based rewards). 5.5. Organizational Finally, structural and organizational measures tend to be significant, even after controlling for country effects. Share ownership organizations, in particular, are large, domestically-owned, and listed. However, there is no clear evidence that either profit sharing or share ownership organizations are better performers. Contrary to suggestions in some of the literature, share ownership firms do not suffer from relatively poor cash-flow (it has been suggested that cash-strapped new economy firms may use share schemes as a substitute for cash remuneration). Not surprisingly, there is some evidence that profit sharing firms have better than average cashflow.
166
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
Finally, the positive and significant coefficients for subsidiary should be noted. It will be recalled that the dataset comprises organizations where the observations apply to the entire company and some where they apply to just one part of the firm. A dummy (equal to one where the observation was not of the entire company) was entered in the analysis to make this difference explicit. As can be seen, those cases where the organization and company were not the same are more likely to have either type of financial participation. This might raise serious methodological problems as these organizations may have substantial but unobservable differences with the remainder of the cases in the study. To deal with this problem we compared the two groups of organizations in respect of each of the organizational-level variables used in the analysis. There were significant differences but primarily in respect of those variables that are already significant (i.e. controlled for) in the multivariate analysis (training, size, stock market listing etc). For this reason, this shortcoming of the dataset does not appear to invalidate the findings of the study.
6. CONCLUSIONS A number of broad conclusions can be derived from this comparative study. First, it is clear that country level factors are a key influence on the incidence of broadly based, all employee financial participation in Europe, and that for the most part these are more important than various measures of organizational characteristics. Both tax and equity market factors are shown to be important. By contrast, management-only schemes appear to be driven by factors other than legislation. In most countries the incidence of narrow-based schemes is higher than broad-based despite the lack of fiscal supports for selective schemes. Second, the other key set of influences on the incidence of financial participation is what might be termed ‘basic’ structural features of organizations, such as size, domestic ownership etc. The measure of company size is significantly and positively related to the presence of share ownership schemes in both estimations. This is entirely consistent with findings from elsewhere in the literature but has ambiguous implications for employee monitoring. Whilst monitoring costs may be proportionally higher in larger organizations, free rider effects are likely to counteract these. Given the strength of the associations shown (employee share ownership is clearly a large-firm phenomenon), it seems likely that administrative economies of scale may be important in explaining the distribution of share schemes. The results are less clear-cut for profit sharing: coefficients are positive but insignificant suggesting
Employee Share Ownership and Profit Sharing in Europe
167
that the balance between monitoring costs and free rider effects are more finely balanced. Third, workforce characteristics and market opportunities (as proxies for monitoring costs) do not appear to be strong influences on the use of financial participation. Nor, contrary to other studies, do various measures of workforce participation and representation. Finally, whilst there is little support for the notion of HRM ‘bundles’ the association between share schemes and training is a potentially important one. This supports a human capital rationale for financial participation. Share schemes provide a pay-off for employees for investments in training, whilst binding them to the firm in the medium term (because of the deferred character of share-based rewards). For researchers, these results support the findings of some previous studies and challenge some theoretical positions, especially those concerning monitoring costs in case of broad-based schemes. Future research could usefully be focused on assessing the relative strengths of these antecedents and whether they change over time.
NOTES 1. An European Union agency based in Dublin. 2. This set of studies was financially supported by the European Commission through the Human Capital and Mobility programme. The results are reported in IPSE (1997), and in numerous papers by members of the country teams. See Biagioli (1995), Biagioli and Curatolo (1997), Carstensen, Gerlach and Hubler (1995), Del Boca and Cupaiuolo (1998), Fakhfakh and Perotin (1993), Mabile (1998), Perotin and Robinson (1998). 3. See Poutsma and Huijgen (1999), Poutsma, de Nijs and Doorewaaard (1999), Festing, Groening, Kabst and Weber (1999). 4. For further details of the methodology of the Cranet survey see Mayrhofer (2000), Brewster et al. (2000), Brewster et al. (1996). 5. The first survey (of five countries) took place in 1989. Since then, the Cranet survey has now been administered in three full rounds: 1992, 1995 and 1999/2000. 6. A further weighting procedure was applied to Germany, as this country was considerably over-represented according to the first procedure. The German cases were reduced by a factor of 2. 7. There are a large number of cases where the measure is zero. 8. Our preliminary analysis confirmed this to be the case. See Pendleton et al. (1991) for further information on this point. 9. This is preferable to dividing the sub-sample at the median as there is a highly skewed distribution. This could result in schemes with high coverage being classed as narrowly based.
168
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
10. All-employee schemes typically do not have 100% participation at any given time because it is usual to have minimum employment periods before eligibility is granted. 11. The diverse pattern of broad-based financial participation throughout Europe is also supported by secondary analysis of a ten country survey data from the EPOC project in 1996. This survey covers smaller establishments (not business units) and asked for the existence of broad based financial participation plans. According to this survey 15% of establishments with more than 200 employees had a share scheme for its largest occupational group. In case of profit sharing the proportion was about 30%. The differences between countries had a similar pattern in this survey. 12. In Germany stock options were generally prohibited until 1998 so cash profitbased bonuses may have substituted for stock options in managerial compensation. 13. We attempted to generate an array of more precise and more objective measures of legislative/fiscal support but this proved extremely difficult as it is not readily apparent which specific aspects of tax/legislative frameworks are most important. Furthermore, most countries have some government regulation of financial participation but it is difficult to grade differences of degree. 14. The accounting-based studies in Financial Economics usually use Tobin’s Q as the relevant measure but, in the absence of accounting and stock market data, our chosen measures seem the most appropriate.
ACKNOWLEDGMENTS The research upon which this paper is based was commissioned by the European Foundation for the Improvement of Living and Working Conditions, based in Dublin. We are grateful to the Foundation for its support. In particular we would like to thank Kevin O’Kelly and Hubert Krieger for their interest and encouragement. The data used in the paper was collected by members of the CRANET network and we are most grateful for their efforts.
REFERENCES Alchian, A., & Demsetz, H. (1972). Production, information costs, and economic organisation. American Economic Review, 62, 777–795. Becker, F. G. (1993). Strategische Ausrichtung von Beteiligungssystemen. In: W. Weber (Ed.), Entgeltsysteme (pp. 313–338). Stuttgart: Lohn, Mitarbeiterbeteiligung und Zusatzleistungen. Becker, B., & Huselid, M. A. (1998). High Performance Work Systems and Firm Performance: A Synthesis of Research and Managerial implications. In: G. R. Ferris (Ed.), Research in Personnel and Human Resources Management, 16, 53–102. Ben-Ner, A., Allen Burns, W., Dow, G., & Putterman, L. (2000). Employee ownership: an empirical explanation. In: M. Blair & T. Kochan (Eds), The New Relationship: Human Capital in the American Corporation (pp. 194–233). Washington, D.C.: Brookings Institution.
Employee Share Ownership and Profit Sharing in Europe
169
Biagioli, M. (1995). Italy: decentralization of wage bargaining and financial participation. In: D. Vaughan-Whitehead (Ed.), Workers’ Financial Participation. East-West Experiences (pp. 85–104). Geneva: International Labour Office. Biagioli, M., & Curatolo, S. (1997). La partecipazione dei lavoratori ai risultati economici delle imprese: una indagine econometrica su un panel di aziende metalmeccaniche di dimensioni medio-grandi. In: M. Biagioli (Ed.), L’Analisi Economica delle Relazioni Industriali. Modelli Teorici e Studi Empirici sull’Esperienza Italiana (pp. 187–228). Naples: Edizioni Scientifiche Italiane. Blair, M. (1995). Ownership and Control: Rethinking Corporate Governance for the Twenty-first Century. Washington, D.C.: Brookings Institution. Blair, M. (1999). Firm-specific human capital and theories of the firm. In: M. Blair & M. Roe (Eds), Employees and Corporate Governance (pp. 58–90). Washington, D.C.: Brookings Institution. Brewster, C. (1999). Strategic human resource management: The value of different paradigms. In: M. Festing (Ed.), Strategic issues in international human resource management. Management International Review, 39(3), 45–64. Brewster, C., Tregaskis, O., Hegewisch, A., & Mayne, L. (1996). Comparative Research in Human Resource Management: A review and an example. International Journal of Human Resource Management, 7(3), 585–604. Brewster, C., Tregaskis, O., Hegewisch, A., & Mayne, L. (2000). Comparative research in human resource management: A review and an example. In: C. J. Brewster, W. Mayrhofer & M. Morley (Eds), New Challenges for European Human Resource Management (pp. 324– 348). London: Macmillan. Carstensen, V., Gerlach, K., & Hübler, O. (1995). Profit sharing in German firms. In: F. Buttler, W. Franz, R. Schettkat & D. Soskice (Eds), Institutional Frameworks and Labor Market Performance. Comparative Views on the U.S. and German Economies (pp. 168–207). London and New York: Routledge. Cheadle, A. (1989). Explaining patterns of profit sharing activity. Industrial Relations, 28(3), 398–400. Commission of the European Communities (CEC) (1996). Report from the Commission: Pepper II: Promotion of Participation by Employed Persons in Profits and Enterprise Results (including equity participation) in Member States. Brussels: Commission of the European Communities, Com (96) 697. Commission of the European Communities (CEC) (2002). Communication from the Commission to the Council, the European Parliament, the Economic and Social Committee, and the Committee of the Regions on a Framework for the Promotion of Employee Financial Participation. Brussels: Commission of the European Communities COM (2002) 364. Conyon, M., Peck, S., & Read, L. (2001). Performance pay and corporate structure in U.K. firms. European Management Journal, 19(1), 73–82. Del Boca, A., & Cupaiuolo, E. (1998). Why do firms introduce financial participation? Economic Analysis, 1(3), 221–237. Fakhfakh, F., & Pérotin, V. (1993). The effects of profit-sharing on firm performance in France. Paper presented at the EALE Conference, Maastrichht (September). Festing, M., Groening, Y., Kabst, R., & Weber, W. (1999). Financial Participation in Europe – Determinants and Outcomes. Economic and Industrial Democracy, 20(2), 295–329. FitzRoy, F. R., & Kraft, K. (1987). Cooperation, productivity and profit sharing. The Quarterly Journal of Economics, 102, 23–35.
170
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
Gregg, P. A., & Machin, S. J. (1988). Unions and the incidence of performance linked pay schemes in Britain. International Journal of Industrial Organisation, 6(1), 69–90. Ichniowski, C., Shaw, K., & Prennushi, G. (1997). The Effects of Human Resource Management Practices on Productivity: A Study of Steel Finishing Lines. The American Economic Review, 87(3), 291–313. IPSE (1997). Le partage de profit en Europe. Institutions et effets comparés. Cahier Travail et Emploi, Paris. Jones, D. C., & Kato, T. (1993). On the Scope, Nature and Effects of Employee Stock Ownership Plans in Japan. Industrial and Labor Relations Review, 46(2), 352–367. Kato, T., & Morishima, M. (Forthcoming). The Nature, Scope and Effects of Profit Sharing in Japan: Evidence from New Survey Data. International Human Resource Management Journal, forthcoming. Kato, T., & Motohiro, M. (2001). The Productivity Effects of Participatory Employment Practices: Evidence From New Japanese Panel Data. Forthcoming in Industrial Relations. Kruse, D. (1996). Why do firms adopt profit sharing and employee ownership plans. British Journal of Industrial Relations, 34, 515–538. Kruse, D., & Weitzman, M. (1990). Profit-sharing and productivity. In: A. Blinder (Ed.), Paying for Productivity – A Look at the Evidence (pp. 95–140). Washington, D.C.: The Brookings Institution. Levine, D., & Tyson, A. (1990). Participation, productivity and the firm’s environment. In: A. Blinder (Ed.), Paying for Productivity: a Look at the Evidence. Washington, D.C.: Brookings Institution. McNabb, R., & Whitfield, K. (1998). The impact of financial participation and employee involvement on financial performance. Scottish Journal of Political Economy, 45(2), 171–189. Mabile, S. (1998). Interessement et Salaries: Complementarite or Substitution? Economie et statistique, 316–317. Mayrhofer, W. (2000) Coordination of research networks: market, bureaucracy and clan in the Cranfield Network on European Human Resource Management (Cranet-E). In: C. J. Brewster, W. Mayrhofer & M. Morley (Eds), New Challenges for European Human Resource Management (pp. 303–323). London: Macmillan. Pendleton, A. (1997). Characteristics of workplaces with financial participation: evidence from the Workplace Industrial Relations Survey. Industrial Relations Journal, 28, 103–119. Pendleton, A., & Perotin, V. (2001). Profit Sharing in Europe: The Characteristics and Impact of Profit Sharing in France, Germany, Italy and the United Kingdom. Cheltenham: Edward Elgar. Perotin, V., & Robinson, A. (1998). A comparative paper on profit sharing and productivity. Evidence from Britain, France, Germany and Italy. Advances in the Economic Analysis of Participatory and Labour-Managed Firms, 6, 135–162. Poole, M. (1988). Factors affecting the development of employee financial participation in contemporary Britain: evidence from a national survey. British Journal of Industrial Relations, 26(1), 21–36. Poole, M. (1989). The Origins of Economic Democracy. London: Routledge. Poole, M., & Jenkins, G. (1990). The impact of economic democracy. Profit-sharing and employee-shareholding schemes. London: Routledge. Poole, M., & Whitfield, K. (1994). Theories and evidence on the growth and distribution of profit sharing and employee shareholding schemes. Human Systems Management, 13, 209–220.
Employee Share Ownership and Profit Sharing in Europe
171
Poutsma, E. (2001). Recent Trends in Employee Financial Participation in the European Union. Luxembourg: Office for Offical publications of the European Commission. Poutsma, E., Huijgen, F. (1999). European diversity in the use of participation schemes. Economic and Industrial Democracy, 20(2), 197–224. Poutsma, E., de Nijs, W., & Doorewaard, H. (1999). Promotion of employee ownership and profitsharing in Europe. Economic and Industrial Democracy, 20(2), 171–196. Ramsay, H. (1977). Cycles of control: worker participation in sociological and historical perspective. Sociology, 11, 481–506. Sesil, J., Kroumouva, M., Blasi, J., & Kruse, D. (2002). Broad-based employee stock options in U.S. ‘new economy’ firms. British Journal of Industrial Relations, 40(2), 273–294. Uvalic, M. (1991). Pepper Report. Promotion of employee participation in profits and enterprise results. In: Social Europe, Supplement 3/91, Commission of the European Communities. Vaughan-Whitehead, D., et al. (1995). Workers’ Financial Participation: East-West Experiences, ILO Labour Management Series No. 80. International Labour Office (ILO), Geneva. Wächter, H., & Koch, T. (1993). Erfolgsbedingungen fr die Kapitalbeteiligung von Arbeitnehmern am Beispiel der ’Employee Stock Ownership Plans (ESOPs). In: W. Weber (Ed.), Entgeltsysteme (pp. 285–312). Stuttgart: Lohn, Mitarbeiterbeteiligung und Zusatzleistungen. Whitfield, K., Delbridge, R., & Brown, W. (1998). Using workplace surveys for comparative research. In: K. Whitfield & G. Strauss (Eds), Researching the World of Work: Strategies and Methods in Studying Industrial Relations (pp. 193–212). Ithaca: Cornell University Press. Yermack, D. (1995). Do corporations award stock options effectively? Journal of Financial Economics, 39(2–3), 237–269.
172
A. PENDLETON, E. POUTSMA, J. VAN OMMEREN AND C. BREWSTER
APPENDIX 1 COMPARISION OF CRANET FINDINGS WITH EPOC FINDINGS As mentioned in the text, the CRANET survey is a unique source due to its high coverage of countries. It is difficult to assess whether the distributions shown above are soundly-based either for specific countries or for Europe as a whole. However, we are able to compare these results with those derived from the Employee Participation in Organisational Change (EPOC) survey conducted in 1996, bearing in mind that this was an establishment-level survey. A comparison of findings is presented in the next table. Financial Participation Schemes by Country. Country
Broad based Profit-sharing (EPOC 1996) > 50 employees N = 46041
Broad based Share ownership (EPOC 1996) > 50 employees N = 46041
Broad based Profit-sharing (CRANET 2000) > 200 employees N = 25062
Broad based Share ownership (CRANET 2000) > 200 employees N = 25062
Austria Belgium Denmark France Germany Greece Ireland Italy Netherlands Portugal Spain Finland Sweden U.K.
– – 10 57 13 – 8 5 14 7 8 – 20 40
– – 6 7 4 – 4 3 4 2 10 – 2 23
25 12 8 84 18 7 24 8 55 17 13 27 19 30
4 11 15 23 10 7 16 2 21 2 5 15 12 30
Total
23
9
36
16
Source 1: Poutsma, 2001. Source 2: Pendleton et al., 2001.
THE USE OF PROFIT SHARING WHEN WORKERS MAKE DECISIONS: EVIDENCE FROM A SURVEY OF MANUFACTURING WORKERS Christopher P. Adams ABSTRACT This paper examines the observed complementarity between the use of profit sharing and the delegation of decision making power to production line workers. The paper makes use of a new economy wide data set which provides information on pay and decision making for a large number of production workers. A principal-agent model is presented to guide the empirical analysis. An empirical model is presented that allows the use of profit sharing and the delegation of decision making power to be explicit complements and for this complementarity to vary with observable and unobservable characteristics of the workers. Two major results are derived from the theoretical model and supported by the empirical analysis. First, workers with greater than two years experience are more valuable decision makers. Second, conditional on the worker having decision making, profit sharing is of a greater value to the firm when demand for the firm’s product is volatile.
The Determinants of the Incidence and the Effects of Participatory Organizations, Volume 7, pages 173–209. © 2003 Published by Elsevier Science Ltd. ISBN: 0-7623-1000-6
173
174
CHRISTOPHER P. ADAMS
1. INTRODUCTION In many modern manufacturing plants, firms delegate important decisions on how their product should be made to their production line workers. Many of these same firms give their production line workers performance based pay schemes such as profit sharing, employee share ownership and employee stock options.1 The empirical literature suggests that the value of delegating decision making power to the shop floor is enhanced by profit sharing schemes, that is, these practices are complements (Appelbaum et al., 2000; Ichniowski et al., 1997; Osterman, 1994). However, there is little research on why this complementarity exists, or how features of the product market affect its magnitude.2 This paper analyzes this complementarity between the delegation of decision making power to production line workers and the use of profit sharing schemes.3 Understanding this complementarity is important because changes in government policy or market conditions that lead to an increase in the use of one (e.g. tax breaks for employee share ownership) will increase the value and the use of the other. To understand how product market conditions and other firm characteristics affect firm decision making structures, it is advantageous to have data on the amount of decision making power that is actually delegated to production line workers.4 This information has recently become available in a number of large economy wide data sets.5 To this author’s knowledge, there has been little empirical analysis of the information these data sets provide on individual employee decision making.6 To guide the empirical analysis, this paper develops a principal-agent model in which the firm (the principal) determines whether decision making power is given to the worker (the agent) and whether profit sharing is used to compensate the worker. This model differs from the standard principal-agent model by assuming that the worker has private information. This assumption follows immediately from the statement that workers are given decision making power. A principal-agent model in which the agent has private information is called a “generalized principal-agent” (i.e. moral hazard and adverse selection) model (Myerson, 1982, 1985). However, the application to worker decision making and pay schemes leads naturally to a substantial restriction on the contract space allowed by Myerson.7 At the heart of this model are two important ideas. First, when the worker has the better information, the firm should delegate the decision on what action the worker should choose to the worker (Adams, 2001). Second, the firm should use profit sharing when it is difficult to monitor the worker and to guarantee the firm’s preferred action (Drago & Heywood, 1995; Jones &
The Use of Profit Sharing When Workers Make Decisions
175
Pliskin, 1997). These two ideas lead by implication to the paper’s two main theoretical results. The first result states that conditional on the type of incentive scheme used, the firm values worker decision making more when the worker has more accurate information regarding the circumstances affecting the action choices. It seems reasonable that giving decision making power to someone with better information is going to improve the expected returns from the decision. The question asked in this paper is how the moral hazard problem is affected by the accuracy of the worker’s information. The answer is that the cost of the moral hazard problem is reduced when the worker has more accurate information. The second result states that conditional on the worker having decision making power, the firm values profit sharing more when it is more difficult to observe the circumstances affecting the worker’s action choices. The incentive scheme leads the worker to choose the appropriate action given the appropriate circumstance. Such a scheme becomes more costly to implement when there is greater noise (random variation) in the information used (Holmstrom, 1979). In the case presented in the paper, the noise can arise either because the worker has difficulty observing the circumstances or because the firm has difficulty observing the circumstances. Tests of hypotheses with regard to the firm’s choice on one margin (use of profit sharing) conditional on its choice on the other margin (the delegation of decision making) can be performed with a generalization of the bivariate probit model. The model developed below allows the choices on both margins to interact and for this interaction to vary across workers in observable and unobservable ways. Importantly, the model permits estimation of the probability of giving the worker profit sharing conditional on giving the worker decision making, while allowing both choices to be made simultaneously. This model allows the two human resource practices to be complements and for this complementarity to vary with measured and unmeasured characteristics of the worker. The analysis shows how the value of profit sharing is dependent upon the delegation of decision making power to production line workers. Previous work has recognized the importance of human resource practices which delegate decision making power, in determining the use of profit sharing schemes (for example see Drago & Heywood, 1995). However, the data and empirical modelling used in previous work has not been rich enough to explain why this complementarity exists. The measure of profit sharing in this data set is similar to the measures used in previous work.8 However this paper uses a more direct measure of the amount of decision making power given to the production line workers. Instead of inferring from the use of self-managed work teams (for
176
CHRISTOPHER P. ADAMS
example) that workers have greater decision making power, the paper uses data based on survey responses of workers themselves. The empirical model is more general than the models used in the previous work. In particular, the model not only allows the choice to use profit sharing and the delegation of decision making power to made simultaneously, it also allows for explicit (and measurable) interaction between the choices. The results show that for workers with and without decision making, characteristics of the firm’s product market are important determinants of the use of profit sharing. For workers without decision making, workers that produce high quality products are more likely to have profit sharing, while workers that produce for volatile product markets are less likely to have profit sharing. For workers with decision making power, both workers that produce high quality products and workers that produce for volatile product markets are more likely to have profit sharing. The rest of the paper is organized as follows. Section 2 presents the principal-agent model and the theoretical results. Section 3 discusses the data which is based on surveys of a large number of production workers and the manufacturing establishments that they work for. The section also discusses measurement issues. Section 4 presents the empirical model and the hypotheses implied by the theoretical model. Section 5 presents the results. Section 6 concludes.
2. THE MODEL AND PROPOSITIONS This section presents a model which formalizes the delegation of decision making power to a production line worker and how this delegation impacts the moral hazard problem. Two propositions summarize the main results of the model. Proposition 1 states that conditional on the payment scheme used, the firm values worker decision making greater when the worker’s information is more accurate. It seems straightforward that worker decision making would be better when workers have more accurate information. However, it is not clear what effect moral hazard has. This result states the moral hazard problem is actually reduced when the worker has more accurate information. Proposition 2 states that conditional on the worker having decision making power, the firm places greater value on profit sharing when it is more difficult to monitor the circumstances of the worker’s action choices. Previous empirical work has shown that firms that produce high quality products are more likely to use profit sharing (Drago & Heywood, 1995). The authors argue that this is because it is difficult to monitor a worker’s actions that relate to improving quality. The result presented below suggests that profit sharing may also be valuable when
The Use of Profit Sharing When Workers Make Decisions
177
firms give their workers decision making power, particularly, in cases when the firm has difficulty monitoring the reasons for the choices made by the worker. Consider a worker who has been given the power to slow or stop the production line when he observes a problem or errors occurring on the line.9 While it is relatively easy for the firm to observe that the line has been slowed, the firm may not observe the reason, that is they may not observe the problem or errors that caused the worker slow or stop the line. 2.1. The Principal-Agent Model The model consists of two players, a firm (the principal) and a worker (the agent). The firm offers the worker a binding contract, which describes the task the worker will choose and how the worker will be paid. The worker either accepts or rejects the firm’s contract offer given a common ex ante belief about the state of the world. The value of the worker’s outside option is 0. If he accepts the contract, the worker then observes his private information regarding the state of the world and chooses a task.10 The firm is risk neutral and the worker is risk averse with a utility function that is separable in money, u : 2 → , such that u(, e) = v() e, v(0) = 0, v > 0 and v < 0, where is the monetary payment and e is the “effort” cost of the task chosen. The following time line summarizes the model: 1. The firm and worker have a common belief regarding the state (s{0, 1}, f = Pr(s = 1)). 2. The firm makes a take-it-or-leave-it contract offer, {t(Ws), (t, Fs, r)}, where t{0, 1} is the worker’s task choice, Ws{0, 1} is the worker’s signal of the state, Fs{0, 1} is the firm’s signal of the state, r{0, 1} is the firm’s revenue, and is the payment to the worker. 3. The worker accepts or rejects the contract offer. 4. The worker observes a signal of the state Ws, and updates his belief regarding the state ( f1 = Pr(s = 1 Ws = 1) and f0 = Pr(s = 1 Ws = 0), such that f1 > f0). 5. The worker chooses a task t(Ws), knowing that the cost of t = 1 is greater than the cost of t = 0 by the amount e > 0. 6. The firm observes t and Fs, receives revenue r and pays the worker (t, Fs, r). However, the payment scheme is either based on t and Fs, or the payment scheme is based on r, and not any other combination. The first part of the firm’s problem is to allocate to the worker a particular task t : {0, 1} → {0, 1} given the private information of the worker (Ws). For example, the tasks may be either to run the production line fast or to run the line
178
CHRISTOPHER P. ADAMS
slow, the worker’s signal may be the number of errors occurring on the production line and the state of the world may be the complexity of an order (in terms of production difficulty). The worker’s information has value to the firm because the firm’s stochastic revenue r{0, 1} has a distribution that depends on both the task chosen by the worker and the state of the world. The probability distribution can be written as follows. Pr(r = 1t, s) = p0 p1 1 p1
if t = 0, if t = 1, s = 1 if t = 1, s = 0
(1)
where p1 > p0 > 1 p1. The firm can get greater expected revenue by asking the worker to choose task t = 1, however this is only true if the state s = 1. If the state s = 0 then the firm’s expected revenue is greater when it asks the worker to choose t = 0. Continuing with the same example, if the line is run slow (t = 0) then the production line can produce the product with few errors irrespective how complicated the product is, however the run takes significantly longer and hence is less profitable (expected revenue is p0). If the line is run fast, the product can be finished quickly as long as it is relatively simple to make (expected revenue is p1 > p0). If the line is run fast and the product is relatively complex, then there will be many errors and re-dos and the firm would lose more revenue than if the line were slowed down (expected revenue is 1 p1 < p0). In order to reduce the number of cases it is also assumed that p0 > (1 f )(1 p1) + fp1. That is, given just the ex ante belief the firm will receive higher expected revenue if t = 0. The firm also observes information about the state of the world denoted Fs. However, this information is assumed to be received “too late” to be valuable for decision making, although this information may still be “early” enough to be useful for monitoring the worker’s decision making. Figure 1 represents the expected return of each task given the firm and the worker’s common ex ante belief f that the state s = 1. The figure and Eq. (1) represent the fact that if the worker chooses task t = 0, then the probability of r = 1 is independent of the state and equal to p0. If the worker chooses t = 1 the expected return depends upon the state and is higher (p1) if the state is s = 1 and lower (1 p1) if the state is s = 0. Normally in a principal-agent problem the worker’s task choice is a scalar variable. However, in this case the worker observes private information about the state before making his choice. Therefore the firm has four choices. It could ask the worker to always choose one task for all observations of Ws (t(0) = t(1) = 1 or t(0) = t(1) = 0), or to choose a different task for each observation of Ws (t(1) = 1 and t(0) = 0, or t(1) = 0 and t(0) = 1). Decision making is assumed to be delegated to the worker if one of the latter choices is
The Use of Profit Sharing When Workers Make Decisions
Fig. 1.
179
Expected Returns to the Firm Given Belief of s.
made, and the worker’s task choice depends upon the worker’s signal of the state. This characterization of decision making highlights the fact that when the worker is asked to make decisions it will end up choosing different tasks under different circumstances. Below it is shown that when the worker chooses the task, profit sharing is of greater value when it is costly for the firm to monitor the exact circumstance of the worker’s choice. Note that it is assumed to be prohibitively costly for the worker’s information to be communicated to the firm.11 The moral hazard problem arises from the difference in the cost to the worker of the two tasks, that is, one task t = 0 involves less effort than the other t = 1. This means that if no incentive is provided by the firm to choose t = 1 the worker will always choose t = 0. In our example, running the production line slowly may be easier than running the line fast. For simplicity it is assumed that e is “small”. This assumption means that the difference in effort costs is such that the incentive costs are never large enough to outweigh differences in revenue. That is, despite the incentive problem the firm always chooses the action with the highest expected revenue (E(r)). The value of giving the worker decision making power depends on the correlation between the worker’s signal and the state of the world, which is denoted , such that = Pr(Ws = 1 s = 1) = Pr(Ws = 0 s = 0) and 0.5 < ˆ < < 1. Let ˆ be the such that (1 f1())(1 p1) + f1()p1 = p0 > 0.5. This minimum value assumption on means that the worker’s information is always of value to the firm.12 The accuracy of the worker’s signal increases in . If is closer
180
CHRISTOPHER P. ADAMS
to 0.5, then the worker’s signal gives little information about the actual state. If the is equal to 1, then the worker’s signal gives the actual state. The firm acts optimally given the moral hazard problem, by offering the worker a contract given the common ex ante beliefs, {t(Ws), (t, Fs, r)}. To simplify matters the firm can choose one of two types of payment schemes. The first is a “profit sharing” contract, which is a function of the firm’s revenue r and constant in the the worker’s task (t) and the firm’s signal state of the world (Fs). The second is a “monitoring” contract, which a function of the worker’s task and the firm’s signal of the state, but constant in the firm’s revenue.13 Intuitively, the profit sharing contract is straightforward representation of a profit sharing or share ownership payment system. The alternative “monitoring” contract requires further explanation. A standard way to provide incentives to a worker is to pay a flat wage and then pay someone to supervise the worker and fire (or otherwise discipline) the worker when the worker is observed choosing the incorrect action. If the worker also has decision making power then the supervisor is required to observe both the worker’s actions and to observe the circumstances of the worker’s decision making. In the example, the supervisor can observe the worker slowing down the production line but does not perfectly observe the number of errors occurring on the line, and thus the value of the worker’s decision to slow the line.14 The ability of the supervisor to monitor the circumstances of the worker’s decision making is represented by the accuracy of the firm’s signal of the state Fs. The correlation between the firm’s signal of the state and the state is denoted , such that = Pr(s = 1 Fs = 1) = Pr(s = 0 Fs = 0) and 0.5 < < 1. The closer is to 1 the more accurate the firm’s signal of the state of the world. Although the firm does not directly observe the worker’s signal of the state of the world, the firm’s own signal of the state of the world is correlated with the worker’s signal. The worker’s supervisor does not directly monitor the worker’s information, but uses his own independent observation of the circumstances to decide on the reasonableness of the worker’s decision making. The model presented here represents a realistic situation in which the worker has information that the firm considers valuable. The firm can use that information to delegate decision making to the worker. To solve the moral hazard problem that arises, the firm can offer a profit sharing contract or a monitoring contract. The monitoring contract represents a situation where the worker’s supervisor is responsible for monitoring both the worker’s choice of task and the worker’s decision making. The firm’s goal is to make the most of the worker’s information while solving the moral hazard problem in the cheapest possible way.
The Use of Profit Sharing When Workers Make Decisions
181
2.2. The Propositions This section presents the two main propositions from the model. The firm can choose along two margins for a total of four contracts. The firm can give the worker decision making power D{0, 1}, which is modelled as requiring the worker to choose different tasks given different observed signals. The firm can also give the worker a profit sharing contract P{0, 1}, which is modelled as payment scheme that is a function of the revenue of the firm. The alternative payment scheme is a function of the worker’s task choice and the firm’s signal of the state of the world. The value of one of these contracts is denoted VDP. To clarify the implications of the simplifying assumptions made in the previous sub-section, the following lemma is presented. The lemma states that if no decision making power is given to the worker then the firm receives the same value from using either profit sharing or monitoring. The reason is that without decision making power, the optimal contract is the degenerate case in which the worker chooses t = 0 and the incentive compatibility constraint holds when the payment scheme is constant. That is, the principal pays a flat wage, and there really isn’t any “monitoring” or “profit sharing.” Lemma 1. V00 = V01. Proof. See Appendix. The proof of this lemma follows from the simplifying assumption that p0 > (1 f )p1 + f p1. First note that the worker’s effort cost is lower for t = 0 than for t = 1 by e > 0. Therefore, if the worker is paid a wage that does not vary with the signal (r for the profit sharing case and (t, Fs) for the monitoring case), then the incentive compatibility constraint is satisfied for t = 0. This means that there are no incentive costs to requiring the worker to choose t(0) = t(1) = 0. However, because choosing t = 1 has an effort cost requiring the worker to choose t(0) = t(1) = 1 would be costly to the firm. The simplifying assumption states that if the worker does not have decision making power, the firm’s ex ante beliefs are such that requiring the worker to choose t(0) = t(1) = 0 has a higher expected revenue than requiring the worker to choose t(0) = t(1) = 1 (p0 > (1 f )p1 + f p1). Therefore, under both the profit sharing contract without decision making power and the monitoring contract without decision making power, the firm will get more revenue at less cost by choosing t = 0.15 The first proposition states that conditional on the payment scheme used, as the worker’s information becomes more accurate ( gets closer to 1), the value of a contract which gives the worker decision making power increases relative to the contract which does not. The proposition shows that not only does the firm’s revenue increase when the worker has better information but the cost
182
CHRISTOPHER P. ADAMS
associated with the incentive problem decreases when the worker has better information. Proposition 1. Given the assumptions defined above, for all (, ˆ 1) (V10 V00) > 0, and (V11 V01) > 0. (ii)
(i)
Proof. See Appendix. Part (i) of the proposition states that as the worker’s information becomes more accurate ( increases) it is becomes relatively more valuable to have the worker make decisions under a monitoring contract than to have the worker make no decisions under a monitoring contract. Part (ii) makes the equivalent statement for the case when the worker has a profit sharing contract. The proof of the proposition is based on the idea that when the worker’s signal of the state is more closely correlated with the state, then the firm’s expected revenue will be greater when the worker’s information is used. While the worker’s information is always assumed to be valuable ((, ˆ 1)) it may not be valuable enough to outweigh the incentive cost associated with giving the worker decision making power. Note however that the cost of the incentive problem is smaller when is higher. This is because giving the worker decision making power means asking the worker to choose different tasks under different circumstances. The incentive scheme will pay greater amounts when certain tasks are matched appropriately with particular circumstances. If the worker’s signal is not highly correlated with the state, then there will be a lot of noise in the incentive scheme. This noise leads to greater risks associated with the contract. As the worker is risk averse, the worker is compensated with higher average payments and thus the contract is more costly. Therefore, this scheme is cheaper to implement when the worker has more accurate information of the circumstances under which he makes choices. Proposition 2 formalizes the second main result of this section. The first part of the proposition states that the relative value of using profit sharing does not change with the accuracy of the firm’s signal of the state. This part follows trivially from Lemma 1. The second part of the proposition states that conditional on the worker having decision making power, as the accuracy of the firm’s signal of the state falls, the value of the profit sharing contract increases relative to the value of the monitoring contract.
The Use of Profit Sharing When Workers Make Decisions
183
Proposition 2. Given the assumptions stated above for all (0.5, 1), (V01 V00) = 0, (V11 V10) < 0. (ii)
(i)
Proof. See Appendix. When the worker has decision making power, the firm only wants the worker to choose the difficult task (t = 1) when the worker’s information suggests that the state is s = 1. Thus the incentive scheme aims to reward matching the task t = 1 with the state s = 1 and punish mismatching. Therefore, when is high, there is less noise in the incentive scheme and the difference between the high payment and the low payment can be reduced (the “power” of the scheme can be reduced). This means that the risk borne by the worker can be reduced and the firm can therefore reduce the expected payment to the worker. These two theoretical results are tested on a data set based on a large survey of British production line workers. The next section discussed the data set.
3. DATA This paper uses a data set based on survey responses in the Workplace Employee Relations Survey (WERS), a large stratified random sample of British workplaces with over 10 employees. The study was undertaken in 1998. In each workplace, WERS 1998 surveyed the human resource manager, a worker representative and a random sample of up to 25 employees. A workplace is defined as ‘the activities of a single employer at a single set of premises’ (Cully et al., 1999). Note that throughout, the terms workplace, establishment and firm are used interchangeably. Unless otherwise stated all cases refer to the above definition. Since the data does not provide information on whether two or more workplaces are members of the same firm, it is not possible to account for correlation across workplaces within the same firm.16 There are 1,880 workplaces that agreed to a survey of the human resource manager and of a random sample of employees, a response rate of 68% of those firms contacted (Cully et al., 1999). There are 28,237 usable responses to the employee survey, which is a response rate of 64% (as a percentage of surveys distributed). For a more detailed description of the survey, see Cully et al. (1999). The sample is based on WERS 1998 and includes information on production workers in private manufacturing firms. The sample is constructed from the
184
CHRISTOPHER P. ADAMS
28,237 usable employee surveys to which information from each employee’s workplace is attached. The attached information comes from the survey of the human resource manager. The sample uses information collected from 1,153 production line workers in 162 private sector manufacturing establishments. The sample is constructed by making the following exclusions. First, the employee must be a self-described operator,17 assembly worker or skilled tradesperson. Since the measure of decision making power is a subjective measure stated by the employee, it is important that the group of employees used in the analysis be as homogenous as possible.18 The sample includes workers in the traditional core of a manufacturing firm, these are the assembly workers, machine operators and the skilled trades people. This criteria excludes 21,112 observations from the data set. Second, the employee’s workplace must be a private sector manufacturing establishment. The interest of this paper is to analyze the human resource practices of manufacturing firms. Public sector manufacturing establishments are excluded because of concern that these establishments are not maximizing profits.19 This criteria excludes another 5,455 observations. The third criteria is that production workers are the largest single group of employees in the workplace, which excludes 158 observations. This criteria insures that manufacturing is the main enterprize of the workplace. Fourth, the employee must be a full-time worker, meaning that his average hours per week must be greater than 35 hours per week. This criteria excludes another 71 observations. Again, the reason for excluding part-time workers is to insure that the sample is homogenous in terms of the type of work performed. Last, the employee is excluded if there is any information missing from the survey questions used in the analysis. This criteria excludes the final 288 observations. The analysis uses two dependent variables: whether decision making power is delegated to the worker and the type of contract offered to the worker. Both are dichotomous variables. The variable DECISION MAKING is 1 if the worker states that he has influence over the range of tasks that he performs, and 0 otherwise.20 DECISION MAKING is a direct measure of how much decision making the worker has, but it is also a subjective measure where the choices are somewhat vague.21 This measure does not vary with the degree of decision making power, rather it simply measures whether the worker has any sort of decision making power over the tasks he performs (from “a little” to a “a lot”). This measure is simple and it captures the important part of the question. The statement that a worker has no decision making power is substantially different to the statement that the worker has some decision making power. It is not quite as obvious what the difference is between “a lot” and “some”. The theoretical model emphasizes that the once the worker has some decision making power,
The Use of Profit Sharing When Workers Make Decisions
185
the incentive problem changes, the degree of decision making power is not as important. Note that another question asked whether the worker had influence over how he performs these tasks. In the data presented below almost all (more than 90%) respondents who had influence over the range of tasks also had influence over how these tasks were performed. Each workplace indicated the proportion of non-managerial employees that received profit sharing or employee share ownership. Each workplace also indicated which categories of employees received profit sharing or employee share ownership. The variable PROFIT SHARING is 1 if the worker’s firm offers either profit sharing or share ownership to more than 80% of its nonmanagerial employees and it offers profit sharing or employee ownership to employees in the worker’s category (operators, assembly workers or skilled trades persons). PROFIT SHARING is 0 if less than 20% of its non-managerial employees have profit sharing or share ownership, or profit sharing and employee ownership are not offered to employees in the worker’s category.22 For example, if the firm states that only skilled trades persons receive profit sharing, then PROFIT SHARING is set equal to 1 if the employee is a skilled trades person, 0 if the employee is not. A concern with PROFIT SHARING is that it is not based on a direct response by the employee, but rather it is based on a series of questions given to the firm. Therefore, there is a possibility that an employee is coded as having profit sharing but actually does not. This would occur when more than 80% of the non-managerial employees in the worker’s establishment have profit sharing but the particular is not part of the 80% that receive it. Similarly, there is a possibility that an employee is coded as not having profit sharing when he actually does if less than 20% of non-managerial employees in the establishment have profit sharing but he is a member of the small fraction that do.23 Three measures are used to describe the firm’s product market. The first measure is CHANGING which is 1 if the manager stated that the market for the firm’s main product is not stable, and is set to 0 otherwise. This measure attempts to capture both volatility in the profits of the firm and volatility on the production floor arising from the types of orders received by the firm. The second measure is MULTI-PRODUCT which is 1 if the establishment produces multiple products, and 0 otherwise. This measure is meant to complement CHANGING by measuring the amount of volatility there is on the production floor. The expectation is that firms that produce multiple products will be more likely to change the types of orders that are on the production line. The third measure of the firm’s product market is QUALITY, which is 1 if the manager stated that the firm has achieved some externally assessed quality standard, and 0 otherwise. An example of such a standard is ISO 9000, in which the firm pays
186
CHRISTOPHER P. ADAMS
for an outside organization to come into the firm and assess whether the firm has achieved the independent benchmarks (set by ISO 9000) in certain quality measures. It is assumed that firms that have achieved such a standard aim to produce a high quality product, or at least compete on quality differentiation. The incentive literature suggests that it will be more difficult to monitor the worker’s actions when those actions include quality margins as well as quantity margins (Drago & Heywood, 1995). The analysis uses two measures of the firm’s characteristics. The first is UNION REP ALLOWED, which refers to the firm’s discipline procedures. UNION REP ALLOWED is 1 if the employee is allowed to be accompanied by union representative during discipline or dismissal procedures, and 0 otherwise. This measure accounts for the ability of the firm to monitor and discipline its workers. The second measure of the firm’s characteristics is SIZE which is measured by the total number of full-time employees in the workplace.24 Previous studies have shown the size of the firm to be an important determinant of both the degree to which decision making power is delegated to the shopfloor (Adams, 2001; Osterman, 1994) and the degree to which profit sharing is used (Jones & Pliskin, 1997; Osterman, 1994). The incentive literature suggests that when there are more workers covered by a particular incentive scheme there is more likely to be “shirking” or free-riding (Holmstrom, 1982). The analysis uses four measures of employee characteristics. The first, 2YEARS is 1 if the worker has been at the firm for more than 2 years, and 0 otherwise. This measure is meant to capture the worker’s knowledge and experience with the production process. The second, UNION MEMBER, is 1 if the employee is a member of a union and 0 otherwise. Previous work suggests that the existence of unions decreases the likelihood that the firm will delegate decision making power to production workers (Adams, 2001; Osterman, 1994). It has also been argued that unions tend to be opposed to profit sharing (Gregg & Marchin, 1988). However, a firm may have greater difficulty disciplining or firing a union member than a non-union member. After all that is one reason employees join unions! If the alternative to profit sharing is a monitoring contract in which the worker is fired (or otherwise disciplined), then unions make this contract more costly and so the relative value of PROFIT SHARING would increase with UNION MEMBER. The a priori “union effect” could positively or negatively affect the probability that the worker receives profit sharing. The third characteristic, MALE is 1 if the employee is male and 0 if the employee is female. The fourth and final characteristic of the employee, SKILLED is 1 if the employee is a skilled trades person and 0 if the employee is an operator or assembly worker. One
The Use of Profit Sharing When Workers Make Decisions
187
concern with the measure DECISION MAKING, is that it may vary systematically across types of work. Though more detail would be preferred, SKILLED is the only measure of the type of work available in the data set. However, this measure may not be exogenous because firms can choose the skill of the worker. If there are unmeasured characteristics of the firm that determine both the skill level of a particular employee and whether the worker will be given DECISION MAKING, then using this measure could bias estimates.25 Table 1 presents the sample means (frequencies) of the variables used in the analysis, weighted to reflect population averages.26 The first column presents the frequencies as a proportion of the workers and the second column presents the frequencies as a proportion of the establishments. 77% of workers have (some) decision making power over the tasks they perform, 10% have profit sharing or share ownership and 8% have both decision making and profit sharing. Importantly, workers tend to come from the larger firms in the sample, with the mean worker coming from a firm with 1,250 full time workers while the mean firm has 401 workers. The larger firms are more likely to face CHANGING markets, produce QUALITY products and have UNION REPRESENTATION ALLOWED. Still, the relatively small difference between the columns suggests that workers from large firms do not dominate the sample. The next section presents the empirical model used to test the hypotheses.
Table 1.
Sample Means by Employee and Firm.
Variable
Mean Employee
Mean Firm
Decision Making Profit Sharing Decision Making and Profit Sharing Changing Multiple Products Size Quality Probability Union Rep Allowed Union Member Male Two Years Skilled
0.77 0.10 0.08 0.67 0.58 1,250 0.73 0.53 0.55 0.78 0.82 0.38
– – – 0.63 0.59 401 0.69 0.46 – – – –
Number in Sample
1,153
162
188
CHRISTOPHER P. ADAMS
4. EMPIRICAL MODEL This section presents the empirical model that is used to test the implications of the theoretical model presented above. The model is a latent profit model. There are three subsections. The first subsection presents the notation for a linear latent profit model. It also describes the implication of assuming that profit sharing and delegating decision making are complements. The second subsection presents the restrictions imposed on the structure of the firm’s latent profits by the theoretical model and the hypotheses to be tested in the empirical section. The third subsection presents the estimated model, including a description of distribution assumptions. 4.1. Linear Latent Profit Model There exist four possible contracts, the latent value of each is presented below. Recall that VDP is the latent value to the firm, where D indicates whether decision making is delegated to the worker and P indicates whether the worker is provided with profit sharing. First, the value of neither delegating decision making nor using profit sharing is denoted by Aij for worker i and firm j. The latent profits of the other contracts will be compared to this one. V00 = Aij
(2)
The value of delegating decision making but not using profit sharing is V10. The relative value of this contract may be a function of the measure of the accuracy of the worker’s information, denoted Xi, of the measure of the accuracy of the firm’s information, denoted Xj, of other characteristics of the worker (Xi), and of the firm (Xj). Value is also affected by unobservable characteristics of the worker and firm ( ijD). V10 = Aij + Xi D + Xj D + Xi iD + Xj jD + ijD
(3)
The value of using profit sharing but not delegating decision making is V01. V01 = Aij + Xi P + Xj P + Xi iP + Xj jP + ijP
(4)
where ijP represents unobservable characteristics that affect the relative value of the contract with profit sharing only. The other variables are defined above. The value of delegating decision making and also using profit sharing is V11. V11 = Aij + Xi DP + Xj DP + Xi iDP + Xj jDP + ijDP
(5)
where ijDP represents the unobservable characteristics that affect the relative value of the contract in which decision making is delegated and profit sharing is used. Again the other variables are defined above.
The Use of Profit Sharing When Workers Make Decisions
189
One of the main objectives of the empirical analysis is to measure the factors that affect the complementarity between delegating decision making power and using profit sharing. If the two practices are complements then they are supermodular in the firm’s latent profits (Athey & Stern, 1998; Milgrom & Roberts, 1990). This implies that, V00 + V11 ≥ V01 + V10.
(6)
To see the implications of this assumption, and measure the factors that affect the complementarity, it is easier to rewrite Eq. (5) in the following way. Where the *DP variables are substituted for equivalent DP D P variables. V11 = Aij + Xi D + Xj D + Xi iD + Xj jD + ijD + Xi P + Xj P + Xi iP + Xj jP + ijP
(7)
+ Xi *DP + Xj *DP + Xi *iDP + Xj *jDP + *ijDP. Therefore, *DP = DP D P, *DP = DP D P, *iDP = iDP iD iP, *jDP = jDP jD jP and *ijDP = ijDP ijD ijP. In this sense, the * coefficients determine the “extra” value of having both practices together. If the two practices are complements for the mean worker then Eq. (6) implies the following inequality.27 Xi *DP + Xj *DP + Xi *iDP + Xj *jDP ≥ 0.
(8)
Equation (8) is derived by substituting Eqs (2), (3), (4) and (7) into Eq. (6). The estimated values of the coefficients in Eq. (8) give us insight into the factors that affect the complementarity between delegating decision making power and using profit sharing. If the estimated value of Eq. (8) is 0 then this is equivalent to stating that delegating decision making and using profit sharing are not complements. 4.2. Hypotheses Recall that Lemma 1 states that V01 V00 = 0, that is, when the worker does not have decision making power, profit sharing gives the firm no added value over monitoring. This lemma implies the following restriction on the empirical model. Hypothesis 1. Xi P + Xj P + Xi iP + Xj jP = 0. This hypothesis states that conditional on the worker not having decision making power, the measurable characteristics of the firm will have no effect on the value to the firm of using profit sharing. This hypothesis is tested in the next section.
190
CHRISTOPHER P. ADAMS
Proposition 1 states that the firm values giving decision making power to workers more highly when workers have more accurate information, conditional on the type of payment scheme used. In terms of the empirical model the appropriate differences are, V10 V00 = Xi D + Xj D + Xi iD + Xj jD + ijD
(9)
for the case where profit sharing is not used, and V11 V01 = Xi D + Xj D + Xi iD + Xj jD + Xi *DP + Xj *DP + Xi *iDP + Xj *jDP + ijD + *ijDP
(10)
for the case where profit sharing is used. Therefore the proposition implies the following hypothesis. Hypothesis 2. (i) (ii)
(V10 V00) = D > 0, and Xi
(V11 V01) = D + *DP > 0. Xi
Part (i) of the hypothesis states that conditional on not using profit sharing, as the measure of the accuracy of the worker’s information increases, the latent profits from giving the worker decision making increases relative to not giving the worker decision making. Part (ii) of the hypothesis makes the equivalent statement conditional on the use of profit sharing. Recall that denotes the accuracy of the worker’s information. The best proxy for in this data set is 2YEARS which is 1 if the worker has at least two years experience at the establishment. Rewriting the hypothesis in terms of this proxy, the following two equations should hold,
2YEARS D > 0
(11)
2YEARS D + *2YEARS DP > 0.
(12)
Thus, conditional on the type of payment scheme and holding all else equal, the firm should value delegating decision making power to a worker more when the worker has at least two years at the establishment. A concern with the proxy, 2YEARS, is bias. This may occur because experience is associated with “tenure” and with tenure comes greater responsibility. This problem is mitigated by the fact that all respondents are at the same level within the firm (production workers). However, experience may also be associated with ability, especially if the firm has some sort of evaluation period. This issue is mitigated to a certain extent by the fact that 2 years is longer than a standard formal or informal evaluation period. Therefore, the sample includes able people with less than 2 years experience.
The Use of Profit Sharing When Workers Make Decisions
191
Part (i) of Proposition 2 states that if the worker does not have decision making power then the value to the firm of using profit sharing is unaffected by measures of the firm’s ability to observe the worker’s information. The appropriate difference is V01 V00 which is discussed in Hypothesis 1. Part (ii) of Proposition 2 states that if the worker has decision making power, then the firm places greater value on using profit sharing when it is more difficult to monitor the worker’s information. In terms of the model the appropriate difference is, V11 V10 = Xi P + Xi P + Xi iP + Xj jP + Xi *DP + Xi *DP + Xi *iDP + Xj *jDP + ijP + *ijDP.
(13)
Therefore, the proposition implies the following hypothesis. Hypothesis 3. (i) (ii)
(V01 V00) = P = 0, and Xi
(V11 V10) = P + *DP < 0. Xi
The first part of the hypothesis states that conditional on the worker having no decision making power, the accuracy of the firm’s information on the circumstances of the worker’s decision making should not affect the firm’s latent profits. The second part of the hypothesis states that conditional on the worker having decision making power, if the measure of the accuracy of the firm’s information on the circumstances of the worker’s decision making decreases, then the latent profits of using profit sharing increases relative to not using profit sharing. The proxy for the ability of the firm to monitor the circumstances of the worker’s decision making () is CHANGING. The proximity of the worker to the production line gives the worker relatively better information than the manager when the production orders are always changing (Adams, 2001). The author argues when the sizes and shapes of production orders are frequently changing, the methods of production also have to change frequently in order to adjust to the changing circumstances. In such situations the worker is able to observe problems on the production line and react quickly to them. For example, the worker may see that the production line is having difficulties with the current order and needs to slow the line so as to reduce errors and re-dos, while the manager may not realize there is a problem until the order would be ready to be shipped. In this case, not only is the manager an inferior decision maker, but is also has inferior information about the circumstances behind the
192
CHRISTOPHER P. ADAMS
worker’s decision to slow the line. For firms that delegate decision making power to their production line workers, changing demand in the product market means greater difficulty in monitoring the circumstances of the worker’s choices, and thus greater value in using profit sharing. Given this proxy, the condition for part (i) of which hypothesis to be supported is if the following equation is satisfied,
CHANGINGP = 0
(14)
and the condition for part (ii) of the hypothesis to be supported is,
CHANGINGP + *CHANGINGDP > 0.
(15)
Note, that when the product market is CHANGING it is more difficult to monitor the circumstances of the worker’s decision making. In other words, is lower and so the firm is more likely to use profit sharing. A major concern with the use of the proxy CHANGING for (the accuracy of the firm’s information with regards to the circumstances of the worker’s decision) is that its use is likely to bias the results against supporting the hypothesis. This occurs if CHANGING demand in the product market is associated with volatility in the firm’s profits, which seems likely. A standard result in the literature is that volatility in the firm’s profits increases the costs of using profit sharing plans when workers are risk averse (Holmstrom, 1979).28 Therefore, coefficients for the use of profit sharing on measures of firms profit volatility are expected to be negative. That is, holding everything else the same, the more volatile is the firm’s profits the less valuable is the use of profit sharing. As the proxy CHANGING is not only correlated with but is also correlated with the firm’s profit volatility, the effect will be to bias the results in the negative direction and against support of the hypothesis. One way around this problem is to use a different proxy for . An alternative proxy for is MULTI-PRODUCT which is 1 if the workplace produces multiple products. It is expected that a firm which produces multiple products will face greater volatility on the production floor than a firm which produces a single product. Increased volatility on the production floor itself is expected to decrease the firm’s information on the circumstances of the worker’s choices. Therefore, the hypothesis is supported if the following equations hold,
MULTI-PRODUCTP = 0
(16)
The Use of Profit Sharing When Workers Make Decisions
193
and
MULTI-PRODUCTP + *MULTI-PRODUCTDP > 0.
(17)
Unlike the proxy CHANGING this proxy is not expected to be correlated with the firm’s profit volatility. It is not clear whether a firm producing multiple products would have more volatile profits or less volatile profits. Therefore, if this proxy is used it is not expected that the results will be biased with regard to supporting Hypothesis 3. 4.3. Estimated Model This subsection presents the empirical model of the firm’s decision to delegate decision making power to the worker and give the worker profit sharing. The model allows the firm’s choice on delegating decision making and using profit sharing to be made simultaneously, and it allows the two choices to interact. The model’s key characteristics is that it allows this interaction to vary from worker to worker in observable and unobservable ways. The error terms { ijD, ijP, *ijDP} are distributed standard tri-variate normal, and are assumed to be independent of Xij, the vector of measurable characteristics of worker i in firm j. The variance-covariance matrix is:
=
1
12 1
13 23 1
.
(18)
The errors are also assumed to be clustered by “workplace”.29 The clustering assumption means that when calculating the standard errors, the observations are only assumed to be independent when they belong to different workplaces. Observations within the same workplace may have some undetermined dependence; by weakening the independence assumption the size of the standard errors is increased. Note that in order to reduce notation and simplify the exposition of the model it is assumed that 13 = 23 = 0. However, in the actual estimation the values of 13 and 23 are determined by the maximum likelihood procedure. To simplify the exposition further, the probabilities over the events D and P (delegating decision making and using profit sharing, respectively) for worker i in firm j can be written as the weighted sum of two probabilities conditional on the level of *DP.
194
CHRISTOPHER P. ADAMS
Pr(D, P Xij ) = (1 ( Xij *DP))Pr(D, P *DP > Xij *DP) + ( Xij *DP)Pr(D, P *DP < Xij *DP)
(19)
where Xij *DP is defined by the left hand side of Ineqaulity (8), and is the standard normal cumulative density function. The case where *DP > Xij *DP occurs when for worker i in firm j, delegating decision making and using profit sharing are complements. That is, conditional on *DP the diagonal events ((0, 0) and (1, 1)) are given greater probability weight than the off-diagonal events ((0, 1) and (1, 0)).30 Figure 2 depicts the state space over D and P conditional on *DP, when *DP > Xij *DP. It can be seen that as Xij *DP + *DP increases, more probability weight that is placed on events (0, 0) and (1, 1) relative to events (0, 1) and (1, 0). The corresponding probabilities are determined by the following three equations. Pr(D = 1, P = 0 Xij, *DP > Xij *DP) =
Xij *DP
(Xij P Xij *DP *DP)
2( Xij D, Xij P Xij *DP *DP, 12)( *DP)d *DP
Fig. 2.
The State Space Conditional on *DP.
(20)
The Use of Profit Sharing When Workers Make Decisions
195
Pr(D = 0, P = 1 Xij, *DP > Xij *DP) =
Xij *DP
( Xij D Xij *DP *DP)
(21)
2( Xij D Xij *DP *DP, Xij P, 12)( *DP)d *DP and Pr(D = 0, P = 0 Xij, *DP > Xij *DP) =
Xij *DP
+
2( Xij D Xij *DP *DP, Xij P, 12)
Xij D Xij D Xij *DP *DP
(22)
Xij D Xij P Xij *DP *DP D(1 + 12) (1 212)0.5
( D)d D( *DP)d *DP
where 2 is the cumulative density of the standard bivariate normal distribution, and is the density function for the standard normal distribution. Equations (20) and (21) show that the probability of the off-diagonal events ((1, 0) and (0, 1)) decreases as Xij *DP + *DP increases. The cumulative distribution that is integrated in Eq. (22) is the standardized cumulative distribution of P conditional upon a particular value of D (the diagonal line in Fig. 2). In the actual estimation, the cumulative distribution that is integrated with respect to *DP is approximated using Monte Carlo methods with R replications. The integral in Eq. (22) is approximated “from below” with a Riemann sum of N steps. If R and N are both converging to infinity as the sample size converges to infinity, then this approximation gives a consistent estimate. For the case where *DP < Xij *DP a simplifying assumption is made. This assumption states that conditional on *DP the probability distribution over events D and P is similar to the distribution for the (seemingly unrelated) bivariate probit, and thus very easy to compute. For these workers, the assumption states that the delegation of decision making and the use of profit sharing are independent as far as observable characteristics are concerned. In terms of Fig. 2 the state space is divided only by the lines Xij P and Xij D, and the events (0, 0) and (1, 1) are given the same weight as the events (0, 1) and (1, 0). The corresponding probabilities are determined by the following three equations.
196
CHRISTOPHER P. ADAMS
Pr(D = 1, P = 0 Xij, *DP ≤ Xij *DP) = (Xij P) 2( Xij D, Xij P, 12)
(23)
Pr(D = 0, P = 1 Xij, *DP ≤ Xij *DP) = ( Xij D) 2( Xij D, Xij P, 12)
(24)
and Pr(D = 0, P = 0 Xij, *DP ≤ Xij *DP) = 2( Xij D, Xij P, 12)
(25)
Equations (23) and (24) show that the off-diagonal events are given no more and no less weight relative to the diagonal events. Summing up, conditional on the value of *DP, there are two cases. In case 1, *DP is relatively large, and extra probability weight is placed on the ondiagonal events of choosing to delegate decision making and use profit sharing, and of using neither practice. In case 2, *DP is relatively small and no extra weight is placed on the on-diagonal events. However, in both cases, the correlation term 12 can skew the distribution either to the on-diagonal or to the off-diagonal. In the computation of Eqs (20) to (25), an *DP is drawn from a standard normal distribution. The simplifying assumption is then imposed by resetting negative values for Xij *DP + *DP to 0. This means that for every firm, the two practices, delegating decision making and using profit sharing, are either complements or the two practices are independent. The greater the estimated value for Xij *DP the greater the proportion of firms for which delegating decision making and using is profit sharing are complements, and the smaller the proportion of firms for which the two practices are independent of each other. The alternative assumption is to allow the two practices to be supermodular in the firm’s latent profits for some firms and sub-modular in the firm’s latent profits for other firms. As it is, I assume that the two practices are supermodular in the firm’s latent profits for some firms and neither sub-modular nor super-modular in the firm’s latent profits for other firms. The assumption is made in order to simplify the computation of the likelihood function.31 Except for the approximations described above, the model is estimated using a standard maximum likelihood procedure. The objective of the empirical project is to use the empirical model to analyze the factors which affect the joint decision of the firm over the complementary choices of delegating decision making power to the worker and giving the worker profit sharing. There are two reasons for using the model presented above. The first reason is that the propositions presented in the earlier section make claims about the
The Use of Profit Sharing When Workers Make Decisions
197
conditional choices of the firm, that is the choice to use profit sharing conditional on the choice of delegating decision making power (for example). These hypotheses are only testable in a model with these characteristics. The second reason is that this model is consistent with the underlying thesis that these practices are complements. The next section presents the results from estimating the model and shows to what extent the hypotheses discussed above are supported by the data.
5. RESULTS Table 2 presents the results for the empirical model described in the previous section.32 The model is estimated on the data described above, weighted to reflect population averages. The interaction includes three terms, CHANGING, MULTI-PRODUCT and SIZE. The number of terms was chosen with idea of keeping the number small while still giving the model enough explanatory power. The more terms that are in the interaction term, the more complicated the model, the greater the requirements on the data set (which is fairly small) and the more difficult it is to estimate the model. In general the estimated value of coefficients is in the range suggested by the theory. However, there is mixed evidence in support of the paper’s main hypotheses. In Table 2 the estimated coefficient values corresponding to equation in Hypothesis 1 are denoted by P. The hypothesis states that Xi P + Xj P + Xi iP + Xj jP = 0
(26)
for all X. That is, the measured characteristics of the workers and firms should have no effect on the latent value of using profit sharing, conditional on the worker having no decision making power. This hypothesis is not supported by the data as the coefficients on CHANGING and QUALITY and the constant term are statistically significant. The coefficient on CHANGING is discussed below. The coefficient on QUALITY is positive which supports the argument that firms value profit sharing more highly when it is difficult to observe the actions of the workers (Drago & Heywood, 1995b). This result suggests that the theoretical model is not capturing an important reason for using profit sharing when the worker does not have decision making power. Hypothesis 2 states that conditional on the type of payment scheme used, firms value giving decision making power to worker more highly when workers have more accurate information. This hypothesis is given support in the data. The coefficient on 2YEARS is 0.37 (Eqs (11) and (12)) for a worker without and with profit sharing.33 The point estimate is consistent with the hypothesis and it is statistically significant.34 To see the “economic significance” consider
198
CHRISTOPHER P. ADAMS
Table 2.
Estimated Model.
Robust SE
Prob x
–0.04 0.10 –0.000015 0.01 0.22 0.37 0.08 –0.31 0.18 0.66
(0.10) (0.10) (0.0000097) (0.12) (0.12) (0.13) (0.10) (0.11) (0.11) (0.16)
–0.01 0.03 –0.00001 0.00 0.06 0.12 0.02 –0.09 0.05 –
–1.13 –0.66 –0.00010 0.88 –0.34 0.22 –0.31 0.37 0.09 –1.19
(0.43) (0.38) (0.00017) (0.38) (0.37) (0.30) (0.21) (0.21) (0.25) (0.40)
–0.11 –0.05 –0.00001 0.05 –0.02 0.01 –0.02 0.03 0.01 –
Changing Multi-Product Size Constant
2.61 0.98 –0.00005 –3.46
(0.68) (0.59) (0.00017) (0.74)
– – – –
12 13 23
–0.06 –0.71 0.01
(0.20) (0.29) (0.23)
– – –
Variable
Decision Making ( D) Changing Multi-Product Size Quality Union Rep. 2 Years Skilled Union Member Male Constant Profit Sharing ( P) Changing Multi-Product Size Quality Union Rep. 2 Years Skilled Union Member Male Constant Interaction ( *DP)
Log likelihood
–1613.96
The Use of Profit Sharing When Workers Make Decisions
199
the final column of Table 2 which presents the unconditional changes in the probability that the firm will choose to delegate decision making power and choose to give profit sharing to the mean worker. A worker with more than 2 years experience is 12 percentage points more likely to be delegated decision making power than an otherwise similar worker with less than 2 years experience.35 Part (i) of Hypothesis 3 states that conditional on the worker having no decision making power, the firm’s value of using profit sharing is not affected by how difficult it is for the firm to monitor the circumstances of the worker’s decision making. Testing the hypothesis (Eq. (14)), the coefficient on CHANGING is 1.13 which is statistically significant. This does not support the hypothesis as the hypothesis states that this coefficient should be 0. However, as discussed above it is expected that using the proxy CHANGING for will bias the estimate downward. This occurs because firms with more volatile profits are less likely to use profit sharing and CHANGING is also likely to proxy for the volatility of the firm’s profits. An alternative proxy is MULTI-PRODUCT. The coefficient (Eq. (16)) is 0.66 which is statistically significant at the 10% level. Therefore the coefficients on both proxies do not support Part (i) of the Hypothesis. To some extent this is because the proxies are capturing other important determinants of the use of profit sharing plans. Part (ii) of Hypothesis 3 states that conditional on the worker having decision making power, the firm values using profit sharing more highly when it is difficult for the firm to monitor the circumstances of the worker’s decision making. As discussed above, this is more likely to be the case when demand in the firm’s product market is changing leading to instability on the production line. This hypothesis is supported by the data. Testing the hypothesis (Eq. (15)), the coefficient on CHANGING is 1.13 + 2.61 > 0 which is consistent with the hypothesis and statistically significant at the 10% level. Some further support for Hypothesis 3 is given by the measure MULTIPRODUCT. The equivalent coefficient is 0.66 + 0.98 > 0 (Eq. 17). The point estimate is consistent with the hypothesis although it is not statistically significant different from 0. There is more likely to be volatility on the production line for a firm that produces multiple products relative to a firm that produces just one product. Therefore, it is likely to be more difficult for the firm to monitor the circumstances of the worker’s decision making in firms that produce multiple products. Most of the results in Table 2 are consistent with previous work by this author and in the literature. If the worker is a member of the union then he is less likely to be given decision making power. The result is consistent with the general findings about the use of employee involvement programs such as self-
200
CHRISTOPHER P. ADAMS
managed work teams. Firms with unions tend to be less likely to adopt such programs and are less likely delegate greater decision making power to the shop floor (Adams, 2001; Osterman, 1994). This result goes further by stating that even within firms, individual union members are less likely to get decision making power. It is expected that skilled workers are more likely to have decision making power. The results suggest that they are 2% more likely to have decision making power than the mean unskilled worker. However, the coefficient is not statistically significantly different from 0. On the use of profit sharing schemes the results are similar to the literature which shows that firms that produce high quality products are more likely to use profit sharing (Drago & Heywood, 1995). Union members are 3% more likely to receive profit sharing than non-union members, supporting the idea that union members are more difficult to fire and so monitoring contracts are relatively less valuable than profit sharing contracts. The coefficient is statistically different from 0 at the 10% level. The coefficient on the measure of firm size is negative, which conforms to the idea that free riding should reduce the value of such firm-wide incentives (Holmstrom, 1982). However, the coefficient is not statistically significantly different from 0. The model states that a 100 person increase in the number of workers decreases the probability of the worker receiving profit sharing by 0.1% which is not very much. The results give mixed support for the hypotheses presented in the previous section. The results suggest that the theoretical model does not capture some important reasons for using profit sharing. However, the results do give support for the explanation for why firms delegate decision making power to production line workers and for why, conditional on the worker having decision making power, firms use profit sharing. The results support Hypotheses 2 and 3 despite concerns with the proxy that is used for volatility in the product market.
6. CONCLUSION By delegating decision making power to a production line worker, the firm asks (expects) the worker to choose different tasks based upon his private information. For example, if the worker sees a large number of errors occurring on the line then he may be expected to slow the line down to improve the precision of the work. If some tasks are harder than other tasks, the worker’s preferences over which tasks he can choose may be different than those of the firm. For example, the worker may prefer to slow the line down because the work is easier when the line is running slow, irrespective of the number of errors occurring on the line. To solve this moral hazard problem the firm could
The Use of Profit Sharing When Workers Make Decisions
201
monitor the worker or it could offer the worker pay contingent on the firm’s profits. The theoretical section of this paper shows that conditional on the type of incentive scheme used, the firm will give decision making to the worker when the worker’s information is more accurate. If there was no moral hazard problem then this would be the case because the worker’s decisions are expected to be more profitable, the more accurate the worker’s information. However, the paper shows that it is also the case that the cost of the moral hazard problem is reduced when the worker’s information is more accurate. The reason is that when the worker has decision making power the worker’s choices are dependent on his understanding of the circumstances. Furthermore, the value of the incentive contract is a function of the firm’s observation of the worker’s choices and the firm’s understanding of the circumstances. It is likely that correlation between the worker’s understanding of the circumstances and the firm’s understanding of the circumstances is increasing in the accuracy of the worker’s information. This paper also shows that when the firm gives the worker decision making power, a contract involving monitoring must involve monitoring both the worker’s action choice and the circumstances affecting the worker’s action choice. Simply observing that the worker has slowed down the line is not enough. If the worker is to be fired for slacking the firm must also observe that there were very few errors occurring on the line and there was no need to run the line slow. In practice we often see monitoring arrangements in which the worker is monitored by his immediate supervisor who is familiar both with worker’s action choices and the circumstances of those action choices. The paper continues with a presentation of an empirical model which generalizes the bivariate probit model by allowing the choice to delegate decision making power and the choice to use profit sharing to be complementary. This model allows the author to determine the characteristics of the firm and worker that affect the complementarity. The empirical results give mixed support for the paper’s main theoretical results. First, the results do not support the hypothesis that conditional on the worker having no decision making power, important characteristics of the worker and firm will have no effect on the firm’s latent value of using profit sharing. In fact, the empirical results support the thesis that firms that produce high quality products place greater value in using profit sharing (Drago & Heywood, 1995b). This suggest that the theoretical model does not adequately characterize the firm’s choice to use profit sharing when the worker does not have decision making power. Second, the empirical results show that conditional on the type of payment system, the firm’s latent profits from giving
202
CHRISTOPHER P. ADAMS
the worker decision making power are higher when the worker has at least two years experience with the firm. It is argued that worker’s experience is a good proxy for the accuracy of the worker’s information. Third the empirical results show that conditional on the firm delegating decision making to the worker, the firm values profit sharing more highly when it faces changing demand in its product market. It is argued that volatility in product demand leads to volatility in the production line itself. This in turn leads to greater differences between the information available to production line workers and the information available to their managers and supervisors. As it is difficult to monitor the circumstances of the worker’s decision making these firms are more likely to offer their workers profit sharing. The results of this paper show how product market conditions determine the willingness of firms to give workers decision making power and access to profit sharing. The results also show how these two human resource practices complement each other. Profit sharing complements the delegation of decision making power to production workers by making easier for the firm to provide the worker with the appropriate incentives, when it is difficult for the firm to monitor the circumstances behind the worker’s choices. This improved understanding allows greater insight into the effects that government policies that encourage the use of profit sharing and share ownership, have on the lives of manufacturing workers.
NOTES 1. In the data used in this paper, 77% of production workers make decisions regarding the tasks that they perform and 10% of production workers receive profit sharing or own stock in their firm. 2. It is argued that profit sharing plans and practices that involve “teams” of workers are complements because there is improved peer monitoring (Che & Yoo, 2002; Pliskin, 2000; Kandel & Lazear, 1992). While this seems like a reasonable explanation, an alternative argument is explored in this paper. 3. In order to simplify the analysis, the paper considers only the use of profit sharing and employee share ownership plans. It does not analyze the use of other performance based pay schemes such as gainsharing and piece rates. For further discussion of these see for example, Drago and Heywood (1995) and Jones and Pliskin (1997). 4. An alternative is to measure the use of human resource practices that delegate substantive decision making power to production line workers, such as self-managed work teams (Adams, 2001; Osterman, 1994; Pliskin, 2000). 5. For reasons discussed below, the data used in this chapter is based on the British Workplace Employee Relations Survey 1998 (Cully et al., 1999). However, information on employee decision making also exists in the Australian Workplace Industrial
The Use of Profit Sharing When Workers Make Decisions
203
Relations Survey 1995 (Morehead et al., 1997), and the National Organizations Survey (Kalleberg et al., 1991). 6. A notable exception is a recent paper analyzing the relationship between the level of employee decision making and other characteristics of hierarchies (Meagher & Temur, 2000). 7. Myerson allows the principal to offer a menu of contracts from which the agent can choose after observing her private information (like a monopoly pricing model). Here, the situation is one in which the agent continuously observes new information and makes frequent new choices given that information. It would be cost-prohibitive to renegotiate the contract each time new information is observed. Therefore, Myerson’s contract space is restricted to one similar to ordinary employment contracts where the payment scheme is agreed to before any information is observed by the agent (like an education signalling model). 8. Although the data is at the establishment level rather than at the firm level. See Jones and Pliskin (1997) for a discussion of this issue. 9. The worker is assumed to be male throughout the paper. 78% of the production workers in the data set are male. 10. The generalized principal-agent problem allows the firm to offer a “menu” of contracts as a function of the worker’s private information (Myerson, 1982, 1985). The restriction made above seems to be more in-line with the observation that firms and workers agree to terms before the worker joins the firm and observes any information. 11. See the team theory literature for a more detailed discussion of communication costs in firms (Radner, 1996). 12. If the effort costs of both tasks were the same then for all allowable values of the firm would have a higher expected revenue when the worker’s task choice is a function of the worker’s information where t(0) = 0 and t(1) = 1. To see this, note that if the worker has no decision making power the firm’s expected revenue is p0. However, if the worker is able to use his information, the firm’s expected revenue is (1 f · )p0 + f · ((1 f1())(1 p1) + f1()p1), where the notation f · = (1 f )(1 ) + f. By assumption this is greater than p0, as (, ˆ 1). 13. For simplicity it is assumed that when the firm uses a monitoring contract it observes the worker’s action choice perfectly. Note it is also assumed that writing a contract which uses the firm’s revenue (r) in conjunction with the firm’s other information is infinitely costly. 14. A familiar example occurs in many tenure decisions, where much of responsibility for the decision is given to the senior faculty member with the best understanding of the junior member’s research area. 15. The reason the assumption is made is that it reduces the number of cases and simplifies the exposition of the model. 16. See Jones and Pliskin (1997) for a discussion of this issue as well as analysis of the use of profit sharing using firm level data. 17. Actually, the respondent had to choose a category and then also describe their major tasks and responsibilities, which was later coded by the survey team. I only use those employees whose self-description matched the coding of the survey team. 18. Meagher and Temur (2000) show that the decision making power increases with the level that the person has in the hierarchy. Therefore it is important to restrict the sample to production line workers only. 19. Profit maximization is a central assumption.
204
CHRISTOPHER P. ADAMS
20. The question states ‘In general, how much influence do you have about the following? The range of tasks you do in your job (“A lot”, “Some”, “A little”, “None”, “Don’t know”)’ (Department of Trade and Industry, Advisory, Conciliation and Arbitration Service, 2000). Note that nobody responded “Don’t know”. 21. It would be preferable to have an objective measure where the questions enquire about specific tasks and responsibilities. For example, “Do you stop the production line if you see an error occurring?”, or “Do you decide which equipment will be used on the line?” 22. Profit sharing and share ownership are two different schemes that are combined in the analysis. For a discussion and analysis of the differences between the two schemes see Jones and Pliskin (1997). 23. Employees in establishments which state that between 20% and 80% of employees have profit sharing are dropped from the sample. This criteria has little effect on results and only a small number of firms are eliminated from the data set. 24. Note again that while I use the terms firm, workplace and establishment interchangeably, all refer to the definition of workplace given above. 25. The results presented below include SKILLED as an explanatory variable. Dropping this variable and (separately) restricting the sample to unskilled workers (operators and assembly workers) does not have a significant effect on the main results of the paper. This suggests that if there is an endogeniety problem with the SKILLED variable, it is not large. 26. The entries not presented are the mean of means for the workers within a firm. These are simply an average that is weighted slightly differently from the entry in the first column. 27. It is assumed below that the mean of *ijDP = 0. 28. Similar comments have been made to the author by people actually involved in implementing profit sharing schemes. 29. See Jones and Pliskin (1997) for further discussion of this issue. 30. See Adams (2001) and Athey and Stern (1998). 31. The interior integral in Eq. (22) is quite difficult to compute. If sub-modularity was assumed in place of independence, another similar integral would also need to be computed (Athey and Stern, 1998). 32. For the results presented above, R = 100 and N = 10. The results are similar for the case where N = 2. 33. As 2YEARS is not in the interaction term, Eqs (11) and (12) are identical. 34. The standard errors are calculated from the diagonal of the variance-covariance matrix. Note again that the observations are assumed not to be independent within a workplace. 35. Note that by Simpson’s paradox this result does not follow from Hypotheses 2.
ACKNOWLEDGMENTS The author thanks in particular Larry Samuelson for all his help and suggestions and also Deena Ackerman, Mark Cully, Phil Haile, and the editors of this publication. All errors are my own. Note that this paper does not necessarily represent the views of the Commission or any individual
The Use of Profit Sharing When Workers Make Decisions
205
Commissioners. The author acknowledges the Department of Trade and Industry, Advisory, Conciliation and Arbitration Service, for use of the Workplace Employee Relations Survey: Cross-Section, 1998 [computer file]. 5th ed. Colchester, Essex: The Data Archive [distributor], 20 April 2000. SN: 3955. The copyright holder, the original data producer, the relevant funding agencies and The Data Archive bear no responsibility for their further analysis or interpretation.
REFERENCES Adams, C. P. (2001). Theory and Practice of Shopfloor Decision Making in Manufacturing. Ph.D. dissertation. Madison: University of Wisconsin. Appelbaum, E., Bailey, T., Berg, P., & Kalleberg, A. (2000). Manufacturing Advantage: Why High Performance Work Systems Pay Off. Cornell University Press. Athey, S., & Stern, S. (1998). An Empirical Framework for Testing Theories About Complexity in Organizational Design. April draft. Che, Y.-K., & Yoo, S.-W. (2002). Optimal Incentives for Teams. American Economic Reviev, 91(3), June, 525–541. Cully, M., Woodland, S., O’Reilly, A., & Dix, G. (1999). Britain at Work: As Depicted by the 1998 Workplace Employee Relations Survey. London: Routledge. Department of Trade and Industry, Advisory, Conciliation and Arbitration Service (2000). Workplace Employee Relations Survey: Cross Section 1998. Computer File April (5th ed.). Colchester, Essex: The Data Archive [distributor]. SN: 3955. Drago, R., & Heywood, J. (1995). The Choice of Payment Schemes: Australian Establishment Data. Industrial Relations, 34, 507–531. Gregg, P. A., & Marchin, S. J. (1988). Unions and the incidence of performance linked pay schemes in Britain. International Journal of Industrial Organization, 6, 91–107. Holmstrom, B. (1979). Moral Hazard and Observability. Bell Journal of Economics, 10(1), 74–91. Holmstrom, B. (1982). Moral Hazard in Teams. Bell Journal of Economics, 13(2), 324–341. Ichniowski, C., Shaw, K., & Prennushi, G. (1997). The Effects of Human Resource Management Practices on Productivity: A Study of Steel Finishing Lines. American Economic Review, 87, 291–313. Jones, D., & Pliskin, J. (1997). Determinants of the Incidence of Group Incentives: Evidence from Canada. Canadian Journal of Economics, 30(4b), November, 1027–1045. Kalleberg, A., Knock, D., Marsden, P., & Spaeth, J. (1991). National Organizations Survey. University of Illinois: Survey Research Laboratory. Kandel, E., & Lazear, E. (1992). Peer Pressure and Partnerships. Journal of Political Economy, 100, 801–817. Meagher, K., & Temur, M. (2000). Aspects of Management Structures: an Individual Characteristics Approach. October. University of New South Wales. Milgrom, P., & Roberts, J. (1990). The Economics of Modern Manufacturing: Technology, Strategy, and Organization. The American Economic Review, June, 511–528. Morehead, A., Steel, M., Alexander, M., Stephen, K., & Duffin, L. (1997). Changes at Work: The 1995 Australian Workplace Industrial Relations Survey. Longman.
206
CHRISTOPHER P. ADAMS
Myerson, R. (1982). Optimal Coordination Mechanisms in Generalized Principal-Agent Problems. Journal of Mathematical Economics, 67–81. Myerson, R. (1985). Bayesian Equilibrium and Incentive-Compatibility: An Introduction. In: L. Hurwicz, D. Schmeidler & H. Sonnenschein (Eds), Social Goals and Social Organization (pp. 229–259). Cambridge University Press. Osterman, P. (1994). How Common is Workplace Transformation and Who Adopts It? Industrial and Labor Relations Review, 47, 173–187. Pliskin, J. (2000). The Incidence of Profit Sharing and Flexible Workplace Practices. Conference Paper for the 10th Conference of the International Association for the Economics of Participation, July. Trento, Italy. Radner, R. (1996). Bounded Rationality, Indeterminacy, and the Managerial Theory of the Firm. Cambridge University Press.
The Use of Profit Sharing When Workers Make Decisions
207
APPENDIX Proof of Lemma 1. The proof has two steps. Step (1) considers V00, and step (2) considers V01. Each step shows that the optimal contract is the degenerate case, because it has lower incentive costs and higher ex-ante expected revenue. Step 1. Consider the contract in which the worker does not have decision making power (t(0) = t(1)) and does not receive profit sharing ((t, Fs)). Given the assumption that the effort cost of t = 1 is greater than t = 0 by e > 0, if the contract specifies t(0) = t(1) = 0, then the incentive compatibility constraint is satisfied if (t, Fs) = 0 for all t{0, 1} and Fs{0, 1}. Given the assumption that the worker’s outside option is 0, the worker will accept this contract (t(Ws) = 0, (t, Fs) = 0). This is the degenerate case which means that there are no incentive costs. This case is optimal because of the assumption on the firm’s ex ante beliefs (p0 > (1 f )(1 p1) + f p1), the firm’s ex ante expected revenue of t(0) = t(1) = 0 is p0 which is greater than the firm’s ex ante expected revenue of t(0) = t(1) = 1 which is (1 f )(1 p1) + f p1. Therefore, V00 = p0.
(27)
Step 2. Consider the contract in which the worker does not have decision making power (t(0) = t(1)) but does receive profit sharing ((r)). As in step (1), the assumption on the effort cost implies that if (r) = 0 for all r{0, 1}, and the contract specifies t(0) = t(1) = 0, then the incentive compatibility constraint is satisfied. This is the degenerate case. This case is optimal because the firm’s ex ante expected revenue of t(0) = t(1) = 0 is p0 which is greater than the firm’s ex ante expected revenue of t(0) = t(1) = 1 which is (1 f )(1 p1) + f p1. Therefore, V01 = p0.
(28)
QED. Proof of Proposition1. The proof has a number of steps. Step (1) considers the proof of part (i), and shows that the difference V10 V00 is increasing in . Step (2) considers the proof of part (ii), and shows that the difference V11 V01 is increasing in . Step 1. Consider part (i) of the proposition. First note that V00 is constant in . The rest of the proof shows that V10 is increasing in . Under this contract the firm’s expected revenue is (1 f · )p0 + f · ((1 f1)(1 p1) + f1 p1)).
(29)
As f1 is increasing in and by assumption (1 f1)(1 p1) + f1 p1 > p0, so the firm’s expected revenue is increasing in . Now consider the firm’s incentive
208
CHRISTOPHER P. ADAMS
costs. The argument that incentive costs are decreased is the same as the argument made in the proof of Proposition 2 replacing with f1. Step 2. Consider part (ii) of the proposition. First note that V01 is constant in . The rest of the step shows that V11 is increasing in . Consider first the firm’s expected revenue. This exactly as it is stated in Step (1) and so is increasing in . Now consider the firm’s incentive costs. Under this contract the incentive compatibility constraint is (1 f1 · p1)v((0)) + f1 · p1v((1)) e ≥ (1 p0)v((0)) + p1v((1)) ≥
(30)
(1 f0 · p1)v((0)) + f0 · p1v((1)) e . For the first inequality it must be the case that v((1)) > v((0)). Given this, the second inequality also holds. By maximization the first inequality will hold with equality and can be rewritten as v((1)) v((0)) =
e . f1 · p1 p0
(31)
Note that by definition as > , ˆ f1 · p1 p0 > 0. As f1 is increasing in , from Eq. (31), the difference v((1)) v((0)) is decreasing in . Using the individual rationality constraint it is then straight forward to show that the firm’s expected payment to the worker decreases as the difference v(1) v(0) decreases. QED. Proof of Proposition 2. Part (i) of the proposition follows trivially from Lemma 1. The rest of the proof shows that Part (ii) holds, that is, V11 V10 is decreasing in . First note that V11 is constant in . The rest of the step shows that V10 increases in . By maximization any decision making contract will specify that t(0) = 0 and t(1) = 1. Given that > 0.5, the alternative will have lower expected revenue than a non-decision making contract. The contract must satisfy the incentive compatibility constraint. (1 · f1)v((1, 0)) + · f1v((1, 1)) e ≥ · f1v((0, 1)) + (1 · f1)v((0,0))
(32)
and · f0v((0, 1)) + (1 · f0)v((0, 0)) ≥ (1 · f0)v(pi(1, 0)) + · f0v((1, 1) e .
(33)
By maximization, (0, 1) = (0, 0) = 0 and the Eq. (32) holds with equality. v((1, 1)) v((1, 0)) =
e 2 · f1 1
(34)
The Use of Profit Sharing When Workers Make Decisions
209
and as f1 > f0( > 0.5), Eq. (33) also holds. By Eq. (34) the difference v((1, 1)) v((1, 0)) decreases as increases. Using the individual rationality constraint it is then straightforward to show that the firm’s expected payment to the worker decreases as the difference v((1, 1)) v((1, 0)) decreases. QED.
SPECIFIC HUMAN CAPITAL AND PARTIAL EMPLOYEE OWNERSHIP: A TRANSACTION COST ANALYSIS Nicholas Wilson, Hao Zhang and Andrew Robinson ABSTRACT In this paper, we develop, analyze, and test the hypothesis that partial employee ownership may be used as an institutional arrangement to economize on the costly problem of ex post opportunism inherent in the investment of specific human capital. Based on a unique survey of 655 British firms, we examine the empirical link between the likelihood of partial employee ownership and the presence of specific human capital, as well as uncertainty (internal and external) and frequency of transaction. Adjusted for possible structural differences, empirical evidence suggests considerable support for our hypothesis. Our results are also broadly consistent with the Transaction Cost Economics idea that institutional arrangements can be analyzed as transaction cost minimizing choices to govern specific transactions by providing economizing degrees of ex ante incentive compatibility and ex post contractual safeguard.
1. INTRODUCTION Transaction Cost Economics (TCE), over the last three decades, has developed into an increasingly mature theoretical framework.1 The main insight is that institutional arrangements can be analyzed as transaction-cost minimizing The Determinants of the Incidence and the Effects of Participatory Organizations, Volume 7, pages 211–226. Copyright © 2003 by Elsevier Science Ltd. All rights of reproduction in any form reserved. ISBN: 0-7623-1000-6
211
212
NICHOLAS WILSON, HAO ZHANG AND ANDREW ROBINSON
choices chosen to govern economic transactions. Moreover, a limited but growing body of empirical evidence appears to be remarkably consistent with TCE predictions.2 Central to the TCE framework is the idea that assets can be idiosyncratic or transaction-specific. Six types of asset specificity have been so far identified: site specificity, physical asset specificity, human capital specificity, dedicated assets, brand name capital, and temporal specificity (see Williamson, 1996). Implicit in the TCE framework is also the recognition that choices of governance, in the form of institutional arrangements, could economize on the costly problem of ex post opportunism arising from transaction-specific investments. By implication, the empirical issue then becomes how transactional attributes affect the relative efficiency of observed institutional arrangements. One way to address the issue would be to examine how well transactional attributes explain a certain choice of institutional arrangement within the framework of TCE. In this paper, we analyze the problem of ex post opportunism inherent in the investment of specific human capital and explore how a possible institutional arrangement, partial employee ownership, may economize on the costly problem. Our analysis is carried out within the TCE framework and focuses on the empirical link between partial employee ownership, in the form of Employee Share Ownership Plans (ESOPS), and a number of transactional attributes including specificity, uncertainty (both measurement and environment), and the frequency of transaction. Ours is the first study to explain partial employee ownership using the framework of TCE. Our study is based on a survey of 655 non-diversified, mainly single-product British firms containing information on the existence of an ESOP and relevant transactional attributes. The paper is organized as follows. Section 2 reviews the theory as well as previous studies on specific human capital and presents our ESOP hypothesis within the TCE framework. Section 3 provides background information on British ESOPS. Section 4 explains the data and posits a logistic model to test the proposed hypothesis. Results and a discussion of them appear in Section V. We conclude in the final section.
2. THEORY, PREVIOUS STUDIES, AND HYPOTHESIS A contracting problem arises whenever some or all the inputs, human or nonhuman, supplied to a producer are transaction-specific. The problem is that should the producer renege on the contract ex post, suppliers of transactionspecific inputs, both human or non-human, cannot put the inputs to alternative use without incurring additional transaction cost. Thus, the more transaction-
Specific Human Capital and Partial Employee Ownership
213
specific the input, the lesser the value of alternative use. In extreme cases, transaction-specific assets, human or non-human, may have no alternative use. However, the problem becomes relevant only in the real (as opposed to ideal) world where all contracts are necessarily incomplete.3 The reason is that contractual incompleteness engenders ex post opportunistic activities among contracting parties. In the world of incomplete contracts, it is moreover likely that presence of uncertainty may compound the contracting problem related to transaction specificity. When associated with transaction specificity, uncertainty encourages and facilitates opportunistic activities. These problems, if not addressed, would lead to under-investment of transaction-specific assets and inefficiency (both organizational and technological) of production. TCE predicts that certain institutional arrangements, individually or together, are designed to mitigate these problems by effecting a governance structure. This governance structure would be capable of providing economizing degrees of ex ante incentive compatibility and ex post contractual safeguard. The choice of an arrangement depends on its relative efficiency in relation to the type of transaction. A wide range of institutional arrangements has been studied empirically (Shelanski & Klein, 1995). In terms of specific human capital, a number of studies (e.g. Monteverde & Teece, 1982; Anderson & Schmittlein, 1984; Masten, 1984; Masten et al., 1989) have empirically examined vertical integration as a means to deal with problems arising from transaction specificity. Generally, this is done through the use of a make-or-buy logistic model with selected proxies (gathered from survey data) for different transaction attributes.4 The evidence consistently suggests specific physical capital as a determinant of vertical integration. The main insight of these studies is that since the market is unwilling or unable to supply this type of capital due to specificity, firms may find it more economical to produce it internally (vertical integration). However, it is well known that in the case of specific human capital (unlike specific physical capital), the problem of opportunistic behavior cannot be eliminated by vertical integration (e.g. Masten, 1984). Vertical integration cannot control vertically integrated specific human capital as it can control specific physical capital. The main reason is that the firm does not own or control human capital as it does physical capital. Rather, it is often argued that the hierarchical nature of the integrated firm could go some distance to resolve disputes and enforce co-operative behavior. The details of how co-operation within a hierarchy would come about remain sketchy and vague. In comparison with specific physical capital, specific human capital may be classified as under the “weak property right regime” (Teece, 1986) since it can
214
NICHOLAS WILSON, HAO ZHANG AND ANDREW ROBINSON
be subject to ex post opportunism. Once invested, the specific component of the human capital is tied to the firm’s transactions or “sunk” in the sense that it is costly to put it into alternative use. Without a measure of ex post contractual safeguard, neither the employee nor the employer would want to invest in specific human capital even when it is efficient to do so. If employees invest specific human capital on themselves, the employer ex post may threaten not to use it in order to extract a greater share of the “rent” or surplus resulting directly from the use of human capital specificity. Foreseeing the likely consequences, employees ex ante would under-invest in specific human capital and over-invest in general-purpose human capital for themselves. If the employer invests specific human capital for employees, the employees ex post may threaten to under-use it in order to extract a greater share of the “rent” or surplus resulting directly from the use human capital specificity. Foreseeing this, the employer would also refrain from investing in specific human capital on employees. The under-investment of specific human capital may aversely affect both the organizational and technological efficiency of production.5 We hypothesize that, when efficiency invites the investment of specific human capital by both employers and employees, firms may use partial employee ownership as institutional arrangement to provide economizing degrees of ex ante incentive compatibility and ex post contractual safeguard both to encourage and protect specific human capital investments. The reason is that partial ownership produces both a measure of residual control to deter employer ex post opportunism and a measure of profit-sharing to discourage employee ex post opportunism.6 The arrangement is obviously not perfect, but it could go someway to economize on the costly problem of “weak property regime” inherent in the investment of specific human capital. Conceptually, employees enjoying equity ownership are under a form of hybrid contracting: the general-purpose fraction of human capital usage is compensated through wages while the specific fraction of the usage is compensated through partial ownership. Thus our hypothesis implies that partial employee ownership may function as a transaction cost economizing institutional arrangement to govern human capital specific transactions. We also hypothesize that both measurement uncertainty and environment uncertainty may explain partial employee ownership. Measurement uncertainty suggests that it is costly to observe and verify the utilization of human capital or effort. The problem is likely to be more acute in the case of specific human capital because capital utilization benchmarks are usually not available at any cost. As a result, the employee on fixed wages tends to under utilize his/her human capital or shirk. Partial employee ownership can be used as an
Specific Human Capital and Partial Employee Ownership
215
institutional arrangement to economize on the measurement cost component of transaction cost by effecting a greater degree of incentive compatibility between the employer and employees. Environment uncertainty suggests that the state of nature (e.g. demand conditions of a product) is difficult to predict. One implication is that environment uncertainty discourages the supply of specific capital, both human and non-human, from the external market. The reason is that the specific capital is difficult to re-deploy. If the efficient gains from specific capital usage are large enough to justify the additional (governance) cost, firms would produce it internally. Internal investment/ production may come in the form of vertical integration for specific physical capital and employee ownership for specific human capital. Thus our hypothesis implies that partial employee ownership may function as an institutional arrangement to economize on measurement and environment uncertainties when specific human capital is present. The TCE framework also predicts that the frequency of transactions makes the firm more willing to internally produce and maintain efficiency enhancing specific assets. Because the governance cost of producing and maintaining specific assets is greater than that of producing and maintaining generalpurpose assets, the efficiency gain from using specific assets must be large enough to justify the additional (governance) cost. Since the efficiency gain resulting of using specific assets is likely to be greater in frequent transactions than infrequent one, firms are more likely to adopt institutional arrangements (e.g. vertical integration, partial employee ownership) designed to produce and maintain specific assets in frequent transactions than infrequent ones. Thus, we hypothesize that partial employee ownership may function as an institutional arrangement to economize on the frequency of transaction. Thus, the operational version of the hypothesis is as follows: within vertically integrated firms, specific human capital, environment uncertainty, measurement uncertainty, frequency of transactions can predict the presence of partial employee ownership (in the form of ESOPS).
3. BRITISH ESOPS Partial employee ownership is nothing new. Economic historians have shown that various employee ownership schemes were introduced in Britain during the late nineteenth century (Hatton, 1988), although these schemes in general covered only about 0.5% of the British labor force (Wilson, 1992). In the latter half of the 1980s, however, a new form of employee ownership emerged in the U.K. The Employee Share Ownership Plan (ESOP) differs from other forms of employee share schemes in that the scale of employee ownership is usually
216
NICHOLAS WILSON, HAO ZHANG AND ANDREW ROBINSON
considerably larger and thus the ESOP structure facilitates a greater degree of active involvement and residual control. Most ESOPs take what is known as the ‘case-law’ form, and are based on trust and profit sharing legislation. Initially, equity is held in an Employee Benefits Trust before transfer to a Profit Sharing Trust. These transfers are financed by profits, and the shares are distributed to employees (on a nominal or no cost basis) in accordance with the 1978 Finance Act. These payments out of profits are used by the Employee Benefits Trust to repay the loan taken out to finance the purchase equity at the outset. The other form of ESOP in the U.K. is known as a “statutory” ESOP as it was established by legislation in the 1989 Finance Act (as amended). In this case a single trust – an Employee Share Ownership Trust (ESOT) or Qualifying Employee Share Trust (QUEST) – acquires, holds and distributes equity to employees. To secure Inland Revenue approval and access to tax concessions it is necessary for at least half of the trustees (a majority prior to 1994) to be elected by a majority of the workforce. The growth of British ESOPs, or the lack of it, has been more evolutionary in that it had not been promoted specifically for tax reasons. Indeed, the “statutory ESOP” in the form of Finance Act 1989 was criticized because of its “lack of meaningful tax incentives” and inflexibility for “imaginative tax planning” (Cornford, 1990). Moreover, the qualification requirements for statutory ESOPS (as outlined above) are often viewed by employers as giving too great a decision-making role to employees. Yet, for employees, it is this very link between share schemes and decision making which is its strength and may be enough to motivate them to make firm specific investments in human capital. Not withstanding these issues ESOPs provide a low-cost, low-risk method for employees to acquire substantial portions of equity and control in their employer. Consequently, it is of interest to examine, from a random sample, the characteristics British firms that do establish ESOPs as against those that do not.
4. DATA AND METHODOLOGY Data for this study is from 655 British firms (out of 4,094) that responded to a mail survey on trade credit practice conducted by the authors in late 1994. The survey collected over 300 items providing detailed data on the firm; its ownership, organizational and governance structure, the size and composition of the labor force, product and market characteristics, supplier relationships, financing and credit and financial management practices. The firms mailed were a randomly selected stratified sample of companies originally drawn from the U.K. FAME database.7 The companies selected all
Specific Human Capital and Partial Employee Ownership
217
had a single primary product category, which facilitates the analysis of the relationship between product characteristics and aspects of organizational design. Holding companies and large companies with diverse product profiles were deliberately excluded from the sample. As a check, respondents were asked to identify their primary product and confirm the percentage of sales it represented. The mean percentage of turnover from the primary product category was 85% with the median being 95%. The survey achieved a 16% response rate, which is in the expected range given the detailed nature of the questionnaire. Most responses (81%) related to private limited companies, 16% to Publicly Listed Companies. The remaining companies were wholly owned subsidiaries, partnerships or overseas companies. Almost half of companies (47%) had a majority stake held by current directors, with a further 7% having the current directors holding a significant minority stake. The median number of employees in the sample was 100, with a maximum of 9,500 and a minimum of 4. Seventy one percent of firms had less than 200 employees. Over a third of firms (36%) were involved in the metal engineering industries; 12% in paper, publishing & packaging; 11% in chemicals and petroleum; 10% in materials production (e.g. glass, ceramics, bricks); 9% in textiles; 7% in producing food and drink; 7% in services; 3% in construction, 3% in wholesale and retail; and 1% in primary industries. The areas covered by the survey, which are relevant to this study, include employee ownership and transaction attributes such as specificity, uncertainty, and frequency of transaction. Excluding observations with missing data leaves a useable sample of 638 companies for this study. Out of the 638 companies, 34 (5.3%) are set up with an employee benefit trust or ESOPS (Table 1, Panel A). To examine the likelihood of having ESOP in a company, we use a logistic response model. BESOP, the binary variable, is set to 1 if a company has ESOP, 0 otherwise. Independent variables (both TCE and structural) are defined in Data Appendix. Specificity It is not possible to obtain direct measures of specific human capital. Previous studies, for example, either use production-related measures or indices of surveyed opinions about the nature assets as proxies for specificity.8 However, it is reasonable to assume that specific human capital is highly and positively correlated with specific production processes. Given this assumption, measures of specific human capital are obtained from a number of proxies related to the production process. The proxies are: (1) whether the product line is customized
218
NICHOLAS WILSON, HAO ZHANG AND ANDREW ROBINSON
(SpCum); (2) whether the product is idiosyncratic or different (SpDiff); (3) whether the product is technical in nature (SpTech); and (4) the portion of job (as opposed to batch and flow) type of production (SpJob) used. The varying Table 1.
Selected Statistics of the Sample of 638 Companies.
Panel A: ESOP Companies with ESOP: Companies without ESOP:
34 (5.3%) 604 (94.7%)
Panel B: Company Characteristics Company Type Raw Material Supplier Intermediate Producer Final Producer Wholesaler Retailer Commercial Services Consumer Services
Percentage (%) 4.9 28.5 52.7 4.7 2.7 4.9 1.7
Panel C: Industry Characteristics Industry Type Primary Industries Food and Drink Chemicals and Petroleum Metal Engineering Textiles and Leather Timber and Bricks Paper Print Publishing Construction Wholesales and Retail Services
Percentage (%) 1.2 5.6 11.1 34.6 9.0 10.2 11.4 2.9 7.1 6.9
Panel D: Organizational Characteristics Legal Status Public Company Private Company Partnership or Proprietorship Ownership Firms with current directors holding majority shares: Firms with current director holding significant minority shares:
Percentage (%) 16.1 81.1 2.8 Percentage (%) 47 7
Specific Human Capital and Partial Employee Ownership
219
degrees of specificity are thus obtained either by qualitative differences (from 1 to 5) as in SpDiff, SpCum, and SpTech or by actual portion as in SpJob. Uncertainty With regard to uncertainty, previous studies tend to rely on directly surveyed opinions (e.g. Shelanski & Klein, 1995). We obtain our two different measures of uncertainty: measurement (internal) uncertainty, UncerMea, and market (external) uncertainty, UncerMkt in the similar way. In both proxies, uncertainty is views from the firm’s perspective (as felt by the management). The proxy for measurement uncertainty takes the form of the degree of difficulty (as felt by the management) in inspecting the quality of the company’s product. The proxy for market uncertainty takes the form of the fast changing nature of the product line (as felt by the management). The varying degrees of uncertainties are obtained by qualitative differences (from 1 to 5). Frequency We use the proportion (from 1 to 100) of sales that represent repeated business as a proxy for frequency of transaction (Frequency). There is information in the survey (Table 1) regarding the structural differences of these firms in different dimensions. They include company type, industry type, legal status, ownership concentration, and the total number of employees in a company. We use dummy variables (defined in the Data Appendix) to adjust these structural differences and use the total number of employees as a proxy for size. All the variables above are hypothesized to explain the likelihood of having an ESOP in a company. To account for the difference in scale, all continuous variables are standardized by subtracting the variable’s mean and then dividing by the standard deviation. Thus, the standardized variables have a mean of zero and a standard deviation of one, making the magnitudes of coefficients comparable. We posit the following logistic model: Z = a + b1(SpJoB) + b2(SpCum) + b3(SpDiff ) + b4(SpTech) + b5(UncerMea) + b6(UncerMkt) + b7(Frequency) + b8(RS) + b9(IP) + b10(FP) + b11(CS) + b12(RT) + b13(WS) + b14(Ind1) + b15(Ind2) + b16(Ind3) + b17(Ind4) + b18(Ind5) + b19(Ind6) + b20(Ind7) + b21(Ind8) + b22(Ltd) + b23(Pop) + b24(OM) + b25(Size).
220
NICHOLAS WILSON, HAO ZHANG AND ANDREW ROBINSON
Z = Ln[P/(1 P)], where P is the probability that the firm has an ESOP (i.e. BEOP = 1).9 As our TCE based hypothesis is that the presence of specificity, uncertainty, and frequency would increase the likelihood of ESOP, the TCE coefficients, b1, b2, b3, b4, b5, b6, and b7, are expected to be positive and statistically significant. It is not clear, ex ante, how the coefficients accounting for structural differences in different dimensions including company type, industry type, legal status, ownership concentration, and size would influence the likelihood of ESOP.
5. RESULTS AND DISCUSSION The results from the logistic model are presented in Table 2. The unrestricted model includes both TCE and structural variables. The restricted model includes structural variables only. The chi-square statistic of 87.182 rejects the hypothesis that the TCE and structural variables together are not statistically different from zero. The Likelihood Ratio index, which can be construed as R-Square for logit regressions, is 0.7 indicating good fit. We further check the robustness of the TCE specification using: (1) the likelihood ratio test for statistical significance, and (2) Akaike (1974) Information Criterion for predictive effectiveness. The likelihood ratio test statistic is c = –178.348 + 205.047 = 26.7. The critical value of Chi-Square distribution with 7 degrees of freedom at 0.01 level is 18.48, rejecting the hypothesis that TCE variables are not jointly significant. AIC = –2 (log likelihood number of independent variables). The value of AIC for the model with TCE variables is 230.348. The value of AIC for the model without TCE variables is 243.047. As the smaller AIC is preferred, the model with TCE variables is more predictive The coefficients of three out of the four measures of specificity are positive and statistically significant. The SpJob coefficient is positive and statistically significant at 5% significance level. This result suggests that the more idiosyncratic the production process (job as opposed to batch or flow productions), the higher the likelihood of ESOP. The SpCum coefficient is positive and statistically significant at 10%. This result suggests that the more customized the product line, the higher the likelihood of ESOP. The SpDiff coefficient is also positive and statistically significant at 10%. This result suggests that the more idiosyncratic the product, the higher the likelihood of ESOP. The SpTech coefficient, however, is positive but not statistically significant. This result suggests that the degree of technology does not automatically increase the likelihood of ESOP.
Specific Human Capital and Partial Employee Ownership
221
Table 2. Maximum Likelihood Estimated Coefficients of ESOP Logistic Response Model Based on 638 Companies (Independent Variable = BESOP). Variable
Unrestricted Model Coefficient t-Statistic
Constant
0.0584
0.05
SpJob SpCum SpDiff SpTech
0.4970 0.3863 0.3819 0.1855
2.16*** 1.82** 1.76** 0.77
UncerMea UncerMkt
0.5247 0.5298
2.50**** 2.60****
Frequency
0.3772
1.51*
RS IP FP CS RT WS
–0.0292 –1.0456 –0.8770 –10.657 –8.9814 1.5600
Restricted Model Coefficient t-Statistic –0.3069
0.34
–0.02 –1.06 –0.93 –0.25 –0.14 1.38
–0.1445 –0.4009 –0.2943 –8.0985 –7.9189 1.0760
0.13 –0.45 0.35 –0.29 –0.21 1.07
IND1 IND2 IND3 IND4 IND5 IND6 IND7 IND8
–8.2505 –0.1552 –0.5395 –1.0118 –1.1823 –8.7532 0.4603 0.2659
–0.09 –0.16 –0.75 –1.69** –1.23 –0.03 0.65 0.24
–7.7494 –0.4709 –0.4506 –1.0854 –1.4356 –7.8747 0.2756 0.1790
–0.14 –0.52 –0.68 –1.81** –1.62 –0.41 0.44 0.18
LTD POP OM Size
–2.6265 –0.3258 –0.8679 –0.1745
–4.40**** –0.37 –1.84** –0.71
–2.0036 –0.3665 –0.8549 –0.0607
–5.01**** 0.46 –1.97*** 0.28
Chi-Square = 87.182**** (25 df), Likelihood Ratio Index = 0.70. –2 Log Likelihood Unrestricted = 178.348. –2 Log Likelihood Restricted = 205.047. –2[Log Likelihood (Unrestricted) Log Likelihood (Restricted)] = 26.70**** (7 df ). Akaike’s Information Criterion (Unrestricted) = 230.348. Akaike’s Information Criterion (Restricted) = 243.047. * indicates statistical significance at 13% level. ** indicates statistical significance at 10% level. *** indicates statistical significance at 5% level. **** indicates statistical significance at 1% level.
222
NICHOLAS WILSON, HAO ZHANG AND ANDREW ROBINSON
Evidence from specificity variables appears to imply a provocative idea that more technical or hi-tech products do not necessarily lead to partial employee ownership. ESOP is likely to happen in companies with: (1) idiosyncratic production processes, (2) customized product lines, and (3) idiosyncratic products. In other words, highly trained but transferable human capital does not explain the existence of partial employee ownership. Rather, non-transferable or specific human capital explains the existence of partial employee ownership. This evidence strongly supports our transaction-cost hypothesis. The coefficients of the two measures of uncertainty are both positive and statistically significant. The UncerMea coefficient is positive and statistically significant at 1%. This result suggests that a higher likelihood of ESOP is induced by the inherent difficulties in inspecting the quality of the product. The UncerMkt coefficient is also positive and statistically significant at 1%. This suggests that fast changing product lines (a measure of market uncertainty) may increase the likelihood of ESOP. Evidence regarding uncertainty is broadly in support of the transaction-cost hypothesis. It is, however, noted that measurement uncertainty is as important as external/market uncertainty in inducing partial employee ownership. This result suggests the importance of measurement uncertainty in relation to partial employee ownership and is entirely consistent with the idea that institutional arrangements are used to minimize the measurement cost component of transaction cost (Alchian & Demsetz, 1972). The result in relation to the effects of frequency of transaction on ESOP is, while supportive of the transaction cost hypothesis, weaker. The Frequency coefficient is positive but statistically significant only at 13% significance level, just outside the 10% level for a two-tailed test but at the 10% for a one-sided test. While there is some evidence to suggest that the higher the frequency of transaction, the higher the likelihood of ESOP, the evidence has to be qualified as marginal or inconclusive. The adjustments of structural differences in company type, industry type, legal status, ownership concentration, and size do not appear to have a strong impact on the likelihood of ESOP. Firms with current directors owning substantial shares seem to have a lower likelihood of ESOP. While partnership and proprietorship do not seem to influence the likelihood of ESOP, limited companies (LTD), relative to public companies (the omitted category), seem to reduce it. One explanation may be that limited companies, partnerships, or proprietorships may not want to go to the trouble to set up statutory ESOP. Privately arrangements may accomplish the same objective and incur less transaction cost. Also, individual industry differences do not appear to affect the likelihood of ESOP except IND4 (metal engineering industry) which,
Specific Human Capital and Partial Employee Ownership
223
relative to the omitted industry (retail/wholesale/services), significantly reduces it. However, an additional likelihood ratio test shows that the industry coefficients are jointly different from zero. On the whole, our evidence suggests considerable support for the hypothesis that specific human capital, uncertainty (both internal and external), and frequency of transaction explain the presence of partial employee ownership.
6. CONCLUSION In this paper, we develop a hypothesis regarding partial employee ownership as an institutional arrangement to economize on the costly problem of ex post opportunism inherent in the investment of specific human capital. Our analysis is carried out within the framework of TCE. Based on a unique data set of 638 companies, we empirically test the hypothesis by investigating the link between ESOPS and a number of transaction attributes: human capital specificity, uncertainty, and frequency of transaction. Our results show a positive and statistically significant relation between ESOP and human capital specificity, uncertainty, and, to a lesser extent, frequency of transaction. We control for company type, industry type, legal status, ownership concentration, and size differences. The evidence is in support of the notion that partial employee ownership may function as an institutional arrangement to economize on the costly problem of ex post opportunism. Our results are also broadly consistent with the intrinsically TCE idea that institutional arrangements can be analyzed as transaction cost minimizing choices to govern specific transactions by providing economizing (as opposed to perfect) degrees of ex ante incentive compatibility and ex post contractual safeguard.
NOTES 1. See, e.g. Coase (1937), Williamson (1975, 1985, and 1996), Klein et al. (1978), Grossman and Hart (1986), Hart and Moore (1990). 2. See surveys on the field by Joskow (1988) and Shelanski and Klein (1995). 3. Contractual incompleteness arises because: (1) not all contingencies can be foreseen ex ante, and (2) some foreseeable contingencies can be, ex post, very costly to observe and/or verify. 4. Our study of partial employee ownership is carried out within this tradition. 5. The problem could conceivably be solved if the employer and employees jointly and equally invest in specific human capital on the employees, provided that the division is measurable, observable, and verifiable. 6. In a related study, Pisano (1990) asks why firms may rely on equity linkage to support certain transactions. He argues that partial ownership will dominate contractual
224
NICHOLAS WILSON, HAO ZHANG AND ANDREW ROBINSON
governance when a transaction involves specific capital and uncertainty. He finds evidence from the biotechnology industry to support his argument. Our argument is broadly similar. 7. Separate samples were drawn from the firms in each manufacturing SIC code and random selections from these were combined to form the mailing file. 8. See, e.g. Teece (1982), Masten (1984), and Masten et al. (1989). 9. See, e.g. Pindyck and Rubinfeld (1991) for descriptions of the logistic model.
ACKNOWLEDGMENTS We wish to thank John Cable and Oliver Williamson for comments and encouragement.
REFERENCES Akaike, H. (1974). A New Look at the Statistical Model Identification. IEEE Transactions on Automatic Control, 19, 716–723. Alchian, A. A., & Demsetz, H. (1972). Production, Information, and Economic Organization. American Economic Review, 62, 777–795. Anderson, E., & Schmittlein, D. C. (1984). Integration of the Sales Forces: An Empirical Examination. Rand Journal of Economics, 15, 385–395. Carnell, D. (1992). ESOPS in Listed and Unquoted Companies Compared. In: N. Wilson (Ed.), ESOPS. Macmillan. Cornford, J. (1990). A Stake in the Company. IPPR. ES No. 3. Coase, R. (1937). The Nature of the Firm. Economica, 4, 386–405. Grossman, S., & Hart, O. (1986). The Costs and Benefits of Ownership: A Theory of Vertical and Lateral Integration. Journal of Political Economy, 94, 367–388. Hart, O., & Moore, J. (1990). Property Rights and the Nature of the Firm. Journal of Political Economy, 98, 1119–1158. Hatton, T. J. (1988). Profit Sharing in British Industry 1885–1913. International Journal of Industrial Organization, 6(1), 69–90. Joskow, P. (1988). Asset Specificity and the Structure of Vertical Relationships: Empirical Evidence. Journal of Law, Economics, and Organization, 4, 95–117. Klein, B., Crawford, R. A., & Alchian, A. A. (1978). Vertical Integration, Apropriable Rents, and the Competitive Contracting Process. Journal of Law and Economics, 89, 297–326. Masten, S. E. (1984). The Organization of Production: Evidence from the Aerospace Industry. Journal of Law and Economics, 27, 403–417. Masten, S. E., Meechan, J. W., & Snyder, E. A. (1991). Vertical Integration in the U.S. Auto Industry: A Note on the Influence of Specific Assets. Journal of Law, Economics, and Organization, 7, 265–273. Monteverde, K., & Teece, D. J. (1982). Supplier Switching costs and Vertical Integration in the automobile Industry. Bell Journal of Economics, 13, 206–213. Pindyck, R. S., & Rubinfeld, D. L. (1991). Econometric Models and Economic Forecasts. New York: McGraw Hill. Pisano, G. P. (1990). Using Equity Participation to Support Exchange: Evidence from the Biotechnology Industry. Journal of Law, Economics, and Organization, 5, 109–126.
Specific Human Capital and Partial Employee Ownership
225
Shelanski, H., & Klein, P. G. (1995). Empirical Research in Transaction Cost Economics: A Review and Assessment. Journal of Law, Economics, and Organization, 11, 335–361. Teece, D. J. (1986). Profiting from Technological Innovations. Research Policy, 15, 285–305. Williamson, O. (1975). Markets and Hierarchies. New York: Free Press. Williamson, O. (1985). The Economic Institutions of Capitalism. New York: Free Press. Williamson, O. (1996). The Mechanisms of Governance. New York: Oxford University Press. Wilson, N. (1992). ESOPS: Their Role in Corporate Finance and Performance. New York: Macmillian.
DATA APPENDIX Transaction cost proxies are obtained from the following survey questions. Variable
Definition Panel A:
Specificity SpDiff
SpCum SpTech SpJob
The product we sell is essentially different in terms of physical properties to other product on the market (1, strongly disagree/5, strongly agree) Our product line can be best described as customized (1, strongly disagree/5, strongly agree) Our product line is technical in nature (1, strongly disagree/5, strongly agree) What proportion (from 1 to 100) of output is accounted for by job type production?
Uncertainty UncerMea The quality of our product line is difficult to determine by inspection only (1, strongly disagree/5, strongly agree) UncerMkt Our product line can be described as fast changing (1, strongly disagree/5, strongly agree) Frequency The proportion (from 1 to 100) of our sales that represent repeated business is: 1 (0); 2 (1–20); 3 (21–40); 4 (41–60); 5 (61–80); 6 (81–100).
226
NICHOLAS WILSON, HAO ZHANG AND ANDREW ROBINSON
Panel B Other Variables: Company Type RS = 1 if the company is a raw material supplier, 0 otherwise. IP = 1 if the company is an intermediate producer, 0 otherwise. FP = 1 if the company is a final producer, 0 otherwise. RT = 1 if the company is a retailer, 0 otherwise. WS = 1 if the company is a wholesaler, 0 otherwise. CS = 1 if the company is commercial service company, 0 otherwise. Industry Type IND1 = 1 if the company is in Primary Industries, 0 otherwise. IND2 = 1 if the company is in Food and Drink, 0 otherwise. IND3 = 1 if the company is in Chemical and Petrol, 0 otherwise. IND4 = 1 if the company is Metal Engineering, 0 otherwise. IND5 = 1 if the company is in Textile and Leather, 0 otherwise. IND6 = 1 if the company is in Timber/Bricks, 0 otherwise. IND7 = 1 if the company is in Paper/Print/Publishing, 0 otherwise. IND8 = 1 if the company is in Construction, 0 otherwise. (Retail/Wholesale/Service industry is omitted). Legal Status Ltd = 1, if the company is registered as a limited company, 0 otherwise. Pop = 1, if the company is registered partnership or proprietorship, 0 otherwise. Ownership Concentration OM1 = 1, if the current company directors hold majority or substantial minority shares, 0 otherwise. Size Total number of employees is used as the proxy for size.
UNION-FIRM BARGAINING OVER LONG TERM BENEFITS Jan Erik Askildsen and Norman J. Ireland ABSTRACT The paper studies bargaining over workers’ benefits which are due some time in the future. A union bargains on behalf of a workforce which may be diverse in the sense that workers’ probabilities of staying with the firm vary. Bargaining structure, rather than the bargaining power of the union, is found to be the driving force in the model for determining the level of benefits. A further key issue is that of whose preferences are represented in the union’s objective function, and thereby in the bargaining process.
1. INTRODUCTION The relationship between a firm and its workers often involves more than just paying a wage for the delivery of a contracted effort. Some workers may stay with a firm for a long period, and during a long term relationship they will receive different benefits from the firm. The most commonly observed kinds of deferred payment are old age pensions and sickness insurance, and in some instances shares or stock options. Such arrangements can be agreed upon individually, or bargained collectively by a union and the firm. In Europe firmrelated pension schemes are generally provided on top of social insurance arrangements, whereas in the USA such pensions and sickness insurance arrangements may be the only insurance held. The role of the firm in providing savings and as insurance vehicles for its workers’ pension plans has many The Determinants of the Incidence and the Effects of Participatory Organizations, Volume 7, pages 229–248. Copyright © 2003 by Elsevier Science Ltd. All rights of reproduction in any form reserved. ISBN: 0-7623-1000-6
229
230
JAN ERIK ASKILDSEN AND NORMAN J. IRELAND
explanations. For example, the firm’s workforce may constitute a pool where adverse selection issues are avoided. Also, the firm may be able to avoid many of the costs of operating personal pension schemes by avoiding commissions and spreading fixed transactions costs over all employees. This latter explanation will be sufficient for our purpose: lower costs make firm pension schemes a preferred savings mechanism. This paper seeks to examine the implications of the bargaining activity between firms and labor unions for such deferred benefits. Wage bargaining may be considered the prime role of unions, but how does bargaining relate to the efficient provision of deferred benefits and pensions? The promise of a future payment always involves some risk for the recipient, see Budden (2000) for some case-specific examples, and Curme and Kahn (1990) and Orr (1998) for more general considerations of the risk involved. Will the firm ever be in a position to deliver the goods? Specifically, the firm may go bankrupt, or may close down part of its operation, leaving the individual employee with a substantially lower pension fund as well as no job. An obvious example is the case of defined benefit pension schemes. Here, the employee is promised a pension linked to his/her final salary. If his/her job is removed in mid-career before retirement age then it is only the credit relating to mid-career salary which determines the contribution of this employment to pension: the enhancement, through income growth to retirement age, of early years of service is lost. Also, the pension may be switched into a defined contribution rather than defined benefit scheme, with essentially the same outcome. Orr (1996) discusses the impact of firms “downsizing” on the type of pension offered in the USA. Bloom and Freeman (1992) show that the reduction in pension coverage rates and the shift to defined contribution schemes are only partly explained by structural changes. Orr argues that the remaining factors are linked to downsizing and the desire to avoid the capital risk of the defined benefit schemes. This is particularly sensitive in situations where mortality rates are falling so that actuarial costings of pension benefits tend to be understated. Certainly considerable concern is currently being voiced in the U.K. at both the capital risk of defined benefit schemes and the opportunities for firms to terminate these schemes in favor of less costly defined contribution schemes. On the other hand, it is not certain whether future compensation will have to be made in full by the firm since some or indeed all workers may leave the firm voluntarily before the time of delivery arrives. The credibility of firm promises, the duration of worker-firm relationships and workers’ seniority may be important characteristics of how to evaluate compensation by means of deferred payment.
Union-Firm Bargaining over Long Term Benefits
231
An important distinction in bargaining theory is between the right-to-manage model, and the efficient bargaining model, see McDonald and Solow (1981) and the survey by Oswald (1985). In the former model only wages are bargained and the firm unilaterally determines employment on the labor demand curve. However, this contract is not efficient. Efficient bargaining involves bargaining over wages and employment simultaneously, yielding an efficient outcome on the union-firm contract curve. Manning (1987) shows that the crucial point is not whether bargaining takes place simultaneously. What matters is that the bargaining power is the same on all variables of importance. If a variable influencing the union utility is not covered by bargaining, the union or the firm will anticipate a reaction from the other, and the outcome is off the contract curve. Thus, Grout (1984) and Hoel and Moene (1988) show that the absence of binding contracts between firm and union leads to inefficient levels of capital, and that the capital stock is influenced by the union bargaining power. A pure efficient bargaining model is represented by the labor-managed firm, see Ireland and Law (1982), and long term contracts are credible since the workers control all variables. Thus workers have themselves paid the cost, and have an incentive to stay with the firm until future benefits are paid (Askildsen & Ireland, 1993). It is clearly more difficult for a profit maximizing firm to engage in long term relationships, since it may have an incentive to deviate and choose a myopic, short term optimal path. Promises, not supported by credible contracts that are verifiable in a court of law, may be dynamically inconsistent.1 Some practical consequences of risky firm behavior relate to costly insurance schemes for pensions, see Targett and Beattie (2000). The key issue of this paper is the analysis of bargaining when the firm can take actions to make the delivery of deferred benefits more likely. Examples include: (i) undertaking less risky investments to avoid significant chance of bankruptcy, (ii) rejection of offers of takeover and merger, (iii) avoiding downsizing at too fast a rate, and (iv) avoiding employment policies which penalize existing workers (for example hiring senior personnel from outside rather than internal promotion). Such management policy choices obviously come under the normal interpretation of right to manage, rather than a factor in the bargaining process. The extent of deferred benefits can be agreed but not the mechanism which ensures their delivery. The question is what then determines the level of deferred benefits. Is it the distribution of bargaining power or the level of the firm’s commitment to delivery? Within the limitations of our model we will argue that it is the latter. Thus the highest levels of deferred benefits will be observed in firms which have a high survival probability and a reputation for cradle-to-grave employment, and not necessarily in those firms where unions
232
JAN ERIK ASKILDSEN AND NORMAN J. IRELAND
have high bargaining power.2 An important assumption of our model is that individual workers have risk aversion increasing with age (see next section). A further issue relates to internal conflicts within a union. It may be difficult to agree on a common goal of the union when workers are heterogeneous. In particular they will differ in terms of their interest in the expected return from future benefits like a pension plan. It is customary to settle such political economy problems applying a median voter approach. However, the power structure of unions is not always so simple. Senior workers may have more power that younger workers, or there may be an elected agenda setter, see Freeman (1985) and Flanagan (1993). Thus, the content of the contract, but not the optimality of the bargaining process as such, is likely to be affected by the internal power structure of the union. An accepted tradition of “last-in-first-out” (LIFO) is a clear example of an allocation of power which may affect both union behavior and the results of bargaining. The question then arises as to the impact of this kind of priority on the bargaining outcome when bargaining includes deferred benefits. Our approach will be to take a very simple stylized model, and this is explained in the next section. The benchmark Pareto optimal solution is presented in Section 3, and equilibrium under different bargaining regimes are compared in Section 4. Heterogeneity in the work force is introduced in Section 5, and a reinterpretation and modification of the model is made in Section 6 to permit the LIFO convention to be analyzed. Section 7 makes some concluding remarks.
2. MODEL Our model will be based on the following assumptions and framework. (1) Individual workers live for two periods. In the first they receive wage income w; in the second they may receive a benefit s on top of a base level of benefit normalized to zero. (2) The benefit is paid to the workers only if they stay with the firm and the firm remains in business; thus, both parties must survive for the benefit to be realized. The survival probability of the firm is p. The probability q of a worker staying with the firm, given that the firm stays in business, represents the type of each worker. In this paper we treat q as exogenously fixed. In the early sections all workers have the same q; later this is generalized. (3) The worker’s lifetime utility is the sum of two period utility functions. Appropriate discount factors are included in the second period utility, but are left implicit since they play no additional role in the analysis.
Union-Firm Bargaining over Long Term Benefits
233
(4) Workers are assumed to have increasing risk aversion in age. This has theoretical, empirical and practical support. The theory applies to the notion of the individual making lifestyle choices in period 1 (for example how hard to work given the wage rates, what consumer durable goods are chosen to buy) and then being left few degrees of freedom when old or ill. Thus, retirement plans are based on a given income, but shortfalls on the income have dire consequences just as overshoots on income have little benefit. For older men and women “the lure of substantial financial gains was not worth the possible loss of money-in-hand.” (Botwinick, 1978, pp. 129–130). Essentially, the inability to make up lost income by working harder in conjunction with a more uncertain lifetime (relative to mean lifetime) makes older people more risk averse. The practical evidence that supports increasing risk aversion is the recommended investment strategies as age changes: on retirement a non-risky portfolio is the standard recommendation. Empirical analysis, for example Bakshi and Chen (1994) relate the data on U.S. savings to this “life-cycle risk aversion hypothesis”, by looking at how aggregate behavior changes with the age distribution. They stipulate a relative risk aversion coefficient as changing linearly with average age, and find a significant positive slope, particularly for post-war USA. (5) To make the model tractable the increasing risk aversion is assumed to take a very simple form. The benefit yields a second period utility u(s) to each worker, where u(s) > 0, u(s) < 0,
u(0) = 0,
u(0) > 1.
First period utility is assumed linear in the wage w. Thus the worker changes from being risk neutral in terms of current income in period 1 to risk averse in income in period 2. The model may be interpreted as a two period model or as a model with consumption in two possible states (for example in sickness and in health). A worker of type q has (the same) reservation utility , and her expected utility is given by U = w + pqu(s)
(1)
since the utility from benefit is only obtained if the firm survives and the worker remains within the firm. (6) The firm’s expected profit is given by = p (R qs) N wN v(p), N ≤ M
(2)
where v(p) represents the costs of lost expected profit opportunities incurred by the firm to increase the survival probability p. We assume that
234
(7)
(8)
(9)
(10)
JAN ERIK ASKILDSEN AND NORMAN J. IRELAND
v(0) = 0, v( p) > 0 and v( p) > 0. In later analysis we will take v( p) as constant. The cost of ensuring survival may be the rejection of an investment project with a high risk of disaster but with high expected profits. Declining such projects decreases the firm’s expected profit but increases its chance of survival. Thus, there is a cost involved in increasing the firm’s survival probability. N is employment to be determined by the firm, or bargained with the workers. M is the capacity of the firm with respect to the labor force. Further workers are assumed unproductive. Thus, there is a simple production function which yields revenue linear in employment, RN, for N ≤ M. All workers are equally productive up to the given level of capacity. The benefit s represents the market value of real resources which the firm would otherwise retain. Therefore the firm’s net revenue is (R qs)N to be received with probability p. A generalized Nash bargain is applied to choose values of all or some of N, p, s and w. The bargain takes place at the start of period 1. No renegotiation is possible. Any remaining variables not included in the bargaining process are determined by the firm alone. The key variable likely to be omitted from the bargaining process is p. We first consider the case where workers are homogeneous. When homogeneous, all workers are of the same type q. Under heterogeneity, the workers are distributed according to the density function f (q) of worker types, still with a given maximum work force M. A further heterogeneity of workers is introduced later by assuming that they differ in terms of their trade off between wages and future benefits. Lastly, we investigate possible effects of applying seniority preferences in firing decisions.
3. BENCHMARK: THE PARETO OPTIMAL SOLUTION The Pareto optimal solution maximizes the total “pie”. Bargaining yields a particular distribution of the pie, which is analyzed in the next section. We first find Pareto optimal solutions in our model. Consider the problem of maximizing the social gain from the firm’s activities, W(p,s,N) = + N(U ). The wage transfer drops out since we have assumed U is linear in w, and so the gain is given by W( p, s, N ) = [ p(R qs) + pqu(s)]N v( p).
(3)
Union-Firm Bargaining over Long Term Benefits
235
Provided maximum profit is non-negative and U (for firm and worker to willingly participate in the activity), the Pareto optimum requires that {p, s, N} satisfy the following first-order conditions: [ p(R qs) + pqu(s)] 0 and N = M
(4N)
[(R qs) + qu(s)]N v(p) = 0
(4p)
pq(u(s) 1) = 0.
(4s)
We describe the solution to these first order conditions as N = M, s = s* and p = p*. (4N) ensures that the social gain is positive. The solution s* is the benefit level which equates the marginal utility of income in periods 1 and 2. The intuition of (4s) is that the expected marginal social cost of an additional unit of benefit is pq and the expected marginal social benefit is pq u(s) since the cost and the benefit transfer are only made if the worker-firm relationship survives. From (4p) the optimal p* has the property that the marginal cost of additional survival probability in terms of profits forgone (v(p)) is equal to the expected social gain from firm survival. Having obtained the biggest total “pie”, it is left to the wage w to allocate the pie between the workers and the firm.
4. BARGAINING WITH HOMOGENEOUS WORKERS We will investigate two different bargaining regimes, which will yield different outcomes. Efficient bargaining represents the optimal solution when all variables can be contracted. On the other hand, a right-to-manage solution is interpreted as a situation where the firm has the discretion to determine at least one variable after the others are settled. In particular for our analysis here, the firm may reserve the right to make decisions on the level of investments in the survival probability. Since these decisions are taken to maximize the firm’s outcome, rather than the total pie, they will not usually lead to a Pareto optimal solution. Efficient Bargaining All relevant variables and conditions of work are determined cooperatively through bargaining between the workers and the firm. Thus, the variables N, w, s and p are bargained simultaneously. The outcome can be represented as the solution to an asymmetric Nash cooperative bargaining programme, with as the exogenously given bargaining power of the firm and (1 ) representing the union’s bargaining power, given the alternative wage for all workers:3
236
JAN ERIK ASKILDSEN AND NORMAN J. IRELAND
= {[ p(R qs) w]N v( p)}*{[w + pqu(s)]N + (M N) M}1 , N ≤ M.
(5)
We take the logarithmic transform of (5): ln = ln{[p(R q s) w]N v(p)} + (1 ) ln{[w + pqu(s) ]N},
(6)
and maximize ln with respect to N, w, s and p. The first order condition with respect to w may be written as
ln / w = 0 ⇒ [w + pqu(s) ]N + (1 ) [(p(R q s) w)N v( p)] = 0,
(7)
or, solved for w, as w = (1 ) [ p(R qs) v( p)/N] pqu(s) +
(8)
or [( p(R qs) w)N v( p)] = {[ p(R qs)N v( p)] + ( pqu(s) )N} = W(p, s, N) [w + pqu(s) ]N = (1 ) {[ p(R qs)N v( p)] + ( pqu(s) )N} = (1 )W( p, s, N) so that gives the proportionate allocation of the surplus to the firm. Substitution into (6) yields ln = ln + (1 ) ln (1 ) + ln W( p, s, N).
(6)
Thus it is simple to prove Proposition 1: In an efficient bargain, the solution for p, s, N is Pareto optimal for any distribution of bargaining power. Only the wage rate w reflects relative bargaining power. Proof: For any given w, (6) shows that the values of p, s and N are chosen to maximize W(p, s, N), and these are the Pareto optimal values p*, s* and M given by (4p), (4s) and (4N ). The wage rate is given by: w = (1 )[p*(R qs*) v(p*)/M] p* qu(s*) + This completes the proof of Proposition 1.
(8)
Efficient bargaining yields a Pareto efficient choice of s* and p*, whatever the level of bargaining power. Having derived s* and p*, and N = M, the wage is the only element of the bargain that depends on . Note that with a pure capitalist
Union-Firm Bargaining over Long Term Benefits
237
firm, defined by = 1 to reflect the property that all bargaining power is held by the firm, the wage level is w* = p* qu(s*). (9) In this case, workers receive only their reservation utility . If workers have some bargaining power, 0 < < 1, they will be able to capture some rent, and this will not affect p* and s*, just w*. The added utility is seen by substituting the expression for w* into the utility function (1). By using the definition of profit, utility will be given by U = (1 )W/M + (10) where the workers’ per capita gain above their reservation utility depends on their bargaining power 1 . At the extreme, a labor-managed firm would choose p, s and N to maximize a worker’s utility when all surplus is divided equally among the workers. Then U = W/M + . Returning to the first-order conditions (4p) and (4s), we may derive the effects of changes in the workers’ probability of staying with the firm. We find that assumption (5) on the form of u(s) gives us the result that dp*/dq > 0 while it is clear from (4s) and (4N ) that ds*/dq = 0 and dN*/dq = 0. Using these in the expression for the wage, (8), we find that dw*/dq < 0. Thus, as the probability of the workers staying with the firm increases, the firm will increase effort to survive. The benefit level will not change but wages will be reduced. The reason is that lower turnover and less frequent changes of jobs (higher q) give more room for expected future job-related benefits to offset current wages. Job security will increase, and for a given bargaining power, which determines the workers’ share of the pie, wages must then be reduced. Similarly, the wage level may increase when mobility increases, and traditional job-attached benefits will to a larger extent be provided individually. Note also that the result is independent of worker control. Under efficient bargaining the firm’s supply of public goods (like pensions) to the workers, depends on the workers’ preferences only. The profit maximizing firm and the labor managed firm behave identically. Right-to-Manage: The Firm Determines p Unilaterally Above we have derived the efficient solution, where the union and the firm simultaneously set all variables to be determined. However, it is not clear how
238
JAN ERIK ASKILDSEN AND NORMAN J. IRELAND
the firm and the union can contract on p, the survival probability of the firm. Therefore, assume that p cannot be verified, which means in effect that the firm can determine p after the wage, benefit and employment levels are contracted. Alternatively, the firm cannot commit to a contracted value of p.4 Thus, at the first stage we solve for w, s and N, and at the second stage the firm chooses p to maximize expected profit. Hence, pˆ (s, N) solves v( pˆ ) = N(R qs)
(11)
which is independent of w. It is easily verified that we still have full capacity employment, N = M. Also w is chosen according to (9) above. The workers know that p is set independently of w at the next stage (see (11) above). On the other hand, from (11) above, we see that the choice of s does affect the optimal p chosen by the firm. Hence, the bargaining involves maximizing W( pˆ (s),s,N,) = [ pˆ (s)(R qs) + pˆ (s)qu(s)]N v( pˆ (s)).
(12)
The first order condition for s is simplified by using (11) to be u(s) dˆp/ds = pˆ (1 u(s)).
(13)
Differentiating (11) for N = M, we find that dˆp/ds = q[M/v(p)] < 0. Since dˆp/ds < 0 and u(s) > 0, we must have u(s) > 1 from (13). Remember that the efficient solution (4s) gives u(s*) = 1. Therefore, due to the concavity of u(s), sˆ < s*. The equilibrium values of the right to manage model are given by (11), (13), N = M, and wˆ = (1 )[ˆp(R qˆs) v( pˆ )/M] ( pˆ q u(ˆs) f ).
(14)
We can compare (11) where s = sˆ and N = M with (4p), where p* solves v(p*) = M(R qs* + qu(s*)). Assumption (5) ensures that the maximum value of q(u(s) s) is at s* and is positive. Hence v(p*) v( pˆ ) = Mq(u(s*) s*) + Mqˆs > 0.
(15)
ˆ ) is the solution in the right-to-manage case and Proposition 2: If (ˆp, sˆ, N ˆ = N* = M, and (i) (p*, s*, N*) that in the efficient bargaining case, then N sˆ < s*, and (ii) pˆ < p*. Proof: (i) follows from (13) and (4s). (ii) follows from (15) as well as the convexity of v( p).
Union-Firm Bargaining over Long Term Benefits
239
There will be a compensating adjustment of wages. Comparing the expressions for w ˆ and w*, i.e. (8) with (14), we see that wˆ > w* because pˆ q u(ˆs) < p* qu(s*) and pˆ (R qˆs) v( pˆ )/M > p*(R qˆs) v( p*)/M > p*(R qs*) v(p*)/M. The latter follows from, (i) sˆ < s*, such that the middle term is greater than the right hand term, and (ii) given sˆ, then pˆ maximizes p(R qˆs) v(p)/M. The right-to-manage solution implies both a smaller survival probability for the firm and a smaller investment in benefits. The firm will choose p to maximize its profit, not taking into account the effect p has on the workers’ utilities via s. The workers will find it more profitable to be paid now rather than in the future, since the deferred payment is received with less certainty when the firm cannot commit to its future behavior. Therefore the reduction in benefits will be substituted by an increased wage, given the distribution of bargaining power and thereby the rule for sharing the total pie to be constant. The firm’s incentives to commit to a particular future action will depend on what happens to its profits. If profit is higher under the efficient bargaining solution, and commitment is costless, the firm might look for mechanisms to truthfully commit to its choice of p. However, profit may increase or decrease when going from an efficient solution to a right-to-manage solution. Compare the expressions ˆ = [ pˆ (R qˆs)M v( pˆ ) + Mˆpq u(ˆs) M] and * = [ p*(R qs*)M v( p*) + Mp* qu(s*) M] where we have substituted in the expressions for wages wˆ and w* respectively. The first two terms are higher in ˆ than in *, whereas the penultimate terms are higher in *. If ˆ > * the firm may not agree to contract over p, even when this is found to be possible. The comparative statics of the right-to-manage case are addressed via (11) and (13). On the assumption that v(p) is constant, ds/dq is found to be negative. Thus, if q is very small, dˆp/ds is very small (from (11) and sˆ s* (from (13)). As q increases, so sˆ declines as the firm finds it more expensive to promise future benefits. The direct effect on p from increased q (via (11)) is negative, but the indirect effect (via a smaller sˆ) is positive. Thus, the effect of more stability on the part of the workers is lower benefits and an ambiguous change in the firm’s survival rate.
240
JAN ERIK ASKILDSEN AND NORMAN J. IRELAND
The key difference between the efficient bargaining case and the right-tomanage case is the fact that, while in the former the benefit level s is at the Pareto optimal level, a lower level of benefits is set if the firm cannot commit to a survival probability.
5. HETEROGENEOUS WORKERS The workers are characterized by their types q with a known density function f(q) with mean qa and median qm. Workers of all types will receive the same wage w, and also the same benefit s. Now the internal power structure of the union will matter. Assume that the median voter carries the balance of power so that her utility drives the union’s bargaining. If f(q) is symmetric, the median voter will be the average worker, and the same results will be derived as in the previous section. Alternatively the distribution may be skewed. The mean, qa, is the statistic that figures in the firm’s expected profit function. Thus, the objective functions are slightly altered to become: Um = w + pqmu(s)
(16)
= p(R q s)N wN v( p), N ≤ M a
(17)
ln = ln{[ p(R q s) w]N v( p)} + (1 ) ln[w + pq u(s) ] a
m
+ (1 ) ln N, N ≤ M.
(18)
Again
(ln )/ N > 0, which implies N = M and
(ln )/ w = M/{[ p(R qas) w]M v( p)} + (1 )/{w + pqmu(s) } = 0 so that the wage is given by w = (1 )[ p(R qas) v( p)/M] [p qmu(s) ].
(19)
The social value added by the firm is only affected by the mean type not the median: W(p, s, N) = [ p(R qas) + pqau(s)]N v( p)
(20)
so that substitution of (19) into (18) yields, using (20): ln = ln + (1 ) ln(1 ) + ln[W( p, s, M) Mu(s)p(qa qm)]. (21) Maximizing (21) is equivalent to maximizing the term in square brackets. Note that any difference between the union’s representative preferences (here the
Union-Firm Bargaining over Long Term Benefits
241
median individual) and the mean worker type produces an imperfection in the way the bargaining outcome seeks the largest pie. If we have efficient bargaining so that p and s are both agreed within the bargaining process, then Pareto efficiency is still not attained because of the agency problem within the union. First-order conditions yield qa + qmu(s) = 0 ⇒ u(s) = qa/qm
(22s)
and R qas v(p)/M + qmu(s) = 0 or v( p) = M[R + qm(u(s) su(s))].
(22p)
Let k = (qa qm)/qm be a measure of skewness. Then, letting s*, p* denote the optimal values under heterogeneity, first note that u(s*) = u(s*)(1 + k). The effects of heterogeneous workers on p* relate to the different s in the two cases. If k ≠ 0, the survival probability will also differ. To investigate this further, hold qm fixed and change k from 0. From (22s) and (22p) respectively: ds*/dk < 0
dp*/dk < 0
Thus, s* and p* move together and opposite to the skewness measure, k. Hence s* s*
and
p* p* as
k 0.
Instead, hold k fixed and change qm, which implies that qa changes in the same proportion. We see from (22s) that ds*/dqm = 0, while from (22p) dp*/ dqm > 0. In the right-to-manage model, from (17), (20) and (21), the firm’s optimal choice of p (to maximize (17)) and the bargaining outcome for s (to maximize (21) given the subsequent choice of p), lead to revised first-order conditions (22s) and (22p): p[qa/qm u(s)] = u(s) dˆp/ds
(22s)
v( pˆ ) = N(R q s)
(22p)
a
The strategic effect dˆp/ds in (22s) is found by differentiating (22p) and is clearly negative as before. Substituting for N = M in equilibrium, the solution is denoted sˆ, pˆ . Then we can prove: Proposition 3: In a right-to-manage model, the equilibrium has smaller s and p than in the efficient bargaining model, for any level of skewness in workers’ survival probabilities.
242
JAN ERIK ASKILDSEN AND NORMAN J. IRELAND
Proof: (i) s < s* holds by comparing (22s) and (22s), and (ii) pˆ < p* follows from subtracting v( p*) in (22p) from v( pˆ ) in (22p), using the respective equilibrium values of p and s. This result is explained as before in terms of the bargaining structure. It is also interesting to consider the internal power structure within the union, and compare the outcome to first best. There are now two sources of inefficiencies. Without credible firm commitment, the union will, as in the homogeneous workers case, unambiguously bargain smaller benefits than would be obtained if the firm could commit to a firm survival probability. The skewed power structure may make things worse or better. It will necessarily be the case that benefits are lower if power distribution is egalitarian. However, if the median worker, who carries the balance of power, has a higher survival probability within the firm than the average worker, the right-to-manage solution may be closer to first best than efficient bargaining. The reason is of course that this median worker has an incentive to vote for too high a level of benefits.
6. DISCRIMINATION BY LAST-IN-FIRST-OUT It is frequently asserted that LIFO (last-in-first-out) is a convention subscribed to by both firms and unions. We will show that a minor adjustment of our model transforms it into one where the number of employees kept on by the firm is random and where the probability of a worker’s survival to take up longterm benefits depends on the order of her hiring. Thus suppose that the firm’s profit is = xN(R qs) wN v(p)
(23)
where x is a random variable reflecting demand and is revealed after workers have been hired and paid the wage w but before production is complete and the workers receive the benefits s. The firm employs N workers (in equilibrium this will be assumed equal to the capacity M) but retains only xN to complete production. The expected value of x is p so that the expected profit of the firm is as before. We will take q as the same for all workers. A worker of index i is the worker who had a proportion i workers hired before her and 1 i after her. The expected utility of the i-indexed worker is given by Ui = w + qIu(s)
(24)
where I is the probability that x > i. If the union reflects the preferences of the median worker (i = 1/2), then let the probability that x > 1/2 be I m, and note that this will in general depend on p (as well as on other parameters of the
Union-Firm Bargaining over Long Term Benefits
243
x-distribution, but these we will ignore for simplicity). Hence I m = I m(p), and the efficient solution to the firm-union bargaining game is to choose w, p, and s to maximize ln = ln{( p(R qs) w)N v( p)} + (1 )ln{w + q I m( p) u(s) }. (25) First-order conditions can be simplified by using the capacity level of employment (N = M) and the optimality condition for w in those for p and s respectively to obtain (R qs)M v( p) + q I m( p) u(s)M = 0
(26p)
pqM + MqI m( p)u(s) = 0
(26s)
and
or u(s) = p/I m( p). In the right-to-manage model the bargaining is over s while the firm chooses p, given s, to maximize ( p(R qs) w)M v( p). The decision over s thus affects the subsequent decision over p with the slope dp/ds = qM/v( p). The first order condition for s is thus
ln / s = pqN + qNI m( p)u(s) + {–qN/v( p)}{(R qs) N v( p) + qu(s)N I m( p)} = 0.
(27s)
Now the choice of p requires (R qs) N v( p) = 0
(27p)
and then (27s) can be solved for p + I m( p)u(s) + {–qN/v( p)}u(s)I m( p) = 0.
(27s)
To compare the equilibria in the two models implies comparing p* and s* which solve (26p) and (26s) with pˆ and sˆ which solve (27p) and (27s). This comparison is not always unambiguous but a local result can be obtained for when Im( p) is small. We can prove the following: Proposition 4: Let Im( p) = C + zJ( p) with z → 0. The efficient bargaining ˆ solution ( p*, s*, N*) yields (i) s* = sˆ and (ii) p* > pˆ with (ˆp, sˆ, N) representing the right-to-manage solution. Proof: Comparative statics of the systems around z = 0 yield qN ds* v( p) dp* = qNJ( p)u(s) dz
(28p)
244
JAN ERIK ASKILDSEN AND NORMAN J. IRELAND
I m( p)u(s) ds* dp* = J( p)u(s) dz
(28s)
from (26p) and (26s) for the efficient bargaining model, and qNdˆs v( p)dˆp = 0
(29p)
I ( p)u(s)dˆs dˆp = [ J( p)u(s) + {qNJ( p)u(s)/v( p)}] dz
(29s)
m
from (27p) and (27s) for the right-to-manage model. As the left-hand-sides of these systems are the same when z = 0, we can subtract (29p,s) from (28p,s) and then use Cramer’s rule to show (i) d(s* sˆ)/dz = 0 (ii) d( p* pˆ )/dz = qNJ( p)u(s) > 0.
Thus if dz > 0 and J( p) > 0 we expect the right-to-manage model to yield a smaller chance of firm survival, while the first-order effects on long-term benefits are negligible. It is harder to make more general comparisons. The reason is the difficulty in assessing how the median worker forms her expectations about firm survival. The signs and magnitude of deviations between the two regimes will depend on how changes in p are perceived, technically represented by the differentials of the I m( p)-function. Also, the I m( p)-function may be evaluated differently in the two bargaining regimes. Lastly, the relationship between the random variable x representing demand conditions and the median worker’s perception of uncertainty need not be monotonic. The LIFO principle is important in demonstrating that it is the median worker’s perception of uncertainty that matters. Thus, what actually matters is the probability of this worker losing her job, or rather the median worker’s consequential utility loss. If the median worker is either very senior, or has good market prospects outside the firm, the worker is likely to be affected by changes in p only to a small degree, as illustrated above for z = 0. With the strategic variable p having no significance to the worker, it will not be possible for the firm to take advantage of its discretion even in the non-commitment case. The outcome is efficient. In contrast, as long as the median worker is concerned with the firm’s redundancy decisions, inefficiency arises in the rightto-manage solution. The deviation from first best depends not on LIFO as such but on how the median worker is affected by the resulting employment uncertainty.
7. CONCLUSION Although we have found that the firm’s inability to commit to a strategy for firm survival leads to lower benefits and a lower survival probability (almost
Union-Firm Bargaining over Long Term Benefits
245
certainly), we have shown that the (s, p) equilibrium is in each case independent of the allocation of power in the bargaining process, . We note that the parties’ bargaining powers and (1 ) are commonly seen as functions of the parties’ relative impatience as measured by the rates for discounting future payoffs, see e.g. Moene, Wallerstein and Hoel (1993). The bargaining power of the union does, however, affect the wage outcome, since the wage transfers utility from firm to workers at once and with certainty. All bargaining power is reflected in what happens in the short term wage bargaining. The technical explanation for this pure result is the worker’s assumed linearity of utility in first-period wage. Some flavor of the result would still hold provided the workers were more risk averse in period 2 than in period 1. Arguments for this assumption relate to a number of sources of evidence and have been considered in Section 1. The result calls into question a number of apparently reasonable hypotheses. Among these would be the argument that long-term benefits are better argued when the union is in a strong position. This is not the case: a weak union may agree the same level of future benefits as does a very powerful union. What matters for the level of future benefits is the credibility of firm commitment. Without the ability to contract on firm survival, the union’s powers to bargain for future benefits are small. The firm controls the main variable of importance for such benefits to be delivered: the firm’s probability of staying in business. It would be reasonable to believe also that workers who are loyal in the sense that they are likely to stay with the firm, would be remunerated by higher benefits. This need not be the case either. With efficient bargaining the benefit level, s, is independent of this probability, q, and in the right-to-manage scenario benefits are decreasing in q. The firm will take advantage of the possibility that they do not need to pay the benefit. It is less costly to give promises of future benefits when workers are likely to leave the firm. For example, an option to buy shares in the firm some time in the future, provided workers stay with the firm, is less reluctantly given to a mobile workforce. It is of course an assumption that the union ranks outcomes according to the median voter, while the firm uses the mean worker in determining the expected profit. If both union and firm decide on the same basis, then skewness, and worker heterogeneity in survival, has no effect on the analysis: we remain with the analysis of Section 4. Alternatively, the union may give most weight, not to the median worker, but rather to the most relevant union member among the workforce. Two possibilities arise here: One is that it is the median union member who counts but this is not the median worker, see also Freeman (1985). This could arise if non-union workers had a different perspective. For example if these were “temporary” workers, in that their q-values were low,
246
JAN ERIK ASKILDSEN AND NORMAN J. IRELAND
then qa < qm*, where qm* is the union relevant statistic, and k < 0 due to the negative skewness. Then we know that the bargaining outcome will be for a higher benefit level and a higher firm survival probability whether the bargaining is efficient or right-to-manage. Essentially, the expected costs of providing benefits should the firm survive, has declined due to the likelihood of non-union workers being unable to collect on their benefit entitlement. A second possibility is that the union does not decide its preferences by majority voting, but rather a representative is selected whose position may differ from the median worker’s. For example, suppose a union hierarchy exists where the leadership takes the position that current wages are a measure of its “strength.” This position is tantamount to qm* being small. Then k > 0 since the distribution is effectively positively-skewed, and low p and s result. This second case emphasizes the fact that although we have been discussing k as a measure of skewness, it is really a measure of the difference between the union’s and the firm’s attitude to the given worker heterogeneity. The union in the hands of an unrepresentative clique can act as if the distribution is skewed, even if it is not. In this vein, a number of scenarios can be investigated within the model. For example, suppose the clique was one with high q-values, investment in control of the union being more important to those with greater likelihood of remaining with the firm. Then our model predicts higher s as benefits are more valued by the union bargainers, but also a lower probability of firm survival.
NOTES 1. In some instances non-verifiable variables can be implicitly bargained, see Malcomson and MacLeod (1993) and Malcomson (1997). However, it is also recognized that some hold-up problems cannot be easily overcome. 2. This line of reasoning follows the convention of modeling bargaining outcomes as the solution to e.g. a Nash cooperative bargaining problem. However, equally important for bargaining outcomes may be what can be bargained, who sets the agenda of bargaining etc. Ours is an example that such other matters are important for the outcome. Note that the resultant wages will still reflect bargaining power in the traditional sense. 3. Alternatively, a sequential bargaining model might be used, cf. Manning (1987). With the same bargaining power on all stages, the same efficient solution will emerge. The bargaining power of the workers, , would in a sequential bargaining model be interpreted as representing their degree of impatience, or time preference. 4. For example, an assurance to the unions that their members should accept a low wage and a bright future (high p) may be time inconsistent. Once the low wage is secured, a high-risk strategy would be optimal for the firm.
Union-Firm Bargaining over Long Term Benefits
247
ACKNOWLEDGMENTS We are grateful for comments from seminar participants at University of Bergen and Norwegian School of Economics and Business Administration, and from participants at the 11th Conference of the International Association for the Economics of Participation. The usual disclaimer applies.
REFERENCES Askildsen, J. E., & Ireland, N. J. (1993). Human Capital, Property Rights and Labour Managed Firms. Oxford Economic Papers, 45, 229–242. Bakshi, G. S., & Chen, Z. (1994). Baby Boom, Population Aging, and Capital Markets. Journal of Business, 67, 165–202. Bloom, D. E., & Freeman, R. B. (1992). The Fall of Private Pension Coverage in the U.S. American Economic Review, 82, 539–545. Botwinick, J. (1978). Aging and Behavior: A Comprehensive Integration of Research Findings (2nd ed.). New York: Springer. Budden, R. (2000). Pensions Adrift as Ship Sinks. Financial Times, 27.9.2000. Curme, M., & Kahn, L. M. (1990). The Impact of the Threat of Bankruptcy on the Structure of Compensation. Journal of Labor Economics, 8, 419–447. Flanagan, R. J. (1993). Can Political Models Predict Union Behaviour? In: R. J. Flanagan, K. O. Moene & M. Wallerstein (Eds), Trade Union Behaviour, Pay Bargaining and Economic Performance, FIEF Studies in Labour Markets and Economic Policy (pp. 6–45). Oxford: Clarendon Press. Freeman, R. B. (1985). Unions, Pensions, and Union Pension Funds. In: D. A. Wise (Ed.), Pensions,Labor, and Individual Choice (pp. 89–118), National Bureau of Economic Research Project Report Series. Chicago and London: University of Chicago Press. Grout, P. A. (1984). Investment and Wages in the Absence of Binding Contracts: A Nash Bargaining Approach. Econometrica, 52, 449–460. Hoel, M., & Moene, K. O. (1988). Profit Sharing, Unions and Investments. Scandinavian Journal of Economics, 90, 493–505. Ireland, N. J, & Law, P. J. (1982). The Economics of Labor-Managed Enterprises. St. Martin’s Press. Malcomson, J. M. (1997). Contracts, Hold-Up, and Labor Markets. Journal of Economic Literature, 35, 1916–1957. Malcomson, J. M., & MacLeod, W. B. (1993). Investments, Hold-Up and the Form of Market Contracts. American Economic Review, 83, 811–837. Manning, A. (1987). An Integration of Trade Union Models in a Sequential Bargaining Framework. Economic Journal, 97, 121–139. McDonald, I. M., & Solow, R. M. (1981). Wage Bargaining and Employment. American Economic Review, 71, 896–908. Moene, K. O., Wallerstein, M., & Hoel, M. (1993). Bargaining Structure and Economic Performance. In: R. J. Flanagan, K. O. Moene & M. Wallerstein (Eds), Trade Union Behaviour, Pay Bargaining and Economic Performance, FIEF Studies in Labour Markets and Economic Policy (pp. 63–131). Oxford: Clarendon Press.
248
JAN ERIK ASKILDSEN AND NORMAN J. IRELAND
Orr, D. V. (1996). The Rise of Contingent Employment and the Decline of Pension Coverage. Review of Radical Political Economics, 28, 126–134. Orr, D. V. (1998). Strategic Bankruptcy and Private Pension Default. Journal of Economic Issues, 32, 669–687. Oswald, A. J. (1985). The Economic Theory of Trade Unions: An Introductory Survey. Scandinavian Journal of Economics, 87, 160–193. Targett, S., & Beattie, A. (2000). Pension Reform Proves a Difficult Balancing Act for Westminster. Financial Times, 16.8.2000.
COMPARATIVE SYSTEMS, DESTRUCTIVE TRADE AND WORLD DISTRIBUTIVE JUSTICE Jaroslav Vanek ABSTRACT Using empirical evidence and theoretical analysis we show the fallacy of globalization through free international trade. A new and more comprehensive destructive trade analysis indicates that free international trade per se is likely to worsen world economic conditions and especially those of the poor majority of mankind. The dramatic gap estimated by UNDP between the rich and the poor of the world, and its continuing worsening over the past forty years, are both the cause and the effect of destructive trade. Thus it appears that free-trade globalization involves a state of instability and explosiveness.
1. INTRODUCTION Perhaps the best way to introduce the subject of this paper is to recall the November 1999 riot and demonstrations in Seattle, Washington trying – and actually succeeding – to obstruct the work of the World Trade Organization meetings there. The fact that the WTO on the one hand and a large number of other organizations, most of them equally serious, can be at odds to such an extreme degree, while all claiming the truth of their respective positions, indicates that
The Determinants of the Incidence and the Effects of Participatory Organizations, Volume 7, pages 249–266. © 2003 Published by Elsevier Science Ltd. ISBN: 0-7623-1000-6
249
250
JAROSLAV VANEK
there must be some fundamental misunderstanding or lack of full understanding of the issues and forces involved. The objective of this paper is to clarify these issues and ultimately to indicate who – and why and how – appears to be closer to the truth of the matter. The disagreement between the globalization advocates and opponents can be observed not only in the streets of Seattle but also on the more academic plane, in the economic literature; on the positive side, in writings such as Krugman’s Pop Internationalism or the Brookings’s Globaphobia and on the negative or critical side in contributions of Daly and Cobb.1 While it is impossible to spell out all the positive or negative distinguishing contributions of this study with respect to these earlier writings, one of them deserves to be pointed out. At its foundation, the free-trade argument can be traced to the Ricardian theory of comparative advantage and its later evolutions into the nonlinear Hecksher-Ohlin theory, sometimes now referred to in a further generalization as the Hecksher-Ohlin-Vanek theory2 as well as the increasing-returns theory. The present paper identifies the theory of destructive trade as a new theoretical interpretation which is quite distinct from the earlier theories of international trade.
2. THE RELEVANT WORLD PARAMETERS The state of the world and of the world economy, the booming and volatile stock markets notwithstanding, at the beginning of the century is not a happy one and can be described by two powerful symbols; a cross and a chalice. They are shown in Figs 1 and 2 respectively. Both are based on United Nations statistics and the stories they tell are related.
Fig. 1.
Rich and Poor Countries: Rich and Poor People (orders of magnitude in billions).
Comparative Systems, Destructive Trade and World Distributive Justice
251
Both figures are intended primarily to represent orders of magnitude and pertain to long-run states, dealing in decades rather than in years of performance. In Fig. 1 the emphasis is on the division of world countries between the rich countries (RC) and the poor countries (PC) and the less emphasized division between rich people (RP) and poor people (PP). The numbers in the four entries of the matrix correspond approximately to the numbers of present world population, approximately 6 billion. The exact dividing line between rich and poor is not important, but its order of magnitude can be easily visualized from Fig. 2, where the richest 20% of the world population receives an overall
Fig. 2. Source: United Nations Development Program, 1992.
252
JAROSLAV VANEK
income (darkened area of the chalice) much larger than the income of the remaining 80% of world population. More exactly, the poorest 20% received an income of 301 dollars per capita in 1988 against the per capita income of the richest 20% of $19,542. And this is when reckoning the situation based on richest and poorest countries (the horizontal distinction in Fig. 1). If actual rich and poor people are considered, irrespective whether they are in rich or poor countries, the corresponding numbers for the richest and poorest 20% are approximately $23,000 and $160 respectively, or a ratio of some 140 to one, or the richest living on incomes some 14,000% higher than those of their poor brethren. For the purposes of our analysis, what is important is the ratio of poor people to rich people rather than that of poor countries to rich countries. As we see from Fig. 1, the real human drama of today’s world lies along the horizontal, or north-south division. Those on one side of the line are much more closely related to each other than to those in their countries on the other side of the line. What is more significant, in the long haul things are not getting any better. In 1960, or more than 30 years ago, the income of the richest 20% per capita was thirty times that of the poorest 20%. Thirty years later at the beginning of the past decade the ratio was twice as pronounced, or some sixty times higher. This tells us a lot, analytically, about the nature of the situation we will see throughout our analysis. In terms of Fig. 2, it is as if the stem of the chalice were remaining approximately the same while the “wealth-containing” top grew wider and fatter over time. At the time of this writing, in 2000, most recent figures are coming in for the decade following the date of our chalice, from the same sources, indicating that the dramatic trend continued, from 60 times to 74 times. Also very significant are the side-conditions – whether causes or effects of the situation. A good deal of the almost perverse differentials depends on the use of energy by the rich. It takes several barrels of the purest form of energy, jet fuel, to visit Europe from the United States. And the member of a rich country – and much more pronouncedly a rich person of that country – does or will consume in his or her lifetime close to 1000 tons of energy equivalent measured in such high quality fuel, or some 17 thousand times his or her body weight. The truly fantastic distribution of responsibility for world pollution from such extreme (mostly fossil fuel) energy consumption differentials needs no further description. Let it only be noted that in their cynicism, those who require the extreme energy levels are having a good deal of their production realized in the poor parts of the world.
Comparative Systems, Destructive Trade and World Distributive Justice
253
Further, the typical working person living in the rich part of the world has his living standard supported by some one to two hundred thousand dollars of productive capital assets. Again, this magnitude can usefully be illustrated by what I refer to as the millenial syndrome. If all of the annual net capital formation of the richest country of the world, the United States (some 600 billion dollars worth in 1999), were allocated to create productive capacity, even with zero population growth, it would take the poor subsistence workers behind Fig. 1 or Fig. 2 some one thousand years to attain the same degree of capitalization as that supporting the jobs of the rich. This is so because 600 billion divided by 200 thousand equals three million. And with some three billion jobs needed to support the 4.5 billion poor in Fig. 1, this amounts to one thousand times the yearly net capital formation noted above. If we realize that demographic projections predict the doubling of world population in a matter of decades the syndrome appears even more serious. Perhaps it ought to be referred to the ten-thousand year syndrome. The UNDP Human Development Report of 1998 projects doubling of world population in 50 years, and what is worse from our point of view, the poor countries should increase their proportion from 78 to 87% of the total. The extreme capital requirements of the rich are matched by equally staggering requirements of human capital – that is, education determined by the industrial and technical needs of the advanced production processes, and not education which would make people more humane. The aggregate costs of students reaching graduate school in the United States are hard to estimate exactly, but they reach into several hundred thousands of dollars, all considered. What is more alarming, once the talented brains are stuffed to this advanced level, their recipients often prefer to stay in the country or countries where they received their stuffing, thus draining some of the best human resources from their often poorer countries of origin.
3. FROM RICARDO TO DESTRUCTIVE TRADE The widely accepted argument for free trade – the cornerstone of the WTO, NAFTA and other policies for world trade liberalization – is the classical and notorious Ricardian trade theory of comparative advantage and gains from trade. It is often associated with the concrete example of trade in wine and cloth and its essentials are well known to students of economics, so that it is unnecessary to restate the theory in any detail. Let it only be recalled that on the all-important assumption of full employment of all resources in the world, profit-maximizing traders can conduct a price arbitrage, exporting relatively less expensive products and importing relatively more expensive ones, as
254
JAROSLAV VANEK
measured by the technical transformation ratios given by technologies and/or factor endowments in different countries. The important conclusion of this theory is that at least some countries in the world must benefit and no countries lose from such transactions. Thence the world as a whole will gain. And such beneficial changes also constitute the theory explaining international trade transactions. While not incorrect on its assumptions, this theory of comparative advantage offers a rather poor description of what is happening in the world around the turn of the 20-to-21st century. Another theory – hereafter to be referred as the DESTRUCTIVE TRADE theory – describes much better the actual case of world trade and explains much better the key characteristics of world trade as well as of world production and income distribution such as those described in section two above. Even at this introductory level, however, it is possible to indicate a fatal blow to the comparative-advantage theory applied to world trade of today – and thus also to the free-trade case for globalization. It is a well known theorem of the theory that the smaller economy becomes a price taker in the world exchange of goods and services, adapting to the relative prices of the larger dominant economy and thus reaping all or most of the benefits. Obviously, the dominant larger economy in the world – in spite of its lesser population – is the advanced industrialized world, represented in Fig. 2 by the fat upper part of the chalice; whereas the skinny stem is the poor part of the world which should have, according to the theory, gained over time. The UNDP reality is just the opposite, and continues to be so!
4. DESTRUCTIVE TRADE Let us first offer the essential distinguishing characteristics of destructive trade and destructive trade theory, and then elaborate on some of the key detail of the analysis. The drama and tragedy of humanity described in Section 2 is both the cause and effect of destructive trade. First of all, the old comparative advantage of full employment cannot be sustained. In the affluent countries, full employment need not hold as a result of destructive trade, and the poor countries – the 80% of humanity in the UNDP chalice – can hardly be thought of as fully employed, unless the millions gathering food in garbage dumps at the equivalent of 20 cents per day are to be thought of as fully employed. It is the differential between rich and poor of the order of one hundred to one that is both the cause and the effect of international arbitrage. The arbitrageurs are not the traditional traders, but rather the ever increasingly powerful
Comparative Systems, Destructive Trade and World Distributive Justice
255
international and multinational corporations. Not everyone can conduct the arbitrage – only those who have the bigness, the capital and the technologies necessary for it. The technology and its capital control become the tickets or tokens for the game played by the powerful. At the turn of the century we were afraid of the Y2K problem threatening to return us back to 1900. While this did not happen with computers, it is happening unwittingly with the world (destructive) trade, exploitation of the poor majority of the world by the rich minority owning most of world capital – whether as countries or as individuals. The theses of Karl Marx are reborn in a new disguise: the only new thing is that the loot of exploitation must be transferred somehow from the poor to the rich, from the losers to the winners, and the vehicle of such transfers is the destructive trade. We in the rich countries export food for the slaves and the technology and machinery to harvest the fruits of ten-cent-an-hour labor and import all that labor can produce. The profits of the arbitrage are enormous, and underlie the staggering explosions of the stock market values. But secondarily the profits also go to appease the workers in the rich countries, who have lost their wellpaying manufacturing jobs and had to move into inferior service jobs, but who can now also buy the low-cost gadgets produced by the sub-standard tencent jobs in China, the Caribbean, Manilla or Calcutta. All this tendency in turn supports and continuously worsens, as confirmed by the data of Section 2, the chalice of Fig. 2 above.
5. FORMAL INTERPRETATION But let us now turn to a more careful analysis of destructive trade. The classical theory of trade based on comparative advantage can be explained in several ways. We will choose the explanation through an act of arbitrage of a hypothetical arbitrageur, because the theory of destructive trade can also be outlined using such an approach. Suppose that there is an arbitrageur, in a world without money who observes real exchange ratios – say, between food and gadgets – that are different in two countries, as described in the familiar Fig. 3. For example, such ratios might be one to one in one country and one to two in another country. He can then, starting with an endowment of one ton of food in the first country, exchange that for two tons of gadgets in that country, then export the two tons of gadgets to the other country and obtain through barter exchange two tons of food, which he can import into the first country and thus realize a gain (profit) of one ton of food in the first country. We may refer to this set of real transactions as the basic act of arbitrage, or BAA hereafter.
256
JAROSLAV VANEK
Fig. 3.
It is usually noted that all these transactions are consistent with full employment, production and resources being reallocated in each country according to the BAA. Moreover, the first country gains from trading in this way at each BAA, and the BAA’s will continue so long as the price exchange differential (1/1 and 1/2) will be maintained. The latter will hold up to the point where resources and output are so reallocated that at least one country specializes in the production of a single output. If both specialize, both will gain from trade: if only one specializes it is the one which will gain. All this is “old hat” to anyone who took basic courses in economics. The destructive trade which we observe in today’s world can also be explained using a BAA, but of a different type. To facilitate our exposition, let us turn to Figs 3, 4 and 5. In Fig. 3 we recognize the diagram corresponding to the classical comparative advantage just discussed, with the pre-trade situations being described by the points c and C on the corresponding production possibility loci of the two trading partners. The post-trade situations are illustrated by points c and C for consumption; and points p and P for production. The world gains from trade are represented by the vector cc, and as we know from a well known theorem, the smaller price-taking country realizes all these gains, while the larger country, the price-giver, gains nothing. There will be a similar interesting theorem for destructive trade below. In Fig. 4 we find a somewhat more complex world – the world of destructive trade – which involves also two countries, but in addition to food and gadgets it also recognizes the all-important factor-commodity of human labor. To define clearly destructive trade and its distinguishing characteristics, it is desirable first to recall the definition of traditional Ricardian comparative
Comparative Systems, Destructive Trade and World Distributive Justice
257
advantage, as trade occurring whenever there is a difference between the transformation (production-possibility) loci of the trading partners; or. viceversa, as trade not occurring (under the ideal conditions usually assumed) if and only if the two transformation loci are parallel, or indicate identical relative prices in the two countries. These two formulations or definitions of classical comparative advantage do not apply for destructive trade and thus imply a different, significant definition. As we will see presently, the necessary condition for destructive trade is a significant differential of real wages in the two trading partners. The slope of the transformation functions is essentially irrelevant. Turning to Fig. 4, underlying the case of destructive trade, we also have two significantly different exchange ratios, also stated in real terms without money, but they reflect differences in real wages between the two countries – or now the two different parts of the world of the rich and the poor. In terms of food, these ratios are indicated by the slopes of (a hypothetical) line fl, and the line (actually drawn) fl respectively. It should be noted that if real world wage rates (of the UNDP chalice) were to be reflected in Fig. 4, the point 1 would find itself some four meters, or twelve feet above L. This may serve to illustrate and underline the “power of arbitrage” involved in destructive trade.
Fig. 4.
258
JAROSLAV VANEK
Fig. 5.
Destructive Trade: Basic Act of Arbitrage [BAA]. BAA = a0 + 0b + bc + cd + da [or da or da]
To isolate the analysis of destructive trade from traditional comparative advantage, let us eliminate entirely any possibility of the traditional alternative, and as we show in Fig. 4, make the transformation functions for the two countries parallel. Conforming to real conditions (see Figs 1 and 2) the highwage country is the larger one in terms of GNP, but smaller in terms of population. The profitable arbitrage can now again be represented by a basic act BAA of an arbitrageur as indicated at the top of the diagram of Fig. 5, in the form of a sum of vectors/segments representing the arbitrageur’s transactions; typically, those of a multinational corporation or an agent of the WTO armed with all its instruments of free trade, intellectual and other property rights, and so forth. Suppose that the arbitrageur has or produces every year the amount Oa of gadgets (G) using Od of labor in the rich country and for simplicity makes zero profit. He can now stop production, that is move from a to 0 (a0), thereby saving a wage-bill equivalent to Od. He can then move from 0 to b using just Ob of labor and immediately transform Ob into Oc of food, the line cb representing the labor productivity of food in the rich country. He now exports the machinery which he used to produce gadgets in the rich country and also exports the amount Oc of food to the poor part of the world. With the exported food Oc he now hires Od of labor in the developing poor country, and using the
Comparative Systems, Destructive Trade and World Distributive Justice
259
exported machinery produces – assuming that the productivity of local labor and the exported machinery is the same abroad as in the rich country (an assumption to be relaxed presently) – the old amount Oa of gadgets which he imports to the rich country. The arbitrageur ends up again with the same number of gadgets Oa. BUT! he now makes a profit equivalent to ba of gadgets, matched by unemployment in the rich country equivalent to bd. Realistically, here the story does not end. (1) The violent nature of the situation is indicated by the fact that in the real world the real wage differentials can be as high as 10,000% (in the diagram the segment od would be, as we have seen already, dramatically longer than actually shown). (2) The BAAs continue as in the classic case as long as wage differentials are present. In the real world they have been increasing in the past half century; that is, in theory anything that is transportable should not be produced in the high-wage part of the world, and all corresponding employment should be lost in the long run. While this situation is not going to happen in any near future, it indicates a state of violent lasting disequilibrium. Such disequilibrium is confirmed by the dramatic shifting of domestic employment from transportable manufacturing to non transportable services. (3) Inefficiencies of resource allocation of many kinds can occur as long as the capital’s and WTO profit motive is satisfied. Worldwide declines or stagnation in productivity confirm such trends. Above all, moving from manufacturing to services is equivalent to moving from high to low productivity growth sectors of the economy. (4) Many losses and real production destruction can occur, with the end point of the BAA moving from a to a. The condition of No. 3 thus is that a be to the right of point b. Dozens of such losses are present through labor inefficiency in the poor parts of the world, corruption, transportation and ecological costs, etc. Most important perhaps is the technological inappropriateness of producing gadgets for the rich in very poor but solarenergy endowed countries. (5) With losses of employment and especially high-wage employment, demand thus can also be reduced to points such as a at the end of the aggregate national BAA. And this in turn can generate (Keynesian) inflationary or deflationary gaps. (6) With fantastic redistribution of income from labor to capital, countries with lots of labor and little capital tend to be losers as a whole; and vice versa for the capital rich countries which tend to gain in the aggregate.
260
JAROSLAV VANEK
(7) Nothing is a better proof of this argument than the crazy and overinflated stock markets, especially in the USA, and the ever-increasing indebtedness of the poor countries which try to keep up by borrowing from the rich, but are not able to repay. This is indeed a realistic theory of trade, with imports of gadgets, matched by export of food and machinery. There is no maladjustment in the balance of payments: everything is done in real terms, and the imports of gadgets are paid for by the exports of food and machinery and the horrendous, exploitative foreign earnings of capital. Just one substantiation will suffice. Quoting the New York Times editorial page, a fancy coat sold on Fifth Avenue for $175. is produced in a maquiladora with about 70 cents worth of labor!3 A de facto coverup of the situation in the rich countries consists in two phenomena. First, the working classes have to substitute for good wages of those who lost jobs (mostly family breadwinners) near-minimum wages of mothers or younger family members in service sectors (mostly selling the foreign gadgets or cooking hamburgers to substitute for food which the mothers now are not in a position to cook). In terms of employment statistics this may even increase employment. Second, the low-cost gadgets from China and elsewhere keep the real wages at tolerable – even if declining – levels for enough people to prevent major political upheavals. But the Seattle reaction to WTO in 1999 and the demonstrations in Washington, D.C. in the spring of 2000 for the poorest countries’ debt cancellation all show the hidden truth of the situation. In my view, so does the paranoid construction of the unreliable vertical-takeoff airplanes to be able to lift off from the multinationals’ centers in the poor countries in case the situation of injustice (the ever-fattening top of the chalice) were to get out of hand.
6. COMPARATIVE ANALYSIS With the traditional trade theory altered to fit modern real conditions of world trade, our comparative conditions are immediately apparent. The results of the preceding five sections are intimately related to the postulate – inherent in the modern world economy – of profit-maximization by the principal decisionmaking agents, the multinational corporations. These corporations, as shown in our essential diagram of Fig. 5, can maximize their profits through the separation of decisions concerning capital allocation on the one hand and concerning labor on the other. By contrast, if we lived in a world of economic democracy where enterprises maximize their own income per worker as democratic entities managing their
Comparative Systems, Destructive Trade and World Distributive Justice
261
own affairs, such results – and in particular the basic acts of arbitrage described in Fig. 5 – would be impossible. We may ask, why should a worker-run and -owned enterprise in the United States move its production to China or the Caribbean? The absurdity of such a move is clear simply by putting oneself into the shoes of such working communities. The objective of maximization of income per worker certainly would not be satisfied – and there are scores of other reasons to negate such a decision. Moreover, in an economic democracy, the capitalist lobbies which now dominate and determine the WTO-type policies would not exist, and political pressures would be far more in the direction of protecting domestic labor in the United States and other advanced economies. On the other hand, democratic firms in the developing countries could maximize their incomes per worker and thus earn not only subsistence-survival wages of the capitalist world, but also the benefits of their product market position, entrepreneurship, and so on. Most of all, however, they could, perhaps with the assistance of the international community, choose a development pattern or strategy appropriate to their location and conditions in the world economy. Their people would develop and keep improving their skills and technologies appropriate for their part of the world, instead of producing without such benefits in maquiladoras, or suffering from continuous “brain drain” of their best human resources to the industrialized world. On the same human plane, those pushed out of manufacturing in the rich countries often represent enormous and irreparable losses of high skills, not matched for the most part by positive developments in the low-productivity service industries.
7. CONCLUDING REFLECTIONS ON GAINS FROM TRADE, SYSTEMS AND THE SECOND BEST As we recalled already, the traditional theory of trade based on comparative advantage leads unambiguously to gains from trade for the world and at least some members of the world community. The theory of destructive trade here explained not only leads to dramatic maldistribution of income and welfare – the rich getting richer and the poor poorer – but there is nothing to guarantee the aggregate gains for the world as a whole. In fact we are facing here a case of what economists call the secondbest phenomenon, a situation involving a major disequilibrium in one market – the worldwide labor market.
262
JAROSLAV VANEK
The term MAJOR is by no means an exaggeration, as the disequelibrium is dominant, violent, quasi-permanent and pathological. It is dominant because it concerns the largest market in the world – the market involving all human beings. It is violent, because it refers to price (wage) maladjustments of the order of 10,000%; this we have pointed out already. It is permanent or quasipermanent because it involves the thousand year syndrome, sketched in No. 2 above. And it is pathological, socially and individually, because it leads to scores of human maladjustments of the type noted in No. 6 above and elsewhere in this paper. To document the fact that destructive trade need not lead to overall gains from trade, let us consider again the BAA of Fig. 5. First of all, on the national level all the BAA’s of all the arbitrageurs cannot be arrested in the long run until all gadgets production moves from the rich to the poor countries, that is until a “corner solution” is reached (as in the classical trade theory), without price-wage equalization, wages remaining roughly at the 10,000% differential. We can thus speak of an aggregate BAA and denote it by BAA. In terms of the world gross domestic product, the first element of the basic act involves elimination of all gadget production in the advanced economy. But in the pure theory, that loss of output is compensated by the production and import of gadgets from the poor countries. In reality, however, there will be scores of inefficiencies in that operation, certainly leading to diminution of world GDP. More food will be produced for export in the rich part of the world to feed the ten-cents labor overseas, but that labor usually comes from the traditional economy, thus diminishing food production there, on the limit by the same amount as the food imported for the maquiladoras. Again there need be no increase in world output of food. The positive gain associated with destructive trade is in theory the reemployment of the workers initially thrown out of work in the advanced economies (the segment ba in the diagram). But that segment in reality, when all is considered, will shrink to an aggregate output equivalent to either more or less than the segment aa. And there are so many elements behind that shrinkage – some discussed in this paper – that it is perfectly possible that the second, lesser alternative will obtain, that is, world aggregate output will decline; and this is on top of the phenomenal redistribution towards the rich that is absolutely necessary under destructive trade. The destructiveness of trade is underscored by the fact that in capitalism and profit motivation it suffices to have the segment ba positive to trigger the destructive arbitrage act. And this leaves a lot of possibilities for a’a being larger than the output of the reemployed – and thus for aggregate losses. This possibility is further made
Comparative Systems, Destructive Trade and World Distributive Justice
263
more likely if we consider some of the more subtle effects of ecology, various externalities etc.4 Those who are interested to learn more about the “lesser alternative” in the United States should consult a recently published book on New Rules for a New Economy.5 My conclusions for the capitalist world are less optimistic, but the data and analysis of Herzenberg, Alic and Wial are most useful and instructive. In essence the authors demonstrate the dramatic losses of wages and product of American labor over the recent decades; in real terms, for the minimum wage, involving declines of some 30%. But they argue that all this harm can be undone by enhancing human capital and focusing on the growing service sector. My comparative pessimism is based on three factors. First, our television and some kinds of computers are inflicting serious damage on productive human capital; second, most of the service jobs involve rather dull and primitive activities, limiting the learning by doing; and third, the deadweight of subsistence wage/income of the “millenial syndrome” is not limited in the poor countries, through competitive forces, only to manufacturing jobs. My own overall conclusion is that nothing short of worldwide democracy in all areas of human existence, including our economic system, can get us out of the difficulties of our times. Moreover, as in all second-best situations, with some markets irreparably out of equilibrium, the second-best (constrained) optimum solution for the world will have to rely on some properly designed policy intervention in some markets.
8. THE THEORY OF COMPARATIVE ADVANTAGE VERSUS THE THEORY OF DESTRUCTIVE TRADE I am writing this section inspired by a reader of an earlier version of this paper, who feels that my argument can be stated in terms of the Ricardian theory. I disagree for several reasons and claim that the destructive trade theory is a distinct one, and especially that it describes much more accurately the situation in our present-day world. It is also important to make the present statement because it is the comparative advantage theory that is used, implicitly or explicitly, by the globalization advocates, in promoting their strategies. We presented in Section 4 the Ricardian model precisely to differentiate it from the destructive trade model; but we presented both using the “dynamic” illustration of an arbitrageur operating on the differential between prices occurring in the world economy, one on the differences of product prices, the
264
JAROSLAV VANEK
other on the differences of real wage or human labor prices. More precisely, we note: (1) The Ricardian transformation functions may have identical slopes and thus indicate a no-trade situation. And yet the destructive-trade forces may be enormous, and lead to international trade flows of several goods and services, such as food, machinery, gadgets and technical expertise quite descriptive of present-day real world conditions. (2) Moreover, the Ricardian comparative advantage is basically an equilibrium theory, and the real essence of destructive trade involves states of disequilibria: in fact, disequilibria on two long-range levels: (3) Level I: the “millenial syndrome” explained in Section 2, showing a very long-range and extremely dramatic differential between price conditions in the world labor market. Not only is such a differential insurmountable over very long periods of time, but it has been increasing over the past half century. (4) Level II: even given such a dramatic differential of some 10,000%, the drama of destructive trade is a disequilibrium state where productive capacity of everything transportable is gradually shifting from the rich to the poor countries, the final state of equilibrium (where everything transportable ceases to be produced in the rich part of the world) never being fully attained. (5) Whereas the Ricardian equilibrium always leads to gains for at least some trading partners and losses for none, the destructive trade in each stage of the disequilibrium arbitrage (see Fig. 5) does not guarantee such gains. The positive profit motive of the arbitrageur may last as long as point a is to the right of point b, together with enormous losses for everyone else in the world disequilibrium economy. (6) As we noted already, under Ricardian equilibrium it is the economically smaller (price-taking) trading partner that should be gaining in the trading situation relative to the economically larger partner. Under destructive trade disequilibrium it is just the other way around, the richer (capitalowning and controlling) growing richer relative to the poorer. This is precisely what is happening in the present world. Recall the UNDP data showing increasing differentials between the richest and poorest quintile of 3, 6 and 7.4 thousand percent over the past 40 years. (7) What may be most interesting and most differentiating characteristic of the destructive-trade mechanism with respect to comparative advantage is its inherently destabilizing – or explosive – nature, confirming the increasing differentials just noted in No. 6. Indeed, large wage and income
Comparative Systems, Destructive Trade and World Distributive Justice
265
differentials are the cause of destructive trade; but destructive trade in turn leads to further increases of such differentials (8) It must be emphasized that the destructive trade theory does not replace the theory of comparative advantage which remains valid for some aspects or segments of international trade. However, if we evaluated the significance of a theory using the index of the relative price differentials over which it is applicable, the destructive trade theory, in the present world with up to 10,000% differentials, would appear as the more powerful explanation of real-world conditions.
9. A HUMAN INSIGHT: CLOSING THE CIRCLE We noted earlier the need of mothers and older children substituting at near minimum wages for job losses of fathers who used to make possible the American-dream way of life. This type of effect of destructive trade in the USA has a perfect analogue in the effects on the poor trading partners. Under NAFTA such effects can even be more dramatic. In the bulletin of the organization Witness for Peace (Summer/Fall, 2000) a report entitled “NAFTA’s toll on the Mexican peasant sector” tells a story that should be remembered. Actually our theoretical multinational’s arbitrageur does not need to worry about producing and exporting food – a fellowmultinational corporation will do it for him. From before NAFTA into the first years of its existence the imports of American Cargill corn have increased by some 500%, decimating traditional small farmers in Mexico, thus producing the necessary “slaves” to work in maquiladoras, or passing directly in search of jobs north of the border. It may be good to end this paper with the concluding paragraph of the Witness for Peace article: In Mexico today the corn farmer, once the backbone of rural Mexico, is standing at a crossroads not knowing where to turn. ‘We are like a body trying to defend itself from a plague. We are being attacked from all sides. Only God knows where we will end up’ . . . This is the contradiction of rural Mexico today, exacerbated by NAFTA: productive land in the hands of peasant farmers who are willing to work it, but which yields as its harvest only misery.
NOTES 1. Krugman, P. (1996). Pop Internationalism. Cambridge, MA: MIT Press; Burtless, G., Lawrence, R. Z., et al. (1998). Globaphobia. Washington, D.C.: Brookings Institution; Daly, H. E. and Cobb, J. (1989). For the Common Good. Boston: Beacon; and Daly (1995). The Perils of Free Trade. Scientific American, November 1993, 50–57.
266
JAROSLAV VANEK
2. See for example: James, A. M. and Elmslie, B. T. (1996). Testing HeckscherOhlin-Vanek in the G-7. Weltwirtschaftliches Archiv, 132(1), 139–159; or Maskus, K. E. (1985). A Test of the Heckscher-Ohlin-Vanek Theorem: the Leontief Commonplace. Journal of International Economics, 19(3–4), 201–212. 3. Herbert, B. (1995). Not a Living Wage. New York Times (op ed), October 9. 4. See for example the arguments offered by Daly, loc. cit., or the effects of increased transportation costs shown in J. and F. Vanek (1999). Systems, Location, Ecology and Society. Economic Analysis, 2(3), 209–221. 5. Herzenberg, S. A., Alic, J. A. and Wial, H. (2000). New Rules for a New Economy. Ithaca, NY: Cornell University Press.
REFERENCES Burtless, G., Lawrence, R. Z., Litan, R. E., & Shapiro, R. J. (1998). Globaphobia. Washington, D.C.: Brookings Institution. Daly, H. E., & Cobb, J. (1989). For the Common Good. Boston: Beacon. Daly, H. E. (1993). The Perils of Free Trade. Scientific American, 11, 50–57. Herbert, B. (1995). Not a Living Wage. New York Times, October 9 (op-ed). Herzenberg, S. A., Alic, J. A., & Wial, H. (2000). New Rules for a New Economy. Ithaca, NY: Cornell University Press. James, A. M., & Elmslie, B. T. (1996). Testing Heckscher-Ohlin-Vanek in the G-7. Welavirtschaftliches Archiv, 132(1), 139–159. Krugman, P. (1996). Pop Internationalism. Cambridge, MA: MIT Press. Maskus, K. E. (1985). A test of the Heckscher-Ohlin-Vanek Theorem: the Leontief Commonplace. Journal of International Economics, 19(3–4): 201–212. United Nations Development Programme (UNDP) (1992, 1998). Human Development Report. New York and Oxford: Oxford University Press. Vanek, J., & Vanek, F. (1999). Systems, Location, Ecology, and Society. Economic Analysis, 2(3), 209–221. Witness for Peace (2000). NAFTA’s toll on the Mexican peasant sector. Newsletter. Summer-Fall.