Printer Friendly

Efficiency wage models of unemployment: a second view.


Efficiency wage models, in which firms find it profitable to pay wages above workers' reservation wages, provide a promising explanation for unemployment and interindustry wage differentials. One criticism of such models is they imply firms should sell jobs by requiring up-front bonds from new workers. However, only some efficiency wage models imply this. Moreover, firms might not require bonds for many reasons. We show that moral hazard and adverse selection models together explain many labor market phenomena. The efficiency wage model conforms well to empirical finding, but certain anomalies suggest the need to consider rent-sharing models.

Carmichael's paper in this volume is an important contribution to the debate over the importance of efficiency wages. He clearly outlines some of the deficiencies in the theory, and thus points out areas needing further research. Nevertheless, his conclusions are far too negative. He greatly exaggerates the theoretical weakness of the efficiency wage model and gives insufficient recognition to the empirical implications generated by this model and to tests of these implications.


There is direct empirical evidence that shirking is a serious problem for firms. Firms spend significant sums monitoring workers. Levine [1987] reviews the evidence that worker shirking is widespread and costly. The evidence in Burroway [1979] and Mars [1982] also demonstrates the importance of shirking and stealing. Carmichael's primary argument against shirking models is that if the payment of efficiency wages to deter shirking led to an excess supply of workers, workers would buy their jobs by posting bonds. Since we do not observe workers posting bonds, he concludes that firms do not pay efficiency wages either.

Given the importance of shirking, the absence of market-clearing bonds needs explaining. One explanation is that the market has found some reasonable substitute for bonds, such as deferred wage payments. It is theoretically feasible to substitute deferred wage payments for up-front bonds, as noted by Akerlof and Katz [1989]. However, the empirical evidence of industries with deferred payments, i.e., where low beginning wages are balanced by high subsequent wages, is mixed. Dickens and Katz [1987b], Krueger and Summers [1988], and Abraham and Farber [1987] indicate that industries which pay high wages to high tenure workers tend to pay high wages to low tenure workers. On the other hand, high wage industries also tend to have steeper earnings profiles, which may be evidence of deferred earnings in these industries.

As Shapiro and Stiglitz [1984] note, a second possible explanation for the absence of market-clearing bonding is that bonding gives firms an incentive to fire workers immediately upon collecting their bonds. However, this explanation assumes that a firm will immediately fire every worker who posts a bond. If there are any reputation costs at all, some firms sometimes will not fire bond-posting workers, and we should observe some bonding. Yet with finite length contracts there must be some reputation costs, because otherwise firms would never pay workers in the last period. Thus, firm incentives to cheat and fire workers cannot explain the complete absence of bonding.

Clearly, the absence of market-clearing bonding is not fully understood, and further theoretical work is needed. Dickens, Katz, Lang, and Summers [1989] review a wide variety of possible reasons that bond-posting is limited. The important point is that there are many potential reasons for limited bonding other than the absence of efficiency wages.

On the other hand, the lack of bonding despite the importance of shirking does not demonstrate that firms pay efficiency wages. There may be reputation or other costs to the worker caught shirking, even in the absence of bonding. As Dickens, Katz, Lang and Summers [1989] demonstrate, in the absence of bonding, whether paying efficiency wages dominates relying solely on increased monitoring depends on the parameters of the model and is therefore an empirical issue.

It should be noted that the distinctions between the bonding model and the efficiency wage model are much smaller than the strength of feeling of each model's supporters would lead one to believe. When the models are set up equivalently (i.e., discrete or continuous time, finite or infinite lifetimes), the wages paid (including pensions) are identical in both models for all time periods, except the first period when the bond is or is not paid. This point is brought out in Akerlof and Katz [1989].

Thus the only distinction between the bonding model and the efficiency wage model (besides the up-front bond) lies in their predictions about unemployment, and even here the level of disagreement is less than at first appears. Efficiency wages lead to involuntary and inefficient unemployment. If bonding is costless, the bonding model leads to the efficient level of unemployment (i.e., where the discounted sum of marginal products is equal to the discounted sum of value of leisure/alternative time uses). However, as Dickens, Katz, Lang and Summers point out, if bonding were costless, we should not observe any monitoring since shirking is deterred more cheaply by increasing the size of the bond. We observe worker monitoring; hence bonding is not costless. Thus the level of unemployment is not first-best efficient even in the presence of bonding. Indeed, Lang [1989] shows that with costly bonding, the equilibrium is not even constrained Pareto efficient.

Carmichael correctly notes that even if worker credit constraints make bonding costly, as long as there are no discontinuities in the cost of bonds (e.g., if workers have no access to credit at all) and as long as the disutility associated with having no assets is sufficiently low, the market will still clear with smaller bonds offset by high payments later. The bonds will be small in terms of dollars, but not in utility. However, although costly bonding leads to only "voluntary" unemployment, it is voluntary only in the narrow sense used by economists. Ceteris paribus, the unemployed are those with the least assets. Given their options, the unemployed have "chosen" not to pay the bonds, yet the choice may have been between paying the bond or feeding their families. Although voluntary," this unemployment is every bit as much a social and economic problem as "involuntary" unemployment.

Carmichael takes his second argument against efficiency wage shirking models from MacLeod and Malcomsom's [1989] demonstration that unemployment leads to efficiency wages, rather than vice versa. MacLeod and Malcomsom's shirking model assumes exogenously-set supplies of labor and capital, fixed factor production, perfect ex-post observability of shirking, infinite horizons and no specific human capital or direct mobility costs. They contrast contingent wage payment at the end of the period ("piece rate") to payment at the beginning of the period. When there is an exogenous excess supply of labor, it is reasonable for capital owners to believe that they can always replace a worker at the going wage. Therefore, capital owners cannot be deterred from cheating (e.g., not paying workers) by the threat of the worker quitting, and worker payment must precede work. In this case, wages must be sufficiently above market-clearing ("efficiency wages") to ensure that workers prefer employment to unemployment and do not shirk. In this sense, unemployment leads to efficiency wages. Conversely, if there is an exogenous excess supply of capital, workers cannot be threatened with unemployment, and therefore contingent worker payments are necessary.

Despite their use of the term, the MacLeod/Malcomson paper is not about the standard efficiency wage model in which a) labor supply is endogenous and labor demand is variable and b) the firm cannot observe perfectly if workers have shirked. With the more realistic assumptions of endogenous supply and variable demand, unemployment can be made the consequence of efficiency wages.

Moreover, the efficiency wage model's assumption of imperfect monitoring by firms creates the need for the efficiency wages. With imperfect monitoring, the timing of payment is not what determines whether efficiency wages are paid. Even when a worker's payment is contingent on his not being caught shirking, if leisure on and off the job are perfect substitutes, firms must always pay efficiency wages, i.e. wages above the reservation wage. If leisure on and off the job are not perfect substitutes, whether making payment contingent on not being caught shirking is sufficient by itself to deter shirking, or whether efficiency wages must also be paid is an empirical issue.

Thus, MacLeod and Malcomson's assumption of perfect monitoring assumes away the basis for efficiency wage models. Yet even under these assumptions, they have demonstrated that efficiency wages may exist.


Carmichael is correct in stating that the sociological and selection models need considerable additional theoretical development, and that they should be examined for empirical implications that could provide tests of these theories. Possibly neither of these theories will withstand this further analysis and so will not provide the primary explanation of rigid wages and unemployment.

However, either or both of these factors could operate simultaneously with the shirking problem. Weiss [1980] shows high wages can serve both a shirking-deterrence and a selection purpose, and Akerlof [1982; 1984] proposes a combined guilt/morale/satisfaction purpose.

Carmichael is correct that the objectives of some shirking models and selection" models conflict. In the shirking model, firms will want to hire workers with the greatest incentive to work hard; ceteris paribus, these will be workers with the worst alternative opportunities. Conversely, in the selection model of Weiss, firms want to hire workers with the best alternatives, since these workers are most productive. However, Carmichael is wrong in concluding that the models therefore cannot be combined.

In fact, when the two models are combined, they explain both the absence of market-clearing bonds and the desire not to hire overqualified workers. Thus, in a model with shirking only, up-front bonds should clear the market. In fact, given the selection model's assumption of heterogeneous workers, there is a stronger case for requiring a bond: with heterogeneous workers, firms will not choose to set (lifetime) wages above everyone's reservation wages. While inframarginal workers will receive rents inhibiting their shirking, the marginal worker to enter the job queue will just be indifferent between taking the job or not and hence will not earn rents. Decreasing lifetime earnings by requiring a bond means that even the marginal worker will receive rents from the post-bond wage stream, and will therefore be discouraged from shirking. However, if there is a positive correlation between worker quality and reservation wages, as in the selection model, Weiss [1980] shows that the level of lifetime wages need not be market-clearing.

Moreover, if shirking and selection are both important, firms will pursue methods in addition to high wages to try to attract workers who are inherently more productive and inherently less likely to shirk or quit. This could explain the two seemingly contradictory empirical facts that Carmichael mentions in his argument against selection models although he does not juxtapose them and thus does not note their seeming contradictions). Firms are reluctant to hire overqualified workers and, at the same time, use elaborate screening methods to weed out bad workers. Firms do not want workers who feel that they have much better opportunities elsewhere, since they are likely to shirk or quit. At the same time, because raising wages is not a perfect device for selecting high productivity workers (since the correlation between productivity and the reservation wage is imperfect), employers seek other ways to evaluate productivity.

While models incorporating an active selection process into the efficiency wage model have yet to surface, a series of recent papers deal with the role of high wages in reducing vacancies including Beaudry [1988], Lang [forthcoming], Montgomery [forthcoming] and Weitzman [1989]. These "recruiting" efficiency wage models share the common element that firms with higher wages are less likely to have vacancies.


The proper basis for evaluating the success of the efficiency wage model relative to the competitive model is to ask whether the efficiency wage model has predicted the discovery of empirical relations which would not have been considered in its absence. We feel that recent empirical work is more extensive and more encouraging in this regard than Carmichael's review suggests.

There have been no attempts to measure marginal wage rigidity directly, but this is not surprising. Only the simplest of efficiency wage models predicts perfectly rigid wages. The shirking model, for example, implies that wages vary cyclically because as unemployment increases, the cost of being fired rises and thus the wage needed to deter shirking falls.

Instead, tests of the efficiency wage models fall into two categories: direct measurement of the wage/productivity tradeoff and studies of the inter-industry wage structure. Measurement of the wage/productivity relation is subject to at least one significant problem: every model in which firms maximize profits (including the competitive model) implies that wages equal the value of marginal product (at least over some appropriate interval or appropriate sample).

To avoid this problem, Leonard [1987] looks at the tradeoff between supervision and wages in 200 high-technology firms in one state. In an efficiency wage-shirking model, if the level of effort required of workers in all firms is the same and all firms (in the narrowly-defined industry) are located along the same (zero) isoprofit curve, the reduction in supervision costs must equal the wage increase. However, a one-to-one relationship between supervision costs and wages would also be consistent with a competitive model where high wage firms can observe the quality of the worker. In fact, Leonard finds that the reduction in supervision costs falls far short of justifying the higher wages. His result thus contradicts the predictions of both the competitive and the efficiency wage models. Of course, there are other reasons that the tests might have failed: firms may not be on the same isoprofit curve (e.g., because of different fixed capital levels or different costs of supervision) or there may be other important benefits of raising wages (perhaps selection benefits) which Leonard did not measure.

Raff and Summers [1987] avoid some of these problems by studying a single firm, the Ford Motor Company, which changed its wage policy. They find that Henry Ford's five-dollar wage policy increased profits, while creating considerable excess demand for jobs at Ford. They conclude that the five-dollar wage constituted a successful implementation of an efficiency wage policy. While avoiding some of the pitfalls of Leonard's "production function" approach, Raff and Summers's study suffers from the obvious disadvantage that it relates to a single and probably atypical firm.

Krueger [forthcoming] reverses the production function approach by asking whether wages are higher where supervision is more difficult. He argues that in the fast food industry, franchise owners are likely to supervise workers more closely than are managers in company-owned outlets. As a result, he predicts that wages will be higher in company-owned outlets than in franchises. This result is confirmed even after considerable effort to control for other potential determinants of wages. In addition, quit rates are lower in company-owned outlets. Of course, the possibility remains that there are other unmeasured factors determining both which outlets become franchised and their wage differentials.

Levine 1987] tries to measure the impact of wage differences for workers of identical quality by using survey data in which firms were asked to report their wages relative to the wages paid by their three leading competitors for similar workers. He finds that value of output (sales minus change in inventories minus advertising) per employee increase one for one with wages. As long as the workers in the different firms are truly similar, the results are evidence either of efficiency wages or rent-sharing. Of course, it is possible that despite the survey's wording, wage differences between respondents simply reflect worker quality differences. Moreover, even if workers are of identical quality, the fact that different firms chose to pay different wages may indicate other differences between firms that are responsible for the observed correlation.

Carmichael criticizes the literature for failing to predict which industry characteristics will be correlated with high wages. However, this argument is misleading. Most theoretical work on efficiency wages has assumed that the wage determines the number of efficiency units a worker provides or, equivalently, that the profit function takes the form:

[Mathematical Expression Omited] (1)

where, following Carmichael's notation, P is price, W is wage, L is labor, f is the production function, p is effective labor units per worker, X is a vector of factors which affect effective labor units, and K is other things that enter the production function. In this case, the well-known result is that the only factors which affect the wage are those which affect the elasticity of p with respect to w, so that K does not affect wage. As Carmichael notes, there is no obvious reason in this case that wages should be positively correlated with such factors as the capital/labor ratio.

While equation (1) is a particularly easy form with which to work, it is in many ways more natural to assume (as all the recruiting models have) that the probability that the worker provides a unit of labor depends on the wage. In other words, the probability that the worker does not shirk, quit or prove incompetent is a function of the wage. In this case the profit function is given by:

[Mathematical Expression Omited]

where L is the number of non-shirking or non-quitting workers, N is the number of workers recruited, [w.sub.1] is the per worker labor costs that are conditional on non-shirking or nonquitting, [w.sub.2] is the per worker labor cost incurred whether or not the worker shirks/quits, and q(L) is the probability that exactly L workers do not shirk/quit. This probability depends on labor costs, [w.sub.1] and/or [w.sub.2]. (In fact, recruiting models will have only conditional labor costs wi while shirking models will have only unconditional costs W2.) When the profit function takes this form, any factor which raises the value of marginal product also raises the optimal wage as shown in Lang [forthcoming].

The second body of empirical testing of the efficiency wage model looks at the interindustry wage structure. Dickens and Katz [1987a; 1987b], Krueger and Summers [1987; 1988], and Murphy and Topel [1987] show that there are very persistent (for periods of nearly a century) industry wage differentials which are similar in all developed economies. They therefore appear to reflect some underlying structural determinants.

Briefly, wage differentials are positively related to the average level of education in the industry (holding own education constant), the capital/labor ratio and measures of profitability. These findings are easily reconciled with either the efficiency wage or the competitive model. One would expect shirking (especially in the form of absenteeism) to be a more serious problem for firms which employ a lot of capital and for firms which are profitable. Thus the results conform to the predictions of the efficiency wage model. Similarly, in a competitive model framework where better workers are paid higher wages, better workers may be required by industries which are profitable or use a lot of capital.

However, the evidence on quit rates supports the efficiency wage rather than the quality difference hypothesis. Lower quit rates are observed in high wage industries, a fact predicted by efficiency wages but not by the competitive model in which wage differences reflect quality differences. Of course, it is possible to construct explanations of the quit finding which are compatible with the competitive quality difference model, e.g., by assuming that high capital industries require not only better workers but more investment in specific capital.

Likewise, evidence from job-changers supports efficiency wages as opposed to quality differences. If high wage industries employ better workers, industry wage differentials should largely disappear when measured only for a sample of job changers while they should be largely unaltered if the efficiency wage model is correct. While the work on this question suffers from some significant methodological problems (in particular the failure to take into account the endogeneity of turnover), it appears that the results are largely consistent with the prediction of the efficiency wage model. Again, it is possible to reconcile this finding with the competitive model, e.g., by assuming that the skills possessed by high ability workers are only productive in high wage industries.

Thus, predictions generated by the efficiency wage model are consistently confirmed. While the competitive model can always be adjusted so that it explains" the new finding, the revision is devoid of new empirical content. No theory is ever definitively proved or disproved. However, the efficiency wage model has predicted the discovery of empirical relations, and explains them with fewer auxiliary hypotheses than are needed by a quality-difference model. These are the criteria for a strong theory.

On the other hand, one empirical regularity of the industry wage differential literature cannot be explained directly by the efficiency wage model and requires auxiliary hypotheses. Industry wage differentials appear to be strongly correlated across occupational groups. An efficiency wage-shirking model provides no obvious reason for this finding, although it can be reconciled with the efficiency wage model either by drawing on the sociological" efficiency wage model so that equity of the internal wage structure is an important determinant of wages, or by positing sufficient complementarity among different kinds of workers.

A third model-rent-sharing by workers-appears to be consistent with all of the empirical evidence presented above. Thus, with rent-sharing, a) profitable firms and firms with high capital/labor ratios (to the extent that the capital is financed by equity and is thus a fixed cost) will have more rents or quasi-rents for workers to capture and will therefore have higher wages; b) since high wages represent rents, workers in high wage firms are less likely to quit; c) workers who change to high wage industries gain rents; and d) industry wage differentials will be correlated across occupations.

A natural test of the rent-sharing model occurs with deregulation. With no rent-sharing by workers, deregulation lowers the output price and increases the derived demand for labor. Wages should rise in response to increased demand in the short run. Rent-sharing, on the other hand, implies that the reduction in rents should lower wages. Casual empiricism and the analysis in Rose [19871 is supportive of the rent-sharing model.

A discussion of why rent-sharing may occur is beyond the scope of this response. However, efficiency wage-type arguments might contribute to or even be responsible for rent-sharing. If firms share rents (i.e., set higher wages) to minimize the threat of collective action and thus maximize long-term profits as in Dickens [19861, unemployment will result. Thus, this model of rent-sharing is essentially an efficiency wage model. Alternatively, the shirking deterrence and/or selection effects of efficiency wage models make it cheaper and therefore more plausible that managers share some of the rents with workers.


Our conclusions about the efficiency wage model are much more favorable than Carmichael's. While there remain some theoretical problems to be resolved, the efficiency wage hypothesis, especially when combined with a model of rent-sharing, provides an explanation for a number of labor market phenomena which are not easily explained otherwise.


Abraham, Katherine and Henry S. Farber. "Job Duration, Seniority, and Earnings." American Economic Review, July 1987, 278-97.

Akerlof, George A. "Labor Contracts as Partial Gift Exchange." Quarterly Journal of Economics, November 1982, 543-69.

Gift Exchange and Efficiency Wages: Four Views." American Economic Review, May 1984, 79-83.

Akerlof, George and Lawrence F. Katz. "Workers' Trust Funds and the Logic of Wage Profiles." Quarterly Journal of Economics, August 1989, 525-36.

Beaudry, Paul. "Entry Wages Signalling Future Wages: A Micro-Foundation to Turnover Models of Unemployment." Photocopy, Montreal, 1988.

Bulow, Jeremy I. and Lawrence H. Summers. "A Theory of Dual Labor Markets with Application to Industrial Policy, Discrimination and Keynesian Unemployment." Journal of Labor Economics, July 1986, 376-414.

Burroway, Michael. Manufacturing Consent. Chicago: Chicago University Press, 1979.

Dickens, William T. "Wages, Employment and the Threat of Collective Action by Workers." National Bureau of Economic Research Working Paper No.1856, 1986.

Dickens, William T. and Lawrence F. Katz. "Interindustry Wage Differences and Industry Characteristics," in Unemployment and the Structure of Labor Markets, edited by Kevin Lang and Jonathan S. Leonard. Oxford: Basil Blackwell, 1987a, 48-89.

Interindustry Wage Differences and Theories of Wage Determination." National Bureau of Economic Research Working Paper No. 2271, 1987b.

Dickens, William T., Lawrence F. Katz, Kevin Lang and Lawrence H. Summers. Employee Crime and the Monitoring Puzzle." Journal of Labor Economics, July 1989, 331-47.

Krueger, Alan. "Ownership, Agency and Wages: An Examination of Franchising in the Fast Food Industry." Quarterly Journal of Economics, forthcoming 1990.

Krueger, Alan and Lawrence H. Summers. "Efficiency Wages and the Inter-industry Wage Structure." Econometrica, March 1988, 259-93.

"Reflections on the Interindustry Wage Structure," in Unemployment and the Structure of Labor Markets, edited by Kevin Lang and Jonathan S. Leonard. Oxford: Basil Blackwell, 1987, 17-47.

Lang, Kevin. "Why Was There Mandatory Retirement? Journal of Public Economics, June 1989, 127-36.

Persistent Wage Dispersion and Involuntary Unemployment." Quarterly Journal of Economics, forthcoming 1990.

Leonard, Jonathan S. "Carrots and Sticks: Pay, Supervision and Turnover." Journal of Labor Economics, October 1987, S136-52.

Levine, David. "Tests of Efficiency Wage and Rent-Sharing Theories: The Pay-Productivity Relation," Photocopy, Berkeley, 1987.

MacLeod, Bentley and James Malcomson. Implicit Contracts, Incentive Compatibility, and Involuntary Unemployment." Econometrica, March 1989, 312-22.

Mars, Gerald. Cheats at Work. London: Unwin Press, 1982.

Montgomery, James. "Equilibrium Wage Dispersion and Inter-Industry Wage Differentials." Quarterly Journal of Economics, forthcoming 1990.

Murphy, Kevin M. and Robert H. Topel. "Unemployment, Risk and Earnings: Testing for Equalizing Differences in the Labor Market," in Unemployment and the Structure of Labor Markets, edited by Kevin Lang and Jonathan S. Leonard. Oxford: Basil Blackwell, 1987, 103-40.

Raff, Daniel and Lawrence H. Summers. "Did Henry Ford Pay Efficiency Wages?" Journal of Labor Economics, October 1987, S57-86.

Rose, Nancy L. "Labor Rent-Sharing and Regulation: Evidence from the Trucking Industry." Journal of Political Economy, December 1987, 1146-78.

Shapiro, Carl and Joseph E. Stiglitz. "Equilibrium Unemployment as a Worker Discipline Device." American Economic Review, June 1984, 433-44.

Weiss, Andrew. "Job Queues and Layoffs in Labor Markets with Flexible Wages." Journal of Political Economy, June 1980, 526-38.

Weitzman, Martin. "A Theory of Wage Dispersion and Job Market Segmentation." Quarterly Journal of Economics, February 1989, 121-38.
COPYRIGHT 1990 Western Economic Association International
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1990 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Lang, Kevin; Kahn, Shulamit
Publication:Economic Inquiry
Date:Apr 1, 1990
Previous Article:Efficiency wage models of unemployment - one view.
Next Article:Export instability and long-term capital flows: response to asset risk in a small economy.

Terms of use | Copyright © 2017 Farlex, Inc. | Feedback | For webmasters