Marija Ugarkovi´c Profit Sharing and Company Performance
GABLER EDITION WISSENSCHAFT
Marija Ugarkovi´c
Profit Shari...
93 downloads
1230 Views
383KB 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
Marija Ugarkovi´c Profit Sharing and Company Performance
GABLER EDITION WISSENSCHAFT
Marija Ugarkovi´c
Profit Sharing and Company Performance
Deutscher Universitäts-Verlag
Bibliographic information published by Die Deutsche Nationalbibliothek Die Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at .
Dissertation Universität Dortmund, 2007
1st Edition November 2007 All rights reserved © Deutscher Universitäts-Verlag | GWV Fachverlage GmbH, Wiesbaden 2007 Readers: Frauke Schindler / Anita Wilke Deutscher Universitäts-Verlag is a company of Springer Science+Business Media. www.duv.de No part of this publication may be reproduced, stored in a retrieval system or transmitted, mechanical, photocopying or otherwise without prior permission of the copyright holder. Registered and/or industrial names, trade names, trade descriptions etc. cited in this publication are part of the law for trade-mark protection and may not be used free in any form or by any means even if this is not specifically marked. Cover design: Regine Zimmer, Dipl.-Designerin, Frankfurt/Main Printed on acid-free paper Printed in Germany ISBN 978-3-8350-0896-0
Preface The present dissertation is a synopsis of most of the research that I undertook during my time as a teaching and research assistant at the Chairs of Economic Policy of the Universities of Duisburg-Essen and Dortmund. Writing this thesis on the impact of profit sharing as a wage incentive taught me a lot about individual and group incentives, about intrinsic and extrinsic motivation not only on a theoretical level but also in the hands-on experience of completing this thesis. But above all it taught me about the importance of cooperation: I am indebted to numerous people and institutions that have contributed to the completion of my dissertation project in many ways. I would like to thank, first and foremost, my academic teacher Prof. Dr. Kornelius Kraft for many inspiring discussions and his continuous encouragement and help through all stages of my work. I also thank my second referee, PD Dr. Uwe Jirjahn, for many helpful comments and suggestions in particular during the final stages of this work, and Prof. Dr. Wolfgang Leininger for his readiness to participate in my PhD committee. My research could not have been realized in its present form without the existence of such an extensive and longitudinal data set as the IAB Establishment Panel. I therefore thank the German Federal Employment Service (BA) and the Institute for Employment Research (IAB) for providing the data and for their hospitality during my visits in Nuremberg. Among the employees of the IAB, I would like to express my special thanks to Dana Mueller for her virtually endless patience in running my statistical programs over and over again.
VI
Preface
My work has further profited from several discussions at conferences, within workshops of the Priority Programme “Flexibility in Heterogeneous Labour Markets” of the German Research Foundation (DFG) whose financial support is gratefully acknowledged, and from discussions with my former and current colleagues in Essen and Dortmund. I would like to express my gratitude to them not only for always being ready to discuss parts of my research or answering a quick question but also for their teamwork, company, and the many good laughs we had. I would like to dedicate the final section of this preface to the most important persons in my life: None of this could have been possible without my parents. Thank you for your love, for teaching me values, for always providing a safe haven, and for letting me have it my way. Thank you also to the “best sister in the world” (although she knows she is the only one), Vesna, for your unconditional love and support, your honesty and optimism, and all the fun we had throughout our lives. You really are a gift to all of us! There is, finally, one person left that went through all stages of this dissertation project with me and probably deserves my apologies more than my thanks for sometimes giving him a hard time testing his comedian skills on me and becoming an involuntary expert on wage incentives through several rounds of proofreading and discussing my research. Your love, friendship, encouragement, intellectual curiosity, and humour are an inspiration, Dubi. Every single day. Marija Ugarkoviü
Contents 1
Introduction
1
2
Definition, Incidence and Determinants of Profit Sharing
7
2.1 2.2 2.3 3
Definition Incidence Determinants The Economic Impact of Profit Sharing
3.1
Theoretical Considerations 3.1.1 Profit Sharing and Productivity 3.1.2 Profit Sharing and Employment 3.2 Prior Empirical Studies on the Impact of Profit Sharing 3.2.1 Productivity 3.2.2 Profitability 3.2.3 Wages 3.2.4 Employment Levels 3.2.5 Employment Stability 3.2.6 Summary 4 4.1 4.2
7 8 11 17 18 18 22
28 29 32 33 34 36 37
Data Set and Econometric Methods
39
Data Set and First Descriptive Results Microeconometric Evaluation Methods 4.2.1 Experimental Evaluation Methods 4.2.2 Non-Experimental Evaluation Methods 4.2.2.1. The Instrumental Variables (IV) Estimator 4.2.2.2. The Heckman Selection Estimator 4.2.2.3. Before-After Estimator 4.2.2.4. Difference-in-Differences 4.2.2.5. Matching
39 43 45 46
46 47 49 50 51
Contents
VIII
4.2.2.6. Conditional Difference-in-Differences 4.2.2.7. Which Estimator? 5
The Output, Employment and Productivity Effects of Profit Sharing 5.1 5.2 5.3 5.4
6
Theoretical Considerations Econometric Model Empirical Study Summary Profit Sharing: Supplement or Substitute?
6.1 6.2 6.3 7
Theoretical Considerations Empirical Study Summary Profit Sharing and the Financial Performance of Firms
7.1 7.2 7.3 8
Theoretical Considerations Empirical Study Summary Profit Sharing and Works Councils: Substitutes or Complements?
8.1 8.2 8.3
Theoretical Considerations and Empirical Evidence Empirical Study Summary
53 54 57 57 59 63 70 71 71 75 77 79 79 80 82 85 86 91 98
Conclusions and Outlook
101
Appendix
107
Bibliography
119
List of Figures Figure 1- Labour Demand in Equilibrium
25
Figure 2- Labour Demand under Negative Demand Shock
26
Figure 3- Frequency Distributions of the Propensity Scores
66
Figure 4 – Frequency Distributions of the Propensity Scores, Firms with Works Councils
92
Figure 5 – Frequency Distributions of the Propensity Scores, Firms without Works Councils
93
List of Tables Table 1 – Incidence of Profit Sharing in Selected Countries
8
Table 2 - Mean Comparison of Firms with and without Profit Sharing, 2000
41
Table 3 - Mean Values of Variables for Firms (not) introducing Profit Sharing, 1998
64
Table 4 - Results of the Probit Estimation
65
Table 5 - Results from Conditional Difference-in-Differences Estimation, Introduction of Profit Sharing
68
Table 6 – Results from Conditional Difference-in-Differences Estimation, Introduction of Profit Sharing, Capital Stock included
70
Table 7 - Mean Values of Variables for Firms (not) introducing Profit Sharing, 1998
76
Table 8 – Results from Conditional Difference-in-Differences Estimation
77
Table 9 - Mean Values of Variables for Firms (not) introducing Profit Sharing, 1998
81
Table 10 – Results from Conditional Difference-in-Differences Estimation
82
Table 11 – Results of the Probit estimation
91
Table 12 – Conditional Difference-in-Differences estimation, Firms with(out) Works Council
94
Table 13 – Mean Values of Variables for Firms (not) introducing Profit Sharing after Matching
95
XII
List of Tables
Table 14 – Conditional Difference-in-Differences Estimation, Firms with(out) Works Councils, 21-100 employees
96
Table 15 – Conditional Difference-in-Differences Estimation, Firms with Works Councils, 101-1000 employees
97
Introduction There is a long tradition of academic research into the nature, causes and consequences of financial participation of employees in their enterprises which can be traced back into the nineteenth century. In the past two decades, financial participation has attracted renewed interest among academics and policy makers alike. In several European countries, legislation that is supportive of financial participation has been introduced, while in others debates have been initiated between the social partners. In Germany, several changes have been made to the Capital Formation Law (Vermögensbildungsgesetz) during the 1990s. Financial participation was also propagated by former Chancellor Gerhard Schröder as “one of the pillars of the Social Market Economy” (Schröder, 2000), and only recently the German President Horst Köhler appealed to the social partners that “the time has come to bring up the issue of a financial participation of employees again” (Köhler, 2006). This upswing in the interest in financial participation is closely related to the fact that during the last decade, financial participation has grown in importance as a policy issue for the European Union. The driving force behind the promotion of this matter was the European Commission. An important landmark was the Commission’s Recommendation Concerning the Promotion of Employee Participation in Profits and Enterprise Results of July 1992 which has also become known as the PEPPER Recommendation. The PEPPER Recommendation which was passed in 1992 by the Council of Ministers of the European Union, urged member states to ensure that national legal structures did not inhibit the use of financial participation, to consider the introduction of fiscal incentives and to encourage the use of such schemes by facilitating the supply of adequate information to all relevant parties (Council of the European Communities, 1992). In the Commission’s Social Policy Agenda 20002005, the importance of financial participation was reaffirmed and a proposal was made to “launch a communication and action plan on the financial participation of workers” (Commission of the European Communities, 2000). The motives of the European Commission for a promotion of the so-called PEPPER schemes which
2
Introduction
encompass profit sharing and employee share ownership cover a wide variety of issues as the following quotation expresses :
“The different benefits of employee financial participation make it an integral element of achieving the Lisbon objectives. Financial participation is an excellent example of a policy which can simultaneously address economic, employment and social objectives in a mutually reinforcing way. When introduced in the right way, financial participation can render enterprises more profitable and competitive, improve the motivation, commitment and job satisfaction of workers, enhance the quality of work and last but not least contribute to a more equitable distribution of income and wealth.” (Commission of the European Communities, 2002) Several facets of employee financial participation are thus stressed. On the company level, financial participation has traditionally been introduced to commit employees to the firm which should lead to higher motivation and productivity, higher investment into firm-specific human capital and lower turnover. It has also been used as a means to enhance the flexibility of remuneration, to attract performance-oriented workers and to improve the company’s equity situation. Overall, financial participation should thus lead to an increase of the firms’ competitiveness. From the employees’ perspective, financial participation encourages greater involvement of employees in the progress of their companies which should not be without consequences on workers’ motivation, job satisfaction and work quality. Finally, on an aggregate level, the assumed company-level effects of financial participation can contribute to an overall increase in productivity which might spur economic growth and increase the economies’ prosperity. Martin Weitzman (1984) in his theory of a share economy further propagates financial participation as a means to increase employment levels and employment stability by substituting part of fixed wages with a variable profitrelated part. From a socio-political perspective, financial participation is, finally, expected to raise social integration and to achieve a wider distribution of the wealth generated by the enterprises which workers have helped to produce (for a comprehensive discussion see e.g. Priewe and Havighorst, 1999).
Introduction
3
A recurrent theme in recent discussions of financial participation among policymakers and the social partners is the need for more research into the outcomes of financial participation schemes. The aim of this thesis is to contribute to this lack of empirical research. In general, financial participation schemes can be classified into profit sharing and employee share ownership schemes. We restrict our analysis to the investigation of profit sharing as a method of payment on various firm-level outcomes.1 Quite a number of studies have investigated the effects of profit sharing. The primary focus of the majority of studies is the impact of profit sharing on productivity. Surprisingly, the consistency of findings is remarkable: Profit sharing is associated with higher productivity levels in every case, regardless of methods, model specification and data used. By contrast, only a limited number of studies analyse how profit sharing affects other outcomes such as employment, wages or profitability, and, moreover, do not find conclusive evidence. One contribution of this study is the rare provision of a comprehensive picture of profit sharing that takes into consideration the effects of profit sharing on output, employment, productivity, wages and profitability. We also investigate the question how works councils as a form of representative participation affect the success of profit sharing as a mode of financial participation, a question that has found very little attention so far. Another contribution of this study is of methodological nature. In light of the remarkably clear result that profit sharing is associated with higher productivity, it seems rather surprising that across the industrialized world only a minority of firms considers introducing such a scheme. Since it appears unreasonable to assume that the majority of capital owners are persistently unaware of the possible productivity effects of profit sharing, there must be other reasons for their reluctance to introduce such an apparently useful remuneration model. One explanation might be that firms have specific advantages or disadvantages with respect to the introduction of alternative incentive systems. Depending on different circumstances such as work organization, work content, verification possibilities of individual performance or industrial 1
The restriction of our analysis to profit sharing schemes only is due to the fact that employee stock ownership is still rather uncommon in Germany. According to the Institute for Employment Research (IAB), only about 2 percent of German establishments had implemented an employee share ownership scheme in 2001 (IAB, 2002).
4
Introduction
relations, firms will introduce the incentive system that best suits their needs. Firms with specific advantages in the application of profit sharing introduce it, while others have no interest in such a kind of incentive and rely in contrast on other motivational instruments such as piece rates, promotions or dismissals as a penalty in the case of insufficient performance. If firms, however, choose an adequate payment scheme conditional on specific circumstances, then selection problems pervade the analysis and a simple comparison of firms with and without profit sharing might be misleading. As a matter of fact, most empirical studies neglect the problem of selectivity thus casting doubt on the reliability of their estimates. We apply (to the best of our knowledge) for the first time in connection with profit sharing the conditional difference-in-differences estimator to account for selectivity due to observable as well as unobservable factors. The remainder of the thesis is structured as follows. Chapter 2 begins by briefly clarifying the concept of profit sharing and presenting information on the incidence of profit sharing in different countries. In this chapter, we also outline the theoretical considerations and empirical evidence concerning the determinants of profit sharing which is of high importance for the methodological part of the study. Chapter 3 presents the theoretical arguments of the two main academic discussions on the possible impact of profit sharing, namely its effects on productivity and employment. Several hypotheses arising from these considerations have been tested empirically and are discussed in the second part of the chapter. Chapter 4 introduces the nationally representative German IAB Establishment Panel where the present empirical analysis is based upon and provides first descriptive results. Since our presumption that selectivity pervades the analysis is confirmed, an overview of the methods that deal with the problem of selectivity is given. Based on the discussion of the advantages and disadvantages of the alternative estimators, we outline our choice of the conditional difference-in-differences estimator. The second part of the thesis which encompasses Chapters 5 to 8 presents the empirical studies that are at the core of this work. Chapter 5 deals with the output, employment and productivity effects of profit sharing. Focusing on all three outcome
Introduction
5
variables enables us to test the different hypotheses concerning a potentially beneficial impact of profit sharing on productivity and employment. Chapter 6 investigates the question whether the introduction of profit sharing is associated with a reduction in base wages as proposed by Weitzman (1984) or whether alternative reasons exist that might induce employers to prefer the payment of a supplementary profit share. Chapter 7 explores how profit sharing influences the firms’ financial performance. The implementation of a profit sharing scheme is associated with different costs that might offset potential gains. A priori, the relationship between profit sharing and profitability is therefore not clear. Chapter 8, finally, picks up on the research question of Chapter 5 relating to the output, employment and productivity effects of profit sharing, but specifically focuses on the question how the existence of a works council as a form of representative participation influences the results obtained in Chapter 5. Although this study certainly contributes to the recent discussion on the merits of financial participation by offering a number of interesting results, it is needless to say that a lot of other important questions remain unanswered. Chapter 8 therefore presents conclusions and an outlook on future research topics.
Definition, Incidence and Determinants of Profit Sharing Definition The concept that employees might be paid in part out of profits is not new. The German economist Johann Heinrich von Thünen proposed theoretical arguments for profit sharing as early as 1850 (Thünen, 1850). First profit sharing plans in the United States were reported in 1794 (Coates, 1991). In a strict sense, profit sharing means that profits are shared between capital owners and employees by giving employees in addition to a fixed wage a variable part of income directly related to profits or some other measure of company performance. As opposed to traditional bonuses which are linked to individual performance, profit sharing is a collective scheme applied to all or most employees in a firm or establishment. The rules determining the level of the profit sharing bonus and its distribution between employees are usually determined in advance (Uvalic, 1991, OECD, 1995). In practice, profit sharing can take different forms. The main differences concern the timing of the benefits and the mode of payment: profit sharing can provide employees with immediate or deferred benefits and it can be paid in cash, company shares or other securities. Cash-based profit sharing links employee bonuses directly to some measure of firm performance (e.g. profits, sales or value added) and most frequently provides an immediate payment. It can, however, also be designed as a deferred scheme where cash bonuses are frozen on special accounts for a determined amount of time or where a certain percentage of profits is allocated to company funds which are then invested in the name of employees. Share-based profit sharing, on the other hand, involves the possibility of employees acquiring shares in the company free or on preferential terms. Usually, they are frozen
Definition, Incidence and Determinants of Profit Sharing
8
in a fund for a certain time period before employees are allowed to dispose of them and can therefore be considered as a deferred scheme. By definition, employee shareholding schemes are only included if the share allocation can be regarded as a profit sharing bonus. Regular distributions of share or stock options which are independent of firm performance are excluded (Uvalic, 1991, OECD, 1995).
Incidence Data on the international incidence of profit sharing is extremely sparse. According to the statistics of the European Commission (Poutsma, 2001) and the OECD (1995), the percentage of firms with a profit sharing system is everywhere quite low except for the cases of France and the United Kingdom which both have a long tradition of encouraging financial participation (Table 1). Historically, the system of financial participation in France has Gaullist right-wing roots. De Gaulle’s vision on the cooperation between capital and labour motivated him to introduce financial participation after World War II. The system was, however, never supported by left-wing socialist ideology and so financial participation was never really associated with the idea of participative structures. Despite the changes of left- and right-wing governments in France, the French legislation has continuously offered a legal framework and generous tax advantages to cash-based and deferred profit sharing since the end of the 1950s (Poutsma, 2001). Table 1 – Incidence of Profit Sharing in Selected Countries Country
profit sharing establishments (in %)
Denmark
9
France
51
Germany
12
Ireland
7
Italy
4
Netherlands
13
Portugal
7
Spain
6
Sweden
19
UK
28
USA***
16
Canada***
15
Source: Poutsma, 2001, ***OECD, 1995.
Incidence
9
As the only European country, France obliges firms with more than 50 employees to share profits by means of a deferred profit sharing plan called participation aux fruits de l’expansion (participation in the benefits of growth). The details of the plan, such as the fraction of profits to be shared out, the method of allocation and the retention period are specified in the legislation. After 5 years, employees can withdraw their income free from income tax or social security contributions. The contributions made by the employers are exempt from corporation or payroll taxes. The other scheme favoured by French legislation is the voluntary, cash-based profit sharing scheme called intéressement des salariés de l’entreprise that has been introduced in 1959. Tax concessions for both employers and employees are similar to those of the participation (OECD, 1995). The French government’s belief in profit sharing as a means to improve efficiency and productivity has further led to the introduction of the participation law in 1994 which includes greater tax advantages for deferred profit sharing and the creation of the Superior Council of Participation that watches over the application of financial participation in firms and reports annually to the Prime Minister and the Parliament on the developments in financial participation plans (for a more extensive discussion see Poutsma, 2001). The British government introduced tax concessions for profit sharing schemes in 1987 which were further extended in 1991. In contrast to the French case, the British government was inspired by Martin Weitzman’s arguments (reviewed in further detail below) that profit sharing will stabilize employment, reduce unemployment and increase flexibility (Bradley and Estrin, 1992). Tax relief was granted on the one hand to Approved Profit Sharing Share (APS) schemes and on the other hand to Profit Related Pay (PRP) schemes. The APS legislation encouraged employee share holdings where ordinary shares of the company were bought by a trust funded by the company and distributed to employees. In order to qualify for income tax relief, employees had to have at least five years of service in the company and the shares had to be invested for a period of at least five years. PRP schemes on the other hand were - similar to the French intéressement - cash-based profit sharing schemes that provided income tax concessions for employees up to a certain limit. The introduction of such schemes had to be approved by the workforce and at least 80 percent of the staff with more than three years of service had to participate (OECD, 1995). To create an incentive for employers to set up a PRP scheme, set-up costs of the scheme were
10
Definition, Incidence and Determinants of Profit Sharing
tax-deductible. Initially, profit sharing was paid as an add-on to regular wages. After tax concessions improved, however, employers set up cosmetic schemes known as ‘salary sacrifice’ where employees’ net income levels remained stable but employers were the main beneficiaries of the tax subsidies (Poutsma, 2001). Arguing that the original aim of a wide diffusion of profit-related pay schemes had been achieved by the mid-nineties, tax relief was therefore phased out with the beginning of 2000. Tax relief for APS schemes also phased out in the year 2001 but was substituted by a more flexible all-employee share ownership plan (AESOP) (HM Treasury, 2000, 2001). Most other countries offer no or only weak incentives for an introduction of profit sharing schemes. Accordingly, the diffusion of such schemes is not very wide. The Institute for Employment Research (IAB) which is the research branch of Germany’s Federal Employment Services estimates that in the year 2001 approximately 9 percent of German establishments had implemented profit sharing which is lower than the numbers reported from the European Commission above (IAB, 2002). The encouragement of employee financial participation in Germany is rooted in the German concept of the Social Market Economy (Soziale Marktwirtschaft) where economic democracy is viewed as a means to diminish conflicts between labour and capital. Since the issue of financial participation is viewed primarily as a matter of the social partners, however, German governments have focused on appeals rather than on providing strong monetary incentives to promote an introduction of such schemes. Monetary incentives are rather weak and basically aim at the group of employees with moderate incomes (Carstensen et al., 1995, OECD, 1995, Priewe and Havighorst, 1999, Poutsma, 2001). The German legislation provides no incentives for profit sharing per se. Cash-based profit shares are considered normal wages and are therefore not exempted from social security contributions. There is a considerable body of regulations, however, which are designed to encourage employee stock ownership and asset accumulation and therefore may encompass deferred and share-based profit sharing schemes as well. These are laid down in a series of Capital Formation Laws (Vermögensbildungsgesetz) and the Income Tax Law (Einkommenssteuergesetz). Under the Capital Formation Law, employees are encouraged to participate in their own firm’s capital and in that of other firms. Concessions are offered to single and married persons if their annual taxable income does not exceed 17.900 and 35.800 Euro respectively and
Determinants
11
if the participation in investment is committed for a retention period of at least seven years. The savings bonus is 18% (22%) for West German (East German) employees’ contributions up to 400 Euro annually into investments in equity, and 9% for contributions up to 470 Euro paid into a home ownership savings plan or used to pay off mortgages on residential property (Bundesministerium der Justiz, 2006a). Additionally, the Income Tax Law grants exemption from tax and social security contributions up to a certain maximum amount if employees are offered free or discounted shares by their employers that are frozen for a minimum period of six years (Bundesministerium der Justiz, 2006b, for a discussion see also Priewe and Havighorst, 1999, Poutsma, 2001).
Determinants A chapter that is of major concern for this study regards the question why firms adopt profit sharing. The previous section indicated that national legislation plays a key role with respect to the incidence and growth of profit sharing. In countries which have a legislation that strongly encourages profit sharing, such as France or the United Kingdom, profit sharing has been widely adopted. Other countries that have not implemented such strong incentives have a much lower percentage of profit sharing firms. From a single firm’s point of view, this is quite rational since the implementation of a remuneration scheme is associated with various costs. Firms will introduce a specific method of pay if the overall gains exceed the costs associated with the scheme. Since tax concessions off-set these costs to a certain degree, the probability of introduction increases. But profit sharing is not only driven by legislation. Other firm-specific circumstances also play a role in a firm’s choice of compensation scheme. To understand these factors, it is helpful to briefly discuss the alternatives to profit sharing (for a more extensive discussion see Milgrom and Roberts, 1992, Lazear, 1995, Prendergast, 1999). The simplest model of neoclassical economic theory assumes competitive labour markets with perfect information and mobility. Since the employer has perfect information on employee effort and output, the mode of compensation plays no role and each employee receives the value of his contribution to output. Abandoning the world of perfect information, however, opens the possibility for employees to shirk and to provide less effort. One remedy to this problem is supervision. If supervision is
12
Definition, Incidence and Determinants of Profit Sharing
costly, different compensation schemes can provide ways to induce appropriate levels of employee effort (Kruse, 1993). Among those are piece rates, deferred compensation, efficiency wages, promotions and group incentive schemes. If monitoring is costly but individual worker output can be easily observed, piece rates may be used which usually tie a worker’s remuneration to physical output. Piece rates work particularly well in settings where tasks are primarily simple and repetitive, output and output quality are easy to monitor, and where skill development and cooperation between workers is of minor importance (e.g. Lazear, 1986, OECD, 1995). The downside to piece rates are high costs of monitoring quality and quantity, the possibility of excessive wear or abuse of capital equipment and the difficulty of setting appropriate norms, specifically in cases where worker collusion is possible. In such circumstances, workers have an incentive to try to establish low norms and resist the implementation of technologies which might potentially raise them (Lazear, 1986, Gibbons, 1987). Another instrument to minimize the risk of employee shirking is deferred compensation. Employees will not shirk if the consequences of being caught are associated with a loss in a significant amount of deferred remuneration – even with less than perfect monitoring. This can be done with a pension plan or by tying compensation to tenure such that employees are “underpaid” early in their tenure, and “overpaid” later (Lazear, 1979). In order to induce employees into such arrangements, overall compensation over the working life needs to be higher than without deferral. The longer the period of deferral, the lower are employees’ incentives to increase effort, specifically if future development is uncertain (Hübler, 1995). In essence, deferred compensation is a dynamic efficiency wage theory. Instead of changing the type or timing of pay, employers may also choose to pay “efficiency” wages that are above the market levels. Wages can thus be set in such a way that employees have an incentive to put forth optimal effort because they fear being caught shirking and losing the wage premium (Yellen, 1984, Akerlof and Yellen, 1986). Another very popular instrument to induce higher effort are promotions. In many situations, employees exert effort in order to get promoted to a better paid position, where the reward tied to the position is fixed and where different employees compete
Determinants
13
for those positions. Baker et al. (1994a, 1994b) and Gibbs and Hendricks (2004) show that firms to a large extent provide incentives through the prospect of promotion. Firms may prefer to use promotions because it is usually easier to observe employees’ relative performance than absolute performance (Lazear, 1995). Problems with promotions can arise, however, from the fact that compared to the number of candidates, usually just few are promoted. Repeated lack of promotion is likely to lead to frustration, lower effort and risky behaviour. Furthermore, since promotions evaluate individuals on how well they do relative to others disincentives for cooperation are created (Lazear, 1989, Prendergast, 1999). Finally, group incentive schemes such as profit sharing are an option for the employer to lessen the principal-agent-problems of costly supervision. By tying a portion of pay to company performance, employee and employer incentives are brought into closer alignment. Profit sharing might be the preferred choice when individual output cannot be easily determined, i.e. when production is complex and interdependent and worker cooperation plays an important role. In such an environment, individual incentives are not useful.2 A second potential benefit of profit sharing is that it introduces a certain degree of wage flexibility. In times of lower firm profitability, profit sharing automatically decreases worker compensation without the need for costly renegotiation of wages or layoffs that come at the cost of a loss of firm-specific skills. Profit sharing might hence be interesting for firms that experience variability in firm performance or new firms that have uncertain prospects. A third motivation is reported particularly from the United States. Supposedly, firms have adopted profit sharing there to discourage unionisation. The idea is that by tying pay to company performance, employees’ identification with the employer rises thus blurring the traditional distinction between capital and labour (Mitchell et al., 1990). Summarizing, firms thus have different methods of pay at their disposal that offer different ways how to encourage workers to provide optimal effort, but carry different costs. Depending on the firm’s characteristics and circumstances, management or the owners will choose the compensation scheme that best suits their needs.3 Our 2
3
Jirjahn (2000) shows that in a work organization characterised by multi-dimensional tasks, profit sharing has advantages over piece rates as well as fixed wages. The various systems, thereby, do not necessarily exclude each other but can also be introduced as complementary instruments.
14
Definition, Incidence and Determinants of Profit Sharing
considerations suggest that profit sharing should be preferably introduced in firms where monitoring costs are high, where the organizational structure is complex and worker cooperation is a key production component. Profit sharing might also be desirable in firms that operate on unstable markets or young firms whose future prospects are yet uncertain. Furthermore, industrial relations may play an important role with respect to the adoption of profit sharing. Several econometric studies try to identify the determinants of profit sharing (for an overview see OECD, 1995, or Pérotin and Robinson, 2003). The results of these studies are quite mixed which might be due to several reasons. First, the range of factors being looked at varies greatly between studies owing to the fact that economic theory offers little guidance. We have also outlined above that national differences play a role. Finally, very few studies observe firms before they introduce a profit sharing scheme. Simply comparing firms that have profit sharing with those that have not introduced such a scheme runs the risk of mistaking a feature that is an effect of profit sharing for a motive for its introduction. Despite these limitations, an interesting pattern seems to emerge. Several groups of variables, namely firm size, organizational structure, industrial relations and the firms’ external environment are found to play an important role in the adoption of profit sharing. Size The impact of size on the introduction of profit sharing is a priori not clear. On the one hand free rider effects should intuitively be greater as company size increases, but on the other hand information asymmetries and monitoring become more difficult implying that profit sharing might be an instrument of associating workers with the aims of their employers and encouraging them to monitor each other. Moreover, larger firms usually have better information and more resources to spend on the design and implementation of various remuneration systems. In a recent paper, Adams (2006) provides an alternative explanation why profit sharing might be found in larger firms by arguing that if the production function is such that worker effort levels are complementary (as is often the case e.g. in manufacturing), then equilibrium effort choices are increasing in the size of the firm and firm-wide incentives become more valuable. There is considerable evidence that profit sharing firms are on average larger than their non-sharing counterparts (see e.g. FitzRoy and Kraft, 1987a, Carstensen et al., 1995, Arranz-Aperte and Heshmati, 2003, Amisano and Del Boca, 2004).
Determinants
15
Organizational Structure In environments where work tasks are complex and interdependent, individual outputs are hard to measure and information asymmetries and monitoring problems could be more intense. In such a situation, individual incentives are not useful and remuneration schemes that are based on the output of specific groups or units come to the fore. A range of measures have been used in empirical studies as approximations, including the existence of innovative work practices such as team work or cost and profit centres, the qualification level of employees or investments into information and communication technology. The evidence confirms that these factors play a significant role in a firm’s decision to implement profit sharing (see e.g. Kruse, 1996, Pendleton, 1997, Pendleton et al., 2002, Heywood and Jirjahn, 2002, Wolf and Zwick, 2002a, Pérotin and Robinson, 2003). Industrial Relations We had argued above that trade unions have often been hostile to profit sharing since they worried that employees that participate in profits might come to identify with capital rather than labour. In recent years though, many union organizations and confederations in Europe have come to the view that profit sharing does not necessarily undermine their position (Pendleton et al., 2002). Another important point arises from the fact that a prerequisite for the success of profit sharing is a consultative environment where the level of trust between employees and management is high. Employees need assurance that an introduction of profit sharing will not lead to unexpected reductions in pay or the manipulation of profit figures by management in order to reduce bonuses. In the German context, the presence of a works council might help to create such a climate and was therefore included into a number of studies. Empirical investigations generate a mixed picture of the relationship between representative participation and profit sharing. Using German data, Hübler and Jirjahn (1998) and Wolf and Zwick (2002a) find that profit sharing is positively associated with the presence of a works council. Hübler and Jirjahn find, however, a negative association between profit sharing and the existence of a collective agreement. Cheadle (1989) and Poole (1989) confirm this negative association of profit sharing and unions, while Gregg and Machin (1988) and Kruse (1993) using UK and U.S. data find that profit sharing is positively associated with union presence.
16
Definition, Incidence and Determinants of Profit Sharing
External Environment When profits are unstable and risky, firms have a greater incentive to set up flexible remuneration systems in order to transfer some of the risk to their employees. Therefore, exporting firms and firms on competitive or unstable markets are assumed to have a higher propensity to introduce profit sharing. Drago and Turnbull (1996) take another perspective why profit sharing should be more prevalent in competitive firms by arguing that only in competitive firms where demand is elastic or uncertain, workers will recognize increasing rewards of cooperative behaviour and thus will increase their effort. In a low competition setting with inelastic demand, an introduction of profit sharing would have no impact on worker behaviour and so firms have no incentive to introduce such a scheme. Several studies confirm these presumptions. Kruse (1996, 1993) e.g. finds that firms with higher variability in profits are more likely to introduce profit sharing. Drago and Heywood (1995) find evidence for a strong positive correlation between high levels of competition and presence of profit sharing schemes and Wolf and Zwick’s (2002a) analysis of German firms indicates that exporting firms are more likely to implement profit sharing.
The Economic Impact of Profit Sharing The discussion on a financial participation of employees has a long history. In Germany, one could virtually speak of an evergreen debate: In the fifties and sixties, the issue of distributive justice was emphasized. The participation of employees in capital and profits was seen as an instrument of “wealth accumulation in workers’ hands” (Oberhauser, 1963, Föhl, 1964) and resulted in the introduction of the first Capital Formation Law in 1961. In the eighties, the focus then shifted to a possible employment effect of profit sharing. Martin Weitzman propagated profit sharing in his Share Economy (1984) as a means to reduce unemployment, thereby sparking a yearlong controversy. During the last decade, financial participation is discussed primarily on two different levels: In the East German context, specifically, it has been argued that profit sharing might help to attenuate the unemployment problem. The idea is that wages in East Germany have converged to the higher West German level too quickly. If part of the wages was tied to the firms’ profits and not paid out but rather invested into the firms, two birds could be killed with one stone: Wage flexibility would increase and the firms’ equity situation could be improved thus leading to an overall increase in the firms’ competitiveness which might then, in turn, have a beneficial impact on unemployment (Sinn and Sinn, 1991, Hübler, 1995, Sinn, 1997, Priewe, 1997). Lately, the discussion of profit sharing as a means to increase wage flexibility has also been extended to the West German context. The other issue concerns the incentive effects of profit sharing which have been brought into discussion at regular intervals not only in Germany, but also internationally. In the following, the theoretical considerations on the two main discussions around profit sharing in academic research – namely its effects on productivity and employment – are outlined.
18
The Economic Impact of Profit Sharing
Theoretical Considerations
Profit Sharing and Productivity “Incentives are the essence of economics.” (Lazear; 1987) Since the beginnings of the Industrial Revolution, firms have experimented with profit sharing for employees. Then as now, advocates of profit sharing claimed that its use helped increase company performance. The channels through which productivity might be stimulated can be diverse: Profit sharing can increase worker effort, it can increase the skills of employees and improve the flow of information within the organisation (Kruse, 1992). Regarding the first link to worker effort, the simplest argument is that, ceteris paribus, paying for observed performance presumably generates more output than paying for time at the job. Since under profit sharing employees’ remuneration is tied to company performance, workers further have a direct monetary incentive to increase their effort. Worker effort could, however, also be affected indirectly. If employees are participating in profits, stronger identification with the company could result in increased effort, but also in a decrease in absenteeism and turnover and a higher diligence with respect to the usage of material and machines (Hübler, 1995, OECD, 1995). The second hypothetical link - that profit sharing increases worker skills - rests upon the assumption that a higher identification of employees with their firms reduces employee turnover and thus promotes investment into firm-specific skills. Profit sharing might also attract highly qualified and performance-oriented workers leading to a worker sorting process which implies a higher average qualification level (Lazear, 1986, 2000, 2005). Finally, better information flows not only between employees but also between employees and management could lead to a higher organizational efficiency. Under a profit sharing scheme, employees have an incentive to share information concerning the optimisation of the production process with their co-workers as well as with
Theoretical Considerations
19
management. Drago and Turnbull (1988) and Drago and Garvey (1998) show theoretically that profit sharing should also increase workers’ incentives to help one another. Employees may further be more willing to accept technological changes which increases profits (Kruse, 1992). Several economists have raised doubts whether profit sharing can induce productivityenhancing effects. The three most commonly applied counterarguments are the 1/N – problem, the possible weakening of capitalist property rights due to employees’ demands for codetermination and issues of risk when employees are exposed to an unacceptable degree of pay variation (Weitzman and Kruse, 1990).4 The most serious problem with profit sharing is that the fruits of any additional individual effort have to be shared with all other colleagues in the plan. With N employees, each receives only 1/N of any additional profits thus leading to a very weak individual incentive as the number of workers grows large. This is called the 1/N (with N being the number of employees) or free rider problem (Alchian and Demsetz, 1972, Samuelson, 1977). The 1/N problem may be seen as a classic example of a prisoner’s dilemma from game theory: Every worker has an incentive to shirk, though all employees could be better off if all put forward higher performance. The free riding story is often used to argue that group incentives will be ineffective for any larger values of N and, therefore, greater emphasis should be put on individual incentives. Under a collective incentive system, the free riding problem is mitigated by the fact that shirking or free riding imposes cost directly on all colleagues, so that employees have an incentive to monitor their co-workers and exert pressure if necessity arises. Horizontal monitoring might also be more efficient than control by a supervisor as workers usually have better information on their colleagues’ effort level. Costs for vertical supervision are thus reduced due to horizontal supervision (FitzRoy and Kraft, 1985a, Kandel and Lazear, 1992). While such considerations may ease the problem by extending the individual’s range for influencing group performance, individual incentives may still be weak. In a repeated game setting, however, which implies a long-term relationship among workers, different equilibria could develop, one of which could be a cooperative agreement of all employees to work harder and share the 4
For an extensive discussion see Jirjahn (1998).
20
The Economic Impact of Profit Sharing
rewards (Fudenberg and Maskin, 1986, Axelrod, 1984). It is, however, not certain how such a solution can be established and maintained, and several factors such as group size, task interdependence, the frequency of monitoring team output, exit costs, employees’ capabilities of peer sanction, history, personal relationships etc. may play a role (e.g. MacLeod, 1988, Che and Yoo, 2001).5 Even if the free riding problem could be overcome, another problem may arise: if workers share more profits, then firm owners of necessity do not get the full share of profits. Alchian and Demsetz (1972) argue that this might dilute the owners’ incentives and weaken their motivation, power and authority. With increased profit sharing, workers may demand more codetermination in decision making. The pace of decision making may be reduced and the cost of achieving consensus may be higher. If workers have the tendency to favour short-term increases in remuneration over longer-term investments, stronger employee participation might also lead to less efficient decisions (Alchian and Demsetz, 1972, Jensen and Meckling, 1979). The basic premises of this “property rights” school which criticizes codetermination revolve around the idea that the “essence of the firm concerns monitoring because otherwise labour does not work well, that capital can efficiently monitor labour and, therefore, efficiency requires that capital be given all residual claims on profits and all decision making power” (Weitzman and Kruse, 1990). The basic assumptions of the property rights school have been challenged by arguing that principal-agent problems between owners and managers are ignored, as is the fact that under profit sharing a decrease in monitoring by management is accompanied by an increase in mutual monitoring of workers. Finally, in less extreme settings than perfectly costless monitoring where workers may have technical information that would be costly for management to obtain or in environments where workers can monitor and motivate each other, profit sharing might to some extent even be desirable (e.g. Putterman, 1984, Nalbantian, 1987). Finally, the issue of risk exposure of employees is often mentioned as a disadvantage of profit sharing schemes. Although higher profit sharing might elicit greater effort from employees, their exposure to fluctuations of firm performance also rises. Firm performance, however, is not only influenced by employees’ effort but also by 5
For experimental evidence see e.g. Nalbantian and Schotter (1997).
Theoretical Considerations
21
external factors such as changes in market conditions, cyclical fluctuations or technological changes. This can lead employees to think that their contribution is only marginal and to attribute higher weight to other factors including “wrong entrepreneurial” decisions made by management. Since employees are usually rather risk averse, profit sharing - especially if accompanied with a reduction of fixed base wages - would be interpreted as a shift of company risk to employees. A rejection of the monetary incentive in such a situation seems probable (Hübler, 1995). This problem might be overcome, however, by compensating workers for their additional risk by increasing average remuneration. Conflicts could also arise because management usually has more information and knowledge on the construction of balance sheets (Meade, 1986a). The motivating effects of profit sharing could be diluted if employees worried that management manipulates the accounts in order to reduce profits and therefore the size of the bonus. An environment which is based on cooperation and trust can help to overcome such problems (OECD, 1995). Arguing from a theoretical standpoint, therefore, no definite answer can be given a priori concerning the beneficial productivity effects of profit sharing. Profit sharing has the potential to increase productivity. It is, however, not enough for management just to implement a profit sharing scheme and expect it to improve performance. Weitzman (1995) points out that “(…) To get the productivity-enhancing effects, something more may be needed – something akin to developing a corporate culture that emphasizes company spirit, promotes group cooperation, encourages social enforcement mechanisms, and so forth.” Before presenting empirical evidence on the impact of profit sharing, the second major discussion on the employment effects of profit sharing initiated by Martin Weitzman is presented.
22
The Economic Impact of Profit Sharing
Profit Sharing and Employment “Share firms ever hungry for labour are always on the prowl - cruising around like vacuum cleaners on wheels, searching in nooks and crannies for extra workers to pull in at existing compensation parameter values. Such an economy inherently resists recession.” (Weitzman, 1984) The negative supply shocks in the 1970s confronted both politicians and researchers with the so far unknown phenomenon of stagflation. Since the traditional instruments of monetary and fiscal policy failed and even aggravated the situation, a blank patch on the research map opened up. Martin Weitzman’s idea of a share economy in which widespread use of profit sharing would help protect the economy against instability and unemployment (1983, 1984, 1985, 1986) thus in a way came right in time and was gratefully taken up. His policy proposal to introduce tax incentives in order to encourage profit sharing has been enacted into law in the United Kingdom and has also been seriously considered by politicians in the United States (Levine, 1989). At the heart of Weitzman’s theory lies the idea that the wage can be decomposed into a fixed and a variable profit-related part. Considering that the total share of profits allocated to the workers is independent of employment, the firm has to provide for the fixed component of pay only when hiring an additional worker. The marginal cost of labour is thus lower than in the case of fixed wages and the firm’s demand for labour should rise. The result is that firms with substantial profit sharing will on the one hand have strong incentives to retain workers when business shocks occur, but will also have an incentive in equilibrium to get every unemployed worker who may appear on the labour market. Weitzman’s theory received a lot of attention in the business press as well as in academic publications in the 1980s and 1990s. The first euphoria, however, was soon replaced by a more realistic assessment of the theory in a flood of publications.6 In the
6
See e.g. Meade (1986a, 1986b), Summers (1986), Estrin et al. (1987), Nuti (1987), Levine (1987, 1989), Nordhaus (1988) or Layard et al. (1991).
Theoretical Considerations
23
following, the focus will be put on a primarily graphical analysis of the fundamentals of the theory and the presentation of its main objections.7 The starting point for Weitzman’s claims about the advantages of the share economy is the argument that profit sharing will boost employment. To illustrate this idea suppose that under a share system the real wage of an employee, w, can be decomposed into a base wage w` and a share Ȝ in the net profits where net profits are defined as revenues R minus base wage costs w`L and L denotes the number of employees. Since labour is the only production factor in Weitzman’s model, capital costs can be ignored and the wage formula can be defined as follows:
w
w`O
R(L) w`L . L
(1)
In his analysis, Weitzman assumes an economy with monopolistic competition on the product market and perfect competition on the labour market. The profit function of the representative firm can then be written as: 3
(1 O)(R(L) w`L).
(2)
Maximizing profits with respect to employment leads to the first order condition that R`(L)
w`
(3)
where R’(L) is the marginal revenue product of labour. If unlimited amounts of labour are available under a sharing regime, the firm will choose to hire workers to the point where the marginal revenue product of labour equals the base wage and not the total wage w as in the fixed wage firm. The profit share is essentially ignored because it acts as a form of “tax” on profits and is hence not relevant in a firm’s maximization procedure.
7
For a full formal explanation refer to Weitzman (1983, 1985, 1987).
24
The Economic Impact of Profit Sharing
Figure 1 illustrates the different incentives for fixed wage and profit sharing firms (adapted from Kruse 1993, Nordhaus 1988). The marginal revenue product (MRP) curve is assumed to be the same in both types of firms and represents the value of worker output from an additional labour hour. A fixed wage firm will hire workers up to the point A where the marginal revenue product equals the fixed wage w, with L1F workers hired. If the firm switched to a sharing system where average pay (AC) is assumed to be the same as under fixed wages, but half of it is paid in a base wage w’ and the other half as a profit share, the marginal compensation curve (MC) slopes downward because labour cost decreases as more employees are hired and the profit share is split among a larger number of employees. The firm thus has an incentive to extend employment up to point B where equation (3) – the equality of the marginal revenue product and the base wage – is fulfilled and the firm wishes to buy L1PS units of labour. While in the short-run equilibrium the employment level under profit sharing is higher than under a fixed wage system, the employment variability is not. In case of an exogenous shock on the product market, the labour demand curve would shift. Since A as well as B are on the same curve, employment would be hit similarly under both systems. The picture changes, however, in the long run where competition on the labour market yields full employment. Under full employment, the firm must operate at point A with L1F workers because it will have problems attracting and retaining workers if
compensation is below the market level. Weitzman (1983) states that all compensation systems are isomorphic in the long run. Point A is therefore the long-run equilibrium of the fixed wage as well as the profit sharing system.
Theoretical Considerations
25
Figure 1- Labour Demand in Equilibrium
w
w wPS
A AC B
w´
MC MRP L1F
L1PS
L
Source: own illustration, adapted from Kruse (1993) and Nordhaus (1988).
A significant difference between both systems arises from the fact that under profit sharing there is excess demand for labour in A amounting to ( L1PS L1F ). This incentive to hire more workers is the driving force in the share economy and has been described by Weitzman vividly as a “vacuum cleaner” in the above quotation. To illustrate the behavioural differences of both systems and hence the key result of this theory – that profit sharing induces greater employment and output stability than a fixed wage system – let us turn to the case when there is reduced demand for the company’s products (Figure 2).
The Economic Impact of Profit Sharing
26
Figure 2- Labour Demand under Negative Demand Shock
w
w
C
A E D
w´
B MRP
MRP L2F
L1F L2PS
L1PS
L
Source: own illustration, adapted from Nordhaus (1988) and Michaelis (1998).
Starting from the long-run equilibrium of both systems, point A, with an excess labour demand of ( L1PS L1F ) under profit sharing, a reduction of the demand for the company’s products shifts the marginal revenue product of labour from MRP to MRP . If the wages remain fixed, the firm will lay off workers and the new reduced
employment level is L2F . A different behaviour is observed in the profit sharing firm. Although the desired employment level falls from L1PS to L2PS , the shock does not lead to lay-offs, because the new desired employment level L2PS is still greater than L1F . Lay-offs will only occur if L2PS is lower than L1F and even if this turns out to be the case, it will be to a lesser extent than under a fixed wage system. Hence, under profit sharing the excess demand for labour but not employment itself is reduced. Adjustment occurs in the profit sharing system through a reduction of wages. As is obvious from equation (1), a negative output shock will reduce the profit sharing component of all employees. Compared to the fixed wage system all employees share the burden, whereas under fixed wages only the dismissed workers are disadvantaged. The excess demand for labour under profit sharing thus acts as a shock absorber in the
Theoretical Considerations
27
short run because employment can be stabilized on the full employment level L1F . In the long run, both compensation systems are again isomorphic and the new long-run equilibrium is in point E. In the case of a positive demand shock, starting from full employment both systems exhibit a similar behaviour. In order to attract additional employees, both types of firms would have to increase pay. Under fixed wages, a positive demand shock would create a temporary excess demand for labour analogously to the permanent excess demand under profit sharing. If the positive demand shock represents the recovery from a negative one, profit sharing firms should hire back fewer workers than fixedwage firms because fewer workers had been dismissed during the downturn (Kruse, 1993). Weitzman’s theory therefore predicts that a profit sharing economy will have more employment stability during negative demand shocks and in periods of recovery from such, whereas under positive shocks the behaviour is similar to a fixed wage economy. As has been alluded to above, the theory has not been free of criticism. In the following, the main objections are presented. First, the entire benefit of profit sharing comes from the fact that the employment of new workers reduces the compensation of old workers. It seems questionable whether workers would stand idly by while the firm tried to employ more workers thus diminishing the profit share of the existing workforce. Rather, they will try to also bargain over employment (Summers, 1986, Meade, 1986a, Layard et al., 1991). Weitzman (1985) admits the lack of a formal theory that determines the compensation parameters. In Weitzman (1987) he picks up on this issue by introducing a bargaining model where unions and employers bargain over base wages but not over the profit share. This modification suffices already to eliminate the excess labour demand and to create unemployment in the long-run equilibrium under profit sharing. The theory also neglects efficiency wage considerations. Levine (1987, 1989) integrates profit sharing into an efficiency wage approach where firms pay fixed wages above the market level in order to induce higher productivity from their workers. Switching from a fixed wage system to profit sharing does not lead to higher
28
The Economic Impact of Profit Sharing
employment, since the existing workforce would respond to the decrease in pay with a disproportionately high productivity decline. This argument has been fully accepted by Weitzman (1988). Once profit sharing is introduced, workers have an incentive to acquire information about the firm which also enables them to negotiate other operations of the firm. This brings us back to the problem of codetermination that has been outlined in section 0 above. The ‘insiders’ might abuse codetermination rights to increase their profit share and employers would be reluctant to invest in new capital since some of the profits would have to be shared with the workers. Weitzman thus pleads for no codetermination rights of employees. Several authors respond that this may not be without consequences on workers’ motivation and productivity (Nuti, 1987, Layard et al., 1991). Finally, Weitzman’s assumption that firms regard base wages as the marginal cost of labour has been criticized. In full employment equilibrium, firms may regard average compensation as the cost of an extra labour hour due to the labour supply constraint. Firms compete for a fixed pool of workers and unless a firm pays the going rate of total compensation, it will lose its workers to other firms. This should eliminate the excess demand for labour (Nuti, 1987, Estrin et al., 1987, Nordhaus, 1988). Two key questions emerge from Weitzman’s theory of a share economy: 1. Is profit sharing associated with higher employment levels? In other words: Do firms view the base wage and not total compensation as the marginal cost of labour? 2. And, do firms with profit sharing experience higher employment stability due to an increased variability in compensation? Several empirical studies test the propositions arising from Weitzman’s theory and are reviewed in the following chapter.
Prior Empirical Studies on the Impact of Profit Sharing A variety of studies ranging from employee and employer studies to comparisons of simple statistics and formal econometric evidence have investigated the impact of
Prior Empirical Studies on the Impact of Profit Sharing
29
profit sharing. While the majority of studies address the relationship between profit sharing and productivity, only limited evidence is available on the effects of profit sharing on employment, employment variability, wages or profitability. The objective of this section is to give a thorough overview of existing research where we restrict ourselves to the presentation of results from formal econometric studies. Before presenting the evidence, however, some general methodological issues should be addressed. The empirical literature on profit sharing has been plagued by several problems which primarily stem from data limitations. Studies are frequently based on small sample sizes or investigate single firms or industries only which raises doubts about the representativeness of the results. Other severe problems arise from the fact that the majority of studies is based on cross-sectional data which causes problems with respect to the endogeneity (or reversed causality) of profit sharing and unobserved heterogeneity which has been claimed to have an important impact on both the choice and performance of an incentive scheme. These issues are closely related to the discussion on the determinants of profit sharing above, where we had argued that conditional on certain circumstances, firms have specific advantages or disadvantages with respect to alternative incentive systems and will introduce the compensation scheme that best suits their needs. This may provide explanation for the rather low diffusion of profit sharing in countries where governments do not intervene into the market process. If firms, however, choose to introduce an adequate payment scheme conditional on specific circumstances and if these factors also have an impact on outcomes, then selection problems pervade the analysis and as Prendergast (1996) puts it: “Controlling for this problem is difficult, but without addressing it, there is little hope that the effect can be truly identified.” As a matter of fact, most empirical studies neglect the problem of selectivity thus casting doubt on the reliability of their estimates. Productivity Recalling the theoretical considerations on the productivity effects of profit sharing, no definite answer could be given a priori. All the more surprising is the fact that the consistency of findings of the numerous empirical studies evaluating the productivity effects of profit sharing is remarkable. The representative survey of the Organization
30
The Economic Impact of Profit Sharing
for Economic Co-Operation and Democracy (1995) summarizes that “Profit sharing is associated with higher productivity levels in every case, regardless of methods, model specification and data used.” Kruse (1993) in his comprehensive review further concludes that out of 265 reported estimations only 8.7 percent take negative values.8 Common to the studies is the use of an objective productivity measure (such as sales or value added per employee) and a measure of profit sharing as an independent variable (dummy variable, absolute value in the respective national currency, or percent of remuneration). Several other independent variables that might also have an influence on productivity are usually included. Most studies use a production function approach to model the determinants of productivity (Hübler, 1995). The purpose of this section is not to give a comprehensive review of all studies that have investigated the productivity effects of profit sharing. This has been done elsewhere (see e.g. Weitzman and Kruse, 1990, OECD, 1995). Rather, selected (longitudinal) studies from several countries are briefly introduced with special emphasis on their approach to address the problem of self-selection of firms.9 An interesting study based on U.S. data has been conducted by Kruse (1993). Kruse uses panel data from 1975-1990 on 253 sharing and 247 non-sharing firms and finds that profit sharing increases sales as well as value-added by more than 4 percent. In order to address selection bias, he uses a variety of techniques such as Heckman’s two-step estimator or the instrumental variable approach which, however, do not substantially change the results. As opposed to the majority of studies, his panel contains not only information on the existence of a profit sharing plan, but also information on when the plan was adopted. Similar results for the U.S. are obtained in studies by Mitchell, Lewin and Lawler (1990) or Shepard (1994). Mitchell, Lewin and
8
The focus of this section is to present studies that evaluate the productivity effects of profit sharing. There exist, however, a number of studies that try to identify the channels through which profit sharing increases productivity such as e.g. its impact on absenteeism and turnover (e.g. Wilson and Peel, 1991, Drago and Garvey, 1998, Brown et al., 1999, Azfar and Danninger, 2001, Heywood and Jijahn, 2004), helping on the job (e.g. Drago and Garvey, 1998, Heywood et al., 2005b) or the reduction of conflict between workers and management (e.g. Heywood et al., 2005a). 9 Different methods that deal with selectivity and their (dis-)advantages are discussed in further detail in chapter 4.
Prior Empirical Studies on the Impact of Profit Sharing
31
Lawler do not account for selectivity bias. Shepard’s study, on the other hand, is based on data from 20 firms in the chemicals industry only, 9 of which had implemented a profit sharing scheme. A recent study by Black and Lynch (2004) that uses data from a unique representative survey of private U.S. establishments using fixed effects estimation finds positive but insignificant results. For the United Kingdom, Wadhwani and Wall (1990) investigate the impact of profit sharing on various outcome variables based on ten-year data of 101 primarily large manufacturing firms where 21 have operated a profit sharing scheme at some time. By means of ordinary least squares and instrumental variable estimation, they find positive effects of profit sharing on productivity. Cable and Wilson (1989) come to a similar conclusion in their analysis of 52 firms in the UK engineering industry. Concerning the magnitude of the productivity effect, both studies arrive at similar results of 3 and 3 to 8 percent respectively which is in line with several other studies. For France, Cahuc and Dormont (1997) apply instrumental variables estimation to a 4-year-panel of manufacturing firms and obtain similar results. So do Pérotin and Fakhfakh (2000). In Germany, a number of studies investigate the productivity effects of profit sharing. First studies by Cable and FitzRoy (1980), FitzRoy and Kraft (1985a, 1987a), Cable and Wilson (1990) and Kraft (1991) come to a quite optimistic assessment. Common to these studies is a small sample size because none of the studies analyses more than 100 firms. The issue of selectivity is first brought up by FitzRoy and Kraft (1995). Based on information for 65 West German metal-working firms, FitzRoy and Kraft apply Heckman’s two-step estimator and show that selectivity is at work which strongly influences their results. Due to the small sample size, however, they describe their results as “illustrative rather than precise measures”.10 With the emergence of the IAB Establishment Panel in the mid-nineties new possibilities opened up. The IAB Establishment Panel is an annual representative survey of establishments employing at least one employee covered by social insurance. Since 1993, several thousand firms have been interviewed on a variety of questions, among them also the financial participation of employees. So far, two
10
For a more comprehensive review of German econometric studies conducted before the year 2000 see Möller (2000).
32
The Economic Impact of Profit Sharing
studies have investigated the productivity effects of financial participation based on the IAB Establishment Panel. Common to both studies is the investigation of the effects of financial participation which encompasses profit sharing as well as employee stock ownership plans. Both incentives may, however, have a differential impact on productivity. The first study was conducted by Möller (2000) based on cross-sectional data for 1998. Despite her findings that firms with financial participation schemes differ significantly from their non-sharing counterparts with respect to several characteristics, the issue of selectivity is ignored. In analogy to most existing studies, Möller finds that the financial participation of employees has a beneficial impact on value added. Wolf and Zwick (2002a) pick up on the issue of selectivity again and follow FitzRoy and Kraft’s (1995) approach methodologically. Using the same data set, they show that controlling for selectivity bias bears important implications: the significance of the positive productivity effect disappears. Besides the lack of distinguishing between profit sharing and employee stock ownership plans, both studies do not exploit the longitudinal structure of the data. Several points therefore appear on the research agenda that will be addressed within this study. Profitability As has been mentioned earlier, most studies focus on the productivity effects of profit sharing. Only few studies explore the impact of profit sharing on profitability and except for one study find a significantly positive influence of profit sharing on the financial performance of firms.11 The studies differ, however, in a number of respects. Several studies are based on cross-sectional data. A first study by FitzRoy and Kraft (1986) uses data on 61 West German metal-working firms and finds quite strong evidence that profit sharing is associated with greater profitability measured as return on capital (ROC). FitzRoy and Kraft also find no feedback from their profitability indicator to sharing in their simultaneous estimates although their essentially crosssectional data preclude them from further untangling the issue of causality. Two other cross-sectional studies are conducted by Blanchflower and Oswald (1988) and McNabb and Whitfield (1998). Both studies use UK data from the Workplace Industrial Relations Survey (WIRS) - although from different years.
11
Blanchflower and Oswald (1988) find a positive but insignificant effect.
Prior Empirical Studies on the Impact of Profit Sharing
33
The major difference of these two studies as compared to all other studies is that the WIRS uses a qualitative, subjective measure for financial performance. Managers are asked how they assess the financial performance of their establishment compared with establishments and firms in the same industry: (i) better than average; (ii) about average; (iii) below average. Although such subjective measures of firm performance provide only reduced information relative to a continuous profit measure and have been criticized for their subjective nature, research has shown that they correlate positively with objective measures thus validating their use (Dess and Robinson, 1984, Venkatraman and Ramanujam, 1987, Wall et al., 2004). Both studies neglect the issues of causality and selectivity. Moreover, a limitation of McNabb and Whitfield’s study is their summary of a financial participation variable which includes not only profit sharing but also share option schemes that are not necessarily profit related. Finally, Mitchell et al. (1990), Bhargava (1994) and Zhuang and Xu (1996) investigate the question using longitudinal data. Mitchell et al. for his study of 495 U.S. business units between 1983 and 1986 finds positive and significant effects of profit sharing on return on investment (ROI) and return on assets (ROA). Bhargava’s study is based on a seven-year panel of 114 large UK (quoted) profit sharing companies that are predominantly from the manufacturing, construction and retailing sector. Using a GMM approach proposed by Arellano and Bond (1991), he finds strong support for a positive impact of profit sharing on profitability measured as return on capital. So do Zhuang and Xu who methodologically follow Bhargava’s approach. Contrary to all other studies though, Zhuang and Xu use data on 480 stateowned and not private enterprises in China between 1986 and 1991 to investigate the effects of the introduction of incentives in employee remuneration as one of the centerpieces of economic reforms in the Chinese state sector. Wages Several studies explore the key propositions emerging from Weitzman’s theory, i.e. whether profit sharing implies lower base wages and whether this results in higher employment levels and higher employment stability. We will in turn review studies that deal with the respective issue. It is important to recall here that we report only firm-level evidence that does not allow to draw conclusions about aggregate levels.
34
The Economic Impact of Profit Sharing
For Germany, to the best of our knowledge only Hart and Huebler (1991) investigate the impact of profit sharing on wages. Based on cross-sectional individual-level data from the German Socio-Economic Panel (GSOEP), Hart and Hübler (1991) apply the Heckman procedure to account for selectivity and find that profit shares act as a complement to wages. The authors point out, however, that their results have to be interpreted cautiously since they are likely to suffer from bias due to omitted firmspecific variables. For the UK, Wadhwani and Wall (1990) and Bhargava and Jenkinson (1995) arrive at a similar conclusion. While Wadhwani and Wall use ten-year data of 101 manufacturing firms where 21 have operated a profit sharing scheme at some time, Bhargava and Jenkinson’s study is based on a ten-year panel of 114 large UK (quoted) profit sharing companies that are predominantly from the manufacturing, construction and retailing sector. Both studies apply instrumental variable techniques to address the issue of selectivity. Studies from the United States, however, find contradictory results. Bell and Neumark (1993) in their study on 204 unionised firms between 1978 and 1987 (17 of which had a profit sharing plan) find that firms in which a profit sharing plan was negotiated had a significantly lower labour cost growth than firms that had not introduced such a plan. This result remains unchanged if firm-specific fixed effects are included. In a similar vein, Black and Lynch (2000) find that profit sharing in U.S. firms leads to lower hourly wages, specifically for technical and clerical/ sales workers. Overall, it is hence not clear whether profit sharing acts as an add-on to regular wages or rather as a substitute as proposed by Weitzman. Employment Levels A similar inconclusive pattern emerges when it comes to the impact of profit sharing on employment levels. For the United Kingdom, Wadhwani and Wall (1990) in their above mentioned study of primarily large UK manufacturing firms find little evidence that profit sharing affects the employment decision. This is not surprising since their earlier reported results on the effects of profit sharing indicated that profit sharing is a supplementary payment that unfolds productivity-enhancing effects which might hint
Prior Empirical Studies on the Impact of Profit Sharing
35
at an introduction of profit sharing as a wage incentive rather than as a measure to increase wage flexibility. Bradley and Estrin (1992) in a case study compare the John Lewis Partnership, a British firm in the retail sector, with four comparable non-sharing “first division” firms between 1970 and 1987. In their simultaneous estimation of a labour demand function with a remuneration equation, they find evidence for higher employment in the firm under consideration. This effect holds, however, for the John Lewis Partnership as against all competitors taken together, but does not hold against each of them separately: in two out of four cases the employment effect is not statistically significant. Caution should therefore be applied in the interpretation of the results. Jones and Pliskin (1989) based on unbalanced UK panel data for 127 firms from 1890 to 1975 also reach rather inconclusive results since the employment effect of profit sharing is sensitive to the measure of profit sharing, the modelling of the dynamics and inclusion of a variable for worker participation in decision making. For the United States, the results of Bell and Neumark (1993) from their fixed effects estimations tend to support the proposition of higher employment levels, but are at best marginally significant. Interestingly, Bell and Neumark also find a significantly lower labour cost growth in sharing firms. This may point to employment effects as proposed by Weitzman and stands in contrast to the results by Wadhwani and Wall (1990) for the United Kingdom. Kraft (1992) investigates the impact of profit sharing on employment growth between 1977 and 1979 for a sample of 56 metal-processing West-German firms. His estimation results indicate no significant impact of profit sharing on employment, a result that is also shared by Cahuc and Dormont (1997) for their panel of French firms. Since Cahuc and Dormont and Kraft (using the same data sets) also find productivity-enhancing effects of profit sharing, their results tend to support Wadhwani and Wall’s results.
36
The Economic Impact of Profit Sharing
Employment Stability Several studies also address the question whether profit sharing is associated with lower employment variability. The majority of studies is based on U.S. data where the evidence seems to lean in the direction of supporting the proposition that profit sharing reduces employment variability. Bell and Neumark (1993) find weak support for a higher employment stability in profit sharing firms which fits into the above mentioned overall picture they draw about profit sharing which seems to lean in the direction of supporting the predictions of Weitzman’s theory. Chelius and Smith (1990) investigate the effects of profit sharing on employment stability in the face of negative demand shocks with two cross-sectional data sets separated in time by a decade: on the one hand with data from an employer survey totalling 3988 firms of which 33% had some form of profit sharing scheme and, on the other hand, by exploring the effects of profit sharing on the probability of being laid off using household data. In both data sets there is weak support for the hypothesis that employees that are partially paid in profit shares have greater security in the face of downturns. Kruse (1991) estimates employment growth equations for 2976 publicly traded companies (1198 of which were profit sharing firms) over the period from 1971 to 1985 and finds that manufacturing firms with profit sharing have smaller employment decreases during downturns than their non-sharing counterparts (2% versus 3.1% decrease in employment). No significant differences were found, however, between sharing and non-sharing firms in the non-manufacturing sector. In a follow-up study, Kruse (1993) explores longitudinal data on 500 firms with public stock, approximately half of them with profit sharing matched against firms from the same industry without profit sharing. The results indicate that firms with profit sharing had stronger employment responses to positive shocks and smaller employment responses to negative shocks than similar non-sharing firms thus indicating higher employment stability in profit sharing firms.
Prior Empirical Studies on the Impact of Profit Sharing
37
Only Florkowski (1994) reports no such effects in his panel of 443 publicly traded profit sharing firms which he compares before and after the introduction of profit sharing with average values of non-sharing firms within the same industry. For other countries, only limited evidence is available. For Germany, Kraft (1991) using the above introduced data base applies a Heckman approach to account for selectivity and finds significantly lower dismissal rates in profit sharing firms. Wadhwani and Wall (1990) for the United Kingdom, however, find little support for the proposition. Indeed, employment in profit sharing firms in their sample seems to respond more (although insignificantly) and not less to demand shocks. Summary Although quite a number of studies have empirically investigated the effects of profit sharing, the picture that emerges is anything but conclusive. The majority of studies examine the impact of profit sharing on productivity. Surprisingly, the results are very clear cut indicating that profit sharing is associated with higher productivity levels. In contrast to the extensive investigation of profit sharing on productivity, however, only limited evidence is available on how profit sharing affects other outcomes. While the results on the profitability effects of profit sharing also seem to lean in the direction of a positive effect, an inconclusive pattern emerges for all other above mentioned outcome variables: it is not clear whether profit sharing is associated with lower base wages, higher employment levels and increased employment stability as proposed by Weitzman. Furthermore, most studies look at the effects of profit sharing on single outcome variables and miss providing a “broader picture” of profit sharing to see how different effects are interrelated: Does profit sharing really increase the motivation and productivity of employees or are increased productivity and profitability only the result of lower wages or slower wage growth? What is the relationship between wages and profit sharing anyway? Do lower base wages imply higher employment levels and employment stability as proposed by Weitzman? And would such a substitute for fixed wages really have an impact on productivity? Or is a supplementary concept of profit sharing necessary to induce employees to “work harder and smarter” which could in turn also have beneficial effects on employment levels? What is the impact of industrial relations with respect to the implementation as well as the success of profit
38
The Economic Impact of Profit Sharing
sharing? Despite the overall number of studies, therefore, several points appear on the research agenda. Finally, part of the inconclusiveness of the results can certainly be attributed to several methodological problems which mainly arise from data limitations. The preponderance of cross-sectional designs raises concerns about the issues of endogeneity and unobserved heterogeneity. Small sample sizes and data on single or primarily large firms, specific industries or worker cooperatives question the results with respect to their representativeness. Most studies also do not observe the introduction of a profit sharing scheme but rather its existence which aggravates the determination of the introductory effects of such an incentive scheme.
Data Set and Econometric Methods The review of the literature that has been presented so far should have made clear that research on the effects of profit sharing is far from being complete and that a number of issues remain to be resolved. The objective of this study is to contribute to the literature in several respects. Using nationally representative German firm panel data, we try to provide a comprehensive picture of profit sharing by investigating four questions: First, what are the output, employment and productivity effects of profit sharing? Second, does profit sharing imply lower base wages as proposed by Weitzman or is it rather paid as an add-on to regular wages? Third, what is the impact of profit sharing on the firms’ financial performance? And finally, fourth, how does the existence of a works council affect the performance of profit sharing? Owing to the availability of extensive longitudinal data, we are further able to handle the majority of methodological problems described. Before presenting the empirical evidence in the following chapter, the data set is briefly introduced followed by a discussion of econometric methods that deal with the issue of selectivity where we base our choice of estimator upon.
Data Set and First Descriptive Results “Selection problems pervade economic analysis, as individuals frequently choose the treatments that they undergo (…) Controlling for this problem is difficult, but without addressing it, there is little hope that the effect can be truly identified (…)” (Prendergast, 1996) This study uses establishment-level data from the German IAB Establishment Panel of the Institute for Employment Research (Institut für Arbeitsmarkt- und Berufsforschung, IAB) of the Federal Labour Office (Bundesagentur für Arbeit). The IAB Establishment Panel is an annual representative survey of establishments
40
Data Set and Econometric Methods
employing at least one employee covered by social insurance. Each year since 1993 (1996) several thousand firms from 16 industries and 10 size classes in WestGermany (East-Germany) have been surveyed. Data are extended regularly to correct for panel mortality, exits or the foundation of new firms thus yielding an unbalanced panel. The interviews are conducted by professionals from the market research group “Infratest Sozialforschung” in personal interviews with owners or senior managers of the establishments. Considering that participation is voluntary, a response rate exceeding 70 percent can be regarded as quite high (Bellmann, 1997, Kölling, 2000). As the panel was created for the needs of the Federal Labour Office, detailed information is given about the demand side of the labour market. This encompasses questions on the number and qualification structure of employees, changes in employment, training, apprenticeship and working time which are included every single year. Also included annually is general information on the establishment as well as information on turnover, wages and profitability. Besides this “core of questions”, specific other topics are covered only irregularly. Information on profit sharing is provided for the years 1998, 2000 and 2001. Due to the strict data confidentiality laws in Germany, external researchers can currently access the data only either directly locally at the Institute for Employment Research (IAB) in Nuremberg, Germany, or indirectly by sending statistical programs to a unit within the IAB which has explicitly been installed to serve the needs of researchers working outside of the Federal Labour Office. Turning to our empirical study, we had argued throughout the work so far that firms will introduce profit sharing if they consider this incentive scheme to be advantageous. Company characteristics such as firm size, the organizational structure, but also industrial relations and the market environment have been found to play a significant role in a firm’s choice of remuneration scheme. To investigate whether this is confirmed for the firms in the IAB Establishment Panel, a simple comparison of profit sharing and non-profit sharing firms in the cross-section may be helpful. In the interview, firms are asked whether they have an existing profit sharing scheme. We created a dummy taking the value of one if firms answered this question with ‘yes’ and zero if they responded with ‘no’.
Data Set and First Descriptive Results
41
Table 2 depicts the results of a mean comparison of both groups of firms in the year 2000.12 The sample includes 10.225 establishments with 1.170 firms reporting to have a profit sharing scheme.13 This is somewhat more than 11 percent of the sample and hence roughly coincides with earlier values from the European Commission and projections of the IAB of 12 and 9 percent respectively.
Table 2 - Mean Comparison of Firms with and without Profit Sharing, 2000 Variable Number of employees Establishment size < 20 Establishment size 20-49 Establishment size 50-249 Establishment size 250-499 Establishment size 500+ Exporter Share qualified employees Collective bargaining Works Council Limited liability Age East German establishment Shift responsibilities Teamwork Profit/Cost Center ICT Investment
Firms with profit sharing (N=1170) 365 0.24 0.21 0.30 0.10 0.15 0.41 0.66 0.62 0.51 0.76 0.31 0.24 0.31 0.18 0.19 0.79
Firms without profit sharing (N=9055) 115*** 0.51*** 0.19 0.20*** 0.05*** 0.05*** 0.21*** 0.59*** 0.50*** 0.28*** 0.53*** 0.39*** 0.42*** 0.17*** 0.10*** 0.08*** 0.55***
*/**/*** indicate that means differ with statistical significance in a two-tailed t-test at the 10%, 5%, or 1% level between the establishments from the two groups. Source: IAB Establishment Panel, wave 2000, own calculations
Several approximations for the four groups of variables are included. Profit sharing firms are on average significantly larger than non-sharing firms. Looking at different size classes, the picture becomes even clearer since the share of medium-sized and large firms is much higher among the profit sharing firms than among their nonsharing counterparts. We had also argued that profit sharing might be preferred in organizations where monitoring costs are high due to a complex and interdependent work organization, where individual performance is hard to observe and cooperation plays an important role. In such a demanding environment one would expect a higher qualification level of the workforce, innovative work practices and high investment into information and 12 13
For more information on the definition of variables see Table A.1 in Appendix A. Firms with missing values for one of the included variables in Table 2 have been excluded.
42
Data Set and Econometric Methods
communication technology. We therefore include the ratio of qualified employees and dummies indicating the shift of responsibilities and decisions to lower levels of hierarchy, the introduction of team work and cost or profit centers as well as a dummy indicating investment into information and communication technology into our analysis.14 The results of the mean comparison between both groups confirm the presumptions. Profit sharing establishments have a significantly higher share of qualified staff and seem to be more innovative concerning their organizational structure. Furthermore, the share of firms investing into information and communication technology is significantly higher among the profit sharing firms. Concerning the industrial relations variables, we find that a significantly larger share of establishments has implemented a works council and is covered by collective bargaining. One interpretation could be that the finding is contradictory to the arguments particularly in the U.S. literature, stating that employers introduce profit sharing to undermine the unions’ position. The evidence may, however, also reflect a size effect only, since larger establishments are both more likely to have a works council and to be covered by a collective agreement. We had argued above that firms have a greater incentive to set up flexible remuneration systems in order to transfer some of the risk to their employees when profits are unstable and risky. We assume that exporting firms have a higher propensity to introduce profit sharing which is supported by our results. The share of exporters is significantly higher in the group of profit sharing firms. Contrary to our presumption that younger firms should be more prone to introducing profit sharing, the share of firms founded after 1990 is significantly lower among the sharing firms. In analogy to Möller (2000), the results indicate that firms located in East Germany are less likely to introduce profit sharing. Profit sharing is more frequently found in companies with limited liability, i.e. joint stock companies (AG) and non-public limited liability companies (GmbH). Concerning the incidence of profit sharing in
14
The question regarding organizational changes refers to the introduction of the measures in the last two years and not the existence of such work practices in general. Although this may seem somewhat problematic, unfortunately we are not able to improve on this issue since the question has been posed only irregularly. The inclusion or exclusion of the three variables had, however, no significant influence on the results in the following chapters.
Microeconometric Evaluation Methods
43
specific industries, Table A.2 in Appendix A provides evidence. Profit sharing seems to be more pronounced in manufacturing. Among services, specifically firms dealing with data processing and retail often have implemented profit sharing schemes. Expectedly, profit sharing is rather uncommon in agriculture and the non-profit sector, but also in the health sector, hotels and restaurants and the construction industry. Summarizing, our initial conjecture that firms self-select into profit sharing schemes is confirmed. A simple comparison of firms with and without profit sharing that does not take account of this selectivity effect would thus be misleading. In fact, the majority of studies ignore this issue. Before presenting new evidence based on data from the IAB Establishment Panel, the following section briefly introduces the problem of selectivity and presents various estimators that deal with this problem.
Microeconometric Evaluation Methods The aim of our analysis is to assess the impact of an introduction of profit sharing on several outcome variables as compared to the hypothetical situation of nonintroduction and is thus a situation typical for an evaluation. The standard framework in evaluation analysis to approach such a problem goes back to Roy (1951) and Rubin (1974) and is also known as the Roy-Rubin-model or potential outcome approach. In the potential-outcome framework, each individual has two potential responses (Y0i, Y1i) for a treatment. Y1i is the outcome if individual i is treated, and Y0i is the outcome if individual i is not treated. According to Rubin (1974), the causal effect of a treatment for individual i is defined as the difference between both outcomes: ¨ i = Y1i – Y0i
(4)
The so-called fundamental evaluation problem arises from the fact that at any time, individuals can be in either one of two potential states but not in both. The fundamental evaluation problem – which can be interpreted as a missing data problem – can be solved if an adequate approximation for the unobserved counterfactual outcome is found (Heckman, Ichimura, Todd, 1997).
44
Data Set and Econometric Methods
Since individual treatment effects cannot be estimated, one has to concentrate on average treatment effects. The parameter that received the most attention in the evaluation literature is the average treatment effect on the treated (ATT) which is defined as: ¨ TT = E(Y1 – Y0 |D=1) = E(Y1 |D=1) – E(Y0|D=1)
(5)
where D is a binary assignment indicator determining whether establishment i is treated (D=1) or not (D=0). Another parameter of interest is the average treatment effect for the entire population (ATE), i.e. the impact of treatment on an individual drawn randomly from the population: ¨ ATE = E(Y1 – Y0)= E(Y1) – E(Y0)
(6)
If we assume homogeneous responses, the ATT and ATE are identical. In a more realistic scenario, however, responses are heterogeneous in the sense that they vary across individuals and the measures can differ. In general, both estimates are of interest. While the treatment effect on the treated indicates the average benefit of participation, the average treatment effect could be of relevance if e.g. the policy interest were focused on making a voluntary program compulsory (Bryson, Dorsett, Purdon, 2002).15 Since firms voluntarily decide whether to introduce profit sharing or not and will do so if the advantages from profit sharing outweigh the costs of implementation, the primary parameter of interest in this study is the average treatment effect on the treated. Obviously, focusing on expected values does not help in overcoming the fundamental evaluation problem and an adequate approximation for the second term on the right hand side of equation (4.2) is still needed. If the condition E(Y0|D=1) = E(Y0|D=0) held, the average outcome of firms that haven’t introduced profit sharing might serve as an estimator for E(Y0|D=1). The literature on the determinants of an introduction of
15
For a discussion on different estimands see also Heckman, LaLonde, Smith (1999) or Imbens (2004).
Microeconometric Evaluation Methods
45
profit sharing emphasizes, however, that numerous observable characteristics influence the decision to introduce profit sharing. This has also been confirmed for our sample in Table 2 above. Several other authors argue that unobservable characteristics such as management quality, better industrial relations or a more active human resources department also play a role concerning both the choice of a human resource strategy as well as outcomes (Huselid, 1995, Wolf and Zwick, 2002b). The decision to introduce profit sharing is thus not random but governed by a sample selection process. There are observable as well as unobservable factors which might have a common impact on both the treatment and the outcome decision and therefore the following independence assumption which is necessary to approximate any treatment effect does not hold: Y(D=0), Y(D=1) A D
(7)
Not accounting for this selection mechanism will lead to a biased estimate of the impact of profit sharing. Several alternative approaches exist that take explicit account of the selection problem. They can be classified broadly into experimental and nonexperimental methods and are described briefly in the following. The central issue that all these methods address is the construction of the unobserved counterfactual. Experimental Evaluation Methods The key concept of experiments is random assignment. Individuals who would otherwise participate e.g. in a social program are randomly assigned to one of two groups: the treatment group which participates in the program and the control group which is excluded from it. If carried out properly, random assignment creates a control group with the same distribution of observed and unobserved characteristics as the treatment group. Since treatment and outcomes are independent and equation (7) holds, a simple comparison of mean outcomes of both groups can be used to produce a consistent estimate of the impact of a program on its participants (Smith, 2000, Blundell and Costa Dias, 2002).
46
Data Set and Econometric Methods
Although experiments are widely viewed as the most convincing method of evaluation, they also have their own limitations and problems.16 Experiments are often costly and require close monitoring in order to ensure effective administration. Ethical issues might arise as potential participants are being denied treatment. This may increase the chances of non-cooperation on behalf of those responsible for delivery of the program. Finally, there are practical problems that can bias the estimates. Randomisation bias can arise because the types of persons participating in a program differ from those that would participate without randomisation. The interpretation of experimental estimates is also complicated if treatment group members drop out of the experiment before receiving treatment or if experimental control group members can participate in close substitutes for the experimental treatment. The result of all these problems is that evaluators use non-experimental methods to analyse the data. This is a second-best solution, however, since experiments are designed specifically to avoid this sort of adjustment (Smith, 2000, Bryson et al., 2002). Non-Experimental Evaluation Methods While experiments can provide a direct estimate of E(Y0|D=1), they are rather scarce in economics and researchers often have non-experimental data at their disposition. Since this study is also based on non-experimental data, stronger emphasis in this chapter will be put on the presentation of non-experimental evaluation methods. The standard approaches to the problem of selectivity are the parametric instrumental variables and Heckman estimator which have been applied by a few studies investigating the effects of profit sharing. If longitudinal or repeated cross-section data are available, non-parametric or semi-parametric approaches which are less restrictive can be applied due to the relative richness of information. Each approach invokes different identifying assumptions to construct the counterfactual outcome. In the following, the various estimators are briefly outlined and discussed. The Instrumental Variables (IV) Estimator
The underlying identification strategy of the IV estimator is to find a variable Z which is related to the treatment but has no influence on the outcome equation. The variable 16
The literature on the advantages of experimental data was developed by Bassi (1983,1984) and Hausman and Wise (1985).
Microeconometric Evaluation Methods
47
Z thus affects the outcome only indirectly through the treatment and hence causal effects can be identified through a variation in this variable. More precisely, the regressor Z has to satisfy the following conditions (Blundell and Costa Dias, 2002): 1. Z determines the treatment and thus has a non-zero coefficient in the treatment decision. 2. A transformation of Z exists which is uncorrelated with the error term in the outcome equation, given the exogenous variables X. 3. Z is not completely determined by X. This variable is called an “instrument” and introduces an element of randomness into the assignment rule which approximates the effect of an experiment. If these conditions are met, the standard IV approach can be applied by replacing the treatment indicator D by the transformation of Z and running a regression. Alternatively, Z and X could be used to predict the treatment. This newly built variable Dˆ is then used in the regression instead of D (Heckman and Navarro-Lozano, 2004). Although a very simple estimator, the IV approach suffers from one main drawback: the instrument choice. It is often difficult to think of a variable that satisfies all three conditions above.17 Blundell and Costa Dias (2002) further point out that in a heterogeneous framework the fundamental hypothesis of the IV estimator of no correlation between the instrument Z and the error term of the outcome equation is violated hence invalidating the application of such an approach. The Heckman Selection Estimator
An approach that has been used widely to address the problem of selectivity is the Heckman selection estimator. The basic idea of this estimator is to control directly for that part of the error term in the outcome equation that is correlated with the treatment dummy variable. Heckman (1976, 1979) suggests a two-step procedure: In a first step, the part of the error term of the outcome equation that is correlated with the treatment is estimated. In the second step, this term is included into the outcome equation and
17
Bound, Jaeger and Baker (1995) further emphasize that using “weak instruments” may lead to inefficiency and also inconsistency of the IV estimates.
Data Set and Econometric Methods
48
the unbiased treatment effect can be obtained. To explain this estimator, the outcome equation is rewritten in linear terms as Yi = Xiȕ0 + ĮDi +Ui.
(8)
We also assume that the treatment decision can be parametrised in the following way. For each individual there is an index IN which depends on a set of variables Z and parameters J, for which treatment D occurs if this index rises above zero, i.e. INi = ZiJ + Vi
(9)
where Vi is the error term and Di=1 if INi > 0 and Di=0 otherwise. If we assume – as commonly done – that the error terms of the outcome and treatment equation Ui and Vi have a bivariate normal distribution, the conditional outcome expectations can be written as: I(Zi J ) ) (Zi J )
(10)
I(Zi J ) , 1 ) (Zi J )
(11)
E(Yi|Di=1) = Xiȕ + Į + ȡ and E(Yi|Di=0) = Xiȕ - ȡ
where I (.) and )(.) denote the density and cumulative density functions of the standard normal distribution respectively. The last term on the right hand side of both equations denotes the expected value of the error term conditional on the treatment variable. This new regressor deals with the fraction of the error term that is correlated with the treatment decision. Including it into the outcome equation enables us to separate the true impact of treatment from the selection process. Heckman thus characterises the sample selection model as a special case of the omitted variable problem with the new regressor being the omitted variable (Puhani, 2000, Caliendo und Hujer, 2006). Although Heckman himself recommended his approach for “exploratory empirical work”, his estimator has become a standard way to obtain final estimation results. This has been strongly criticized for various reasons. Blundell and Costa Dias (2002)
Microeconometric Evaluation Methods
49
stress that this method is more robust than the instrumental variables approach, but is also more demanding on assumptions about the structure of the model. In case independent variables are the same in both the outcome and selection equation (no exclusion restriction), the standard Heckman selection model is identified solely on its distributional assumptions. Instruments similar to the IV approach are thus needed that have an impact on treatment but do not appear in the outcome equation. As has been argued in the section above, it is quite hard in practice to find such variables. Another line of criticism emphasizes the sensitivity of the estimated coefficients with respect to the distributional assumptions imposed on the error terms in both equations (Little and Rubin, 1987, Puhani, 2000). Before-After Estimator
A common method in the empirical evaluation literature is the before-after estimator which compares a person with himself/herself. This strategy is based on the idea that individuals can be in both states at different times. The outcome variable of treated individuals before the treatment can thus serve as a proxy for the no-treatment case. The identifying assumption of the before-after estimator is: E(Y0t|D=1) = E(Y0t’|D=1)
(12)
where time is incorporated into the analysis through t and t’ indicating post- and pretreatment periods respectively. If this assumption holds, the before-after estimator is given by ¨ BA = E(Y1t – Y0t’ |D=1).
(13)
The before-after estimator does not necessarily require longitudinal data. As long as the identifying assumption holds, repeated cross-sections are sufficient to construct a before-after estimate. An advantage of this approach is that it only requires information on the participants and their pre-treatment outcome to evaluate the average treatment effect on the treated. A major drawback of this estimator, however, is its reliance on the assumption that the participants’ mean outcome in the notreatment case is the same before and after the treatment, i.e. in t and t’. This assumption excludes changes in the overall state of the economy or changes in the life
50
Data Set and Econometric Methods
cycle position of participants. Changing economic conditions, however, are likely to have an influence on the impact of various treatments and might also influence the introduction of profit sharing. There might also be cases where treatments have an impact on pre-treatment outcomes thus resulting in a violation of the identifying assumption as shown by Ashenfelter (1978). Ashenfelter observed that participants experience a decline in earnings before the start of the training program. This phenomenon which has become known as Ashenfelter’s Dip was confirmed by later research. If this decline in earnings is only temporary, then the before-after estimator will overstate the average treatment effect on the treated. Another problem of the before-after estimator is that any trend in outcomes which is due to macro or lifecycle factors is attributed to the treatment (for a discussion see Ashenfelter, 1978, Heckman, LaLonde, Smith, 1999, or Smith and Todd, 2005). Difference-in-Differences
The difference-in-differences estimator can be regarded as an extension of the beforeafter estimator. Whereas the before-after estimator compares the outcomes of participants after treatment with their outcomes before treatment, this estimator subtracts the before-after change of non-participants from the before-after change of participants – hence the name “difference-in-differences”: ¨ DiD = E(Y1t – Y0t’ |D=1) - E(Y0t – Y0t’ |D=0)
(14)
The identifying assumption of the difference-in-differences estimator states that the average change in the no-treatment outcome measure is the same for participants and non-participants: E(Y0t – Y0t’ |D=1) = E(Y0t – Y0t’ |D=0)
(15)
The difference-in-differences estimator is more data demanding than the before-after estimator since it additionally requires pre- and post-treatment information on nonparticipants. The identifying assumption, however, is more plausible as it can cope not only with time-invariant individual specific effects but also with changes in the overall state of the economy or changes in the lifecycle position that affect both groups similarly. Although this estimator improves on the defects of the before-after
Microeconometric Evaluation Methods
51
estimator, it does not eliminate the problem of Ashenfelter’s Dip because of the transitory nature of the latter (Heckman, LaLonde, Smith, 1999, Blundell and Costa Dias, 2002). Matching
Both the before-after and difference-in-differences estimator are primarily concerned with selection on unobservables. Matching methods, however, which have gained a lot of popularity recently focus on selection due to observable factors. The popularity of matching can be ascribed to its intuitively attractive idea to solve the fundamental evaluation problem in equation (4) by constructing a control group which is as similar as possible to the treatment group with respect to all relevant pre-treatment characteristics X. The difference in mean outcomes between both groups can then be attributed to the treatment. In a way, matching thus tries to mimic an experiment (Heckman, Ichimura and Todd, 1998). In order to solve the fundamental evaluation problem, Rubin (1977) introduces the conditional independence assumption (CIA) which states that treatment status and potential outcome are independent for individuals with the same set of observable characteristics X: Y(D=0), Y(D=1) A D | X
(16)
It is also assumed that for all X there is a positive probability of either participating (D=1) or not participating (D=0), i.e. 0 < Pr(D=1|X) < 1
(17)
which implies that a match can be found for all treated individuals (common support condition). If both assumptions are satisfied, then the matched control group’s mean outcome can serve as an approximation for the unobserved counterfactual in equation (5): E(Y0|D=0, X) = E(Y0|D=1, X)
(18)
52
Data Set and Econometric Methods
and the causal effect can be estimated using the means of the two groups (Lechner, 1998): ¨ MAT = E(Y1| D=1, X) – E(Y0| D=0, X)
(19)
Practically, ensuring the validity of the CIA imposes a very high obstacle since with every additional exogenous variable in the vector X the probability of finding an adequate control group decreases. Fortunately, Rosenbaum and Rubin (1983) were able to find a remedy to this dimensionality problem by introducing the concept of propensity score matching. Instead of matching on a large set of covariates, they propose to match on the propensity score, i.e. the probability to receive treatment given a set of individual characteristics of a firm pr(D=1|X). They are able to show that if the CIA is fulfilled one need not condition on all covariates contained in X, but only on the propensity score. Aside from the CIA, another assumption that is required for the identification of the causal effect is the so-called SUTVA assumption. SUTVA means “stable unit treatment value assumption” and says that the individual treatment effect may not be influenced by the treatment status of other units (Rubin, 1980, 1986, Holland, 1986). Considering that profit sharing is one among a number of different incentive schemes that are primarily introduced to increase workers’ efficiency and that, moreover, has been introduced by a minority of firms only, we assume the SUTVA to hold. Several practical issues are involved in the implementation of matching. These will be discussed in more depth in chapter 5 where the empirical study is presented. A drawback of the matching estimator is that matching assumes the existence of a sample selection process due to observable factors only. Unobservable factors, however, can play an important role with respect to many treatments and have also been emphasized in the profit sharing literature. The success of the matching estimator also depends on the availability of extensive data. The CIA requires that all observable characteristics that influence the outcome and treatment decision simultaneously are known and available. Although the CIA cannot be tested formally, making this assumption plausible in practice requires access to very rich data. The reader may wonder what the advantage of matching over simple regressions is. After all, running a regression of outcomes on a treatment indicator and the covariates
Microeconometric Evaluation Methods
53
X produces an impact estimate that conditions on X. Three important points should be highlighted. First, matching is non-parametric and thus avoids the functional form restrictions implicit in the linear regression model. Dehejia and Wahba (2002) show that this can be important in reducing the bias. Secondly, matching emphasizes the “common support problem”. Imposing the common support implies the identification and elimination of poor matches among the non-treated. The idea is to ensure that any combination of characteristics seen among those in the treated group may also be observed among the control group members. Finally, Lechner (2002) notes that nonparametric matching leaves the individual causal effect unrestricted thus allowing for heterogeneous individual effects. This is not true for the regression approach, although at times it might provide close approximations of the average treatment effect on the treated (Angrist, 1998, Blundell, Dearden, Sianesi, 2004). Conditional Difference-in-Differences
The CIA is quite strong once it is acknowledged that unobserved factors may play a role. A difference-in-differences matching strategy - also denoted in the literature as conditional or matched difference-in-differences - combines the advantages of both the matching and difference-in-differences approach by allowing for temporally invariant differences in outcomes between the treated and the control group (Heckman, Ichimura, Todd, 1997). The identifying assumption is now stated in terms of the before-after evolution instead of levels and the conditional difference-indifferences estimator requires that E(Y0t – Y0t’ |D=1, X) = E(Y0t – Y0t’ |D=0, X)
(20)
This implies that controls have evolved from a pre- to a post-treatment period in the same way the treated group would have done had it not been treated. The treatment effect on the treated can then be estimated according to: ¨ CDiD = E(Y1t – Y0t’ |D=1, X) - E(Y0t – Y0t’ |D=0, X)
(21)
Although conceptually quite appealing, the conditional difference-in-differences estimator is clearly very data demanding. It requires not only pre- and post-treatment
54
Data Set and Econometric Methods
information on the outcomes of both groups, but also rich information about exogenous characteristics of both groups. Which Estimator?
The previous sections have outlined the main estimators dealing with the problem of selectivity and the question thus arises which estimator to choose. This question cannot be answered unambiguously and a lot depends on the data available (Heckman et al., 1997). Heckman’s two-step estimator has been criticized heavily for its restrictive distributional assumptions. A common drawback of the Heckman and the IV estimator is further the need of an instrument that determines treatment but has no impact on the outcome equation. Clearly, finding a plausible instrument in practice imposes a problem. Blundell and Costa Dias (2002) further point out that an application of the IV estimator in a heterogeneous framework is invalidated because its fundamental hypothesis of no correlation between the instrument and the error term in the outcome equation is violated. If rich longitudinal or repeated cross-section information is available, less restrictive non-parametric or semi-parametric estimators might be preferable. The matching estimator has gained a lot of popularity recently. As opposed to the before-after estimator and the difference-in-differences approach which have a strong focus on selection due to unobservables, matching assumes that selection is solely due to observable factors. This seems to be a quite restrictive assumption. Combining matching with differences-in-differences, however, attenuates this problem. In light of the difficulty to choose the right non-experimental evaluation strategy, a literature has arisen, starting with LaLonde (1986), that evaluates non-experimental methods by using experimental data as a benchmark. Much of this literature implicitly tries to find the non-experimental estimator that will always solve the selection bias problem inherent in non-experimental studies. LaLonde concluded that standard nonexperimental estimators are either inaccurate relative to the experimental benchmark or sensitive to the specification used. His results renewed the debate on experimental versus non-experimental evaluations but also spurred the search for alternative
Microeconometric Evaluation Methods
55
estimators. Two recent contributions by Dehejia and Wahba (1999, 2002) have drawn attention to the matching estimator. They apply matching to the same experimental data analysed by LaLonde and find very low biases. Their findings have contributed substantially to the popularity of matching since the late 1990s. Other studies do not confirm these results (e.g. Heckman, Ichimura, Smith and Todd, 1998, Smith and Todd, 2005). Several recent studies indicate that combining matching with differencein-differences proves to be a very effective tool in controlling for both selection on observables and unobservables, and performs better than the matching or differencein-differences estimator (Heckman, Ichimura, Smith and Todd, 1998, Blundell and Costa Dias, 2002, Smith and Todd, 2005). In light of these results, the conditional difference-in-differences estimator seems to be a reasonable choice. Since the IAB Establishment Panel is a rich and extensive data set and information before as well as after the introduction of profit sharing is available for both groups of firms, we believe that the hard requirements imposed on the data can be met.
The Output, Employment and Productivity Effects of Profit Sharing Theoretical Considerations In the theoretical section on the relationship between profit sharing and productivity, we had argued that the underlying idea of an introduction of profit sharing is the employer’s intention to pass on part of the profits to the employees in order to align their interests with her own. The expected result is an increased effort and willingness to cooperate on the employees’ behalf which raises the firm’s overall efficiency (Kruse, 1992). From the employees’ perspective, the incentive effect of profit sharing is basically a trade-off between the positive effect of a variable, performance-related pay and the negative one of sharing the result of any additional effort with all other employees. Employees will increase their effort if their additional individual effort is rewarded by extra income. A prerequisite is of course that profit sharing is regarded as a fair system, balancing the interests of labour and capital. As long as the balance sheets are not manipulated, the overall profit is a clear performance indicator. Profit sharing systems are also usually not changed in response to a higher than expected performance of the workforce which is frequently observed with piece rates.18 Hence, profit sharing is a reliable and verifiable claim on a significant part of the real return to the production factors capital and labour. The counterargument is also quite intuitive. If more than one person is employed, the productivity effect of any additional effort of an individual has to be shared with every other worker. Unless the firm is not very small, there is a considerable asymmetry 18
Adjusting the terms of the piece rate scheme to favour the firm might result in “ratchet effects”. The standard ratchet effect implies that workers may be unwilling to work hard today because they fear that the employer may infer that the workers’ cost of effort is low and thus will offer a lower wage in future periods (see Lazear, 1986, Gibbons, 1987).
58
The Output, Employment and Productivity Effects of Profit Sharing
between the disutility from effort and the monetary reward for it. This is called the 1/N (with N being the number of employees) or free rider problem. Individual incentive systems like e.g. piece rates could circumvent this problem, but considering that the organization of work moves more and more away from the “tayloristic” system towards a system where work is flexibly organized and tasks are overlapping, individual output is hardly measurable and the performance of work groups comes to the fore. In such a situation piece rates or any other individual incentive systems are not useful. If a flexible reward system is considered at all, it must be some kind of remuneration scheme that is based on the output of a specific group or unit, like for example profit sharing. The free rider problem is also mitigated by the fact that workers usually know more about the effort of their co-workers than supervisors. If employees are aware of the negative externality and if they can effectively monitor and punish their shirking colleagues, a reaction is probable. Hence, costs for vertical supervision are reduced due to horizontal supervision (FitzRoy and Kraft, 1985a, Kandel and Lazear, 1992). Aside from the productivity effect, another intensive discussion centred around the employment effects of profit sharing. As had been elaborated in the previous chapter, Weitzman (1984) started a discussion on a positive employment impact. The basic idea of Weitzman is that the originally fixed wage is decomposed into a reduced base wage and a variable profit-related part. The firm then determines its level of employment on basis of the equality of marginal productivity and the base wage - not the variable profit share. Hence, there is an incentive to increase employment beyond the level that would have been chosen in a traditional firm with fixed wages, as long as it is possible to reduce the base wage and compensate the workers with part of the profits. Weitzman did not include productivity effects of profit sharing in his models. Perhaps, however, productivity and flexible remuneration have to be considered simultaneously. If profit sharing increases productivity, unit costs will decrease and if the firm can sell more on a competitive market it will make a greater profit by expanding output and employment.19 With a declining marginal product of labour, however, the estimated productivity level might not differ very much from the one
19
In case of an imperfectly competitive market the firm may still increase profits by expanding if the corresponding price decline is small (Pérotin and Robinson, 2003).
Econometric Model
59
before the introduction of the sharing system. But both output and employment would be larger. This argument might provide an explanation for a number of studies that find insignificant productivity effects of profit sharing. Using a very simple theoretical model, Kraft and Ugarkovic (2005) indeed show that assuming profit sharing is a supplementary payment and enhances workers’ efficiency, firms will grow with respect to employment and output. Productivity, however, if measured as output divided by sales, would not be affected. Before empirically testing the different hypotheses that arise from these considerations, several practical issues that are associated with the implementation of the conditional difference-in-differences approach are outlined.
Econometric Model A first analysis of our data in the previous chapter indicated that selectivity problems pervade the analysis of profit sharing which have to be accounted for methodologically. Several non-experimental estimators that deal with this issue have been introduced. Based on the discussion of the different estimators in chapter 4 and the data available, the conditional difference-in-differences estimator emerged as the most advantageous approach. To the best of our knowledge, this estimator has not been used to evaluate the impact of profit sharing so far. The intuitive idea of the conditional difference-in-differences approach is to solve the fundamental evaluation problem by constructing for the group of establishments that have introduced profit sharing a control group that is as similar as possible before the introduction of profit sharing and to conduct a before-after comparison for both groups. Differences in outcomes can then be attributed to the measure introduced (Heckman et al., 1998). Since the matching estimator, the difference-in-differences estimator and the conditional difference-in-differences estimator have all been outlined in section 0, we refrain from outlining the methodology again and focus instead on the practical issues involved with conditional difference-in-differences estimation (see e.g. Bryson et al., 2002, Caliendo and Kopeinig, 2006).
The Output, Employment and Productivity Effects of Profit Sharing
60
The general estimation procedure can be divided into several steps: 1. Separation of the observations with respect to the information whether a profit sharing system was introduced or not. 2. Estimation of the propensity score. 3. Choice of the matching algorithm. 4. Verification of the overlap of the distributions of propensity scores and common support condition. 5. Verification of the matching quality and estimation of the treatment effect. Step 1 involves dividing the firms into two groups based on the information whether they have introduced profit sharing or not. In order to match the establishments, we refer to the results by Rosenbaum and Rubin (1983) who were able to show that instead of matching on a high-dimensional vector of exogenous characteristics X, one can also match on the propensity score, i.e. the probability to introduce profit sharing given a set of individual characteristics of a firm pr(D=1|X). The propensity scores can be obtained e.g. via probit regression where the decision to introduce (D=1) or not introduce (D=0) profit sharing serves as the dependent variable (Step 2). Although the index commonly used for matching is the probability of treatment, it need not necessarily be. One can also use the predicted linear index instead of the probability (unbounded propensity score). The advantage of the unbounded propensity score is that it differentiates more in the tails of the distribution (Lechner, 2000b). Coming to Step 3, in the literature several matching methods have been proposed. The major difference between the various matching estimators lies in the weights attached to the control group observations. Typically, the more similar an untreated unit j is compared to a treated unit with respect to the covariates (or propensity score) the higher the weight attached to it in constructing the match. The estimated treatment effect for a firm i can generally be assessed as follows: Yi1 -
6 wN0 (i, j) Yj0 j {D=0}
where wN0 (i, j) denotes a weighting function and N0 is the number of non-treated firms with Dj=0 (Heckman, Ichimura and Todd, 1998). The two extremes among the
Econometric Model
61
matching estimators with respect to the weighting function are the nearest neighbour and the kernel matching estimator. The nearest neighbour matching estimator can be described as a pair-wise matching because it tries to find for every treated unit i the most similar (ideally a “twin”) nontreated unit and, thus, leads to the following weighting function: 1 if ||Xi – Xj|| j = min ||Xi – Xj|| wN0 (i,j) = 0 else Nearest neighbour matching can be performed with and without replacement. Matching with replacement means that control units can be used more than once as a matching partner. Matching without replacement might run into problems in regions where the frequency of probabilities is very low for the control group compared to the treatment group. This means that in case of only few comparison units similar to the treated unit, we may be forced to match treated units to control units that are different in terms of their probability. Allowing for replacement avoids this problem as long as there is sufficient overlap in the distributions. While matching with replacement hence potentially leads to an increase in the average quality of matching and a decrease of bias, it also comes at the cost of higher variance because fewer observations are being used for the control group (Dehejia and Wahba, 2002, Smith and Todd, 2005). In the case of the kernel matching estimator, on the other hand, the comparison group consists of all non-treated observations and the weighting function is specified as: Gij wN0 (i, j) = 6 Gik K {D=0}
where Gik= G((Xi – Xk)/h) is a kernel that downweights distant observations and h is a bandwidth parameter (Heckman, LaLonde and Smith, 1999). The major advantage of kernel matching is a lower variance due to the fact that more information is used. On the other hand, however, kernel matching runs the risk of potentially using bad
62
The Output, Employment and Productivity Effects of Profit Sharing
matches. The common support condition is therefore very important when considering the application of kernel matching. Several other estimators that lie between these two extremes have been proposed in the literature. Since nearest neighbour matching runs the risk of bad matches if the neighbour is far away, caliper and radius matching have been suggested. Caliper matching imposes a tolerance level on the maximum propensity score distance. The control unit that lies within this caliper and is closest in terms of the propensity score is chosen as a matching partner. Radius matching can be regarded as a variant of caliper matching. Instead of using the closest unit in terms of the propensity score as matching partner only, the idea of this estimator is to use all units within the caliper as matching partners (Smith and Todd, 2005). Another approach is stratification or interval matching. The idea is to divide the units into a set of intervals (strata) according to their propensity scores and to calculate the treatment effect within each interval by taking the mean difference in outcomes between both groups of observations. The average treatment effect on the treated can then be obtained as a weighted average of the separate effects where the fraction of treated units within each interval are used as weights. The question which method to select in practice depends in particular on the degree of overlap between the treatment and control group in terms of the propensity score. When there is substantial overlap, most of the matching algorithms yield similar results (Dehejia and Wahba, 2002). Heckman et al. (1997) stress that in small samples the choice of estimator is important. A reasonable approach might therefore be to consider more than just one matching estimator. Another important requirement apart from the conditional independence assumption is the common support condition which was described earlier in equation 17 (Step 4). Common support ensures that units with the same observable set of characteristics have a positive probability of being both treated and control units (Heckman et al., 1999). In the literature, several ways have been suggested to examine the overlap and the region of common support between treatment and control group. A straightforward way is the visual analysis of the frequency distributions of the propensity scores in both groups followed by a drop of all control group observations
Empirical Study
63
whose propensity score is smaller than the minimum and larger than the maximum of the treatment group (Lechner, 2000a). Finally, in Step 5 the matching quality has to be assessed. The reliability of all matching estimators strongly depends on the balancing property of the propensity score. After matching, a simple comparison of both groups with respect to all relevant exogenous characteristics should indicate no significant differences in means. Having outlined all practical issues that are involved with the conditional difference-indifferences estimation, we now turn to the empirical investigation of the relationship between profit sharing and output, employment and productivity.
Empirical Study Most empirical studies dealing with the effects of profit sharing observe the existence of a profit sharing system only. An advantage of our study is that we make use of the longitudinal structure of our data by identifying those establishments that actually introduce this scheme. In order to investigate the effect of an introduction of profit sharing on output, employment and productivity, in a first step we identify all establishments that were surveyed continuously from 1998 to 2002. Of these establishments all non-profit firms and firms not reporting turnover or with missing values for the variables used are excluded. In addition, establishments from the forestry, agriculture and fisheries, and banking and insurance sector are omitted (see e.g. Bauer, 2003, Zwick, 2003).20 The treatment group is then constructed by identifying all establishments that reported to have no profit sharing in 1998, but had introduced it by 2000 and still reported it in 2001 – summing up to 73 establishments. The control group on the other hand comprises all those establishments that reported to have no profit sharing in either 1998, 2000 or 2001 and contains 1545 establishments. Table 3 presents mean values of exogenous characteristics of both groups of establishments. In comparison to Table 2, an additional variable is included that was available in the survey of the year 1998, but not in the year 2000. Firms were asked to
20
Establishments from the forestry, agriculture and fisheries sector are omitted due to a differing production process, whereas banks and insurances report balance sheet totals and volume of insurance contributions instead of turnover.
64
The Output, Employment and Productivity Effects of Profit Sharing
rate the degree of competitive pressure they are exposed to on a 4–point-scale ranging from ‘no’ to ‘high’ competitive pressure. A dummy variable is created which takes the value of one if a firm reports to have ‘medium’ or ‘high’ competitive pressure. In analogy to our initial comparison above, we find that firms introducing profit sharing differ significantly from their non-sharing counterparts with respect to several exogenous characteristics. This has obviously an impact on outcomes as well. In the lower part of the table mean comparisons of both groups with respect to the outcome variables of interest are presented. Firms that introduce profit sharing have significantly higher sales and employment levels. They also exhibit higher productivity levels measured as sales per employee than their non-sharing counterparts. More details concerning the dependent variables will be given below as we proceed with our estimations.
Table 3 - Mean Values of Variables for Firms (not) introducing Profit Sharing, 1998 Variable Establishment size <20 Establishment size 20-49 Establishment size 50-249 Establishment size 250-499 Establishment size 500+ Shift responsibilities Teamwork Independent work groups Share qualified employees Collective bargaining Works council Competition Exporter ICT investment Limited liability Age East German establishment Total Sales (in DM) Number of employees Sales per employee (in DM) Propensity score
Firms with profit sharing .21 .15 .37 .11 .16 .36 .22 .21 .69 .77 .58 .92 .44 .71 .74 .40 .41 2.02e+08 423 416794 -1.23
Firms without profit sharing .56*** .20 .17*** .04* .03*** .17*** .11** .07*** .58*** .54*** .21*** .86* .14*** .42*** .45*** .44 .56** 1.98e+07** 78*** 201661** -2.09***
Matched firms without profit sharing .17 .15 .41 .13 .15 .36 .20 .23 .71 .68 .57 .93 .45 .73 .77 .42 .48 8.82e+07 339 284030 -1.30
*/**/*** indicate that means differ with statistical significance in a two-tailed t-test at the 10%, 5% or 1% level between the establishments from the two groups. Source: IAB Establishment Panel, wave 1998, own calculations.
As has been explained already, the first task in matching establishments from the two groups is to reduce the problem of dimensionality by estimating propensity scores on
Empirical Study
65
the basis of a probit model where the decision to introduce (D=1) or not introduce profit sharing (D=0) serves as the dependent variable. Table 4 presents the results of the probit estimates which reveal that rather young and medium-sized or large exporting firms with a high share of qualified employees that invest into information and communication technologies introduce profit sharing. Moreover, the existence of a works council seems to foster the introduction supporting the theory that profit sharing is more likely to be introduced in an environment characterized by mutual trust. Firms that are located in the Eastern part of Germany have a lower probability to introduce profit sharing and, finally, profit sharing seems to be flanked by other organizational measures, especially the existence of independent work groups.21
Table 4 - Results of the Probit Estimation Variables Establishment size 20-49 Establishment size 50-249 Establishment size 250-499 Establishment size 500+ Competition Exporter Share of qualified employees Collective bargaining Works council Limited liability Age East German Establishment Shift of responsibilities Teamwork Independent work groups Ict investment intercept Number of observations Pseudo R2
Coefficient .05 .52 .43 .77 .16 .28 .62 .16 .32 .18 .40 -.46 .15 -.06 .32 .25 -3.28 1618 0.20
t-value 0.25 2.39** 1.42 2.37** 0.72 1.64 2.29** 0.99 1.65* 1.16 2.54** -2.91*** 0.95 -0.35 1.71* 1.75* -5.52***
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Values for industry dummies are not included and are available from the author on request. Source: IAB Establishment Panel, wave 1998, own calculations.
Lechner (1998) stresses that a fundamental requirement for a successful implementation of the matching procedure is a sufficiently large overlap between the
21
To test the robustness of our results, we tried alternative specifications of the probit model. This had no significant impact on our results presented in the following chapters.
The Output, Employment and Productivity Effects of Profit Sharing
66
distributions of the conditioning variables in both subsamples. Figure 3 shows the overlap for the unbounded propensity scores.22
Figure 3 - Frequency Distributions of the Propensity Scores 90 number of observations
80 70 60 50 40 30 20 10 -0,2
0,01
-0,5
-0,7
-0,9
-1,2
-1,4
-1,6
-1,9
-2,1
-2,3
-2,6
-3
-2,8
-3,3
-3,5
-3,7
0
unbounded propensity score with profit-sharing
without profit-sharing
Source: IAB Establishment Panel, wave 1998, own calculations.
Since both distributions overlap to a great extent, this important assumption is fulfilled. Following Heckman et al. (1998) who stress that a successful implementation of matching methods is only possible inside the range of common support of the distribution of the propensity scores of the profit sharing and non profit sharing group, we impose the common support restriction.23 We apply nearest neighbour matching in all of the following estimations. Due to the similar number of profit sharing and non-profit sharing establishments in the right tail
22
We used the unbounded propensity score x ' Eˆ because of its preferable distribution properties
(Hujer et al., 1997). Using the bounded propensity score ) (x ' Eˆ ) changed our results only marginally. 23 This leads to a drop of 219 observations (approximately 14%) of the control units as they are poor matches, i.e. their propensity score falls short of the minimum propensity score of the treated units. We do not consider this as being problematic as we are still left with more than 1300 control units for 73 treated units.
Empirical Study
67
of the distributions, we decided to match with replacement. In addition to matching on the propensity score, we match establishments from the two groups according to their number of employees to improve the quality of the match. In order to control for the robustness of our results, we additionally apply kernel matching.24 We can now proceed with our estimations. Several dependent variables are considered in order to test the predictions of our model. We investigate whether firms introducing profit sharing have a higher growth of labour productivity, employment and sales between 1998 and 2002 as compared to the non-profit sharing control group. Growth in labour productivity is measured by the percentage change in sales per employee. A more conventional measure of output is value added which can be computed using the IAB Establishment Panel data by subtracting input costs (material and intermediate inputs) from sales. Although conceptually certainly superior, we decided to stick to sales per employee due to the high rate of item non-response and our impression that the available data are rather rough guesses than precise values.25 Employment and sales growth are measured by the percentage change in the total number of employees and the volume of sales respectively. We take account of the sampling variability in the estimated propensity score by applying a bootstrap method to construct the standard errors of the estimated treatment effects (Heckman et al., 1998).26
24
Matching was performed in Stata 8 using the psmatch2 command (Leuven and Sianesi, 2003). Approximately one quarter of the establishments in the sample reported no data for input costs and almost two thirds reported data falling in 5%- increments, i.e. 5%, 10%, etc. Addison et al. (2003a) report similar problems. 26 The means of the outcome variables of the comparison group are based on an estimation and are not result of random sampling thus leading to biased t-statistics. By bootstrapping a random sample is repeatedly drawn from the original sample and the matching procedure is performed. Thereby, the distributions of the mean outcomes of the comparison group are simulated and unbiased t-statistics can be obtained (see e.g. Greene, 2003).
25
68
The Output, Employment and Productivity Effects of Profit Sharing
Table 5 - Results from Conditional Difference-in-Differences Estimation, Introduction of Profit Sharing Nearest neighbour matching outcome variable productivity growth employment growth sales growth
treaties .21 .12 .32
controls .19 -.06 .10
att .02 .18 .22
t-value 0.26 2.93*** 2.27**
treaties .21 .12 .32
controls .19 -.03 .11
att .02 .15 .21
t-value 0.38 3.78*** 2.83***
Kernel matching outcome variable productivity growth employment growth sales growth
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations.
The results we get from nearest neighbour and kernel matching support the implications of the theoretical model by Kraft and Ugarkovic (2005) and hence the hypothesis that employment can be stimulated indirectly through increases in productivity. Establishments introducing profit sharing seem to grow substantially faster with respect to sales and employment. Productivity, however, measured as sales over employment, is only insignificantly higher. In order to ensure that the matching method was successful in balancing the two groups and thereby reducing the selection bias, a comparison of means of the exogenous variables after matching has to be conducted. Comparing the differences in the means of the exogenous variables of the matched and unmatched control group members with the profit sharing establishments in Table 3 above indicates that the application of the matching procedure leads to a substantial reduction of bias. After matching there are no statistically significant differences in the means of the exogenous variables anymore. The results we get are quite astonishing as the empirical literature on profit sharing mainly comes to the conclusion that profit sharing increases productivity, whereas its impact on employment is not clear cut. Our results, on the other hand, indicate a substantially higher sales and employment growth of the profit sharing establishments thus levelling out any productivity effect.
Empirical Study
69
In order to make sure that this result is not to be attributed to the rather small number of establishments in our sample that actually do introduce profit sharing, we conduct the same estimations for the existence of profit sharing. This increases our number of matched treated establishments to 226, but does not change our results in essence as is shown in Table B.1 in Appendix B. Whether nearest neighbour or kernel matching is applied, the treated establishments still have a significantly higher growth in sales and employment and only insignificantly higher productivity growth.27 So far, information on the capital stock of firms has not been included in our estimations. The IAB Establishment Panel contains no direct information on capital, so the factor would have to be approximated by replacement investments. The idea is that the known amount of replacement investments is expected to be proportional to the unknown amount of capital stock (e.g. Möller, 2000, Wolf and Zwick, 2003). A lot of firms, however, report no data or report a value of zero which doesn’t seem trustworthy to us (see also Addison et al., 2003a). To control for the robustness of our results, we nonetheless conduct the estimations including replacement investments as a proxy for capital. This reduces our sample by more than thirty percent, leaving us with 1041 control and 57 treated establishments. The results are reported in Table 6 indicating that the estimates we obtained above are fairly robust as we find a significantly higher sales and employment growth in the case of sharing firms. Similar results are obtained for the existence of profit sharing where the number of treated firms drops to 177 (Table B5 in the Appendix).
27
Further information on the estimation procedure is provided in Tables B1-B4 in Appendix B.
70
The Output, Employment and Productivity Effects of Profit Sharing
Table 6 – Results from Conditional Difference-in-Differences Estimation, Introduction of Profit Sharing, Capital Stock included Nearest neighbour matching outcome variable productivity growth employment growth sales growth
treaties .20 .12 .32
controls .19 -.07 .03
att .01 .19 .30
t-value 0.10 2.10** 2.94***
treaties .20 .12 .32
controls .20 -.02 .13
att .00 .14 .19
t-value -0.03 2.91*** 2.83***
Kernel matching outcome variable productivity growth employment growth sales growth
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations.
Summary Among the empirical studies that investigate the effects of profit sharing, the relationship between profit sharing and productivity is by far the one that received most attention. In the meantime, one could virtually speak of common knowledge that profit sharing increases productivity. Using the German IAB Establishment Panel, we investigate how profit sharing affects output, employment and productivity and find differing results. Our estimations indicate that an introduction of profit sharing is associated with significantly higher output and employment growth. Productivity, however, defined by the ratio of output to employment is not affected. This finding of a neutral productivity effect is in line with several other studies, including the one by Wolf and Zwick (2002a) who use the same data set. The theoretical model developed by Kraft and Ugarkovic (2005) provides an explanation: If an introduction of profit sharing is associated with an increase in efficiency, the representative firm will extend output and employment. With a declining marginal product of labour, however, the estimated productivity level might not differ very much from the one before the introduction of the sharing system.
Profit Sharing: Supplement or Substitute? Theoretical Considerations In their theoretical model, Kraft and Ugarkovic (2005) assume that profit sharing is a supplementary payment that induces higher worker productivity. The results we presented so far could, however, also be the result of lower wages or slower wage growth. We therefore investigate in the following an important question that has been the focus of only very few studies presented in Chapter 3.2.3, namely the question how profit sharing affects wages and hence the question how employees fare when a profit sharing scheme is implemented. Theoretically, this question is closely related to the discussion on the employment effects of profit sharing which was initiated by Martin Weitzman with his book The Share Economy (1984) and had been formulated earlier by Vanek (1965). As has been outlined above, Weitzman claims that widespread profit sharing would generate excess demand for labour and thus could cure unemployment. His basic idea is that the originally fixed wage is decomposed into a reduced base wage and a variable profit-related part. Considering that the total share of profits allocated to the workers does not vary with the number of employees, the firm determines its level of employment on basis of the equality of marginal productivity and the base wage - not the variable profit share. The base wage is thus the “true” marginal cost of labour implying a higher level of employment compared to a traditional firm with fixed wages. Weitzman’s model hinges, however, on two crucial assumptions that have been disputed heavily. On the one hand, it is arguable whether management can determine employment as it pleases. If employees have a say on employment decisions, they might resist additional hirings to avoid a reduction of their own profit-related wage component (Layard et al., 1991, OECD, 1995). More interesting in our context,
72
Profit Sharing: Supplement or Substitute?
however, is the question whether the profit-related share indeed replaces part of the basic remuneration or rather tends to supplement wages. Productivity Effects and Wages The focus of Weitzman’s theory is primarily centred around the firms’ incentives with respect to a more flexible pay. His theory does not consider productivity effects of profit sharing. The standard motivation for an introduction of profit sharing, however, is the employer’s intention to align the employees’ interests with her own. This should result in an increased effort and willingness to cooperate on the employees’ behalf thus raising the firm’s overall efficiency (Kruse, 1992). Quite a number of empirical studies have investigated the productivity effects of profit sharing and come to a remarkably clear result which attests profit sharing a significantly positive effect on productivity (OECD, 1995, Doucouliagos, 1995). It seems questionable whether such a strong effect could be generated by merely substituting part of employees’ fixed pay with a variable share which, moreover, is based on the profits of the entire firm and thus is highly uncertain (Prendergast, 1999). Standard reasoning of the free riding problem suggests that agents should respond only weakly to such incentives, since they receive only 1/N of the returns to effort in a firm with N workers (Holmström, 1982). Further explanations are thus needed to account for the unambiguous empirical results. The most popular counterarguments to free riding are peer pressure and mutual monitoring. Workers usually know more about the effort of their colleagues than supervisors. In a profit sharing environment where an individual’s effort is tied to all other colleagues’ utility, employees will exert pressure if they can effectively monitor shirking colleagues and are aware of negative externalities (FitzRoy and Kraft, 1985a, Kandel and Lazear, 1992). Such peer pressure will hardly exist, however, if the base wage has been reduced and the workforce feels treated unfairly by the management.28 Several other approaches - ranging from efficiency wages over worker sorting to compensating differentials - might provide another very plausible explanation for the
28
This is supported by Freeman et. al (2004) who show that employees respond more to shirking if labor-management relations are good.
Theoretical Considerations
73
clear incentive effects associated with profit sharing. Common to these approaches is the suggestion that firms might not reduce base wages, but rather pay a supplementary profit share. Efficiency Wage Theory. “Effective profit sharing must ordinarily presuppose the payment of the full going rate of wages to participants (…) The object (…) is to induce a special degree of effort, efficiency, cooperation. Or some other desirable result, not usually obtainable by the payment of a flat wage. Obviously these special results cannot be expected unless the rewards which call them forth can be counted on to exceed the regular and usual wage.” (Burritt et al., 1918) This early statement by Burritt et al. resembles very much to what was to be subsumed under the term efficiency wage theory later on. Efficiency wage theory suggests that employers can maximize profits by paying workers a premium because the resulting increment in productivity yields the highest profits. Channels through which the productivity is increased are (among others) higher effort due to a greater cost of a job loss, increased loyalty and morale as well as lower turnover and recruitment costs (for an overview see e.g. Yellen, 1984). Arguing from an employer’s perspective, paying the premium in the form of a supplementary profit share has quite some appeal. On the one hand, it can replace permanent pay increases which firms might be reluctant to grant, while on the other hand signalling the employer’s intention to be generous and share profits in good times (OECD, 1995). The theory on monetary incentives also suggests that profit sharing may have advantages in work settings where individual output is hard to measure, where employees have discretion about their working methods and where the cooperation of employees is an elementary determinant of success. In such an environment, a supplementary profit share as a group incentive presumably further intensifies the communication and cooperation within the firm.29 An important implication of the
29
Supplementary profit sharing, however, is no perfect substitute for efficiency wages as it is not the result of individual effort only, but is also dependent on other factors such as e.g. demand fluctuations.
74
Profit Sharing: Supplement or Substitute?
payment of profit shares as a supplement to fixed wages is that employees might respond not so much to profit sharing itself but rather to an increase in pay. Worker Sorting. Besides eliciting the productivity of existing workers, productivity could also increase because of the type of worker the firm can attract due to a switch to a higher and performance-related pay. Introducing a supplementary profit-related share might lead to a rising stress level and work pace and increased workloads as a result of closer monitoring and higher requirements. People who are unwilling to accept such burdens leave the firm in question and are substituted with performanceoriented workers. A higher productivity can then be partially attributed to “worker sorting” (Lazear 1986, 2000, 2005). In addition to sorting, skill requirements might change as well. Profit sharing is frequently accompanied by an increase in employee involvement which requires more general skills as the complexity of tasks rises. This may increase the demand for more educated workers. As employee turnover rates fall and the importance of communication within the firm rises (in the long run), specific human capital is increased as well. According to human capital theory, workers with higher human capital also receive higher compensation. Compensating Differentials theory assumes that employees that face undesirable working conditions should be compensated accordingly (Rosen, 1986, Groshen, 1991, Handel and Levine, 2004). As has been indicated above, profit sharing introduces the concepts of mutual monitoring and peer pressure by tying every individual’s effort to all other colleagues’ utility. While the problem of free riding is thus alleviated and vertical monitoring costs are reduced, there is also a downside from the employees’ perspective. Closer monitoring presumably leads to a higher stress level due to increased workloads and a faster work pace. Higher remuneration may thus act as a compensation for the more demanding working conditions. This argument is further strengthened if profit sharing comes along with a higher degree of employee involvement. A priori it is unclear whether the higher remuneration is only paid out as shared profits or also by an increased base wage. It is sometimes argued that profit sharing is a means to shift risk to workers. Considering that most employees are risk averse, they would ask to be compensated
Empirical Study
75
for sharing this risk with the firm. This argumentation makes sense if profit sharing and the base wage are alternatives. In this case, the employer must trade off the incentive effect of a profit-related pay to the compensation of the risk if remuneration is variable. This trade-off will affect the inclination to reduce the base wage. If profitsharing is an add-on, however, this consideration is unnecessary. A priori, the predictions of economic theory are thus ambiguous and the need for empirical evidence arises. In light of the vast amount of empirical studies on the effects of profit sharing on productivity, it is rather surprising that only a small number of studies investigate whether the productivity gains are partially offset by a growth in labour costs. Section 0 above presented the existing evidence on the relationship between profit sharing and wages. So far, an inconclusive pattern emerges. In light of the empirical scarcity and the contradictory results that are obtained, further research is needed. We try to add another piece of evidence using German nationally representative firm panel data.
Empirical Study In analogy to the study in the previous chapter, we first identify all establishments that have been surveyed continuously from 1998 to 2002. The group of firms that have introduced profit sharing thereby encompasses all firms that reported to have no profit sharing in 1998, but had introduced it by 2000 and still reported it in 2001 – summing up to 82 establishments. The group of non-sharing firms on the other hand consists of all those establishments that reported to have no profit sharing throughout the period from 1998 to 2002 and contains 1649 establishments. Table 7 presents descriptive evidence on both groups with respect to the relevant
exogenous characteristics. Since the sample deviates only slightly from the one in the previous chapter, we find similar results concerning the differences in means between both groups. The decision to introduce profit sharing is obviously driven by specific factors which also seem to have an impact on outcomes, i.e. wages. Based on a probit estimation the results of which are reported in Table C.1 in Appendix C, the propensity scores for both groups of establishments are determined. Since Table C.2 in Appendix C shows that both distributions overlap to a great extent,
Profit Sharing: Supplement or Substitute?
76
we can proceed with the conditional difference-in-differences estimation. In analogy to our procedure in the previous chapter we impose the common support restriction.30
Table 7 - Mean Values of Variables for Firms (not) introducing Profit Sharing, 1998 Variable Average monthly base wage per employee (in DM) Establishment size <20 Establishment size 20-49 Establishment size 50-249 Establishment size 250-499 Establishment size 500+ Number of employees Shift responsibilities Teamwork Independent work groups Share qualified employees Collective bargaining Works council Competition Exporter Ict investment Limited liability Age East German establishment
Firms with profit sharing
Firms without profit sharing
Matched firms without profit sharing
4075 0.26 0.13 0.35 0.10 0.16 370 0.38 0.21 0.22 0.67 0.78 0.57 0.91 0.45 0.71 0.72 0.39 0.41
2774*** 0.57*** 0.19 0.16*** 0.04 0.04*** 83*** 0.16*** 0.11** 0.07*** 0.58*** 0.54*** 0.22*** 0.86* 0.14*** 0.43*** 0.45*** 0.45 0.56**
3866 0.24 0.12 0.38 0.12 0.14 287 0.28 0.14 0.20 0.67 0.84 0.57 0.88 0.45 0.62 0.76 0.37 0.41
*/**/*** indicate that means differ with statistical significance in a two-tailed t-test at the 10%, 5% or 1% level Source: IAB-Establishment Panel, wave 1998, controlled remote data access via the Research Data Center (FDZ) of the Federal Employment Service (BA) in the Institute for Employment Research (IAB). Note: Industry dummies are included in the analysis but are not reported.
To investigate whether the variable profit-related part is paid on top of the previously fixed wage, we make use of the payment information available in the data set which refers to overall gross wage payments for the workforce in June of the respective year. Divided by the number of employees, we use this information to approximate base wages. More precisely, base wage growth is specified as the difference between the logarithmic values of monthly base wages per employee in 2002 and 1998. If the variable profit-related share indeed substituted part of the initially fixed wage as was proposed by Weitzman, then a significantly lower growth of average monthly base wages would be expected in firms that have introduced profit sharing. 30
This leads to a drop of approximately 13% of the control units as they are poor matches, i.e. their propensity score falls short of the minimum propensity score of the treated units. We do not consider this as being very problematic as we are still left with more than 1400 control units for 82 treated units.
Summary
77
Table 8 – Results from Conditional Difference-in-Differences Estimation treaties
controls
att
t-value
Kernel matching base wage growth
0.077
0.052
0.026
0.64
Nearest neighbour matching base wage growth
0.077
0.037
0.040
0.56
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations (controlled remote data access via FDZ).
Table 8 contains the results of our estimations which are quite clear indicating that profit sharing firms had an even insignificantly higher growth of base wages between 1998 and 2002. The right column of Table 7 above reports the results of a comparison of means of the exogenous variables after matching. Since no significant differences after matching can be detected, the matching method was successful in balancing the two groups and thereby reducing the selection bias. On the basis of the conditional difference-in-differences estimates we thus conclude that profit sharing seems to be a supplementary payment and is not substituting part of the fixed wage as was proposed by Weitzman.
Summary Most studies evaluating the effects of human resource management practices focus on their impact on company performance. In the case of profit sharing, a vast amount of empirical literature evaluates the productivity effects of profit sharing and comes to a quite optimistic assessment. Only little attention, however, has been given to the costs that are associated with the implementation of a profit sharing scheme, most notably higher wages, which might outweigh the productivity benefits. We thus focus in our study on the effect of an introduction of profit sharing on wages. Weitzman (1984) suggests that the profit-related part should substitute part of the previously fixed wage which corresponds to a reduction in base wages. A number of other authors, on the other hand, doubt whether such a profit sharing scheme could generate the strong incentive effects found in empirical studies. Prendergast (1999) e.g. brings forward the argument that profit sharing might come along with a rise in compensation thus implying that the productivity effects of profit sharing found in the studies are not the result of the scheme itself but rather of an increase in pay. A number of approaches indeed provide reasonable explanations why firms do not
78
Profit Sharing: Supplement or Substitute?
reduce base wages but rather pay a supplementary profit share. The ambiguity of economic theory calls for an empirical investigation of this question. The results of this study indicate that firms introducing profit sharing do not reduce base wages, but rather pay the profit-related part as an add-on to regular wages which conflicts with Weitzman’s assumption. By considering the growth of base wages over a four-year-period as our outcome variable of interest, we take account of the possibility that firms that introduce profit sharing might have a slower growth of base wages over time as was stressed in prior studies. We find no evidence for such a behaviour. On the contrary, our findings suggest that profit sharing firms have an even higher (although statistically insignificant) growth of base wages after the introduction of the scheme. Summarizing, the results thus provide evidence that the efficiency effects found in the previous chapter are not the result of lower wages or slower wage growth but rather come at the cost of increased labour expenses.
Profit Sharing and the Financial Performance of Firms Theoretical Considerations Our previous findings take us straight to the next question. The empirical results in chapter 5 pointed to significantly higher output and employment growth due to the introduction of profit sharing. Productivity growth, however, was only insignificantly affected. Kraft and Ugarkovic’s (2005) theoretical model provided an explanation: Profit sharing leads to an increase in productivity which induces firms to extend output and employment. With a declining marginal product of labour, however, the estimated productivity level might not differ very much from the one before the introduction of the sharing system.31 Central to the success of a profit sharing scheme thereby seems to be the fact that employees are rewarded for their additional effort by extra income: Our estimations in chapter 6 showed that an introduction of profit sharing does not come along with lower wages or a slower wage growth as proposed by Weitzman, but is rather of a supplementary character. If employees, however, are rewarded financially for their increased effort, it seems plausible to assume that owners would like to see the implementation of such a scheme pay off as well. This is the more so as the implementation of an adequate sharing scheme is associated with further costs in terms of time, effort and other resources (Bhargava, 1994). Despite the importance of the issue, our overview of existing studies in chapter 3.2.2 showed that the impact of profit sharing on the financial performance of firms and thus the question whether a more efficient
31
This chapter is an extension of the work published in Kraft and Ugarkovic (2006).
80
Profit Sharing and the Financial Performance of Firms
utilization of labour in the production process is also reflected in a firm’s higher profitability has rarely been explored in the literature.
Empirical Study The objective of this study is to add another piece of evidence on the effect of profit sharing on the financial performance of German firms. In analogy to the previous chapters, we therefore construct the treatment group by identifying among all establishments that were surveyed continuously between 1998 and 2002 those firms that reported to have no profit sharing in 1998 but had introduced it subsequently. Non-sharing firms on the other hand include all those establishments that reported to have no profit sharing scheme implemented throughout the whole period. This leaves us with 87 sharing and 1831 non-sharing firms.
Table 9 depicts a familiar picture. The group of profit sharing firms significantly differs from their non-sharing counterparts with respect to several exogenous characteristics. This result extends to the profitability measure as well, since sharing firms report to be more profitable.
In analogy to Blanchflower and Oswald (1988) and McNabb and Whitfield (1998), the profitability measure used in this paper is a categorical measure reported by the firm’s top management and can take five values ranging from ‘1 - very good’ to ‘5 very bad’. Subjective measures of company performance of this kind have been applied widely in empirical research. Despite the reduced information they provide relative to a continuous profit measure and the potential problems arising from their subjective nature, they possess valuable information. Research has provided evidence that they correlate positively with objective measures thus validating their use (Dess and Robinson, 1984, Venkatraman and Ramanujam, 1987, Wall et al., 2004).
Empirical Study
81
Table 9 - Mean Values of Variables for Firms (not) introducing Profit Sharing, 1998 Variable Number of employees Shift responsibilities Teamwork Independent work groups Share of qualified employees Collective bargaining Works council Competition Exporter Investment in ICT Limited liability Age East German establishment Profitability
Firms with profit sharing
Firms without profit sharing
359 0.37 0.21 0.22 0.66 0.78 0.55 0.91 0.45 0.70 0.71 0.40 0.44 2.83
79*** 0.16*** 0.11** 0.07*** 0.58*** 0.54*** 0.21*** 0.86 0.14*** 0.41*** 0.45*** 0.46 0.55** 3.05*
Matched firms without profit sharing 263 0.28 0.14 0.14 0.69 0.81 0.56 0.87 0.47 0.59 0.69 0.40 0.50 3.01
*/**/*** indicate that means differ with statistical significance in a two-tailed t-test at the 10%, 5% or 1% level Source: IAB-Establishment Panel, wave 1998, controlled remote data access via the Research Data Center (FDZ) of the Federal Employment Service (BA) in the Institute for Employment Research (IAB). Note: Industry dummies are included in the analysis but are not reported.
Tables D.1 and D.2 in Appendix D show the results of the probit estimation as well as the distribution of propensity scores. Since both distributions overlap to a high degree, we can proceed with the conditional difference-in-differences estimation. The results are reported in Table 10.32
The outcome variable is defined as the change in profitability between 1998 and 2002 and takes a value of one if the firm reported an improvement (deterioration) in its profit situation and zero otherwise. Kernel as well as nearest neighbour matching results indicate a significantly positive effect of profit sharing on profitability: Approximately forty percent of the firms that have introduced a profit sharing scheme report an improvement of their profit situation in 2002 as compared to 1998. This share is significantly lower in the case of non-sharing firms.
32
Imposition of the common support condition results in the drop of less than 12% of control observations. We do not consider this as being problematic, since we are still left with more than 1600 control observations for 87 treated units.
Profit Sharing and the Financial Performance of Firms
82
Table 10 – Results from Conditional Difference-in-Differences Estimation Change in profit situation Firms with Matched firms (dummy: 1= improvement) profit sharing without profit sharing
att
t-value
Nearest neighbor matching
0.40
0.23
0.17
2.15**
Kernel matching
0.40
0.26
0.14
2.43**
Nearest neighbor matching
0.29
0.41
-0.13
-1.34
Kernel matching
0.29
0.40
-0.11
-2.26**
Change in profit situation (dummy: 1= deterioration)
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations (controlled remote data access via FDZ).
Since employees’ efforts to improve their firms’ situation can be counterbalanced by external events, an overall deterioration in terms of profitability despite an introduction of profit sharing cannot be excluded. The lower part of Table 10, however, indicates that the fraction of sharing firms that report a deterioration of their profit situation would have been significantly higher had profit sharing not been introduced. Before finally concluding on the positive impact of profit sharing on profitability, the quality of matching needs to be insured. We therefore compare the group of sharing and matched non-sharing firms after matching. The results which are reported in the right column of Table 9 indicate no differences in means between both groups. The matching procedure was therefore successful in reducing selection bias.
Summary In light of the virtual flood of investigations on the productivity effects of profit sharing, it seems rather astonishing that evidence on the important question of its impact on the financial performance of firms is scarce. We therefore contribute to the literature by analysing this question based on longitudinal German firm panel data. Our results indicate that an introduction of profit sharing is associated with higher profitability: Among the group of profit sharing establishments, a significantly higher share reports an improvement of profitability, whereas the share of firms that report a deterioration is also significantly lower.
Summary
83
This finding fits into the overall picture of profit sharing we have obtained so far which indicates that both parties benefit from profit sharing (at least in monetary terms): Employees benefit from profit sharing through increased pay. Employers, on the other hand, are better off because the efficiency gains from profit sharing obviously outweigh the costs associated with the implementation of the scheme and materialize in the form of higher profitability.
Profit Sharing and Works Councils: Substitutes or Complements? A key factor determining the success of a profit sharing scheme which has so far only been touched upon briefly in this study, is the climate between management and employees. A cooperative and trustful climate contributes to a greater sense of identification of the work force with the company’s interests thus encouraging higher motivation and commitment (Levine and D’Andrea Tyson, 1990). This is where alternative modes of worker participation come into play which explicitly aim at establishing a cooperative climate. A distinct feature of the German system of industrial relations that has found a lot of attention internationally is that information, consultation and participation rights are formally prescribed by law and institutionalised in the body of works councils. According to the Works Constitution Act:
“(…) the works council safeguards the interests of the employees in dealing with the employer. The works council and the employer shall work together in a spirit of mutual trust and in co-operation with the trade unions and employers’ associations for the good of the employees and of the establishment. (…)”. (Federal Ministry of Labour and Social Affairs, 1990) The rent-creating effects of works councils are thus stressed. Works councils exert a collective voice (Freeman and Medoff, 1979, FitzRoy and Kraft, 1987b), they foster trust and cooperation and improve the flow of information between management and employees (Freeman and Lazear, 1995). By analogy with the union institution, however, works councils have two faces, the other one being characterized by its monopolistic power. Works councils may use the power conveyed by knowledge and
86
Profit Sharing and Works Councils: Substitutes or Complements?
legal entitlement to modify and postpone profit-maximizing decisions and to redistribute the joint surplus in favour of labour, thus leading to inefficiencies (Freeman and Medoff, 1979, Addison et al., 2004). A priori, it is therefore not clear how both types of employee participation interact with each other with respect to company performance. One should also bear in mind a major difference of the two modes of participation raising questions about their efficiency: while profit sharing is usually induced by management, employees take the initiative for the set-up of a works council (FitzRoy and Kraft, 1987b). From a theoretical standpoint thus both positive and negative interaction effects can be deduced. The fact that empirical research on this question is virtually absent is an important motivation to empirically address the question whether financial and representative participation complement or substitute each other with respect to output, employment and productivity. In this chapter, we explore this question using longitudinal information on profit sharing, works councils and measures of corporate performance.
Theoretical Considerations and Empirical Evidence This chapter is closely related to chapter 5 where we investigated the output, employment and productivity effects of profit sharing. Our results in chapter 5 indicated that there is a close link between the productivity- and employmentenhancing effects of profit sharing. A firm’s demand for labour can also rise as a result of an increased productivity. Intuitively, unit costs will drop if profit sharing increases productivity and so the firms can realize a higher profit by expanding output and employment (Pérotin and Robinson, 2003). The fact that firms obviously pay profit sharing as an add-on to regular wages excludes employment effects as proposed by Weitzman. A crucial factor for the success of a profit sharing scheme has been disregarded so far. The incentive effect of profit sharing hinges strongly on the level of information, trust and voice of a firm’s employees. Incentive effects are materialized only if the work force is informed extensively and on a regular basis about firm performance and its determinants, and if the relationship between employees and management is built on
Theoretical Considerations and Empirical Evidence
87
trust. Moreover, because profits depend not only on employee effort but also on numerous other factors, the incentive effect could be lost if employees felt they were not participating in decisions which have a key impact on profit, such as strategic or investment decisions, but were instead bearing the risk (Pérotin and Robinson, 2003). The importance of employees’ information and voice for a materialization of incentive effects of profit sharing takes us within the German context to the question on the role of works councils. The German Works Constitution Act grants workers in firms with five or more employees the option to set up a works council. According to the law, the works council’s duties are to cooperate trustfully with the employer for the good of the employees and of the establishment. For that purpose, works councils are equipped with rights that reach from information over consultation and - what distinguishes it from its continental European counterparts - to codetermination rights. Beginning with the information rights of works councils, employers have to provide the works council with timely and comprehensive information in order to enable it to meet its general duties. Information rights further increase with establishment size. In establishments with more than 20 employees with a voting right, the works council has to be informed in good time and extensively on establishment changes that might have substantial negative consequences for the firms’ employees or at least a substantial part of the workforce. This includes the closing or relocation of (a substantial part) of the establishment, the merger or division of the establishment, but also fundamental changes in the establishment organization or the introduction of new working methods. In plants with more than 100 employees, an economic committee (Wirtschaftsausschuss) must be set up which has to be informed timely and comprehensively on the current and future economic matters of the establishment. Consultation rights cover topics such as structural establishment changes, planned alterations concerning equipment or working methods, decisions relating to personnel planning and individual dismissals. Finally, works councils have codetermination rights with respect to a set of issues under the heading of “social matters” which are precisely laid down in the legislation. These include regulations concerning the beginning and termination of working times, temporary work reductions or overtime, leave arrangements, the introduction and application of technical devices to monitor employees’ performance, the introduction of new compensation methods and
88
Profit Sharing and Works Councils: Substitutes or Complements?
teamwork, and health and safety issues. If no agreement can be reached, a decision is made by the conciliation panel (Einigungsstelle). In firms with more than 20 employees with a voting right, works councils, moreover, have to consent to the hiring, (re)grading and transfer of workers. In case the number of employees exceeds 500, the works council may request the implementation of guidelines on these issues. Works councils can also negotiate works agreements (Betriebsvereinbarung) and, as a special case, social plans (Sozialplan) as a compensating measure for workers that were dislocated due to a closing of the plant or major changes in organization in firms with more than 20 employees. Unless explicitly authorized for, such work agreements cannot cover the negotiation of remuneration or other conditions of employment that are usually determined in industry-level collective agreements.33 Important in the context of our analysis is the fact that works councils have codetermination rights with respect to remuneration issues. More specifically, works councils have a codetermination right with respect to the introduction, application and changes of new remuneration methods.34 It is not clear a priori how managementinduced measures such as profit sharing and the employee-induced participatory instrument of works councils which enjoy codetermination rights with respect to an introduction and application of such a remuneration scheme interact. On the one hand, works councils exert a collective voice by providing employees with a means of expressing discontent other than through quitting the establishment. This implies a reduction of hiring and training costs and higher firm-specific investment (Freeman and Medoff, 1979, FitzRoy and Kraft, 1987b). Works councils also reduce information asymmetries which produce inefficient outcomes by collecting and communicating information on workers’ preferences and increasing management’s credibility (Freeman and Lazear, 1995). At first glance, it seems that both measures are to a certain extent substitutes as they both aim at the improvement of information flows and communication between labour and management. The existence of a works council can, however, positively influence the effects of profit sharing in a number of ways. Information on workers’ preferences
33 34
Bundesministerium der Justiz, 2006. See also Addison et al., 1997, or Addison et al. , 2001. This includes profit sharing schemes (e.g. Richardi, 2002).
Theoretical Considerations and Empirical Evidence
89
and the works council’s codetermination right with respect to remuneration can help to design and implement a compensation scheme that is appropriate from the employees’ standpoint therefore unfolding its full incentive effect. This effect can be further enhanced if the works council succeeds in creating a cooperative climate based on trust where management is prevented from behaving opportunistically and employees are provided with a voice and thus have no incentive to withhold neither necessary information nor additional effort (Jirjahn, 1998, Weitbrecht and Mehrwald, 1999). Cooke (1994) points out that works councils have a political stake if they have negotiated profit sharing and therefore have an incentive to encourage employees to pursue improvement in firm performance. Conversely, a central prerequisite for an effective collective voice is cohesion among workers which can be expected to increase through the introduction of profit sharing due to its collective nature (Levine, 1991, Zwick, 2004). On the other hand, however, by analogy with the union institution works councils have a monopolistic side (Freeman and Medoff, 1979). Works councils can abuse their knowledge and legal entitlement to redistribute the joint surplus in favour of labour and to modify and delay profit-maximizing decisions, thus creating inefficiencies. The German law counteracts these forces by imposing a peace obligation on works councils and by the implementation of the dual industrial relations system where wages are typically determined on an industrial or sectoral level (Addison et al., 2004). By exploiting their codetermination rights in areas where formally they have a say, however, works councils can extract concessions and informally bargain with management over issues which are outside of their sphere of competence formally (Müller-Jentsch, 1995).35 Another relevant objection which has become known as the management competence hypothesis was formulated by FitzRoy and Kraft (1987b). FitzRoy and Kraft point out that competent managers are aware of the importance of effective communication and participation and hence will provide for both without the imposition of regulations or the establishment of formal institutions. According to this hypothesis, works councils as an external restraint are superfluous if not even hindering in the process of an efficient internal organization of a firm.
35
Indeed, most empirical studies find a positive effect of works councils on the level of wages (see e.g. Jirjahn and Klodt, 1998).
90
Profit Sharing and Works Councils: Substitutes or Complements?
Several other authors argue that the proliferation of different management-induced measures – among them profit sharing as well – blurs the traditional distinction between management and labour (see e.g. Arnal, Ok and Torres, 2001). A number of Anglo-Saxon studies report that unions are often hostile towards the introduction of new measures by management, specifically if unions feel that the plan erodes their status or authority within the bargaining unit (Florkowski, 1987). This argument is consistent with other studies that report that management introduces certain measures on purpose to align the employees’ interests with those of the company and thereby undermine the union’s position (see e.g. d’Art, 1992, Poutsma, 2002). Although we believe this can be prevented in the German system of industrial relations by equipping works councils with codetermination rights and by incorporating profit sharing into the existing process of pay determination and collective bargaining, such non-cooperative behaviour cannot be excluded and thus gives room for neutral or even negative interaction effects. Summarizing, it is not clear a priori how profit sharing and works councils interact with respect to company performance. What is clear, however, is the fact that the outcome depends crucially on the mechanisms how information is exchanged, consensus can be reached and trust generated (Heywood et al., 1998). FitzRoy and Kraft (1985b, 1987b, 1990) investigated the effects of works councils on firm performance in a first series of papers and drew a quite pessimistic picture. Recent investigations with large-scale longitudinal data, however, seem to indicate that works councils appear in more favourable light (for an overview see Addison, 2003). Only a small number of studies have investigated empirically the interaction effects of works councils and profit sharing. FitzRoy and Kraft (1995) find that the existence of a works council has a positive and significant effect on the value added in sharing firms, but a negative effect in non-sharing firms. Frick and Möller (2003) find significantly positive effects of profit sharing and works councils on value added, but an insignificantly negative interaction effect. Similar international studies have been conducted by Black and Lynch (2004) or Cooke (1994). While Black and Lynch (2004) using a nationally representative sample of U.S. establishments report no significant interaction effects of unions and profit sharing on labour productivity,
Empirical Study
91
Cooke (1994) reports higher performance effects of profit sharing in non-unionised than in unionised firms. In light of the rising importance of human resource management, it is surprising that empirical evidence on the interaction of financial with traditional participatory mechanisms has attracted rather little attention. The aim of this study is to contribute to a closing of this gap.
Empirical Study Recalling Table 4 in Chapter 5 which contained the results of the probit estimation, we found the quite interesting result that the existence of a works council has a significantly positive influence on the introduction of a sharing scheme (Table 11 reproduces Table 4 for reasons of convenience). This could reflect the works councils’ success in creating a cooperative and thus favourable climate for an introduction of profit sharing. It could, however, also simply reflect size effects due to the fact that most large establishments have set up a works council. Nevertheless, although the data indicates that the existence of a works council promotes the introduction of profit sharing, the more interesting question is how performance is influenced by its existence.
Table 11 – Results of the Probit estimation Variables Establishment size 20-49 Establishment size 50-249 Establishment size 250-499 Establishment size 500+ Competition Exporter Share of qualified employees Collective bargaining Works council Limited liability Age East German Establishment Shift of responsibilities Teamwork Independent work groups Ict investment intercept Number of observations Pseudo R2
Coefficient .05 .52 .43 .77 .16 .28 .62 .16 .32 .18 .40 -.46 .15 -.06 .32 .25 -3.28 1618 0.20
t-value 0.25 2.39** 1.42 2.37** 0.72 1.64 2.29** 0.99 1.65* 1.16 2.54** -2.91*** 0.95 -0.35 1.71* 1.75* -5.52***
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Values for industry dummies are not included and are available from the author on request. Source: IAB Establishment Panel, wave 1998, own calculations.
Profit Sharing and Works Councils: Substitutes or Complements?
92
One way to answer this question is to subdivide the treatment as well as the control group further according to the information whether a works council exists or not. For this aim, we first construct both groups in analogy to chapter 5, leaving us with 73 treated and 1545 non-treated establishments. The groups are then further split according to the information whether a works council exists or not. This leaves us with 42 profit sharing and 315 non profit sharing firms which have a works council, on the one hand, and with 31 profit sharing and 1230 non-profit sharing establishments with no works council on the other hand. By matching both groups of treated establishments to their non-treated counterparts, we expect to gain new information on the role of works councils in the implementation phase of profit sharing, specifically with regard to performance. As we match establishments with and without works councils separately, sufficient overlap has to be assured in both cases. Figures 4 and 5 show the overlap of frequency distributions for the unbounded propensity scores. Since in both cases the distributions overlap to a great extent, this important assumption is fulfilled.
Figure 4 – Frequency Distributions of the Propensity Scores, Firms with Works Councils 45 40
number of observations
35 30 25 20 15 10 5 0 -2,7 -2,6 -2,5 -2,4 -2,2 -2,1 -2
-1,9 -1,7 -1,6 -1,5 -1,4 -1,2 -1,1 -1 -0,9 -0,7 -0,6 -0,5 -0,4 -0,2 -0,1 0
unbounded propensity score
without profit-sharing
with profit-sharing
Empirical Study
93
Figure 5 – Frequency Distributions of the Propensity Scores, Firms without Works Councils 200
number of observations
180 160 140 120 100 80 60 40 20 0 -3,7 -3,5 -3,3 -3,1 -2,9 -2,7 -2,5 -2,3 -2,1 -1,8 -1,6 -1,4 -1,2 -1 -0,8 -0,6 -0,4 -0,2 unbounded propensity score
without profit-sharing
with proft-sharing
Source: IAB Establishment Panel, wave 1998, own calculations (controlled remote data access via FDZ).
We can now proceed with presenting Table 12 which contains the results of our matching estimations. In analogy to chapter 5, we consider several dependent variables. We investigate whether firms introducing profit sharing have a higher growth of labour productivity, employment and sales between 1998 and 2002 as compared to the non-profit sharing control group for the groups of firms with and without a works council separately. Growth in labour productivity is measured by the percentage change in sales per employee. Employment and sales growth are measured by the percentage change in the total number of employees and the volume of sales respectively.36
36
We take account of the sampling variability in the estimated propensity score by applying a bootstrap method to construct the standard errors of the estimated treatment effects (Heckman et al., 1998).
94
Profit Sharing and Works Councils: Substitutes or Complements?
Table 12 – Conditional Difference-in-Differences estimation, Firms with(out) Works Council Establishments with works councils
Employment growth Sales growth
treaties 0.15 0.10 0.25
Productivity growth
treaties 0.31
controls 0.16
att 0.14
t-value 0.75
Employment growth Sales growth
0.16 0.44
0.03 0.16
0.13 0.28
1.04 1.32
Productivity growth
controls 0.19 -0.09 0.08
att -0.05 0.19 0.18
t-value Wilcoxon-Test -0.59 -0.05 3.86*** 3.69*** 2.07** 1.87*
Establishments without works councils Wilcoxon-Test -0.47 2.17** 1.57
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations (controlled remote data access via FDZ).
The results we obtain are puzzling. The introduction of profit sharing leads to an increase in sales and employment in firms that have set up a works council. This picture coincides with our earlier results in the first part of this study where we argued that initial productivity effects induced by profit sharing lead to lower unit costs. If firms can sell more on a competitive market they will realize a higher profit by expanding output and employment. Labour productivity defined as sales per employee, however, remains unaffected. What is striking, however, is that the positive effect on sales and employment is significant in firms with a works council only. The t-values indicate in the case of firms without a works council in all cases that the null hypothesis of no difference in means cannot be rejected, assuming a normal distribution. Owing to the small sample size and the possibility of important outliers we additionally applied the Wilcoxon signed-rank test. This nonparametric test is supportive of our result that in noncodetermined firms an introduction of profit sharing does not lead to the expected output effects (although we do find a significant impact on employment). Before drawing an overhasty conclusion, however, by stating that profit sharing unfolds its output- and employment-enhancing effect only in establishments where a works council has been set up, the quality of the matching procedure needs to be verified. Comparing the differences in the means of the exogenous variables of the matched and unmatched control group members with the profit sharing establishments in Table 13 indicates that the application of the matching procedure leads to a substantial reduction of bias. After matching there are no statistically significant
Empirical Study
95
differences in the means of the exogenous variables anymore. The matching procedure was thus successful in balancing the relevant groups. What draws our attention, however, is the fact that establishments with a works council differ from those without - first and foremost with respect to size. The average profit sharing firm with a works council is (with several hundred employees) quite large, whereas their counterparts without a works council have on average less than forty employees and are thus rather small or medium-sized. This is a very interesting finding since it implies that small firms have either of the two participatory mechanisms but hardly both and that the existence of both measures can be found primarily in large establishments.
Table 13 – Mean Values of Variables for Firms (not) introducing Profit Sharing after Matching Firms with works councils Firms without works councils Variable Number of employees Shift responsibilities Teamwork Independent work groups Share of qualified employees Collective bargaining Works council Competition Exporter ICT investment Limited liability Age East German firm Propensity score
708 .45 .31 .19
matched controls 567 .33 .22 .31
.70 .91 1 .93 .64 .81 .79 .26 .31 -.93
.75 .83 1 1 .56 .70 .83 .25 .47 -1.02
treaties
38 .21 .10 .21
matched controls 37 .21 .10 .21
.67 .55 0 .90 .17 .59 .66 .55 .59 -1.71
.67 .72 0 .86 .14 .59 .59 .66 .62 -1.71
treaties
*/**/*** indicate that means differ with statistical significance in a two-tailed t-test at the 10%, 5% and 1% level between the establishments from the two groups. Source: IAB Establishment Panel, wave 1998, own calculations (controlled remote data access via FDZ).
The difference in establishment size, however, prevents us from drawing a clear-cut conclusion regarding the influence of works councils and the estimation results that we have obtained above thus have to be refined. Arguing along the lines of Freeman and Lazear’s (1995) works council/employee involvement model, Addison et al. (2000) state that the optimal level of employee involvement differs among firms and is likely to be much higher in larger firms. Hence, tailoring a mandated level of
96
Profit Sharing and Works Councils: Substitutes or Complements?
involvement to fit both large and small firms turns out to be very complicated. Addison et al. (2000) thus presume that any mandated level of involvement will disadvantage small establishments with respect to performance. In order to empirically account for this heterogeneity, they suggest to conduct estimations separately for establishments with 21-100 and 101-1000 employees. A further advantage of a separate consideration of establishments with 21 to 100 employees is that works council rights and bureaucratisation increase with plant size but are a datum within this size class. Finally, the majority of large establishments has set up a works council. By restricting the estimations to a size class where enough variation in works council status is given, it can be avoided that the measured works council effect potentially reflects a size effect only (Addison, 2003). Obviously, the sample sizes of our treatment groups get unacceptably low if further subdivided according to these values.37 We therefore relax our definition of a treatment and focus additionally on the existence of a profit sharing scheme. In chapter 5 we showed that the effects of profit sharing remain more or less unaffected by the modification of the treatment definition. The advantage is that the sample sizes increase overall to 124 (102) in the case of sharing firms with (without) a works council. Restricting the sample to establishments with 21 to 100 employees leaves us with 29 codetermined treated and 45 non-codetermined treated units. Table 14 – Conditional Difference-in-Differences Estimation, Firms with(out) Works Councils, 21-100 employees Establishments with works councils Productivity growth
treaties 0.11
controls 0.06
att 0.05
t-value 0.46
Wilcoxon-Test 0.35
Employment growth Sales growth
0.05 0.15
0.04 0.10
0.02 0.05
0.17 0.35
0.69 0.46
Productivity growth
treaties 0.44
controls 0.15
att 0.30
t-value 1.92*
Wilcoxon-Test 1.41
Employment growth Sales growth
0.00 0.40
0.03 0.12
-0.03 0.28
-0.20 1.86*
-0.08 1.31
Establishments without works councils
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations (controlled remote data access via FDZ).
37
Our treatment samples reduce to 8 (14) treaties with (without) a works council if we restrict our sample to establishments with 21-100 employees only.
Empirical Study
97
Table 14 displays the results of the estimations for the groups of 21 to 100 employees for both subgroups separately. The control firms in both samples, i.e. firms with(out) a works council that have not introduced profit sharing, perform similarly well. We thus find no indication of a negative impact of works councils on the firms’ performance. Interestingly, however, the positive impact of profit sharing in codetermined firms that we found earlier does not seem to hold for small and medium-sized establishments. The results in the group of non-codetermined firms also indicate weak positive effects of profit sharing only thus raising the question whether profit sharing affects all size classes equally or whether larger establishments benefit more from an introduction of profit sharing. In the case of establishments over 100 employees a comparison is problematic due to data limitations. Our sample contains – depending on the definition of the treatment variable – 26 (65) codetermined establishments, but only 2 (5) non-codetermined establishments. This is not surprising in light of the findings of Addison et al. (2003b) who report that in the year 2000 approximately 70 percent of firms with 101-200 employees had a works council. In the size class of 201-500 employees the share was even higher with 80, and in firms with 500 and more employees over 90 percent of establishments. We thus have to refrain unfortunately from investigating the effects of profit sharing in larger non-codetermined establishments. Table 15 reports the results from the matching estimation for codetermined establishments with 101-1000 employees only. The results are very clear-cut indicating that an introduction as well as the existence of profit sharing in larger codetermined firms leads to a significant increase in sales and employment. Table 15 – Conditional Difference-in-Differences Estimation, Firms with Works Councils, 1011000 employees Establishments with works councils Productivity growth
treaties 0.23
controls 0.21
att 0.02
t-value 0.21
Employment growth Sales growth
0.10 0.36
-0.04 0.16
0.14 0.20
2.28** 1.81*
Wilcoxon-Test 0.79 2.21** 2.15**
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations (controlled remote data access via FDZ).
Although economic theory predicts that individual incentive effects are stronger in smaller firms, the findings suggest that large establishments benefit from an introduction of profit sharing whereas there is no effect or a weak effect only in small
98
Profit Sharing and Works Councils: Substitutes or Complements?
and medium-sized firms. One explanation might be that information asymmetries and monitoring costs are higher in larger firms. The benefits from an introduction of an incentive system that aims at reducing these inefficiencies should therefore also be higher among the group of larger firms. Another relevant argument is certainly that larger firms usually have better information and more resources to spend on the design and implementation of a remuneration system. It is conceivable that the costs of implementation can outweigh the productivity effects of profit sharing in small establishments. Large establishments might also be able to create a corporate culture which promotes an unfolding of performance-enhancing effects of profit sharing. Although data limitations prevent us from drawing a final conclusion on the impact of works councils in reaching this aim, our findings are in line with recent research on the effects of works councils by Addison et al. (2000) who find that larger establishments benefit from the existence of a works council, whereas there is no significant impact in the case of smaller firms.
Summary In light of the recently blossoming research on human resource management practices on the one hand and the evergreen discussion on the merits and drawbacks of mandated works councils in Germany on the other hand, it seems natural to ask how both kinds of measures interact. The aim of this paper was to specifically investigate how profit sharing as a management-induced measure as opposed to mandated works councils as an employee-induced measure that both aim at involving employees, jointly affect company performance. From a theoretical standpoint, positive as well as negative interaction effects are possible and the need for empirical investigation arises. First estimations suggest that profit sharing unfolds significantly positive output- and employment-enhancing effects in codetermined firms only, thus implying positive interaction effects. Taking a closer look at the firms introducing profit sharing, however, reveals some very interesting findings. Profit sharing is introduced primarily by large codetermined or small non-codetermined establishments implying that both measures are substitutes in small establishments, whereas they might be complementary in large establish-
Summary
99
ments. Obviously then, firm size seems to play a prominent role and the results have to be further refined. We therefore conduct our estimations separately for establishments with 21 to 100 and 101 to 1000 employees as is suggested in the literature on German works councils. Interestingly, the significantly positive effects of profit sharing disappear for codetermined establishments with 21 to 100 employees. Profit sharing also unfolds weak effects only in the group of small non-codetermined establishments, while the expected effects seem to hold for the group of establishments with 101 to 1000 employees. This raises questions about heterogeneous effects of profit sharing in different size classes, a question that (to the best of my knowledge) has been neglected in research so far. It is not the objective of this thesis to finally answer this question. Instead, this issue should be addressed by future research. Concerning the relationship between profit sharing and works councils, we find no evidence that the existence of a works council has any significant effect on the performance of profit sharing in small and medium-sized firms. In the case of firms with 101 to 1000 employees, profit sharing unfolds significantly positive output- and employment effects in codetermined firms. Due to the lack of large non-codetermined establishments that have introduced profit sharing in our sample, we are unfortunately not able to finally conclude on the complementarity of profit sharing and works councils in large firms. Our results might be in line, however, with some recent studies on works councils that claim no significant effects of works councils in small, but a significantly positive impact in large establishments. In spite of the contributions of our study, shortcomings could unfortunately not be avoided and are primarily due to data limitations. One major issue are certainly the small sample sizes owing to the very data-demanding methods and the fact that the variables used are not surveyed annually. We do believe, however, that our approach not only takes account of the major methodological problems but also reveals some interesting and new findings and thus should be reapplied and extended as more data or other more extensive datasets become available.
Conclusions and Outlook Although not a new topic, employee financial participation has gained renewed interest in Europe in the last decade. To a large extent, this development can be ascribed to the promotion of the issue by the European Commission since its so-called PEPPER Recommendation in 1992 which called upon member states to encourage the use of such schemes by providing adequate information to the relevant parties and to consider introducing fiscal incentives. The expectations of financial participation are thereby manifold, reaching from economic over employment and social objectives. A recurrent theme in recent discussions of financial participation among policymakers and the social partners is the need for more research into the outcomes of financial participation schemes. The objective of this thesis was to contribute to this lack of empirical research by investigating the effects of profit sharing on various measures of company performance based on large-scale German firm panel data. Quite a number of studies have analysed the productivity effects of profit sharing and come to the surprisingly clear result that across countries profit sharing is associated with higher productivity levels. Only limited and, moreover, inconclusive evidence is available on how profit sharing affects other outcomes such as employment, wages or profitability. One contribution of this study is the provision of a comprehensive instead of so far mostly patchy picture of profit sharing. In the first part of the study, we find that an introduction of profit sharing is associated with a significantly higher sales and employment growth. Productivity, however, measured as the ratio of sales to employment is not affected. At first glance, this result seems somewhat surprising. It is, however, in line with several other studies that find neutral productivity effects. Using a simple theoretical model, Kraft and Ugarkovic (2005) explain this result by arguing that profit sharing leads to an increase in efficiency which might result in lower unit costs. Firms can increase their profits by expanding output and employment. With a declining marginal product of labour, however, the estimated
102
Conclusions and Outlook
productivity level might not differ very much from the one before the introduction of the remuneration system. But both output and employment would be larger. A priori, it is not clear whether these findings are the result of increased productivity or rather lower wages or a slower wage growth. In the second part of the study, we therefore inspect how wages are affected by an introduction of profit sharing. In accordance with the assumption in the theoretical model, we find that profit sharing is a supplementary payment and does not substitute fixed wages as proposed by Weitzman. Obviously then, profit sharing is introduced by firms as a wage incentive that pays for both employers and employees. Presumably, employers will introduce such an incentive only if the benefits outweigh the costs associated with the scheme. The third part of the study hence deals with the impact of profit sharing on firm profitability and shows that profit sharing is indeed associated with an improvement in firms’ financial performance. Finally, we pose the so far neglected question on the role of works councils as a form of representative participation on the success of profit sharing. From a theoretical perspective, positive as well as negative interaction effects are possible. Several interesting results emerge from our empirical study: First, both measures are primarily found in large establishments, whereas small and medium-sized firms have usually only either one of them. This implies that firm size obviously plays a prominent role which has to be accounted for empirically. Conducting our estimations for establishments with 21-100 and 101-1000 employees separately reveals that profit sharing yields beneficial effects in large establishments only. In small and mediumsized firms, profit sharing has no significant influence on the observed outcome variables which raises the question of heterogeneous effects of profit sharing in different size classes. With respect to the impact of works councils on the effects of profit sharing, the results show no significant impact among the group of smaller establishments. In the case of larger establishments, unfortunately, we do not observe enough large co-determined firms to finally conclude on this issue. Our results might be in line, however, with some recent empirical literature on works councils which finds positive effects of works councils in large and neutral effects in small and medium-sized firms. It would go beyond the scope of this thesis to further investigate these questions, but future research should certainly address them in more depth.
Conclusions and Outlook
103
Before giving an outlook on future research questions, a note should be made on another contribution of this study. The existing empirical literature on profit sharing has been flawed in several respects which can be mainly attributed to data limitations. Most studies suffer from small sample sizes, the lack of representative data and the observation of the existence rather than an introduction of profit sharing. The crosssectional design of the majority of studies further brings along problems with respect to endogeneity and unobserved heterogeneity. Our analysis indeed shows that firms which introduce profit sharing differ significantly from their non-sharing counterparts. Selectivity effects are hence at work that need to be accounted for methodologically. The majority of studies neglects this issue which raises doubts about the reliability of their estimates. To improve on these drawbacks of the existing literature, we apply (to the best of our knowledge for the first time in connection with profit sharing) the conditional difference-in-differences estimator to the nationally representative German IAB Establishment Panel. As just mentioned, our findings particularly in the last part of the study provide directions for future research. On the one hand, our estimations suggest that larger establishments are the ones that primarily introduce and also benefit from an introduction of such a remuneration scheme. The relationship between profit sharing and firm size has been only touched upon in a few studies. Another hitherto neglected point worth further investigations is the impact of representative participation, i.e. unions and works councils, on the effects of profit sharing. Due to data limitations, we are not able to finally conclude on this question, but believe that our approach is worth a repetition as more extensive data becomes available. A new related research strand further argues that human resource management practices such as performance-related pay, teamwork, quality circles, training etc. are not introduced separately but rather in bundles of complementary measures. Several studies try to identify how different (bundles of) measures interact, but at present no conclusive picture emerges. This might be due to several reasons: Different studies investigate different measures depending on data availability and, contrary to profit sharing, most other human resource management practices are not ‘standardized’. We, nonetheless, believe this is another important research area that needs to be pursued in the future. Returning to the initial chapter where we defined profit sharing, we had argued that profit sharing can take different forms, specifically concerning the timing and the mode of payment. Presently, only very limited information is available on the incidence, determinants
104
Conclusions and Outlook
and effects of different forms of profit sharing. Despite the contributions of this study, there are thus several points on the research agenda that need to be addressed by future research. What policy implications emerge from the results presented in this study? The main rationale for economists to intervene in markets are positive or negative externalities. In both cases, economic agents do not capture the full impact of their behaviour. Public policy might thus correct an existing market failure. The results emerging from the present study indicate that an introduction of profit sharing is associated with higher sales and employment growth (via higher productivity), higher wages and an improved financial performance of firms. If profit sharing obviously yields such positive effects, why is it then that only a minority of firms considers introducing such a compensation scheme in countries where governments do not intervene into the market process by providing tax incentives? A review of the literature on monetary incentives provides an explanation for this phenomenon which is also confirmed by our data: Firms view profit sharing as one among several alternative incentive schemes. Depending on different circumstances such as e.g. establishment size, qualification level of the work force, work content, verification possibilities of individual performance, industrial relations or market conditions, firms will introduce the compensation scheme that best suits their needs. Firms with specific advantages in the application of profit sharing introduce it, while others have no interest in such a kind of incentive and rely in contrast on other motivational instruments like piece rates, promotions or efficiency wages. Since all benefits from an implementation of profit sharing accrue to the firm concerned and no obvious externalities stem from its use, there is no argument for outlays from the public purse. One could make a case, however, for limited government support to inform firms about the potential benefits of profit sharing and the circumstances that are necessary for its implementation to be successful. As to the macro-level employment effects of profit sharing proposed by Weitzman which might be worthy of subsidy, the existing empirical literature is limited and inconclusive with respect to both the impact on employment variability as well as employment levels. In this study, we investigate the effect on employment levels only and find higher employment growth in firms introducing profit sharing. Our analysis indicates, however, that employment is increased via higher productivity and not
Conclusions and Outlook
105
higher pay flexibility as suggested by Weitzman. More research particularly for Europe is therefore needed on this question to produce reliable information for policy. While this study addresses profit sharing and its effects primarily in terms of a wage incentive, the European Commission’s quotation in the introductory passage to this study emphasizes that its implications are wider than that: In times of increased competition, the cooperative efforts of the workforce as a whole are gaining in importance. Profit sharing might be a way of creating a stronger association of employees with the matters and fortunes of their firms which may be an important ingredient in achieving the Lisbon objectives.
List of Appendices A Appendix to Chapter 4
108
B Appendix to Chapter 5
110
C Appendix to Chapter 6
113
D Appendix to Chapter 7
115
E Appendix to Chapter 8
117
A Appendix to Chapter 4 A.1
Variable Definition
Variable Number of employees Establishment size <20 Establishment size 20-49 Establishment size 50-249 Establishment size 250-499 Establishment size 500+ Exporter Share qualified employees Collective Bargaining Works Council Limited liability
Age East German establishment Shift responsibilities
Teamwork Profit/Cost Center ICT Investment
Competition
Variable Definition Total number of employees reported by the establishment Dummy=1 if establishment reports to have less than 20 employees Dummy=1 if establishment reports to have between 20 and 49 employees Dummy=1 if establishment reports to have between 50 and 249 employees Dummy=1 if establishment reports to have between 250 and 499 employees Dummy=1 if establishment reports to have more than 500 employees Dummy=1 if part of revenues is generated outside Germany degree Dummy=1 if establishment reports to be covered by an industry or company wide collective wage agreement Dummy=1 if establishment reports to have a works council Dummy=1 if establishment reports to be a joint stock company (AG) or a non-public limited liability company (GmbH) Dummy=1 if establishment was founded after 1990 Dummy=1 if establishment is located in East Germany Dummy=1 if establishment reports to have shifted responsibilites and decisions to lower levels of hierarchy in the last two years Dummy=1 if establishment reports to have introduced teamwork in the last two years Dummy=1 if establishment reports to have introduced profit or cost centers in the last two years Dummy=1 if establishment reports to have invested into communication technology and data processing in the last business year Dummy=1 if establishment reports to operate unter medium or high degree of competition
Appendix to Chapter 4
A.2
109
Mean Industry Comparison of Firms with(out) Profit Sharing
Variable
Firms with profit sharing (N=1170)
Firms without profit sharing (N=9055)
Agriculture, Forestry and Fishing
0.01
0.03***
Mining and quarrying, energy and water supply
0.01
0.01
Food and Tobacco
0.03
0.04**
Manufacture of Textiles, Wearing Apparel, Paper, Publishing, Printing
0.04
0.05**
Manufacture of Wood, Chemicals, Rubber, Plastic Products, Glass, Ceramics, Non-metallic Mineral Products
0.11
0.12
Manufacture of Metals, Machinery, Recycling, Vehicles, Electrical Machinery, Precision and Optical
0.23
0.14***
Construction
0.09
0.14***
Wholesale and Retail Trade; Repair
0.20
0.17**
Transport and communications
0.05
0.05
Banking and Insurance
0.01
0.00
Hotels and Restaurants
0.02
0.04***
Education
0.01
0.01*
Health
0.03
0.06***
Data Processing
0.03
0.01***
Research and Development
0.01
0.00
Consulting, Advertising
0.03
0.03
Real estate and housing
0.01
0.01
Renting, other services
0.08
0.05***
Other community, social and personal services Non-Profit Sector
0.03 0.00
0.04 0.00***
*/**/*** indicate that means differ with statistical significance in a two-tailed t-test at the 10%, 5%, or 1% level between the establishments from the two groups. Source: IAB Establishment Panel, wave 2000, own calculations
B Appendix to Chapter 5 B.1
Results from Conditional Existence of Profit Sharing
Difference-in-Differences-Estimation,
Nearest neighbour matching outcome variable productivity growth employment growth sales growth
treaties .22 .09 .29
controls .19 -.01 .15
att .03 .10 .14
t-value 0.46 1.97** 1.99**
treaties .22 .09 .29
controls .21 -.02 .15
att .01 .11 .14
t-value 0.28 3.18*** 2.86***
Kernel matching outcome variable productivity growth employment growth sales growth
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations.
B.2
Mean Comparison of Firms with(out) Profit Sharing
Variable Number of employees Competition Exporter Ict investment Shift responsibilities Teamwork Independent work groups Share qualified employees Collective bargaining Works council Limited liability Age East German establishment Establishment size <20 Establishment size 20-49 Establishment size 50-249 Establishment size 250-499 Establishment size 500+
firms with profit sharing 402 .93 .41 .67 .37 .28 .23 .69 .76 .55 .76 .39 .41 .23 .18 .32 .10 .17
firms without profit sharing 78*** .86*** .14*** .42*** .17*** .11*** .07*** .58*** .54*** .21*** .45*** .44 .56*** .56*** .20 .17*** .04*** .03***
matched firms without profit sharing 228 .94 .37 .68 .29 .24 .15 .67 .67 .42* .75 .41 .42 .24 .24 .33 .08 .11
*/**/*** indicate that means differ with statistical significance in a two-tailed t-test at the 10%, 5% or 1% level between the establishments from the two groups. Source: IAB Establishment Panel, wave 1998, own calculations.
Appendix to Chapter 5
B.3
111
Frequency Distribution of Propensity Scores 110 100
number of observations
90 80 70 60 50 40 30 20 10
0,
9 89 0,
63
6
45
95
45
95
4
1
37
11 0,
50 ,1 -0
,4 -0
0,
1
6
6
13
1 ,6 -0
-0
,9
75
38
6
1
00
63
,2
,4 -1
-1
1
6 25 ,7
,9 -1
,2 -2
-1
50
88
6
1
6
13 ,5
,7 -2
-2
38
00
,0 -3
,3 -3
75
1
6
0
u n b o u n d e d p r o p e n si ty sc o r e gb= 0
B.4
gb= 1
Results of Probit Estimation, Existence of Profit Sharing
Variables Establishment size 20-49 Establishment size 50-249 Establishment size 250-499 Establishment size 500+ Competition Exporter Share of qualified employees Collective bargaining Works council Limited liability Age East German Establishment Shift of responsibilities Teamwork Independent work groups Ict investment intercept Number of observations Pseudo R2
Coefficient .099 .342 .273 .516 .326 .236 .521 .162 .289 .298 .412 -.495 .188 .055 .439 .171 -1.886 1923 0.20
t-value 0.79 2.34** 1.27 2.32** 2.11** 1.99** 2.91*** 1.52 2.23** 2.93*** 3.89*** -4.70*** 1.79* 0.45 3.43*** 1.82* -3.76***
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Values for industry dummies are not included and are available from the author on request. Source: IAB Establishment Panel, wave 1998, own calculations.
Appendix to Chapter 5
112
B.5 Results of Conditional Difference-in-Differences-Estimation (incl. Capital Stock), Existence of Profit Sharing Nearest neighbour matching outcome variable
treaties
controls
att
Productivity growth
0.19
0.24
-0.06
-1.00
Employment growth
0.11
-0.04
0.16
2.85***
Sales growth
0.28
0.16
0.12
1.72*
t-value
Kernel matching outcome variable
treaties
controls
att
Productivity growth
0.19
0.20
-0.01
-0.36
Employment growth
0.11
0.00
0.11
2.11**
Sales growth
0.28
0.16
0.12
1.80*
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations.
t-value
C Appendix to Chapter 6 C.1
Results of Probit Estimation
Variables Establishment size 20-49 Establishment size 50-249 Establishment size 250-499 Establishment size 500+ Competition Exporter Share of qualified employees Collective bargaining Works council Limited liability Age East German Establishment Shift of responsibilities Teamwork Independent work groups Ict investment intercept Number of observations Pseudo R2
Coefficient 0.11 0.37 0.18 0.46 0.08 0.36 0.44 0.24 0.33 0.18 0.35 -0.36 0.27 -0.08 0.33 0.21 -3.15 1731 0.18
t-value 0.59 1.83* 0.66 2.63*** 0.39 2.21** 1.81* 1.58 1.77* 1.26 2.41** -2.49** 1.88* -0.45 1.92* 1.56 -5.44***
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Values for industry dummies are not included and are available from the author on request. Source: IAB Establishment Panel, wave 1998, own calculations (controlled remote data access via FDZ).
Appendix to Chapter 6
114
C.2
Frequency Distributions of Propensity Scores
Propensity Score Firms with Profit Sharing
Propensity Score Firms without Profit Sharing
.0154
.0021
10 Percentile
.0242
.0036
15th Percentile
.0374
.0048
20th Percentile
.0410
.0062
25th Percentile
.0439
.0079
5th Percentile th
th
30 Percentile
.0537
.010
35th Percentile
.0577
.0122
40th Percentile
.0712
.0148
45th Percentile
.0883
.0172
50th Percentile
.1014
.0209
55 Percentile
.1296
.0248
60th Percentile
.1530
.0296
65th Percentile
.1595
.0353
70th Percentile
.1824
.0428
75 Percentile
.1913
.0516
80th Percentile
.2385
.0635
85th Percentile
.2625
.0866
90th Percentile
.3339
.1163
95th Percentile
.3717
.1635
100th Percentile
.4358
.5814
th
th
Source: IAB Establishment Panel, wave 1998, own calculations (controlled remote data access via FDZ).
D Appendix to Chapter 7 D.1
Results of Probit Estimation
Variables Establishment size 20-49 Establishment size 50-249 Establishment size 250-499 Establishment size 500+ Competition Exporter Share of qualified employees Collective bargaining Works council Limited liability Age East German Establishment Shift of responsibilities Teamwork Independent work groups ICT investment intercept Number of observations Pseudo R2
Coefficient 0.039 0.310 0.270 0.436 0.015 0.403 0.265 0.288 0.266 0.150 0.326 -0.256 0.223 -0.094 0.327 0.221 -3.077 1918 0.17
t-value 0.22 1.56 1.01 1.49 0.08 2.56** 1.16 2.00** 1.50 1.13 2.32** -1.86* 1.63 -0.56 1.96** 1.76* -5.63***
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Values for industry dummies are not included and are available from the author on request. Source: IAB Establishment Panel, wave 1998, own calculations.
Appendix to Chapter 7
116
D.2
Frequency Distributions of Propensity Scores Propensity Score Firms with Profit Sharing
Propensity Score Firms without Profit Sharing
5th Percentile
0.010
0.003
10th Percentile
0.020
0.004
15 Percentile
0.029
0.006
20th Percentile
0.036
0.007
25th Percentile
0.042
0.009
30th Percentile
0.052
0.011
35th Percentile
0.057
0.013
40 Percentile
0.068
0.015
45th Percentile
0.089
0.018
50th Percentile
0.104
0.021
55th Percentile
0.119
0.024
60 Percentile
0.134
0.029
65th Percentile
0.152
0.034
70th Percentile
0.170
0.041
75th Percentile
0.199
0.049
80th Percentile
0.228
0.061
85th Percentile
0.248
0.080
90 Percentile
0.308
0.107
95th Percentile
0.336
0.156
100th Percentile
0.409
0.519
th
th
th
th
Source: IAB Establishment Panel, wave 1998, own calculations (controlled remote data access via FDZ).
E Appendix to Chapter 8 E.1
Results of Conditional Difference-in-Differences-Estimation, Existence of Profit Sharing, 21-100 Employees Establishments with works councils
Productivity growth
treaties 0.10
controls 0.18
att -0.08
t-value -0.43
Wilcoxon-Test -0.04
Employment growth Sales growth
-0.06 0.02
0.00 0.16
-0.06 -0.14
-0.92 -0.90
-1.04 -0.75
Productivity growth
treaties 0.36
controls 0.22
att 0.15
t-value 0.68
Wilcoxon-Test 0.87
Employment growth Sales growth
0.01 0.36
-0.03 0.12
0.04 0.24
0.65 1.65*
0.95 1.90*
Establishments without works councils
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations (controlled remote data access via FDZ).
E.2
Results of Conditional Difference-in-Differences Existence of Profit Sharing, 101-1000 Employees Establishments with works councils
Productivity growth
treaties 0.26
controls 0.12
att 0.14
t-value 1.81*
Wilcoxon-Test 2.29**
Employment growth
0.05
-0.04
0.09
2.38**
1.60
Sales growth
0.32
0.05
0.27
3.29***
2.78***
***/**/* indicates statistical significance at the 1%, 5% and 10% level. Source: IAB Establishment Panel, waves 1998-2002, own calculations (controlled remote data access via FDZ).
Estimation,
Bibliography Adams, C.P. (2006): Optimal Team Incentives with CES Production. Economics Letters 92, 143-148. Addison, J.T. (2003): The Course of Research into the Economic Consequences of German Works Councils, IZA Discussion Paper No. 878, Bonn. Addison, J.T., C. Schnabel and J. Wagner (1997): On the Determinants of Mandatory Works Councils in Germany. Industrial Relations 36, 419-445. Addison, J.T., W.S. Siebert, J. Wagner and Xiangdong Wei (2000): Worker Participation and Firm Performance: Evidence from Germany and Britain. British Journal of Industrial Relations 38, 7-48. Addison, J.T., C. Schnabel and J. Wagner (2001): Works councils in Germany: their effects on establishment performance. Oxford Economic Papers 53, 659-694. Addison, J.T., T. Schank, C. Schnabel, and J. Wagner (2003a): German Works Councils in the Production Process. IZA Discussion Paper No. 812, Bonn. Addison, J.T., L. Bellmann, C. Schnabel and J. Wagner (2003b): German Works Councils Old and New: Incidence, Coverage and Determinants. Schmollers Jahrbuch 123, 339-358. Addison, J.T., L. Bellmann, C. Schnabel and J. Wagner (2004): The Reform of the German Works Constitution Act: A Critical Assessment. Industrial Relations 43, 392-407. Akerlof, G. and J. Yellen (1986): Efficiency Wage Models of the Labor Market. New York: Cambridge University Press. Alchian, A. and H. Demsetz (1972): Production, information costs, and economic organization. American Economic Review 62, 777-795. Amisano, G. and A. Del Boca (2004): Profit related pay in Italy. International Journal of Manpower 25, 463-478. Angrist, J. (1998): Estimating the Labor Market Impact of Voluntary Military Service using Social Security Data on Military Applicants. Econometrica 66, 249-288.
120
Bibliography
Arellano, M. and S. Bond (1991): Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations. Review of Economic Studies 58, 277-297. Arnal, E., W. Ok and R. Torres (2001): Labour Market and Social Policy, Occasional Papers No. 50, Paris: Organisation for Economic Co-operation and Development. Arranz-Aperte, L. and A. Heshmati (2003): Determinants of Profit Sharing in the Finnish Corporate Sector. IZA Discussion Paper No. 776. Bonn. Ashenfelter, O. (1978): Estimating the Effects of Training Programs on Earnings. Review of Economics and Statistics 60, 47-57. Axelrod, R. (1984): The Evolution of Cooperation. New York: Basic Books. Baker, G., M. Gibbs, and B. Holmström (1994a): The Wage Policy of a Firm. Quarterly Journal of Economics 109, 921-955. Baker, G., M. Gibbs, and B. Holmström (1994b): The Internal Economics of a Firm: Evidence from Personnel Data. Quarterly Journal of Economics 109, 881-919. Bassi, L. (1983): The effect of CETA on the post-program earnings of participants. Journal of Human Resources 18, 539-556. Bassi, L. (1984): Estimating the effects of training programs with non-random selection. Review of Economics and Statistics 66, 36-43. Bauer, T. (2003): Flexible Workplace Practices and Labor Productivity. IZA Discussion Paper No. 700, Bonn. Bell, L.A. and D. Neumark (1993): Lump-Sum Payments and Profit-Sharing Plans in the Union Sector of the United States Economy. The Economic Journal 103, 602-619. Bellmann, L. (1997): The IAB-Establishment Panel with an Exemplary Analysis of Employment Expectations. IAB Topics No. 20, Nürnberg. Bhargava, S. (1994): Profit Sharing and the Financial Performance of Companies: Evidence from U.K. Panel Data. The Economic Journal 104, 1044-1056. Bhargava, S. and T. Jenkinson (1995): Explicit Profit Sharing and the Determination of Wages: Microeconomic Evidence from the UK. Labour 9, 73-95.
Bibliography
121
Black, S.E. and L.M. Lynch (2000): What’s Driving the New Economy: The Benefits of Workplace Innovation. NBER Working Paper No. 7479. Cambridge. Black, S.E. and L.M. Lynch (2004): What’s Driving the New Economy: The Benefits of Workplace Innovation. The Economic Journal 114, F97-F116. Blanchflower, D.G. and A.J. Oswald (1988): Profit-Related Pay: Prose Discovered? The Economic Journal 98, 720-730. Blundell, R. and M. Costa Dias (2000): Evaluation Methods for Non-Experimental Data. Fiscal Studies 21, 427-468. Blundell, R., L. Dearden, and B. Sianesi (2004): Evaluating the Impact of Education on Earnings in the UK: Models, Methods and Results from the NCDS, Working Paper No. 03/20, The Institute for Fiscal Studies, London. Bound, J., D. Jaeger, and R. Baker (1995): Problems with Instrumental Variables estimation when the Correlation between the Instruments and the Endogenous Explanatory Variable is weak. Journal of the American Statistical Association 90, 443-450. Bradley, K. and S. Estrin (1992): Profit Sharing in the British Retail Trade Sector: The Relative Performance of the John Lewis Partnership. Journal of Industrial Economics 40, 291-304. Brown, S., F. Fakhfakh and J.S. Sessions (1999): Absenteeism and Employee Sharing: An Empirical Analysis on French Panel Data, 1981-1991. Industrial and Labor Relations Review 52, 234-251. Bryson, A., R. Dorsett, and S. Purdon (2002): The Use of Propensity Score Matching in the Evaluation of Active Labour Market Policies. Working Paper Number 4, Department for Work and Pensions. Bundesministerium der Justiz (2006): Betriebsverfassungsgesetz, http://bundesrecht. juris.de/betrvg/index.html (date of download: March 2, 2006). Bundesministerium der Justiz (2006a): Fünftes Gesetz zur Förderung der Vermögensbildung der Arbeitnehmer, http://bundesrecht.juris.de/bundesrecht/vermbg_2/ gesamt.pdf (date of download: March 11, 2006). Bundesministerium der Justiz (2006b): Einkommenssteuergesetz, http://bundesrecht. juris.de/bundesrecht/estg/gesamt.pdf (date of download: March 11, 2006).
122
Bibliography
Burritt. A.W. et al. (1918): Profit Sharing: Its Principles and Practice. New York: Harper and Brothers. Cable, J. and F. FitzRoy (1980): Productive Efficiency, Incentives and Employee Participation: Some Preliminary Results for West Germany. Kyklos 33, 100-121. Cable, J. and N. Wilson (1989): Profit Sharing and Productivity: An Analysis of UK Engineering Firms. The Economic Journal 99, 550-555. Cable, J. and N. Wilson (1990): Profit Sharing and Productivity: Some further Evidence. The Economic Journal 100, 550-555. Cahuc, P. and B. Dormont (1997): Profit-sharing: Does it increase productivity and employment? A theoretical model and empirical evidence on French micro data. Labour Economics 4, 293-319. Caliendo, M. and R. Hujer (2006): The Microeconometric Estimation of Treatment Effects – An Overview. Journal of the German Statistical Society/ Allgemeines Statistisches Archiv 90, 197-212. Caliendo, M. and S. Kopeinig (2006): Some Practical Guidance for the Implementation of Propensity Score Matching, forthcoming in Journal of Economic Surveys. Che, Y.-K. And S.-W. Yoo (2001): Optimal Incentives for Teams. American Economic Review 91, 525-541. Coates, E.M. (1991): Project Sharing Today: Plans and Provisions. Monthly Labor Review, April, 19-25. Carstensen, V., K. Gerlach, and O. Hübler (1995): Profit Sharing in German Firms, in: Buttler, F., W. Franz, and R. Schettkat (eds.): Institutional Frameworks and Labour Market Performance. Comparative Views on the US and German Economies. London: Routledge, 168-207. Cheadle, A. (1989): Explaining patterns of profit sharing activity. Industrial Relations 28, 387-400. Chelius, J. and R.S. Smith (1990): Profit Sharing and Employment Stability. Industrial and Labor Relations Review 43, 256-273. Commission of the European Communities (2000): Communication from the Commission to the Council, the European Parliament, the Economic and Social Committee and the Committee of the Regions. Social Policy Agenda. COM
Bibliography
123
(2000) 379 final, Brussels, http://europa.eu.int/comm/employment_social/social_ policy_agenda/com379_en.pdf (date of download: Feb 13, 2006). Commission of the European Communities (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. COM (2002) 364 final, Brussels, http://europa.eu.int/comm/employment_social/labour_law/docs/ com2002_364_ en.pdf (date of download: Feb 13, 2006). Council of the European Communities (1992): Council Recommendation of 27 July 1992 concerning the promotion of participation by employed persons in profits and enterprise results (including equity participation). Official Journal of the European Communities No. L245/53, 26 August 1992, 53-55. Cooke, W.N. (1994): Employee Participation Programs, Group-Based Incentives, And Company Performance: A Union-Nonunion Comparison. Industrial and Labor Relations Review 47, 594-609. D’Art, D. (1992): Economic Democracy and Financial Participation. A Comparative Study. London: Routledge. Dehejia, R. and S. Wahba (1999): Causal Effects in Nonexperimental Studies: Reevaluating the Evaluation of Training Programs. Journal of American Statistical Association 94, 1053-1062. Dehejia, R. and S. Wahba (2002): Propensity Score Matching Methods for NonExperimental Causal Studies. Review of Economics and Statistics 84, 151-161. Dess, G.G. and R.B. Robinson (1984): Measuring Organizational Performance in the Absence of Objective Measures: The Case of the Privately-held Firm and Conglomerate Business Unit. Strategic Management Journal 5, 265-273. Doucouliagos, C. (1995): Worker Participation and Productivity in labor-managed and participatory capitalist firms: A meta-analysis. Industrial and Labor Relations Review 49, 58-78. Drago, R. and G.K. Turnbull (1988): Individual versus Group Piece Rates under Team Technologies. Journal of the Japanese and International Economies 2, 1-10.
124
Bibliography
Drago, R. and J. Heywood (1995): The Choice of Payment Schemes: Australian Establishment Data. Industrial Relations 34, 507-531. Drago, R. and G.K. Turnbull (1996): On the Incidence of Profit Sharing. Journal of Economic Behavior & Organization 31, 129-138. Drago, R. and G.T. Garvey (1998): Incentives for Helping on the Job: Theory and Evidence. Journal of Labor Economics 16, 1-25. Estrin, S., P. Grout, and S. Wadhwani (1987): Profit-sharing and employee share ownership. Economic Policy 2, 13-62. Federal Ministry of Labour and Social Affairs (1990): Co-determination in the Federal Republic of Germany. Bonn: Bundesminister für Arbeit und Sozialordnung. FitzRoy, F. and K. Kraft (1985a): Mitarbeiterbeteiligung und Produktivität: Eine ökonometrische Untersuchung. Zeitschrift für Betriebswirtschaft 55, 21-36. FitzRoy, F.R. and K. Kraft (1985b): Unionization, wages, and efficiency: Theory and evidence from the U.S. and West Germany. Kyklos 38, 537-554. FitzRoy, F. and K. Kraft (1986): Profitability and Profit Sharing. The Journal of Industrial Economics 35, 113-130. FitzRoy, F. and K. Kraft (1987a): Cooperation, Productivity and Profit sharing. Quarterly Journal of Economics 102, 23-35. FitzRoy, F. and K. Kraft (1987b): Efficiency and Internal Organization: Works Councils in West German Firms. Economica 54, 493-504. FitzRoy, F. R. and K. Kraft (1990): Innovation, rent-sharing and the organization of labor in the Federal Republic of Germany. Small Business Economics 2, 95-103. FitzRoy, F. and K. Kraft (1995): On the Choice of Incentives in the Firm. Journal of Economic Behavior and Organization 26, 145-160. Florkowski, G.W. (1987): The Organizational Impact of Profit Sharing. Academy of Management Review 12, 622-636. Florkowski, G.W. (1994): Employment Growth and Stability under Profit-Sharing: A Longitudinal Study. British Journal of Industrial Relations 32, 303-318.
Bibliography
125
Föhl, C. (1964): Kreislaufanalytische Untersuchung der Vermögensbildung in der Bundesrepublik und der Beeinflussbarkeit ihrer Verteilung. Tübingen: Mohr. Freeman, R.B. and J.L. Medoff (1979): The two faces of unionism. The Public Interest 57, 69-93. Freeman, R.B. and E.P. Lazear (1995): An Economic Analysis of Works Councils, in: Rogers, J. and W. Streeck (eds.), Works Councils – Consultation, Representation and Cooperation in Industrial Relations. Chicago: Chicago University Press, 2752. Freeman, R.B., D. Kruse and J. Blasi (2004): Monitoring Colleagues at Work: Profit Sharing, Employee Ownership, Broad-Based Stock Options and Workplace Performance in the United States. CEP Discussion Paper No. 647, London School of Economics and Political Science. Frick, B. and I. Möller (2003): Mandated Works Councils and Firm Performance: Labor Productivity and Personnel Turnover in German Establishments. Schmollers Jahrbuch 123, 423-454. Fudenberg, D. and E. Maskin (1986): The Folk Theorem in Repeated Games with Discounting or with Incomplete Information. Econometrica 54, 533-554. Gibbons, R. (1987): Piece-rate Incentive Schemes. Journal of Labor Economics 5, 413-429. Gibbs, M. and W. Hendricks (2004): Do Formal Salary Systems really matter?. Industrial and Labor Relations Review 58, 71-93. Greene, W.H. (2003): Econometric Analysis. Upper Saddle River: Pearson Education. Gregg, P.A. and S.J. Machin (1988): Unions and the Incidence of Performance Linked Pay Schemes in Britain. International Journal of Industrial Organization 6, 91107. Groshen, E.L. (1991): Five Reasons Why Wages Vary Among Employers. Industrial Relation 30, 350-381. Handel, M.J. and D.I. Levine (2004): Editor’s Introduction: The Effects of New Work Practices on Workers. Industrial Relations 43, 1-43. Hashimoto, M. (1979): Bonus payments, on-the-job training, and lifetime employment in Japan. Journal of Political Economy 87, 1086-1104.
126
Bibliography
Hart, R.A. and O. Hübler (1991): Are Profit Shares and Wages Substitute or Complementary Forms of Compensation? Kyklos 44, 221-231. Hausman, J.A. and D.A. Wise (1985): Social Experimentation. Chicago: University of Chicago Press for National Bureau of Economic Research. Heckman, J.J. (1976): The Common Structure of Statistical Models of Truncation, Sample Selection and Limited Dependent Variables and a Simple Estimator for such Models. Annals of Economic and Social Measurement 5, 475-492. Heckman, J.J. (1979): Sample Selection Bias as a Specification Error. Econometrica 47, 153-161. Heckman, J.J., H. Ichimura, and P. Todd (1997): Matching As An Econometric Evaluation Estimator: Evidence from Evaluating a Job Training Program. Review of Economic Studies 64, 605-654. Heckman, J. J., H. Ichimura, J.A. Smith, and P. Todd (1998): Characterizing Selection Bias Using Experimental Data. Econometrica 65, 1017-1098. Heckman, J.J., H. Ichimura, and P. Todd (1998): Matching as an Econometric Evaluation Estimator. Review of Economic Studies 65, 261-294. Heckman, J.J., R.J. LaLonde, and J.A. Smith (1999): The Economics and Econometrics of Active Labour Market Programs, in O. Ashenfelter and D. Card (eds): Handbook of Labor Economics 3a. Amsterdam: Elsevier, 1865-2097. Heckman, J.J. and S. Navarro-Lozano (2004): Using Matching, Instrumental Variables, and Control Functions to estimate Economic Choice Models. Review of Economics and Statistics 86, 30-57. Heywood, J.S., O. Hübler and U. Jirjahn (1998): Variable Payment Schemes and Industrial Relations: Evidence from Germany. Kyklos 51, 237-257. Heywood, J.S. and U. Jirjahn (2002): Payment Schemes and Gender in Germany. Industrial and Labor Relations Review 56, 44-64. Heywood, J.S. and U. Jirjahn (2004): Teams, Teamwork and Absence. Scandinavian Journal of Economics 106, 765-782. Heywood, J.S. U. Jirjahn, and G. Tsertsvadze (2005a): Does Profit Sharing Reduce Conflict with the Boss? Evidence from Germany. International Economic Journal 19, 235-250.
Bibliography
127
Heywood, J.S. U. Jirjahn, and G. Tsertsvadze (2005b): Getting Along with Colleagues – Does Profit Sharing Help or Hurt?. Kyklos 58, 557-573. HM Treasury (2000): Finance Bill, http://www.hm-treasury.gov.uk/consultations_ and_legislation/financebill2000/consult_finance_clause49_2000.cfm (date of download: January 5, 2006). HM Treasury (2001): Budget 2001: Press Notices, http://www.hmtreasury.gov.uk/ budget/budget_2001/press_notices/bud_bud01_pressshare.cfm (date of download: January 5, 2006). Holland, P.W. (1986): Rejoinder. Journal of the American Statistical Association 81, 968-970. Holmström, B. (1982): Moral Hazard in Teams. Bell Journal of Economics 13, 324340. Hübler, O. (1995): Produktivitätssteigerung durch Mitarbeiterbeteiligung in Partnerschaftsunternehmen?. Mitteilungen zur Arbeitsmarkt- und Berufsforschung 28, 214-223. Hübler, O. and U. Jirjahn (1998): Zeit-, Leistungs- und Gruppenentlohnung. Empirische Untersuchungen mit Betriebsdaten zur Entlohnungsart, in: Gerlach, K., O. Hübler, and W. Meyer (eds.): Ökonomische Analysen betrieblicher Strukturen und Entwicklungen: Das Hannoveraner Firmenpanel. Frankfurt am Main: Campus, 148-172. Hujer, R., K.O. Maurer, and M. Wellner (1997): The Impact of Training on Unemployment Duration in West Germany. Discussion Paper No. 74, Department of Economics, University of Frankfurt. Huselid, M. (1995): The Impact of Human Resource Management Practices on Turnover, Productivity, and Corporate Financial Performance. Academy of Management Journal 38, 635-672. IAB (2002): Die Verbreitung von Kapital- und Gewinnbeteiligungsmodellen, http://doku.iab.de/betriebspanel/ergebnisse/ 2002_09_01_05_mitarbeiter beteiligung2001.pdf (date of download: Feb 28, 2005). Imbens, G.W. (2004): Nonparametric Estimation of Average Treatment Effects under Exogeneity: A Review. The Review of Economics and Statistics 86, 4-29.
128
Bibliography
Jensen, M. and W. Meckling (1979): Rights and Production Functions. An Application to Labour-Manages Firms and Codetermination. Journal of Business 52, 469-506. Jirjahn, U. (1998): Effizienzwirkungen von Erfolgsbeteiligung und Partizipation, New York: Campus. Jirjahn, U. (2000): Incentives for Multitasking: Fixed Wages or Profit-sharing? Economic Analysis 3, 137-148. Jirjahn, U. and T. Klodt (1998): Betriebliche Determinanten der Lohnhöhe, in: Gerlach, K., O. Hübler and W. Meyer (eds.): Ökonomische Analysen betrieblicher Strukturen und Entwicklungen. Frankfurt: Campus Verlag, 91-118. Jones, D.C. and J. Pliskin (1989): British Evidence on the Employment Effects of Profit Sharing. Industrial Relations 28, 276-298. Kandel, E. and E.P. Lazear (1992): Peer Pressure and Partnership. Journal of Political Economy 100, 801-817. Köhler, H. (2006): Gewinn- und Kapitalbeteiligungen für Arbeitnehmer (Interview). Der Stern 01/2006, http://www.stern.de/presse/vorab/ 552132.html (date of download: January 5, 2006). Kölling, A. (2000): The IAB-Establishment Panel. Applied Social Science Studies 120, 291-300. Kraft, K. (1991): The Incentive Effects of Dismissals, Efficiency Wages, Piece-Rates and Profit Sharing. Review of Economics and Statistics 73, 451-459. Kraft, K. (1992): Profit-Sharing and Employment. Jahrbuch für Sozialwissenschaft 43, 349-363. Kraft, K. and M. Ugarkoviü (2005): The Output, Employment and Productivity Effects of Profit Sharing: A Matching Approach, mimeo, University of Dortmund. Kraft, K. and M. Ugarkoviü (2006): Profit sharing and the financial performance of firms: Evidence from Germany. Economics Letters 92, 333-338. Kruse, D.L. (1991): Profit Sharing and Employment Variability. Microeconomic Evidence on the Weitzman Theory. Industrial and Labor Relations Review 44, 437453.
Bibliography
129
Kruse, D.L. (1992): Profit Sharing and Productivity: Microeconometric Evidence from the United States. The Economic Journal 102, 24-36. Kruse, D.L. (1993): Profit Sharing – Does it make a Difference?. Kalamazoo, Michigan: W.E. Upjohn Institute for Employment Research. Kruse, D.L (1996): Why Do Firms Adopt Profit-Sharing and Employee Ownership Plans? British Journal of Industrial Relations 34, 515-538. LaLonde, R. (1986): Evaluating the econometric evaluations of training programs with experimental data. American Economic Review 76, 604-620. Layard, R., S. Nickell, and R. Jackman (1991): Unemployment: Macroeconomic Performance and the Labour Market. Oxford: Oxford University Press. Lazear, E.P. (1979): Why is there Mandatory Retirement?. Journal of Labor Economics 87, 1261-1284. Lazear, E. P. (1986): Salaries and Piece Rates. Journal of Business 59, 405-431. Lazear, E. P. (1987): Incentive Contracts, in J. Eatwell, M. Milgate, and P. Newman (eds): The New Palgrave: A Dictionary of Economics, Vol. 2 (E-J). London: The Macmillan Press Limited, 744-748. Lazear, E.P. (1989): Pay equality and industrial politics. Journal of Political Economy 97, 561-580. Lazear, E.P. (1995): Personnel Economics. Cambridge: The MIT Press. Lazear, E.P. (2000): Performance Pay and Productivity. American Economic Review 90, 1346-1361. Lazear, E.P. (2005): Output-Based Pay: Incentives, Retention or Sorting. Research in Labor Economics 23, 1-25. Lechner, M. (1998): Training the East German Labour Force – Microeconometric Evaluation of Continuous Training after Unification, Heidelberg: Springer. Lechner, M. (2000a): A Note on the Common Support Problem in Applied Evaluation Studies, Discussion paper No. 2001-01, University of St. Gallen. Lechner, M. (2000b): An Evaluation of Public Sector Sponsored Continuous Vocational Training Programs in East Germany. Journal of Human Resources 35, 347-375.
130
Bibliography
Lechner, M. (2002): Some practical issues in the evaluation of heterogeneous labour market programmes by matching methods. Journal of the Royal Statistical Society, A 165, 59-82. Leuven, E. and B. Sianesi (2003): PSMATCH2: Stata module to perform full Mahalanobis and Propensity Score Matching, version 1.2.1, (http://fmwww.bc.edu/ repec/bocode/p/psmatch2.ado). Levine, D.I. (1987): Efficiency Wages in Weitzman’s Share Economy. Economics Letters 23, 245-249. Levine, D.I. (1989): Efficiency Wages in Weitzman’s Share Economy. Industrial Relations 28, 321-334. Levine, D.I. (1991): Cohesiveness, Productivity, and Wage Dispersion. Journal of Economic Behavior and Organization 15, 237-255. Levine, D.I. and L. D’Andrea Tyson (1990): Participation, Productivity, and the Firm’s Environment, in: A.S. Blinder (ed.): Paying for Productivity. A Look at the Evidence. Washington D.C.: Brookings Institution, 183-243. Little, R.J.A. and D.A. Rubin (1987): Statistical Analysis with Missing Data. New York: John Wiley & Sons. MacLeod, W.B. (1988): Equity, efficiency and incentives in Teams, in: D.C. Jones and J. Svejnar (eds.): Advances in the economic analysis of participatory and labor-managed firms, London: JAI Press, 5-23. McNabb, R. and K. Whitfield (1998): The Impact of Financial Participation and Employee Involvement on Financial Performance. Scottish Journal of Political Economy 45, 171-187. Meade, J. (1986a): Alternative Systems of Business Organization and of Workers’ Remuneration. London: Allen and Unwin. Meade, J. (1986b): Different Forms of Share Economy. London: Public Policy Centre. Michaelis, J. (1998): Zur Ökonomie von Entlohnungssystemen. Tübingen: Mohr Siebeck. Milgrom, P. and J. Roberts (1992): Economics, Organization and Management. Englewood Cliffs, NJ: Prentice-Hall International.
Bibliography
131
Mitchell, J.B., D. Lewin, and E.E. Lawler III (1990): Alternative Pay Systems, Firm Performance, and Productivity, in: A.S. Blinder (ed.): Paying for Productivity: a Look at the Evidence, Washington D.C.: The Brookings Institution, 15-94. Möller, I. (2000): Produktivitätswirkung von Mitarbeiterbeteiligung. Mitteilungen aus der Arbeitsmarkt- und Berufsforschung 4/2000, 565-582. Müller-Jentsch, W. (1995): Germany: From Collective Voice to Co-Management, in: J. Rogers and W. Streeck (eds.): Works Councils – Consultation, Representation and Cooperation in Industrial Relations. Chicago: Chicago University Press, 5378. Nalbantian, H.R. (1987): Incentives, Cooperation and Risk Sharing, Totowa: Rowman and Littlefield. Nalbantian, H.R. and A. Schotter (1997): Productivity under Group Incentives: An Experimental Study. The American Economic Review 87, 314-341. Nordhaus, W. (1988): Can the Share Economy Conquer Stagflation? The Quarterly Journal of Economics 103, 201-217. Nuti, D.M. (1987): Codetermination and profit-sharing, in: J. Eatwell et al. (eds.): The New Palgrave. A Dictionary of Economics, Vol. 1, 465-469. Oberhauser, A. (1963): Finanzpolitische und private Vermögensbildung – eine theoretische und empirische Untersuchung, Köln: Westdeutscher Verlag. OECD (1995): Profit Sharing in OECD countries. Employment Outlook, 139-169. Pendleton, A. (1997): Characteristics of workplaces with financial participation: evidence from the Workplace Industrial Relations Survey. Industrial Relations Journal 28, 103-119. Pendleton, A., E. Poutsma, J. van Ommeren, and C. Brewster (2001): Employee share ownership and profit-sharing in the European Union. Luxembourg: Foundation for the Improvement of Living and Working Conditions. Pendleton, A., E. Poutsma, J. van Ommeren, and C. Brewster (2002): The incidence and determinants of employee share ownership and profit sharing in Europe, in: T. Kato and J. Pliskin (eds.): The Determinants of the Incidence and Effects of Participatory Organizations. Advances in the Economic Analysis of Participatory and Labor Management, Volume 7, Greenwich, CT: JAI Press, 1-37.
132
Bibliography
Pérotin, V. and A. Robinson (2003): Employee participation in profit and ownership: A review of the issues and evidence, European Parliament Working Paper SOCI 109 EN, Luxembourg: European Communities. Poole, M. (1989): The Origins of Economic Democracy: Profit Sharing and Employee Shareholding Schemes. London: Routledge. Poutsma, E. (2001): Recent Trends in Employee Financial Participation in the European Union. Luxembourg: Office for Official Publications of the European Commission. Poutsma, E. (2002): Developments and prospects of profit sharing and employee share ownership in Europe. Transfer 1/02, 22-46. Prendergast, C. (1996): What happens within firms? A survey of empirical evidence on compensation policies. NBER Working Paper No. 5802, National Bureau of Economic Research, Cambridge. Prendergast, C. (1999): The Provision of Incentives in Firms. Journal of Economic Literature 37, 7-63. Priewe, J. (1997): Wege aus dem ostdeutschen Lohndilemma – Investivlöhne und Vermögenspolitik. Wirtschaftsdienst, vol. 1997/VI, 332-339. Priewe, J. and F. Havighorst (1999): Auf dem Weg zur Teilhabergesellschaft?. Bonn: Friedrich Ebert Stiftung. Puhani, P. (2000): The Heckman Correction for Sample Selection and its Critique. Journal of Economic Surveys 14, 53-68. Putterman, L. (1984): On Some Recent Explanations of Why Capital Hires Labor. Economic Inquiry 22, 172-187. Richardi, R. (2002): Betriebsverfassungsgesetz. München: Beck. Roy, A.D. (1951): Some Thoughts on the Distribution of Earnings. Oxford Economic Papers 3, 135-146. Rosen, S. (1986): The Theory of Equalizing Differences, in: O. Ashenfelter and R. Layard (eds.): Handbook of Labor Economics, Vol. 1, Amsterdam: NorthHolland, 641-692. Rosenbaum, P.R. and D.B. Rubin (1983): The Central Role of the Propensity Score in Observational Studies for Causal Effects. Biometrika 70, 41-55.
Bibliography
133
Rubin, D.B. (1974): Estimating Causal Effects of Treatments in Randomized and Non-Randomized Studies. Journal of Educational Psychology 66, 688-701. Rubin, D.B. (1977): Assignment to Treatment Group on the Basis of Covariate. Journal of Educational Studies 2, 1-26. Rubin, D.B. (1980): Comment. Journal of the American Statistical Association 75, 591-593. Rubin, D.B. (1986): Comment. Which Ifs have Causal Answers. Journal of the American Statistical Association 81, 961-962. Samuelson, P.A. (1977): Thoughts on Profit Sharing. Zeitschrift für die Gesamte Staatswissenschaft, Special Issue. Schröder, G. (2000): Mitarbeiterbeteiligung als ein Weg zur Teilhabegesellschaft. Gewerkschaftliche Monatshefte 06/2000, 321-327. Shepard, E.M. (1994): Profit Sharing and Productivity: Further Evidence from the Chemicals Industry. Industrial Relations 33, 452-466. Sinn, G. and H.W. Sinn (1991): Kaltstart. Volkswirtschaftliche Aspekte der deutschen Wiedervereinigung. Tübingen: J.C.B. Mohr Paul Siebeck. Sinn, H.-W. (1997): Kapitalbeteiligung und Lohndifferenzierung: ein Vorschlag zur Lösung der Beschäftigungskrise. Mitteilungen zur Arbeitsmarkt- und Berufsforschung 30, 822-827. Smith, J. A. (2000): A Critical Survey of Empirical Methods for Evaluating Active Labour Market Policies. Schweizerische Zeitschrift für Volkswirtschaft und Statistik 136, 1-22. Smith, J.A. and P. Todd (2005): Does matching overcome LaLonde’s critique of nonexperimental estimators?. Journal of Econometrics 125, 305-353. Summers, L. (1986): On the Share Economy. Challenge, November/December, 4750. Thünen, J.H.v. (1850): Der naturgemäße Arbeitslohn und dessen Verhältnis zum Zinsfuß und zur Landrente. Rostock: Leopold. Uvalic, M. (1991): The PEPPER Report: Promotion of Employee Participation in Profits and Enterprise Results in the Member States of the European Commu-
134
Bibliography nity. Social Europe, Supplement 3/91, Luxembourg: Commission of the European Communities.
Vanek, J. (1965): Workers’ Profit Participation, Unemployment and the Keynesian Equilibrium. Weltwirtschaftliches Archiv 94, 206-214. Venkatraman, N. and V. Ramanujam (1987): Measurement of Business Economic Performance: An Examination of Method Convergence. Journal of Management 13, 109-122. Wadhwani, S. and M. Wall (1990): The Effects of Profit Sharing on Employment, Wages, Stock Returns and Productivity: Evidence from UK Micro Data. Economic Journal 100, 1-17. Wall, T.D., J. Michie, M. Patterson, S.J. Wood, M. Sheehan, C.W. Clegg, and M. West (2004): On the Validity of Subjective Measures of Company Performance. Personnel Psychology 57, 95-118. Weitbrecht, H. and S. Mehrwald (1999): Mitbestimmung, Human Resource Management und neue Beteiligungskonzepte, in: B. Frick, N. Kluge and W. Streeck (eds.): Die wirtschaftlichen Folgen der Mitbestimmung, Frankfurt: Campus, 89127. Weitzman, M.L. and D.L. Kruse (1990): Profit Sharing and Productivity, in: A.S. Blinder (ed.): Paying for Productivity. A Look at the Evidence. Washington D.C.: The Brookings Institution, 95-141. Weitzman, M.L. (1983): Some Macroeconomic Implications of Alternative Compensation Systems. Economic Journal 93, 763-783. Weitzman, M.L. (1984): The Share Economy: Conquering Stagflation. Cambridge: Harvard University Press. Weitzman, M.L. (1985): The Simple Macroeconomics of Profit Sharing. The American Economic Review 75, 937-953. Weitzman, M.L. (1986): Macroeconomic Implications of Profit Sharing. NBER Macroeconomics Annuals 1986, 291-335. Weitzman, M.L. (1987): Steady State Unemployment under Profit Sharing. Economic Journal 97, 86-105.
Bibliography
135
Weitzman, M.L. (1988): Comment on ‘Can the Share Economy Conquer Stagflation?’. Quarterly Journal of Economics 103, 219-223. Weitzman, M.L. (1995): Incentive Effects of Profit Sharing, in: H. Siebert (ed.): Trends in Business Organization. Tübingen: J.C.B. Mohr, 51-87. Wilson, N. and M.J. Peel (1991): The Impact on Absenteeism and Quits of ProfitSharing and other Forms of Employee Participation. Industrial and Labor Relations Review 44, 454-468. Wolf, E. and T. Zwick (2002a): Produktivitätswirkung von Mitarbeiterbeteiligung: Der Einfluss unbeobachteter Heterogenität. Mitteilungen aus der Arbeitsmarktund Berufsforschung 1/2002, 123-132. Wolf, E. and T. Zwick (2002b): Reassessing the Impacts of High Performance Workplaces. ZEW Discussion Paper No. 02-07, Mannheim. Wolf, E. and T. Zwick (2003): Welche Personalmaßnahmen entfalten eine Produktivitätswirkung? Zeitschrift für Betriebswirtschaft, Supplement No. 4, 43-62. Yellen, J. (1984): Efficiency Wage Models of Unemployment. American Economic Review 74, 200-205. Zhuang, J. and C. Xu, (1996): Profit-Sharing and Financial Performance in the Chinese State Enterprises: Evidence from Panel Data. Economics of Planning 29, 205-222. Zwick, T. (2003): Works Councils and the Productivity Impact of Direct Employee Participation. ZEW Discussion Paper No. 03-47, Mannheim. Zwick, T. (2004): Employee participation and productivity. Labour Economics 11, 715-740.