The effects of democratic determination of wages : theory and evidence from self-managed firms

Policy Research WORKING PAPERS Transitlon and Macro-Adjustment Country Economics Department The World Bank August 1992 WPS 971 The Effects of Democratic Determination of Wages Theory and Evidence from Self-Managed Firms Milan Vodopivec Should workers participate in decisionmaking about pay? Demo- cratic decisionmaking about pay - if divorced from substantive participation of workers in other areas-decreases productivity. PolicyRearchWodngPapeisdissaemnatethefindings ofwork in press and enouage the exchangeof ideas amongBank staffand allothersinterested indevelopmentisues.Thesepapes,sdistnbutedby theResearvhAdvisoiy Staff, cany thenames of the authon, reflect onlytheirviews,and shouldbused and ited accordingly. T ofindings, ira tions,a o an o is rethe autr o'own. ey should not be attnibuted to the World Bank, its Board of Dirwwtrs. its management, or any of its member countries.  Policy Research Transitlon and Macro-Adjustment WPS 971 This paper- aproduct of the Transition and Macroadjustment Division, Country Economics Department - is part of a larger effort in the department to investigate the labor market in transitional economies. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact CECTM', room N6-045, extension 37188 (August 1992, 26 pages). Some assert that when efficiency requires structure affects productivity in a sample of cooperation, effectiveness is increased by an Yugoslav firms. He finds that democratic egalitarian pay structure resulting from warkers' decisionmaking about pay - if divorced from participation in decisionmaking about pay. But it substantive participation of workers in other can also be argued that equalizing pay reduces areas - decreases productivity. the morale of highly productive workers, and thus more than offsets the positive effects of One implication of this finding for policy- cooperation. makers, particularly in Eastern Europe and the former Soviet republics, is that programs de- To shed light on this controversy, Vodopivec signed to allow workers to participate in pay explores both theoretically and empirically how decisions must be consistent with the workers' productivity is affected when workers determine general involvement in decisionmaking. If relative pay differences democratically (by participation is limited to decisions about pay, or referendum). The median voter model suggests if external control is imposed on intrafirm wage that this kind of decisionmaking process pro- differentials (which has effects on wage distribu- duces an egalitarian wage structure. Using tion similar io those of worker participation), the alternative assumptions about worker incentives, resulting compressed wage structure is likely to Vodopivec formalizes and empirically tests two produce negative effects on productivity. competing views about how an egalitarian wage The Policy Research Working Paper Series dissemninate the findings of work under way in the Bank. An objective of the series is to get these findings out quickly, even if presentations are less than fully polished. The f,ndings, interpretations, and conclusions in these papers do not necessarily represent official Bank policy. Produced by the Policy Research Dissemination Center  THE EFFECTS OF DEMOCRATIC DETERMINATION OF WAGES: THEORY AND EVIDENCE FROM SELF-MANAGED FIRMS Milan Vodopivec The World Bank Helpful comments from Alan Gelb, Dennis Mueller and Peter Murrell on earlier versions of the paper are gratefully acknowledged.  TABLE OF CONTENTS 1. MODELING THE EFFECTS OF WORKER PARTICIPATION IN PAY DECISIONS .... .... 2 1.1 The Cooperation Argument .................................... 2 1.2 The Lowering-of-Morale Argument ............................. 3 2. EMPIRICAL RESULTS . .................................................... 9 3. CONCLUDING REMARKS ...... ........... ................................. 16 APPENDIX 1: How Workers' Earnings were Determined in Yugoslav Self-Managed Firms ............ ..................... 18 APPENDIX 2: Proofs of Propositions .................................. 20 APPENDIX 3: Data and Variables Used in the Empirical Analysis ........... 22 REFERENCES ........ 24  Part of the controversy about whether workers should participate in decisionmaking involves the question of whether workers should participdie in decisionmaking about pay.I Both theory (median voter models) and evidence (producer cooperatives, the unionized sector) support the conclusion that when workers participate in pay decisions, the range of wages in a firm narrows.2 Some assert that "when efficiency requires efficient cooperation, almost any movement toward a democratic egalitarian structure increases effectiveness" (the "cooperation" argument).3 According to this argument, more equal pay increases cohesiveness, which increases productivity by improving the flow and use of information through peer pressure (Levine and Tyson 1990) and by reducing "sabotage" -- workers hindering the activity of their co-workers (Lazear 1989). But it can also be argued that equalizing pay hurts the morale of highly productive workers, which more than offsets the positive effects of cooperation (the lowering-of-morale argument). To shed light on the controversy, this paper explores both theoretically and empirically how productivity is affected when workers determine relative pay differences democratically (by referendum). The median voter model suggests that such a determination produces an egalitarian wage structure. Using alternative assumptions about worker incentives, I formalize and 1There is a profound controversy about the effects of worker participation in decisionmaking on productivity. Advocates of worker participation (notably, Vanek 1970) argue that it induces workers to work harder by boosting team spirit, improving morale, and reducing absenteeism, thus reducing costs of monitoring and conflict resolution. Worker participation is also believed to improve firm-specific human capital by both increasing Job longevity and enhancing training of workers (Ireland and Law 1982). Among critics, Jensen and Meckling (1979) point out a control problem and Williamson (1975) point out the transaction costs arising from collective decisionmaking. 2See, for example, Freeman and Medoff (1984) on evidence of how unions affect wage equality. 3morton Deutsch, cited in Levine and Tyson (1990, p. 211). 1  empirically test two competing views about how an egalitari - wage structure affects productivity, in a sample of Yugoslav firms. My major finding: that democratic decisionmaking about pay -- if divorced from substantive participation of workers in other areas -- decreases productivity. After formalizing the two competlng views about how worker participation in pay decisions affects productivity, and deriving their empirical implications (section 1), I present the estimating framework and the empirical results (section 2). I conclude with the discussion of results and their implications for policy. 1. MODELING THE EFFECTS OF WORKER PARTICIPATION IN PAY DECISIONS To test how worker participation in pay decisions affects productivity, I assume that labor input (effective labor) depends on the incentives workers face.4 I derive the function of effectively provided labor under two alternative hypotheses about the effects of egalitarian wage structure: (1) that such a wage structure increases cohesiveness and hence productivity, and (2) that such a wage structure hurts morale and thus decreases the effort of the more productive workers, and hence the firms's productivity as a whole. 1.1 The cooperation argument This line of reasoning emphasizes cooperation as an important feature of the firm. It assumes two things: (1) that a narrower wage dispersion increases cohesiveness ("the propensity to obey group norms because the approval of the group is valued" -- Levine 1991, p. 237), and (2) that greater cohesiveness increases effective labor, and hence productivity. Formally, L = L(C(D)), witn 4This is a standard assumption in the literature on Incentives -- for example, see Sen's pioneering contribution (1966). 2  L' ( ) positive and C' ( ) negative (L is effective labor, C is cohesiveness, and D Is wage dispersion). Note that this argument implicitly assumes that a worker's utility depends not only on effort (e) and earnings (w) -- the standard arguments -- but on the approval of co-workers, that is, U = U[w, e(C), C]. Akerlof (1982) used a similar utility function, with work norms substituting for cohesiveness. This reasoning predicts that the narrower the wage dispersion, the larger the provision of effective labor. 1.2 The lowering-of-morale argument By contrast, the second line of reasoning emphasizes individualistic behavior and derives the labor supply from a standard utility maximization over income and leisure. It explicitly takes account of the democratic determination of lntrafirm relative pay differences, and thus endogenously generates wage structure. The empirically testable prediction is that effective labor depends negatively on the ratio of the firm's mean to median earnings. The setting Suppose a firm has N workers who differ In productivity but are otherwise identical. Following Mueller's (1976) argument (that a significant part of a worker's effort is non-contractual), assume also that the firm's labor input, L, is the sum of contractual labor input, Lc , and non-contractual labor input, N N L . Thus, L =L + L where Lxe n L c xe. The ability -nc c nc c n i=l i=l of worker i Is denoted by xl, the provision of the minimum effort needed to 5Recognizing the non-contractibility of effort as an essential characteristic of the firm is particularly important for socialist firms, for which "stick" incentives are Intentionally designed out of the system (see Vodopivec 1991). 3  retain a job is ec, and the provision of an above-minimum effort is ei. (I sometimes call the product xiei the productivity of the i-th worker.) Let the firm's value-added, Y, correspond to labor input except for N the unit of measurement -- that Is, Y = Yc + Ync, where Yc = Z xiec N ~~~~~~~~~~~~~i=l N and Ync = xiei. Moreover, to reflect the compensation policies in i=l participatory firms, assume that the pre-tax earnings of the worker 1, wi, comprise two components: (1) an egalitarian, lump-sum component, c, equal across workers, and (2) an income-sharing component, y-, equal to xiei (that is, equal to the worker's productivity).6 Accordingly, W= c + Yi = c + xiei (1.1) where N c = xie (1.2) N E i c(12 i=1 so that the overall earnings exhaust the firm's value-added. The egalitarian wage tendency that characterizes the democratic wage-setting environment is modeled in the following way.7 The income-sharing 'This mirrors the wage-setting practiced by Yugoslav firms in 1986: c corresponds to the earnings-in-kind component, and xIeI to the cash-earnings component of a worker's compensatior. Moreover, a worker's effort, eI, and ability, xI, correspond to his/her index of the quality of work and the complexity of work, respectively (see appendix 1). 7There is a well-documented tendency to compress differences in earnings among workers within the Yugoslav firm. According to Prasnikar and Svejnar (1988), "[v]arious measures suggest that the distribution of personal income Is relatively egalitarian in Yugoslav firms," and "Yugoslav skilled workers and managers earn less relative to unskilled workers than their counterparts in 4  component of the worker's pre-tax earnings, xiei, is "taxed" at the rate t (O s t < 1) and used for financing a lump-sum subsidy, r, for each worker. The subsidy is paid out of the tax revenues, and all tax revenues ara redistributed. In other words, r = ty, where y (the average of the income-sharing component of pre-tax earnings) equals Ync IN. The redistribution mechanism is the same as for a progressive linear income tax. The net subsidy per worker is equal to t(y - xJeI), which means that the net subsidy is positive for workers who work below the mean level of productivity and negative for workers who work above it. One can justify such a concept on the grounds that it captures two basic features of real-world intrafirm redistribution: (1) the redistribution is from richer to poorer (and so compresses the structure of personal earnings), and (2) it preserves the order of personal earnings. Taking account of intrafirm redistribution, the earnings of the worker with ability xi are equal to: Wi = wp + r - tx e = r + c + (1- t)xie1 (2) The effects of compressed wage differentials on effective labor (value-added) To model the effects of compressed wage differentials on productivity within a standard utility maximization framework, I modify the median voter model of Meltzer and Richard (1983), which employs a Stone-Geary utility function: 8 capitalist firms" (p. 279). Similarly, Flakierski (1989) finds that, in Yugoslavia, "interskill differentials are quite small by Western standards and smaller than in many socialist countries" (p. 75). 8Like Meltzer and Richard, I assume that the distribution of income is positively skewed. s  u(w, l) = in (w + w) + a ln(e + A); a, w, A > 0 (3) where e is leisure, and w and A are the parameters of the utility function. Each worker takes the parameters r, t, and c as given and chooses ei to maximize his utility. Substituting (2) into (3), maximizing (3), and setting e 1- e, yields max u(wI, 1 - ei) = max[ln(w + r + c + (1 - t)xiei) + a ln(1 - e1 + A)]. ei (4) Differentiating (4) with respect to ei we get x I(1 - t)a w + r + c + (1 - t)x IeI 1 - ei + A - (5) and solving for e we have (1 + A)(1 - t)xi - (w + r + c)a ei = (a + 1)(1 - t)xi for x > x (6.1) ei =0 for x s x (6.2) where x = (w + r + c)a (7) o (A + 1)(1 - t) The ability of the most productive worker, among those who opt not to provide more than the minimum effort, e, is denoted by x . I call this worker the i -th worker. 0 On the basis of the supply-of-effort functions (6), we can obtain the following comparative statics result:9 Proposition 1: Increasing the intrafirm tax rate t has a negative 9Proofs of this and the following proposition are given in appendix 2. 6  effect on the firm's effectively provided labor (value-added) -- that is, decreases the size of the pie. Unfortunately, this intuitively appealing result does not lend itself to empirical verification: the intrafirm tax rate, t, is unobservable, By exploiting the fact that relative earnings (and hence the tax rate) are determined by majority rule, however, w can link the tax rate to an observable variable -- the task we turn to now. Linking ir.trafirm redistribution to an observable variable The modeling of intrafirm redistribution as a simple tax-subsidy scheme reduces the issue of redistribution to two dimensions: a per capita subsidy, r, and an intrafirm tax rate, t. If we require that tax revenues equal the subsidies paid out, we can determine only one of these variables independently. I use the tax rate t as the choice variable. Exploiting democratic wage setting (such as prevailed in Yugoslav firms -- see appendix 1), we then invoke a median voter theorem, that the median voter decides. 10 Note the balanced budget, r = ty, remember the budget constraint (2), and denote the income-sharing component of the median worker's pre-tax earnings with Yd (yd = xded). This yields the following maximization for the median worker: max u(w do1 e - ed max ln(w+ty+c+ (l-t)ydl +aln(i-e+ A)] (8) The first-order condition is: y + t dy d = O (9) '0For a discussion of the median voter theorem, see, for example, Mueller (1989). 7  since by the envelope theorem the coefficient of the ded/dt term, Su/Sedo is zero (ed is the outcome of utility maximization). To balance the budjet, higher taxes are needed to increase the subsidy. The median voter thus faces the following trade-off (embedded in equation 9): by imposing a higher tax rate, he/she is better off through the higher subsidy r (which is proportional to t). But Jncreasing the tax rate has also an indirect effect: decreasing the size of the pie, as stated in proposition 1, which makes the median voter worse off. By using the equilibrium condition (9), w- arrive at the following relationship (see appendix 2), linking the tax rate to an observable variable: Proposition 2: The higher the mean of the income-sharing component of the workers' pre-tax earnings, compared with the income-sharing component of the median worker's pre-tax earnings, the higher the equilibrium tax rate, t. This proposition can be reinterpreted as follows. With the intrafirm redistribution scheme adopted in the model, the structure of post-tax earnings corresponds to the structure of pre-tax earnings, so it also reflects the structure of workers' productivity. The proposition thus says that the lower the productivity of the median worker, compared with the firm's mean productivity, the more the median worker can gain by imposing a higher tax rate (or, loosely speaking, a greater redistribution of income within the firm). Combining propositions 1 and 2 produces the following corollary. Corollary 1: The higher the mean of the income-sharing component of the workers' pre-tax earnings, compared with the income-sharing component of the median worker's pre-tax earnings, the lower the firm's effectively provided 8  labor (value-added). Finally, what was the policy instrument through which the median voter implemerted redistribution in the Yugoslav firms? It was the wage scale, determined by a referendum of workers under majority rule (see appendix 1). With the optimal intrafirm tax rate in mind, the median worker set the scale so as to produce the desired redistribution. The worker accomplished that by compressing the wage scale -- shrinking the index of work complexity (Ic, see appendix 1) for workers whose productivity was above-average, and increasing the index for workers with below-average productivity. 2. EMPIRICAL RESULTS To test the two competing hypotheses about the effects of worker participation in pay decisions, I estimated an augmented production function for a sample of Yugoslav firms, with the augmenting variables that determine effective labor dictated by the above theory. I postulated that effective labor depends not only on the number of workers, but also either positively on wage dispersion (the cooperation argument) or negatively on the ratio of the firm's mean to median earnings (the lowered-morale argument). Note that tThe resulting post-redistribution distribution of cash earnings was thus D = {eIxI + (y - xIeI)t a eIxIa, i=1, ..., NE, compared with the distribution without redistribution, D = ieIxI, i = 1, ..., N}, where iis the compression coefficient associated with the wage index cf the worker with ability xl. The compression coefficients, cr-s, are defined as v1= I + - xe1 , i 2, ..., N (note that a, = 1 asy - xe). e x < < Ii In the empirical test, wage dispersion is normalized by a dispersion of skills, to control for differences in skill dispersion across firms. I also take the ratio of the mean to the median actual (that is, after-tax) personal earnings as a proxy for the ratio of mean to median income-sharing component 9  the theory provides a rationale for using the augmented variables (wage dispersion and the ratio of mean to median earnings) in a labor-embodied iashion. By estimating a production function, of course, the effects of the compressed wage structure on effectively provided labor are identified only indirectly, by how they affect value-added, so other determinants of value-added are carefully controlled for. In general, the production function to be estimated can be written as Y = f(K, L(N, D, m, 2))g(Z) (10) where Y is value-added; K is capital; L is effectively provided labor, N is the number of workers; D is wage dispersion (normalized by skills dispersion), m is the ratio of the firm's mean to median earnings; and ZI and Z2 are embodied and disembodied control variables, respectively. The choice of the functional forms for the functions in (10) was based on the following considerations. First, the complexity of the term representing effectively provided labor dictated the use of the simplest functional form for the function f( ) -- the Cobb-Douglas one. Second, the functional form for the effectively provided labor was selected on an experimental basis, choosing between the form where the number of workers interacted with other variables in a multiplicative way, and the form where that interaction was additive. (The first form was selected; the estimation of the second did not of pre-tax earnings (the rationale being that the intrafirm redistribution scheme preserves the correspondence between the structure of pre-tax and after-tax earnings). 13Prasnikar, Svejnar and Klinedinst (1992) find Cobb-Douglas specifications of production function superior to CES and trans-log one for Yugoslav firms during 1975-9. 10  converge.) Third, as is standard in the literature, an exponential function was adopted for disembodied control variables, that is, for the function g( ). As for the choice of the control variables, the following two were used in the labor-embodied fashion: (1) the average number of years of education of the workforce (EDUC), to adjust for differences in skills acquired through education, and (2) labor turnover (TURN), to control for differences in skills because of "asset specificity" -- the longer the workers stay in the firm, the more firm-specific knowledge they acquire, and (other things being equal) the more productive they become. A more complex version of the above model allows for another interpretation of the link between labor turnover and value-added, the interpretation based on cooperation among workers. Assuming that a worker's effort depends on his co-workers's effort, one can show that the equilibrium effort of each worker is positively associated with the "team spirit" -- more precisely, with the sum of changes in co-workers' effort in response to a change in effort of one of them. 14 If one is willing to interpret labor turnover as a proxy for the team spirit (presumably, the higher the team spirit, the lower the turnover), the cooperative link between labor turnover and value-added follows. Two groups of disembodied control variables were included in the equation. One group consisted of variables that affected value-added through output and input prices: (1) the firm's market share (MSHARE), as proxy for monopoly power; (2) the share of exports in the firm's output (EXPSHARE), reflecting exchange rate pdticies; and (3) industry dummies (DUM), to control 14The proof of this proposition is available to the Interested reader, on request. 11  for different price regimes across Industries (which might have affected both input and output prices). (The positive association of the export share with value-added could also indicate that export-oriented firms were more efficient.) The other group of disembodied control variables addressed efficiency in setting the "boundary of the firm" (in the sense of Coase 1937) -- that is, efficiency of production organized as a hierarchy compared with production coordinated through prices. Yugoslav firms differed in organizational structure regarding vertical integration, and these differences might have contributed to differences in their performance.i Therefore, two additional control variables were included in the estimated equation: the unit's membership in a so-called Working Organization of Associated Labor (WOAL) and its membership in a so-called Composite Organization of Associated Labor (COAL). Based on the above, the following augmented production function was estimated in its logarithmic transformation: Y = AK(N(l + aID + a2m + a3EDUC + a4TURN))e 1=26 a EXPSHARE + a7WOAL + a8COAL + Z aDUM( 1=9 15Under the system of Associated Labor that prevailed in Yugoslavia during 1974-88, decisionmaking within the firm involved up to three layers of the hierarchy. The lowest-level units were Basic Organizations of Associated Labor (BOALs). The mid-level units were Working Organizations of Associated L*s.bor, typically consisting of several BOALs. Under certain circumstances, hcwever, Working Organizations of Associated Labor did not comprise BOALs (the so-called Uniform Working Organizations); in those units, decisionmaking was not delegated to a lower level. Top-level units were Composite Organizations of Associated Labor, with Working Organizations of Associated Labor as members (membership was voluntary). See Prasnikar and Svejnar (1988) for details. 12  (A, a, 13 and a are parameters to be estimated; other symbols are explained above). To repeat, the main variables are wage dispersion (D) and the ratio of firm's mean to median earnings (m). The empirical results show that worker participation in pay decisions has a negative effect on productivity (see the first two columns -- representing alternative specifications -- of table 1). Above all, the ratio of the firm's mean earnings to its median personal earnings negatively affects value-added (highly significantly), thus supporting the lowering-of-morale argument. Moreover, the positive (though only weakly significant) value of the coefficient of wage dispersion not only refutes the cohesiveness argument but provides additional support that the compression of the wage structure is counterproductive. Estimates of elasticities suggest that the economy operates at decreasing returns of scale (the sum of capital and effectively provided labor elasticities is .82), which is similar to the result found by Prasnikar, Svejnar, and Klinedinst (1992). Among the control variables, only the market share (apart from industry dummies) proved to be significant and of the expected positive sign, suggesting that monopoly affects prices. Other variables were insignificant (except for education, they are dropped in the second column), but the estimates do suggest a weak positive effect from skill specificity (through the variable TURN) and a negative effect from the existence of a COAL-level hierarchy (possibly through restrictive trade practices). 16Nonlinear estimation was used, with capital and the number of workers instrumented by other right-hand-side variables and lagged values of capital and number of workers. Definitions of the variables, their summary statistics, and data sources are given in appendix 3. 13  TABLE 1 ESTIMATES OF THE AUGMENTED PRODUCTION FUNCTION (1) (2) (3) Wage dispersion (D) .151+ .119k -.177 (1.84) (1.78) (-.67) Mean-to-median earnings (m) -.888 -.848 -.970 (-4.26) (-5.10) (-2.08) Education (EDUC) .045 .028 .129 (.92) (.85) (.76) Labor turnover (TURN) -.019 -.036 (-1.53) (-.88) 5 ~~~** ** ** Market share (10 2*MSHARE) 3.565 3.652 3.655 (4.47) (4.55) (4.17) Export share (102*EXPSHARE) -.070 .160 (-.46) (.71) Membership in a WOAL (WOAL) .037 -.063 (.61) (-.90) Membership in a COAL (COAL) -.090 -.094 (-1.61) (-1.30) Effective labor (L) .551 .554 .517 (7.72) (7.80) (6.06) Capital (K) .273 .265 .298 (5.28) (5.11) (5.35) INTERCEPT 3.817 3.968 3.641 (9.55) (10.88) (5.94) R2 .849 .846 .906 Number of observations 394 394 213 Notes: Among the variables in the first column appears also L, the effectively provided labor -- the expression in big brackets of (11), with f3 as the corresponding coefficient. The values in parentheses are t-statistics. Variables significant at 1 percent are marked with two asterisks, those significant at 5 percent with one asterisk, and those at 10 percent with a plus. Industry dummies are included, but their coefficients are not reported (the excluded industry is printing). Only four of them -- drawn and rolled steel, cast metal products, paper and paper products, and the manufacture of cotton fabrics -- deviated from a common intercept at 1 percent significance. 14  These results suggest that worker participation in pay decisions, in general, decreases productivity. It is possible that for a subset of firms -- call them "participatory" firms -- such effects may have been less pronounced. One attempt to identify such firms is to hypothesize that worker decisionmaking about pay increases cohesiveness only if workers are also substantially involved in other areas of decisionmaking (the "guaranteed individual rights" argument of Levine and Tyson 1990). By assuming that workers in such firms spend less time on sick leave (so we use sick leave as a proxy for absenteeism), I identified such firms as those with below-average rates of sick leave. The estimation results in the third column of table 1 refer to this group of firms. In contrast to the estimates for the whole sample, the wage dispersion coefficient is negative (though insignificant), thus suggesting that cooperation may have indeed played a larger role for this subsample. Calculations show that the value-added forgone because of the negative effects on productivity of an egalitarian wage structure amounts to 4.34 percent.17 Compared with estimates of welfare loss from monopoly, this is a significant effect, providing support to Leibenstein's (1966) claim that X-inefficiency losses are much larger than losses arising from monopoly. 18 1 This result was obtained from the equation (11), by comparing value-added generated under the true value of the mean-to-median earnings with the value- added under the hypothetical case where mean-to-median earnings takes the value of one (in that case, the median voter model suggests that there would be no wage compression). Calculations were performed at the sample mean (except for K, for which the sample mean for log(K) was used; industry dummies were dropped). 1"With few exceptions, studies of the U.S. economy -- including the pioneering Harberger's (1954) study -- find welfare losses attributable to monopoly of a very low magnitude (about 0.1 percent of GNP). 15  3. CONCLUDING REMARKS In a recent overview of evidence on the effects of worker participation in decisionmaking on productivity, Levine and Tyson (1990) cautiously note that "tp]articipation usually has a positive, often small, effect on productivity, sometimes a zero or statistically insignificant effect, and almost never a negative effect" (p. 183). Literature on organization development (for example, Lawler 1981) also takes an optimistic view of the effects of worker participation in pay decisions. But researchers agree that many questions still await an answer: For example, is successful participation representative or direct? Consultative or substantive? Formal or informal? Does it cover a wide range of issues (layoffs, investment, pay system, safety), or focus narrowly on organization of the work and quality control? And, most important, do various aspects of participation interact? This paper provides evidence that worker decisionmaking about pay can have a negative impact on productivity. So what went wrong in the Yugoslav case? Levine and Tyson (1990) argue that the recognition that workers have a say In control of the firm is a basic feature that explains the success of participatory firms (p. 205). One plausible interpretation of this paper's empirical results -- corroborated by the above result on "participatory firms" -- is that in Yugoslav firms some essential ingredients of successful participation were missing. Indeed, many agree that under the Yugoslav self-management, because of extensive government intervention in firms' decisionmaking and because of the hierarchical distribution of power within firms, workers controlled the enterprise on paper, but much less so in reality; possibly because of detailed, highly formalized procedures, decisionmaking about the relative wage structure was the exception, not the 16  rule. 19 Workers' direct decisionmaking was limited to a few areas, and representative participation might not have allowed for the effective participation of most workers. 20 One implication of the paper for policymakers, particularly in Eastern Europe and the ex-Soviet republics, is that programs designed to allow workers to participate in pay decisions must give them full participation in decisiomaking. If participation is limited to decision about pay, or if external control is imposed on intrafirm wage differentials (having similar effects on wage distribution as worker participation), the resulting compressed wage structure is likely to produce negative effects on productivity. 21 19On the workers' lack of substantive control of the firm workers see Obradovic (1978), Prasnikar and Svejnar (1990), and Kraft and Vodopivec (1992), among others. 20Tyson and Levine (1990, pp. 205-14) identify four characteristics of successful participatory schemes: profit- or gain- sharing; job security and long-term employment; measures to build group cohesiveness (under which they emphasize pay equality); and guaranteed individual rights (as a means to achieve a substantive participation in decisionmaking). While job security, pay equality, and, to some degree, profit-sharing can be found in Yugoslav firms, the absence of independent judicial system and independent trade unions worked against the fourth above-mentioned characteristic -- guaranteed individual rights. 21 In Slovenia, for example, 1991 legislation limits wage differences in social enterprises to a 15-to-1 range. Even in the U.S.A., there has been a suggestion to narrow intrafirm wage differences administratively (recently congressman Martin 0. Sabo proposed taxing that part of managers' salaries that exceeds 25 times the wages of the least paid worker -- Business Week, March 30, 1992). 17  APPENDIX 1: HOW WORKERS' EARNINGS WERE DETERMINED IN YUGOSLAV SELF-MANAGED FIRMS The Law of Associated Labor (LAL), which regulated wage determination in Yugoslav firms in 1986, specified that workers share the firm's income, and that their earnings consist of two components: cash earnings and earnings in kind (the second, typically smaller, component included such items as subsidized housing and free meals). The purpose of the two-component compensation was to distribute firm income according to work (the principle used in distributing the firm's cash earnings), as well as on the basis of "solidarity" (the principle used in distributing the firm's collective consumption fund -- LAL, Art. 126). Workers' cash earnings were determined by both the quality and quantity of their work. 22 The quality of work was evaluated either "objectively" (by work norms and standards) or "subjectively" (by a supervisor's evaluation of an individual's performance). The quantity of work was determined, above all, by the complexity of labor provided, which was calculated on the Marxian concept of "simple" and "complex" types of labor. 23 An index of complexity (defined as the number of units of simple labor required for a unit of complex labor) was assigned to each job in a firm, thus creating a wage scale for the firm. 24 (In the model, ei corresponds to the quality, and xi to the complexity of labor.) 22See, for example, Schrenk (1981). 23Workers typically worked the same number of hours because of rigid employment rules (Vodopivec 1991). This made duration of work, the other determinant of the quantity of work, unimportant. 24Four criteria were taken into account: the worker's capability, the amount of responsibility involved, the physical strain required of the worker, and environmental effects on the worker (for more details, see Vodopivec 1989). 18  Workers determined the firm's wage scale by referendum -- a feature that is central to the theoretical model constructed above. The proposed wage scale was subject to a debate in the firm and was voted on by simple majority rule (LAL, Art. 463). Evaluations of quality of work among workers did not vary a lot, so the wage scale was the most important determinant of earnings for all workers, including managers. The following wage scale of a Slovenian manufacturing firm shows the compression of wage differentials in the Yugoslav firm in 1986 (for example, a plant manager's pay was only 3 times the pay of a plastic-filer):25 Description of Work Position Degree of Complexity 1. Simple filing off the excess plastic from 1.00 finished products 2. Producing of metal foil cans 1.22 3. Cutting foil 1.25 4. Running the printing and painting machine 1.28 5. Transporting 1.31 6. Producing intermediate products 1.34 7. Carrying out less demanding installation 1.38 8. Driving electric fork lift 1.42 9. Administering a warehouse 1.46 10. Collecting and manipulating data 1.52 11. Simple product accounting 1.60 12. Simple metalwork 1.70 13. Complicated invoicing 1.83 14. Demanding metalwork 1.96 15. Supervising of a work group 2.09 16. Very demanding metalwork 2.24 17. Supervising a shift 2.38 18. Preparing and controlling technological processes 2.53 19. Buying on foreign markets 2.68 20. Most demanding product design 2.83 21. Managing a bigger plant 2.99 22. Managing a group of designers 3.15 23. Researching and designing construction elements 3.34 24. Managing a production division 3.54 25Wage range in Mondragon cooperatives has been similar to the range presented here. Bradley and Gelb (1985) report a range of 1:4.5 (after including special bonuses), significantly below the range outside the cooperatives. 19  Description of Work Position Degree of Complexity 25. Managing a division 3.75 26. Managing a Basic Organization of Associated Labor 3.95 27. Managing a Working Organization of Associated Labor 4.54 Technically, earnings in self-managed Yugoslav firms were determined in the following way. First, the performance score of the i-th worker (T ) was obtained as T = Icx IQx H where I is the index of the complexity of labor of the i-th worker as determined by the job scale, IQ is the index of the quality of labor of the i-th worker, and H is the number of hours worked by the i-th worker. Second, the i-th worker's cash earnings (W ) were obtained as T W = x WBILL' where WBILL is a the firm's fund earmarked for cash earnings, and the i-th worker's share in It. APPENDIX 2: PROOFS OF PROPOSITIONS Proof of Proposition 1 i Let us denote F(x ) = N- as the share of workers who do not provide a N non-contractual contributions of effort and information. Note that 0 s F( ) < 1 (in a non-trivial case when at least some of the workers are contributing non-contractual effort). Modifying the result from Meltzer and Richard (1983, p. 408)) -- the budget constraint in the model above includes an additional 20  lump-sum term, c -- we have dy a(1 - -F)(y + c + W) (Al) dt (I - t)l + a)(1 - t) + a(l - F)tM This expression is negative because all of the multipliers of the ratio on the right-hand side are positive. QED Proof of Proposition 2 Meltzer and Richard (1983, p. 409) show that the tax rate, t, can be approximated by t - (1 + a) (m - 1) (A2) a(l - F) (l .. g) where m = , and g = w > (Y =xe,where subscript d refers to the Yd d~~~~ d decisive worker). The latter term takes account of the modified budget constraint. We want to show that dt > 0. de dt Let us first show that dt s 0. For xi s x , it follows from (6.2) that de d = 0. For x1 > x , total differentiation of (6.1) gives de d (1 + A)(1 - t)x1 - (w + ty + c)a dtr -dat (a + 1)(1 - t)x J a(y + tdtY) - t) + a(w + c + ty) a(fy + t(1 - t dt + +C) (1 + a)x I(1 - t)2 (1 + a)x I(1 - t)2 Substituting (Al) in the above expression, a(l1-F(y + c+ ) de a(y + w + c - t(1 - t) (1 - t)[(1 + a)(1 - t) + a(1 - F)tt) dt (1 + a)x (I t)( (1 + a)(1 - t) + a(i - F)t - a(1 F)t - (1 + a)(1 - t) + a(i - F)t a(y + c + w) (1 + a)x (1 - t)2 21  (1 + a)(1 - t)a( c+ ) (1 + a)(1 - t) + a(l - F)t a(y + c + w) (1 + a)x(1 t a(y + c + w)< tA (1 - t)H(1 + a)(1 - t) + a(1 - F)t]x.