Efficiency Wage Models: Explaining Wage Differentials and Unemployment, Study Guides, Projects, Research of Literature

Download Efficiency Wage Models: Explaining Wage Differentials and Unemployment and more Study Guides, Projects, Research Literature in PDF only on Docsity! NBER WORKING PAPER SERIES EFFICIENCY WAGE THEORIES: A PARTIAL EVALUATION Lawrence F. Katz Working Paper No. 1906 NATIONAL BUREAU OF ECONOMIC RESEARCH 1060 Massachusetts Avenue Cambridge, MA 02138 April 1986 This is a revised version of a paper presented at the NBER Conference on Macroeconomics held in Cambridge, Massachusetts on March 7 and 8, 1986. 1 am indebted to William Dickens for numerous discussions and for comments on previous drafts of this paper; a significant part of this paper grew out of our joint work. I thank George Akerlof, Joe Altonji, Stan Fischer, Kevin Lang, Lawrence Summers, Janet Yellen, and the conference participants for helpful comments. I am grateful to Phil Bokovoy and Elizabeth Bishop for expert research assistance. The Institute of Industrial Relations at U.C. Berkeley provided research support. All remaining errors are my own. The research reported here is part of the NBER's research program -in Labor Studies. Any opinions expressed are those of the author and not those of the National Bureau of Economic Research. NBER Working Paper #1906 April 1986 Efficiency Wage Theories: A Partial Evaluation ABSTRACT This paper surveys recent developments in the literature on efficiency wage theories of unemployment. Efficiency wage models have in common the property that in equilibrium firms may find it profitable to pay wages in excess of market clearing. High wages can help reduce turnover, elicit worker effort, prevent worker collective action, and attract higher quality employees. Simple versions of efficiency wage models can explain normal involuntary unemployment, segmented labor markets, and wage differentials across firms and industries for workers with similar productive characteristics. Deferred payment schemes and other labor market bonding mechanisms appear to be able to solve some efficiency wage problems without resultant job rationing and involuntary unemployment. A wide variety of evidence on inter—industry wage differences is analyzed. Ef- ficiency wage models appear useful in explaining the observed pattern of wage differentials. The models also provide several potential mechanisms for cy- clical fluctuations in response to aggregate demand shocks. Lawrence Katz School of Business Administration University of California Berkeley, Ca. 94720 (415) 642—4732 The paper is organized as follows. Section 2 presents a rudimentary efficiency wage model and discusses some of the basic implications of the efficiency wage hypothesis. Alternative models with efficiency wage structures, their empirical predictions, and theoretical shortcomings are analyzed in section 3. Efficiency wage models in which firms are assumed to only be able to utilize simple wage schemes as compensation mechanisms are shown to be capable of generating a number of important labor market phenomena including involuntary unemployment, dual (segmented) labor markets, and wage distributions for workers with identical productive characteristics. The same problems, such as the inability of firms to monitor worker performance costlessly and costly turnover, that give rise to efficiency wage payments above the market clearing level create incentives for the use of alternative incentive devices and the development of internal labor markets and long—term contractual relationships in the labor market. Alternative forms of labor contracts, typically involving the (implicit) posting of perform- ance bonds, can eliminate the job rationing that arises in versions of the models in which firms are limited to the use of simple wage policies. Practical problems arising from capital market imperfections and moral hazard problems on the part of firms may limit the potential for alternative compensation arrangements to eliminate efficiency wage problems. A wide variety of evidence on inter—industry wage differences is analyzed in section 4. Efficiency wage models make strong predictions concerning the exist- ence of wage differentials arising from differences across industries in the wage—productivity relationship. Important systematic wage differentials across industries are present that do not appear to be easily explained by the standard competitive rationales of differences in labor quality, compensating differen- tials, or transitory disturbances. Although no single efficiency wage model seems Efficiency Wage Theories: A Partial Evaluation 4 consistent with all the facts, efficiency wage models do appear useful in ex- plaining the observed pattern of wage differentials. The consistency of effi- ciency wage theories with evidence on the cyclical behavior of labor markets and on labor market discrimination is also discussed in section 4. Section 5 discusses the mechanisms through which efficiency wage models may help explain wage rigidity and cyclical fluctuations. The models explain why firms may not lose much if they fail to adjust wages to shocks. The addition of small costs of changing prices and wages, as emphasized by Mankiw (1985) and Blanchard and Kiyotaki (1985), or of near—rational inertial behavior, as analyzed by Akerlof and Yellen (1985a,b), to efficiency wage models leads to a potential model of cyclical fluctuations in response to aggregate demand movements. Concluding remarks concerning the usefulness of the efficiency wage approach are presented in section 6. 2. The Basic Efficiency Wage Hypothesis Some of the primary implications of efficiency wage models can be illuminated in a simple model in which a worker's physical health and therefore productivity is assumed to depend positively on the real wage paid. This formulation was advanced by Leibenstein (1957) to highlight the linkages among wages, nutrition, and health in less developed countries. Firms, in this context, get healthier, more pro- ductive workers if they pay higher wages. Solow (1979) formulates a formally similar model for developed economies in which increased wages improve morale and this directly affects productivity through an increase in worker effort. Efficiency Wage Theories: A Partial Evaluation 5 Following Yellen (1984), I consider an economy with identical, perfectly compet- itive firms, each possessing a short—run production function of the form Q = aF(e(w)L) where e is the effort (or efficiency) level of a worker, L is the number of employees, w is the real wage, a is a productivity shifter, and Q is output. The price of output is taken to be the numeraire. All workers are assumed to have identical wage—productivity relationships of the form e(w) with e'>O, e(O)O, and the elasticity of e(w) with respect to w declining in A profit—maximizing firm, able to hire all the labor it wants at the wage it chooses to offer, solves the following problem (1) max aF(e(w)L)-wL w, L The solution to the problem yields (2) eI(w*)w*/e(w*) = 1 and (3) e(w*)aFI(e(w*)L) = w. The optimal wage w* satisfies the condition that the elasticity of effort with respect to the wage is unity. The wage w* is known as the efficiency wage since it minimizes wage costs per efficiency unit of labor. Each firm hires labor up to the point where its marginal product equals w*. Efficiency Wage Theories: A Partial Evaluation 6 creates an incentive for workers not to shirk.. Models in which the need of firms to elicit effort from their workers can lead to the payment of wages in excess of market clearing and generate equilibrium involuntary unemployment have re- cently been examined by Bowles (1985), Bulow and Summers (1986), Calvo (1979,1985), Calvo and Wellisz (1978), Eaton and White (1982, 1983), Foster and Wan (1984), Gintis and Ishikawa (1983), Jones (1985), Shapiro and Stiglitz (1984), and Stoft (1982). In the Shapiro and Stiglitz (1984) version of the model, firms can only imperfectly monitor their workers' job performance, and workers make a discrete choice of whether to work or shirk. Workers and firms are assumed to be homogenous. If all firms pay the same market clearing wage, there is full em- ployment and no cost to shirking since workers can immediately find another job at the sane wage if fired. This strong result of no costs of shirking requires no job switching or search costs and no adverse reputational effects on workers in the labor market if they develop a poor employment history. The homogenous workers assumption eliminates reputational effects on workers because all workers are assumed to act the same given the same incentives. If effort is costly, all Thus, it pays each firmworkers shirk under these full employment conditions to increase its wage to eliminate shirking. When all firms do this, the average wage rises and employment is reduced. In equilibrium, all firms pay a wage above the market clearing level creating unemployment. Since jobs are scarce and ra— tioned, the loss of a job can involve a lengthy unemployment spell. The reserve army of the unemployed acts as a worker discipline device making shirking costly. Although some unemployment is optimal in this model since it plays a required role in creating work incentives, Shapiro and Stiglitz show that the equilibrium un— employment rate is not Pareto optimal.9 Equilibrium unemployment is involuntary Efficiency Wage Theories: A Partial Evaluation 9 in this model since identical workers are treated differently and since the un- employed strictly prefer to be be employed. The shirking model postulates a variety of factors that affect the firm's ability to extract effort from workers and consequently yields some potentially testable predictions concerning the nature of wage differentials and unemployment. Firms should pay higher wages to a given quality worker where monitoring is costly and/or difficult so that the probability of detecting shirking is low. Higher wages may be required for positions in which poor employee performance can cause a great deal of damage. In fact, the job evaluation systems used in the design and maintenance of wage structures in many industries rate positions on a re- sponsibility factor that is directly related to the probable damage that can be caused by improper job performance (Milkovich and Newman, 1984). Workers in po- sitions of trust and responsibility should receive wage premiums (Eaton and white, 1982). The value of a worker maintaining a job is reduced if the likelihood of a future separation is great. This means firms with monitoring difficulties should avoid hiring workers from observable groups believed to exhibit high turnover and should attempt to maintain long—term employment relationships per- haps through the use of worksharing or a layoff—recall process to deal with tem- porary downturns. Increased variability in labor demand across sectors (greater sectoral shift activity) directly increases unemployment through more separations to facilitate labor reallocation and indirectly raises the structural unemploy- ment rate by requiring firms to pay higher wages to prevent shirking since it induces a greater likelihood of a future separation.1° The expected income form unemployment affects the wage needed to induce proper worker behavior. A higher unemployment insurance benefit raises the required wage Efficiency Wage Theories: A Partial Evaluation 10 and reduces employment. A higher unemployment rate and hence longer expected duration of unemployment for a worker sacked for shirking reduces the needed wage. Finally, the level of wages offered by other firms affect the prospects of a discharged worker. The model suggests firms should be concerned with their po- sition in the wage hierarchy (Bulow and Summers, 1986). 3.1.2. Segmented Labor Markets The shirking model also provides a rationale for dual labor markets with a utility differential for similar workers across the primary and secondary sectors and rationing of primary sector jobs. The dual labor market hypothesis states that the labor market can be roughly be divided into a primary sector that offers jobs characterized by high wages and internal labor markets and a secondary sector that offers low paying, menial jobs with little room for advancement (Doeringer and Piore, 1971). Dickens and Lang (1985a,b) find in two different micro data sets that the esti- mation of a switching model of wage determination with unknown regimes yields two distinct wage equations. The two equations closely resemble the predictions of dual market theory for the characteristics of earnings functions in the primary and secondary sectors. The equation to which most workers are associated yields significant returns to experience and education. The other equation indicates little or no returns from human capital variables. The estimation technique al- lows the simultaneous determination of the probability of each worker's attachment to each sector and each sector's earnings equation. The procedure allows a hy- pothesis test that can be interpreted as a test of the rationing of primary sector Efficiency Wage Theories: A Partial Evaluation 11 at these firms. The threat of forfeiting a bond or of paying a new employment fee to gain employment can create work incentives and enable the market for jobs to clear eliminating involuntary unemployment. Instances of workers posting direct performance bonds or purchasing their jobs are rare, although not entirely unknown. Employment arrangements that may im- plicitly perform bonding functions such as upward sloping age—earnings profiles, pensions and other deferred compensation schemes, and internal promotion ladders are observed in a large segment (the primary sector) of the labor market. These mechanisms appear in large establishments where monitoring problems are likely to be important. Lazear (1979,1981) demonstrates that seniority wage systems in which workers post a bond against cheating by accepting wages below their marginal product initially and have it returned in the form of wages above the value of their marginal product later in their careers and/or in the form of a pension upon retirement can solve the effort elicitation problem. Practical objections arise to the use of complete bonding schemes in the labor market. In the first place, workers, particularly early in their working lives, face capital market constraints and lack the liquidity required to post large bonds. If the probability of detecting shirking is low, the required bond or em- ployment fee may be substantial. Carmichael (1985) argues that even if capital markets are imperfect, firms can charge a fee sufficient to make the expected utility of the job offer equal to the value of workers' reservation wages. Al- though this may eliminate the directly involuntary nature of unemployment, it does not lead to an efficient level of unemployment (Shapiro and Stiglitz, 1985). Efficiency Wage Theories: A Partial Evaluation 14 Since outside party verification of detected shirking is difficult, firms have an incentive to falsely claim that workers are shirking and claim the bonds. Firms may collect employment fees and then dismiss workers. The firm's concern for its reputation as an employer may be able to overcome this problem (Lazear; 1979, 1981). The difficulty of potential workers in verifying the honesty of a firm's behavior means that the reputation mechanism is quite fragile and may be a far from perfect enforcement mechanism (Kreps, 1984). The likelihood of firm default on the bond can be reduced if the firm does not expect to gain from falsely claiming that the worker is shirking. For example, if the firm claims that a worker has shirked and discharges the worker, it is possible for the bond to be paid to a third party instead of the firm. The worker is disciplined and the firm does not gain from cheating. We do not see the direct use of third party systems like this in practice. Carmichael (1983) argues that seniority promotion rules with a fixed wage hierarchy can play this role. Additionally, tournament schemes (Bhattacharya, 1983 and Malcoimson, 1984) may permit the firm to commit itself to a wage plan that creates the proper incentives for workers. Reputational considerations are likely to be important precisely for the large, visible employers, such as IBM or General Motors, that provide high paying, pri- mary sector jobs. These large firms offer exactly the type of jobs that the shirking model indicates should pay efficiency wages and be rationed. Smaller, less visible secondary firms which may not stay in business long are unlikely to be able to get workers to trust that they will not renege on agreements concerning deferred compensation. Thus, the implication of considering possible bonding mechanisms and where they may be effective is that the secondary sector needs to pay efficiency wages while the primary sector can utilize deferred payments and pay market wages. This unrealistic prediction of the model suggests that further Efficiency Wage Theories: A Partial Evaluation 15 work must be done on why full bonding schemes are not practicable for visible employers quite likely to value their reputations for keeping (implicit) promises to their employees. If capital markets were perfect and third party verification of shirking always possible, firms would spend next to nothing on raising the probability of de- tecting shirking and demand large bonds since monitoring is costly and the posting of bonds would be costless under these circumstances (Becker and Stigler, 1974; Dickens, Katz, and Lang, 1986). The empirical observation that firms devote substantial resources to monitoring workers suggests that a full bonding solution to the shirking problem is unattainable.' Thus, firms must be choosing one or both of two second best alternatives to bonding: monitoring workers intensively and/or paying efficiency wages. The likely outcome is that firms utilize bonds to the extent possible, and then pick the optimal combination of efficiency wages and monitoring required to prevent shirking in the presence of limited bonding ability. Lazear (1979) argues that the existence and observed pattern of mandatory re- tirement provisions indicates that some bonding must be utilized in the labor market. Adverse selection problems provide an alternative rationale for the use of mandatory retirement policies. If individual performance is hard to observe, wage cuts for older workers may lead to the better workers moving to other jobs and the "lemons" remaining. Additionally, Medoff and Abraham (1980) provide ev- idence that experience—earnings and tenure—earnings profiles cannot be fully ac- counted for by rising productivity. The typical finding in cross—section wage equation estimates of a positive effect of years of tenure at current job (sen- iority) on earnings is often pointed to as evidence of these types of bonding Efficiency Wage Theories: A Partial Evaluation 16 3.3. The Adverse Selection Model Imperfect. information by firms about the abilities of workers may provide a se- lection rationale for efficiency wage payments. If workers are heterogeneous in ability and if ability and reservation wages are positively correlated, firms which offer higher wages will attract higher quality job applicants. If firms cannot observe applicant quality and lack devices to induce workers to reveal their true abilities, random hiring from the applicant pool must be utilized. A higher wage increases the expected ability of a worker hired randomly from the applicant pool. A wage above the market clearing level may minimize costs per efficiency unit of labor under these circumstances (Stiglitz, 1976; and Weiss, 1980). Institutional, legal, or sociological constraints preventing firms from differentiating wages across workers with different productive characteristics can lead to similar results.'4 A basic objection to the model is that firms are likely to eventually learn a worker's ability. In this case, performance bonds can solve the adverse selection problem. The same moral hazard problem on the part of firms and capital market imperfections that limit the use of bonding for shirking problems apply in this context as well. If firms can measure performance on the job, pay—for—performance schemes eliminate the problem. Employment contracts with self—selection incen- tives (nonlinear wage— employment contracts) can also potentially ease adverse selection problems (Mookherjee, 1984a). 3.4. Sociological Models Efficiency Wage Theories: A Partial Evaluation 19 Workers effort levels may significantly depend on the extent to which they feel they are being treated fairly by their employers. The perceived justness of the wage may affect worker productivity if effort leVels are linked to worker morale and feelings of loyalty to the firm.'5 Akerlof (1982,1984) and Solow (1979) argue that wage rigidity in the face of unemployment may be due to the importance of social wage norms and other behaviors not well—captured by traditional individualistic utility functions. Akerlof (1984) discusses a variety of evidence from sociological studies indicating that a worker's effort level depends on the norms of his or her work group (peer pressure) and posits a number of sociologically based models with efficiency wage implications. Akerlof (1982) develops a model in which firms can raise group work norms by offering wages above the level necessary to attract a labor force. The firms "gift" of high wages is rewarded by the "gift" of improved work norms and increased individual effort. Wages in excess of market clearing may be the outcome when wages play a dual role of both allocating labor across firms and of satisfying interpersonal and intertemporal wage norms that matter for worker performance. Most firms pay careful attention to the perceived fairness and consistency of their internal wage structures. Doeringer and Piore (1971) find that firms devote far more resources to and place more weight in their wage policies on job evalu- ation programs designed to maintain and br justify their internal wage structures than to market wage surveys utilized to keep wages in line with those of product or labor market competitors. Richard Wing (1984, p. 9/18), a former compensation director at Eastman Kodak, notes There is always some degree of conflict between internal and external pay equity. The position taken by most salary administrators is that internal relationships should be given first priority, and external pay relationships for certain jobs must be compromised on occasion. Efficiency Wage Theories: A Partial Evaluation 20 If certain wage relationships matter to a firms workforce, it is in the firm's interest to take those considerations into account. The sociological models indicate that efficiency wages are likely to arise where work groups and teamwork are important. These models may also explain the direct impact of product market factors and the firm's "ability—to—pay" on wages. Worker morale and loyalty and consequently productivity may depend on the extent that the firm shares its rents with its employees. 3.5. The Union Threat Model Firms are likely to face important diseconornies of scale in the hiring and training of workers. If there are costs to job search and/or relocation, a firm may have to increase its wage offer or reduce the quality of accepted applicants to replace a large number of workers quickly. If one or a few workers leaves a firm, their co—workers are likely to have overlapping firm— and job—specific knowledge and to be able to train replacements. When many workers quit in a short time period, more valuable knowledge is lost per worker and no one may be left capable of training replacement employees. Since the costs of turnover to the firm rise rapidly as the number of workers needed to be replaced in a given in- terval increases, collective action can provide workers with more bargaining power than they have acting individually. Social norms such as the willingness of customers and suppliers to boycott a struck firm, may in many settings prevent firms from doing business even if they can replace workers who attempt to act collectively. Collective action by workers may enable them to shut a business down. This bargaining power can potentially be used by workers to claim for Efficiency Wage Theories: A Partial Evaluation 21 Table 1: A Synapsis of Alternative Efficiency Wage Theories Theory Problems Leading to Efficiency Wage Payments Benefits to Firm of High Wages Shirking Turnover Imperfect observability of worker effort level and performance; monitoring is costly Firms must bear part of turnover costs (hiring and training costs) Imperfect observability of worker quality and performance Raise cost of job loss encouraging good performance; economize on monitoring costs High wages reduce turnover costs if quit rate is decreasing function of wages higher quality applicants productive have better opportunities Sociological Morale and worker feelings of loyalty to firm depend on perceived fairness of wages Improved work norms, morale, feelings of loyalty to firms which raise productivity Union Threat Costs of replacing existing workforce gives employees bargaining power Maintain industrial peace and/or prevent unionization Adverse Selection Attract pool of if more workers outside within long—term employment relationships may be able to solve some efficiency wage problems without the resort to job rationing. The empirical relevance of efficiency wage theories is examined in more detail in the next section. 4. Some Empirical Evidence Relating to Usefulness f the Approach 4.1. Inter—industry Wage Differences 4.1.1. Some Implications of Efficiency Wage Models for Wage Differentials It has long been noted that there are large differences in wages across industries for apparently similar work. Substantial industry wage differentials remain even after controlling for union status and observed worker and Job characteristics (Bloch and Kuskin, 1978; Dickens and Katz, 1986 ; Krueger and Summers, 1986; and many others). In fact, large wage differences for essentially the same type of work in the same locality have invariably been uncovered by area wage surveys. Slichter (1950) observes that hiring rates paid for common labor by 85 plants in Cleveland in February 1947 ranged from $.50 to $1.09 an hour. Re notes that this spread persisted despite the fact that rates had been compiled and distributed to firms throughout the city by the Cleveland Chamber of Commerce for over twenty years. John Dunlop (1985; p. 18) summarizes the typical finding of studies of local wage variation: It is a well established fact that wage rates or average hourly earnings for a defined job classification, such as maintenance electrician or key punch operator, show very wide variation in a locality, particularly in a community with a variety of industries. The top wage rates for the same job classification are often two or three times the low ones. Differences in fringe benefit programs enlarge these differences. Efficiency Wage Theories: A Partial Evaluation 24 These wage differentials provide an empirical challenge to alternative labor market models. The ability of competitive and efficiency wage models of the labor market to meet this challenge is the focus, of this section which draws heavily upon the review and analysis presented in Dickens and Katz (1986). A standard competitive labor market model implies that persistent industry wage premiums require industry—related differences in labor quality (skill) and/or non—wage dimensions of work requiring equalizing differences. Reder (1962, p. 276) summarizes these predictions: In the long run, under competitive conditions, any industry will pay the same price for a given grade of labor as any other industry hiring in the same location. This remark must be qualified for differences in nonpecuniary attractions of different industries and locations . . . Therefore, in the long run, the real wage differentials among industries will reflect differences in the skill mix. Alternatively, industry wage differences at any given point of time for similar work may reflect transitory differentials related to shifts in labor demand across sectors and imperfect short—run labor mobility. On the other hand, a basic implication of efficiency wage models is that if the conditions necessitating efficiency wage payments differ across industries, then the optimal wage will differ among industries. This means that workers with identical productive characteristics are paid differently depending on their in- dustry affiliation. These wage differences for similar workers reflect industry characteristics that do not directly affect the utility of workers and thus do not require compensating differentials. Intra—industry wage distributions for Efficiency Wage Theories: A Partial Evaluation 25 TABLES Table 2: Estimated OLS Log Wage Differentials for One—Digit Industries and Union Status 1983 CPS — Nonagricultural Private Sector Workers Variable Coefficient (Standard Errors) Mining .289 (.009) Construction .127 (.005) Nondurables .050 (.004) Durables .098 (.004) Transportation & Public Utilities .154 (.005) Wholesale Trade .042 (.005) Retail Trade —.161 (.004) FIRE .052 (.005) Servtces — .064 (.003) Union .192 (.003) Sample Size 134928 .546 Controls included are education (years of schooling) and its square; experience (age—education—5) and its square; married; sex; race; part—time work; SMSA; interaction terms for both experience and its square with married, race, sex, education, part—time work and SMSA; 11 occupation dummies; and 50 state dummies. Tables state of residence. The industry variables have a sizeable impact on wages. For example, workers in mining and transportation and public utilities earn approxi- mately 45 and 32 percent more than observationally equivalent workers in retail trade. The industry wage effects are comparable in magnitude to the effect of union status. Krueger and Summers (1986) provide evidence on industry wage differentials for a more disaggregated industry breakdown. They report industry wage premiums that range from 38 percent above the average industry for the petroleum industry to 37 percent below the average for private household workers in estimates from the 1984 May Current Population Survey that include a wide variety of controls for individual characteristics, union status, and occupation. Kreuger and Summers find the employment—weighted standard deviation of industry wage premiums for two digit industries to range from 10 to 15 percent for different years of the CPS from 1974 to 1984. Dickens and Katz (1986) find that the industry wage differentials persist when union and nonunion workers are analyzed separately. The patterns of industry wage premiums are extremely similar for union and nonunion employees. The raw corre- lation of three digit industry wage premiums for union and nonunion workers in the combined 1983 CPS sample is .65.19 4.1.2. Competitive Explanations for Wage Differentials A number of explanations consistent with standard competitive labor market models are possible for the large impact of industry affiliation on wages even after controlling for measured human capital variables. The first posits that differ— Efficiency Wage Theories: A Partial Evaluation 28 ences in technology across industries make it profitable to hire higher quality workers (and hence pay higher wages) in some industries. Estimated industry wage premiums in a cross—section may primarily reflect individual—specific components of earnings capacity that are unobservable (to the econometrician) and correlated with industry status. If unmeasured ability is highly correlated with observed labor quality variables, such as years of schooling and labor market experience, then unmeasured labor quality cannot provide an explanation for the large esti- mated industry wage effects, Dickens and Katz (1986) show that the size of in- dustry wage effects is not much altered if wage equations are first estimated without industry variables and then the residuals are used to determine the in- dustry impacts. This approach credits observed quality variables with the impacts of unobserved variables correlated with both measured quality variables and in- dustry status. Longitudinal data provides a potential vehicle to control for time invariant, unmeasured labor quality. If high wage industries simply have workers of high unobserved ability and if workers of a given quality are paid equally in different industries, wage changes should not systematically be limked to changes in in- dustry status. Longitudinal data allow one to examine the wages of a given in- dividual as he or she switches industries. First difference (or fixed—effects) estimation allows one to eliminate the impacts of unchanging unobserved ability components (that are rewarded equally in all industries) on the industry wage effects estimates. Krueger and Summers (1986) estimate large effects of industry switches (for broadly defined industries) on wages in first differenced re- gressions utilizing a pooled sample of matched May CPS data for 1974—75, 1977—78, and 1979—80. The estimated industry effects from the first differenced regression are similar in direction and magnitude to pooled regression estimates. Thus, Efficiency Wage Theories: A Partial Evaluation 29 ticular grades of labor were fairly stable. Slichter (1950) finds the rank cor- relation of males unskilled average hourly earnings for 20 manufacturing industries between 1923 and 1946 to be •7324 Strong stability in inter—industry wage rankings is also evident for the postwar United States. Montgomery and Stockton (1985) report that the rank correlation of mean hourly wages for 20 2—digit manufacturing industries between 1951 and 1981 was .675. Bell and Freeman (1985) find strong stability in the rankings for a group of 53 industries (both manufacturing and non—manufacturing) from 1948 to 1982. Both Bell and Freeman and Montgomery and Stockton note that the dispersion (log standard deviation) of industry mean wages increased substantially during the 1970's. On the other hand, Krueger and Summers (1986) find that the estimated two digit wage premiums using individual data from the CPS and controlling for individual characteristics did not appear to have grown from 1974 to 1984. Krueger and Summers find the correlation of the estimated industry wage premiums between 1974 and 1984 to be .970. These results indicate that the rising dispersion in average industry wages since the early 1970's may largely reflect changes in the composition of labor forces across industries and possibly also changes in union wage impacts. A further possibility is that industry wage differences arise from differences in patterns of human capital accumulation across industries. Krueger and Summers (1986) find that industry wage differentials are approximately equal in magnitude and highly correlated for young (20 to 35 years old) and older (50 to 65 years old) workers. Furthermore, the 1979 Current Population Survey contains informa- tion on job tenure (years with current employer). I utilized this data set to determine if industry wage differences vary with job tenure. Estimates of log Efficiency Wage Theories: A Partial Evaluation 32 earnings equations for separate tenure groups for nonunion, private sector workers indicated that the industry differentials are quite similar for different tenure groups. For example, the standard deviation of one-digit industry wage differen- tials for workers with less than one year of tenure and for workers with greater than 10 years of tenure are .11 and .12 respectively.25 The correlation (corrected for sampling error) for the differentials of the two tenure groups is .74. Large industry wage effects are apparent for entry level workers that are close in size and highly correlated with those of long—term employees. One exception is that the differentials of all other industries versus retail trade appear to be sub- stantially larger for workers with long job tenure than for entry level workers. 4.1.3. The Occupational Structure of Industry Wage Premiums Although most explanations for wage differentials provide reasons why one would expect particular occupational groups to be high paid in some industries relative to other industries, they do not lead one to expect the pattern of wage premiuns to be the sane across industries for diverse occupational groups. For example, a standard competitive model suggests that an industry with dangerous production jobs may pay its blue collar workers high wages to compensate them for the risks their jobs entail, but it does not also suggest that secretaries in this industry should earn a pay premium. Working conditions, skill requirements, and monitoring problems are quite likely to differ across occupations in a firm or industry. Dickens and Katz (1986) estimate industry wage differentials by occupation (for a 12 occupation and three digit Census of Population industry breakdown) for nonunion, private sector workers from the combined 1983 CPS sample. The effects of human capital variables, demographic characteristics, and locational variables Efficiency Wage Theories: A Partial Evaluation 33 were constrained to be the same across occupational groups. Industry wage effects were allowed to vary by occupation. This involves the estimation of an earnings function of the following form: (4) Wj = X.. + a. + where is log(hourly wage) of individual i in industry—occupation cell j, isa vector of individual and locational variables for individual i, is a vector of parameters, a is a fixed effect (or differential) for industry—occupation cell j, and a.. is an error term. This equation is equivalent to a wage equation with industry dummies, occupation dummies, and a full set of interaction terms between the industry and occupation dummies. The large number of industry—occupational cells implies that the feasible approach to estirnatin the industry differentials for each occupation is to first run a de—meaned regression in which the industry—occupation cell means are subtracted off for the dependent variable and all the independent variables: (5) = (X-X) + u. where W is the mean of the log of hourly earnings for workers in cell 3, is the the vector of the means of the individual and locational variables for workers in cell 3, and u3 is a regression error. This regression, assuming that the in equation (4) are uncorrelated with the yields a consistent estimate of . The mean residual for each cell 3 is then a consistent estimate of the industry—occupation 3 fixed effect: (6) j =ij Efficiency Wage Theories: A Partial Evaluation 34 4.1.4. Industry Characteristics and Industry Wage Patterns An understanding of the empirical relevance of alternative theories of wage de- termination requires knowledge of the industry characteristics associated with high wages amd with low wages after controlling for worker characteristics. A considerable amount of empirical research has focused on the relationship between wages and industry structure. These studies (partially surveyed by Long and Link (1983) and Kwoka (1983)) have focused on the influences of product market power, firm (or plant) size, and extent of unionization on wages. The industry charac- teristics that affect wage levels and the extent to which these variables matter appear to be quite sensitive to the specification (e.g. other control variables included) and to the particular sample analyzed (e.g. time period and use of only manufacturing vs. wider variety of industries). This suggests the effects are not uniform across industries and that multicollineari-ty is important for many industry variables. Nevertheless, some patterns emerge from these previous studies. Industry wage levels are strongly positively correlated with industry concen- tration when no labor quality variables are utilized (Weiss, 1966). The re- lationship is far more ambiguous when detailed labor quality controls are included. Pugel (1980) and Hodson and England (1985) find strong positive effects of industry profitability measures on average industry wages even with controls for average worker characteristics, extent of unionization, and other industry variables including the rate of employment growth. Dickens and Katz (1986) find that prpfits as a percentage of sales are strongly positively related to industry wage premiums for nonumion workers. Kwoka (1983), Long and Link (1983) and Mellow Efficiency Wage Theories: A Partial Evaluation 36 (1982) find a positive and significant effect of industry concentration on wages utilizing individual level data on earnings and worker characteristics combined with other industry level variables. This contrasts with Weiss's (1966) finding that concentration does not matter once individual worker controls are taken into account. Overall, industry wage differences appear to be related to product market power (ability—to—pay) although measurement problems in variables such as concentration and accounting profits mean these conclusions should be be viewed as somewhat tentative. Proportion of workers in an industry in large plants or average establishment size have typically been found to be positively related to industry wage levels even in the presence of detailed control variables (Kwoka, 1983; Long and Link, 1983; Pugel , 1980 and many others). Although establishment size and firm size appear to have quite important effects on wages within industries, they cannot explain much of inter—industry wage differentials. The May 1979 CPS contains a special survey including questions on establishment and firms size. Krueger and Summers (1986) find in analyzing this data set that the inclusion of plant size and firm size controls barely affects the estimates of industry wage differentials. They find the employment weighted standard deviation of two—digit industry log wage differentials falls only from .104 to .99 when plant and firm size controls are added to a log earnings equation with controls for occupation, region, union status and individual characteristics. The raw correlation of the estimated in- dustry differentials with and without employer size controls is .96. In re- gressions (not reported here but available upon request) utilizing the 1979 CPS, I find that estimated industry differentials are on slightly affected by the inclusion of plant size and firms size dummies when nonunion workers are analyzed in isolation. These results correspond to the conclusion of Brown and Medoff Efficiency Wage Theories: A Partial Evaluation 37 (1985) that most of the employer size effect on wages occurs within detailed in- dustries. Percentage of workers covered by collective bargaining in an industry has a strong positive effect on average industry wages.27 Dickens and Katz (1986) find that extent of industry unionization has a strong positive effect on both union and nonunion wages. They also find that regional union density has a strong positive effect on nonunion wages and a much weaker impact on union wages. Dickens (1986) argues that this is the pattern of union density impacts that arises from an im- portant role of a union threat in wage determination. The impacts of industry variables on wages remain a bit of puzzle. The findings of most studies are fairly consistent with some role for union threat effects since product market power and extent of unionization seem to explain a fair portion of inter—industry wage differentials for nonunion workers. Sociological models of the Akcerlof (1984) variety also seem to have some support. The findings of Hodson and England (1985) and Lawrence and Lawrence (1985) that capital in- tensity (capital to labor ratio) has a positive effect on industry wages provides some support for the shirking model since the cost of worker malfeasance is likely to be greater in capital intensive industries. Capital—skilled labor complementarity suggests that the capital—labor ratio finding may simply proxy for unmeasured labor quality. The strong linkages of wages to product market variables even after controlling for a large nunber of individual and locational variables appear difficult to reconcile with a strict unobserved ability inter- pretation of industry wage differentials. 4.1.5. Direct Evidence on the Benefits to Firms of High Wages Efficiency Wage Theories: A Partial Evaluation 38 workers. The long—term stability of industry wage differences through periods with substantial differences in the extent of union organization indicate only a more limited role for union threats and direct union impacts. Large wage dif- ferentials across industries for occupations with little threat of collective action suggests other factors must also be important. Industry differentials are strongly correlated across occupational groups. Sociological models in which industry wage contours or wage norms gain normative significance appear consistent with these similarities in the industry wage patterns across occupations. Eco- nomic efficiency wage rationales arising from monitoring, selection, and/or turnover problems can provide reasons why certain jobs in an industry require wage premiums. The concerns of firms with the perceived fairness of their internal wage structure may mean that these differentials come to permeate the entire wage structure in the industry. A combination of the economic efficiency wage models with the sociological (normative) efficiency wage models provides a fairly con- sistent though far from elegant account of the observed pattern of wage differ- entials. Much more empirical work is needed trying to uncover direct evidence of the gains to firms of high wage policies. Better measures of the factors that economic efficiency wage models indicate should be important for wage differences (e.g. monitoring costs) are required to better determine the relevance of these models for understanding the apparent large impact of industry and firm affil- iation on earnings. 4.2 Labor Market Discrimination28 Efficiency Wage Theories: A Partial Evaluation 41 Efficiency wage models provide several reasons for persistent discrimination by race and sex in a competitive labor market. Group differences unrelated to pro- ductivity can potentiaJly generate wage differences and occupational segregation. If two identifiable labor market groups differ in their rates of turnover or labor force withdrawal, the group with the higher turnover propensity will have a shorter horizon on a job and is likely to require greater inducement not to shirk. Bulow and Summers (1986) show that in the dual labor market version of the shirking model that the higher turnover group must be overrepresented in the secondary sector since if the wage is the same for both groups in the primary sector, the shorter horizon group is more likely to shirk unless the chances of being able to get a primary job in the future are lower than for the long horizon group. Since Poterba and Summers (1984) estimate a much higher rate of labor force withdrawal for women aged 25 to 59 than for men in the same age group, this yields a prediction of occupational and industrial segregation by sex with women tending to be found in lower paying jobs that are easier to monitor. Bulow and Summers (1986) demonstrate that if labor market discrimination arises from a re- sult of differences in separation probabilities by groups, that anti— discrimination policies such as affirmative action can raise welfare under a utilitarian welfare criterion. Johnson and Solon (1984) estimate that the earnings of both males and females are negatively related to the proportion of females in their occupation even when a wide variety of individual control variables are included. Additionally, Johnson and Solon that most of the difference in male and female earnings after control- ling for differences in individual characteristics is related to differences in the industrial distribution of employment by sex. If industry wage differentials Efficiency Wage Theories: A Partial Evaluation 42 arise form efficiency wage considerations, then this is strong evidence that these factors play a major role in differences in earnings by sex. Goldin (1986) presents evidence on widespread sex segregation across jobs within manufacturing requiring similar training and ability in 1890. She also finds that 50 percent of female operatives were paid by piece rates as opposed to 13 percent of male operatives. Monitoring costs are found to be cheaper for piece rates than for time rates in female dominated industries, while the opposite for male domi- nated industries. Goldin argues that deferred payment systems, such as discussed in Lazear (1979, 1981), conserved on monitoring costs and were feasible for males but not for females because of their shorter work horizons. Females were confined to jobs utilizing more expensive piece rates and thereby received lower wages than males. The same argument would apply to the use of efficiency wages rather than deferred payments for male dominated jobs if full bonding were not feasible. Goldin also presents evidence showing that the feminization of the clerical oc- cupations occurred only with standardization and division of tasks that made monitoring easier. - Occupational and industrial segregation arises in the shirking model because of the inability of high turnover groups to post performance bonds. This suggests that groups facing capital market imperfections are more likely to be affected. Furthermore, even if groups don't differ in turnover propensities, if dtsadvan— taged groups are liquidity constrained and unable to post bonds (accept deferred payment schemes), they will be less able to get primary sector jobs. Dickens and Lang (1985a,b) find that nonwhites are overrepresented relative to their observed characteristics in secondary sector jobs. Efficiency Wage Theories: A Partial Evaluation 43 Efficiency wage theories suggest that firms may find it profitable to pay workers wages above the market clearing level since such wage premiums can help reduce turnover, prevent worker malfeasance and collective action, attract higher qual- ity employees, and facilitate the elicitation of effort by creating feelings of equitable treatment among employees. Simple versions of efficiency wage models can explain involuntary unemployment, segmented labor markets with queues for primary sector jobs, and large differences among firms in the wages paid for what appear to be similar workers. Suitably modified these models can parsimoniously explain many of the major stylized facts concerning the behavior of labor markets over the business cycle. The primary criticism of efficiency wage models is that bonding mechanisms can solve effort elicitation, turnover, and adverse selection problems in an efficient manner. Additionally, the primary rationales for the use of efficiency wages also are all arguments for the emergence of long—term contracts and long—term employer—employee attachments. Such long—term relationships appear quite inipor— tant in the primary sector of the labor market where efficiency wage consider- ations are typically viewed as most important. These long—term attachments help facilitate bonding through the use of deferred payment mechanisms. It is an open empirical question whether seniority wage systems and pensions provide full bonding or whether they are only partial solutions leaving room for efficiency wages. Future theoretical work is required attempting to analyze efficiency wage problems in an explicitly contractual setting.32 The reasons why the contract market fails to clear need to be more fully explicated. The presence of more than one efficiency wage consideration may mean that bonding mechanisms solving one Efficiency Wage Theories: A Partial Evaluation 48 efficiency wage problem exacerbate others. For example, Shapiro and Stiglitz (1985) argue that employment fees to clear the market in presence of the shirking problem may exacerbate adverse selection problems. Complicated contracts re- quired to perform implicit bonding functions may create misunderstandings and lead to feelings of inequity which harm morale and productivity. Evidence on industry wage differences indicates that large differentials remain that are quite difficult to explain in terms of differences in labor quality or differences in important nonpecuniary aspects of work requiring compensating differentials. The persistence of industry wage premiums for long time periods implies that they are not just transitory differentials arising to facilitate the sectoral reallocation of labor in a dynamic market economy. Large, persistent wage differentials for similar workers and types of jobs provide strong evidence in favor of the importance of some type of efficiency wage behavior by many firms. The complex pattern of differentials is difficult to reconcile with individual variants of the efficiency wage argument. Further empirical research is required to isolate the primary benefits to firms of high wages, the determinants of the uses of alternative compensation systems (piece—rate vs. time rate etc.), and the industry characteristics associated with large wage premiums. Efficiency Wage Theories: A Partial Evaluation 49 FOOTNOTES 1. If efficiency wage considerations are equally important in all sectors of the economy, involuntary unemployment, in the sense that similar workers are treated differently with some being employed and others being unemployed and the unem- ployed preferring to be employed, can arise. If efficiency wage problems are not important in some sectors, jobs may always be available in these sectors. Jobs in the efficiency wage sector will still be rationed and offer a positive utility differential. Equivalent workers are treated differently even if there are always some (typically lousy) jobs available. Unemployment may occur from workers searching and waiting for the better rationed jobs. See Mookherjee (1984b) for an interesting discussion of alternative concepts of involuntary unemployment. 2. Freeman and Medoff (1981) and Lewis (1982) provide detailed surveys of empir- ical studies of union relative wage impacts 3. Foulkes (1980) tam high wages at Freeman and Medoff on unionization on presents numerous examples least partially as part of (1984, chapter 10) provide nonorganized labor in the of large nonunion firms which main— explicit union avoidance strategie. a detailed discussion of the effects United States. 4. See, for example, Akerlof and Yellen (1984), Rulow and Summers (1986), Jones (1985), Stiglitz (1984), and Yellen (1984). 5. Stiglitz (1984) compares and contrasts the inplicit contract and the efficiency wage theoretical literatures. Akerlof and Yellen (1984), Calvo (1979), and Yellen (1984) present excellent surveys of work on efficiency wage models. This paper differs from these earlier surveys in that it discusses new developments in the literature, analyzes the similarities Of efficiency wage and union threat effect models, and focuses more explicitly on empirical evidence that can help in de- termining the consistency of the predictions of efficiency wage models with actual labor market behavior. 6. Akerlof and Yellen (1984) provide economic interpretations of these conditions required on the e(w) function for a sensible solution to the firm's maximization problem 7. Solow (1979) shows that wage rigidi wage enters the production function in a run production function of the form Q = gidity with respect to these types of shocks 8. Lazear (1983) and Penacavel (1977) analyze the major issues arising in the choice of a piece—rate as opposed to a salary or time—rate compensation system. is generally not efficient since firms fail to ade— the impacts of their wage and monitoring levels on the utilize to prevent shirking by employees (Shapiro and Footnotes 50 ty of this type only arises when the real labor—augmenting way. A general short— F(w,L) need not generate real wage ri— 9. qua pol Sti The market equilibrium tely take into account icies other firms must glitz; 1984, 1985). the May 1979 CPS. Workers without tenure data and with earnings less than $1 an hour and greater than $250 an hour were deleted from the sample. The sample size for the less than one year of tenure group is 2770 and for the 10 years or more of tenure group it is 1912. 26. Dickens and Katz (1986) provide further details on the estimation technique and describe the procedure for correcting the correlation coefficients for sam- pling error. Since the industry differentials are estimated and since some of the industry—occupation cells are small, sampling error can be an important problem leading to an upward bias in the standard deviations of industry wage effects within an occupation and a downward bias in the correlations of industry premiums across occupations. The raw correlations (uncorrected for sampling er- ror) of the industry wage differentials across occupations lead to similar results with all of them quite positive. 27. Lewis (1983) provides a comprehensive survey of estimates of extent of unionization on wages in industry level studies. 28. This section is intended as a brief discussion of some of the implications of efficiency wage models for the analysis of labor market discrimination. I make no attempt to survey the vast empirical and theoretical literature on discrimi- nation in the labor market. See Cain (1984) for an excellent survey of the dis- crimination literature. 29. Okun (1973) discusses in detail empirical regularities observed in the cy- clical behavior of labor markets. Bernanke and Powell (1984) empirically analyze the differences in similarities in the cyclical behavior of industrial labor markets in the pre—war and post—war United States. 30. This follows directly from the structure of the model presented in Section II. Stiglitz (1984) and Akerlof and Yellen (1985a, 1985b) develop this point in detail. 31. Adjustment costs may be greater for output and employment than for prices and wages. The menu costs or "near—rationality" argument appears equally consistent with inertial output and employment policies with fluctuating wages and prices in response to small shocks. 32. Mookherjee (19Mb) and Bester (1985) provide some interesting initial attempts at linking efficiency wage and implicit contract theories. Stiglitz (1984) dis- cusses future directions for research in this area. Footnotes 53 REFERENCES Abraham, K.G. and H.S. Farber. "Job Duration, Seniority, and Earnings." MIT De- partment of Economics Working Paper No. 407, January 1986. Abraham, K.G. and L.F. Katz. "Cyclical Unemployment: Sectoral Shifts or Aggregate Disturbances ?" Journal of Political Economy; forthcoming 1986. Akerlof, G.A. "Labor Contracts as Partial Gift Exchange." Quarterly Journal of Economics 87 (November 1982): 543—69. . "Gift Exchange and Efficiency Wages: Four Views." American nomic Review 74 (May 1984): 79—83. Akerlof, G.A. and '(ellen, J. "Efficiency Wage Models of the Labor Market: Intro- duction." Mimeo, U.C. Berkeley, 1984. Near Rational Model of the Business Cycle, with Wage and Price Inertia." Quarterly Journal of Economics 100 (August 1985a). . 'Can Small Deviations from Rationality Make a Sig- nificant Difference to Economic Equilibria?" American Economic Review 75 (Sep- tember 1985b) Allen, S. "Trade Unions, Absenteeism and Exit—Voice." Industrial and Labor lations Review 37 (April 1984): 331—345. Altonji, J. and J. Ham. "The Collective Impact of Sectoral Shocks on Aggregate Employment Fluctuations." Mimeo, September 1985. Altonji, J. and R. Shakotko. "Do Wages Rise With Job Seniority?." NBER Working Paper #1616, May 1985. Antel , J.J. "Costly Labor Contract Negotiation and the Labor Mobility of Young Men. " American Economic Review 75 (December 1985): 976—991. Becker, G.S. and G.J. Stigler. "Law Enforcement, Malfeasance, and the Compensation of Enforcers." Journal of Legal Studies 3 (January 1974): 1—18. Bell, L. and R.B. Freeman. "Does a Flexible Industry Wage Structure Increase Em- ployment? The U.S. Experience." NBER Working Paper #1604, April 1985. Bernanke, B. and J. Powell. "The Cyclical Behavior of Industrial Labor Markets: A Comparison of Pre—War and Post—War Eras." Mimeo, February 1984. Bester, H. "Long—term Wage Contracts and Dual Labor Markets." Mimeo, University of Bonn, February 1985. Bhattacharya, S. "Tournaments and Incentives: Heterogeneity and Essentiality." Mimeo, March 1983. References 54 Bus, M. 3., "Real Wages over the Business Cycle: Evidence from Panel Data." Journal of Political Economy 93 (August 1985): 666—89. Blanchard, 0.J. And N. Kiyotaki. "Monopolistic Competition, Aggregate Demand Externalities, and the Real Effects of Nominal Money." NBER Working Paper #1770, December 1985. Bloch, F. and M. Kuskin. "Wage Determination in the Union and Nonunion Sectors." Industrial and Labor Relations Review 31 (January 1978): 183—92. Bowles, S."The Production Process in a Competitive Economy: Walrasian, Neo— Hobbesjan and Marxian Models." American Economic Review 75 (March 1985): 16—36. Brown, C. and 3. Medoff. "The Employer Size Wage Effect." Mimeo, November 1985. Brown, C. "Standard—Rate Wage Setting, Labor Quality, and Unions." NBER Working Paper #1717, October 1985. Bulow, J. I. and Summers, L. H. "A Theory of Dual Labor Markets with Application to Industrial Policy, Discrimination and Keynesian Unemployment." Journal of bor Economics, forthcoming 1986. Cain, C. "The Economic Analysis of Labor Market Discrimination: A Survey" In- stitute for Research on Poverty Special Report No. 37, 1984. Calvo, C. "Quasi—Walrasian Theories of Unemployment." American Economic Review (May 1979): 102—7. . "The Inefficiency of Unemployment: The Supervision Perspective." Quarterly Journal of Economics 100 (May 1985): 373—87. Calvo, C. and S. Wellisz. "Supervision, Loss of Control, and the Optimum Size of the Firm." Journal of Political Economy 86 (October 1978): 943—52. Carmichael, L. "Firm—specific Human Capital and Promotion Ladders." Bell Journal of Economics 14 (Spring 1983): 251—58. . "Can Unemployment Be Involuntary?: Comment." American Economic Review 75 (December 1985): 1213—14. Cullen, D. "The Interindustry Wage Structure, 1899—1950." AER 46 (June 1956): 353—69. Dickens, W.T. "Wages, Employment and the Threat of Collective Action by Workers." Mimeo, January 1986. Dickens, W.T. and L.F. Katz. "Industry Wage Patterns and Theories of Wage Deter- mination." Mimeo, U.C. Berkeley, March 1986. Dickens, W.T. , L.F. Katz, and K. Lang. "Are Efficiency Wages Efficient?." Mirneo, January 1986. Dickens, W.T. and K. Lang. "A Test of Dual Labor Market Theory." American nomic Review 75 (September 1985a): 792—805. References 55 W. "Emp 1 aye r References 58 Long, J.E. and Link, A.N. "The Impact of Market Structure on Wages, Fringe Bene- fits, and Turnover." Industrial and Labor Relations Review 36 (January 1983): 239—50 Malcolrnson, J. "Work Incentives, Hierarchy, and Internal Labor Markets." Journal of Political Economy 92 (June 1984): 486—507. Mankiw, N.G. "Small Menu Costs and Large Business Cycles: A Macroeconomic Model of Monopoly." Quarterly Journal of Economics 100 (May 1985). Medoff, J. and KG. Abraham. "Experience, Performance and Earnings." Quarterly Journal of Economics 95 (December 1980): 703—36. Mellow, Size and Wages." RESTAT 64 (August 1982): 495—501. Milkovich, 0. and J. Newman. çpjggpsation Plano, Texas: Business Publications Inc., 1984. Montgomery, E. and D. Stockton. "Evidence on Causes of the Rising Dispersion of Relative Wages." Mimeo, April 1985. Mookherjee, D. "Involuntary Unemployment and Worker Self—Selection." Research Paper #741, Graduate School of Business, Stanford University, April 1984a. . "Invol untary Unempl oyment and Worker Moral Hazard." Research Paper #745, Graduate School of Business, Stanford University, June 1984b. Murphy, K.M. and R. Topel . "Unemployment, Risk, and Earnings: Testing for Equal- izing Wage Differences in the Labor Market." Mimeo, UCLA, March 1986. Oi , W. 'Heterogeneous Firms and the Organization of Production." Economic Inquiry 21 (April 1983): 147—71. Okun, A. "Upward Mobility in a High—Pressure Economy." Brookings Papers on nomic Activity (1973:1): 207—52. Pencavel J. An Analysis of the Quit Rate in American Manufacturing Industry. Princeton, N.J.: IR Section Princeton, 1970. . "Work Effort, On—the—Job Screening and Alternative Methods of Remuneration." in R. Ehrenberg, ed. , Research in Labor Economics. Greenwich, Ct. JAI Press, 1977a. . "Industrial Morale." in 0. Ashenfel ter and W. Uates , eds. , Essays in Labor Market Analysis. New York: Halstead Press, 1977. Poterba, J. and L.H. Summers. "Adjusting the Gross Changes Data: Implications for Labor Market Dynamics." NBER Working Paper No. 1436, August 1984. Pugel, T. "Profitability, Concentration and the Interindustry Variation in Wages." RESTAT 62 (May 1980): 248—53. Reder, M. "Wage Differentials: Theory and Measurement" in Aspects of Labor nomics. Princeton, N.J.: Universities—NBER, Princeton U. Press, 1962. Reynolds, L. Labor Economics and Labor Relations, 7th edition. Englewood Cliffs, N.J.: Prentice—Hall, 1978. Rosen, S. "The Theory of Equalizing Differences.' NORC Discussion Paper No. 85—3,January 1985. Ruback, R. and M. Zimmerman. "Unionization and Profitability: Evidence from the Capital Market." Journal of Political Economy 92 (December 1984): 1134—57. Salop, S. "A Model of the Natural Rate of Unemployment." American Economic Review 69 (March 1979): 117—25. Shapiro, C. and J.E. Stiglitz. "Equilibrium Unemployment as a Worker Discipline Device." American Economic Review 74 (June 1984): 433—44. . "Can Unemployment Be Involuntary?: Reply." ican Economic Review 75 (December 1985): 1215—17. Slichter, S. "Notes on the Structure of Wages." RESTAT 32 (February 195D): 80-91. Solow, R. "Another Possible Source of Wage Stickiness," Journal of Macroeconomics 1 (1979): 79—82. . 'Insiders and Outsiders in Wage Determination." Scandanavian Journal of Economics 87, no.2 : 411—28. Stoft, S., "Cheat Threat Theory: An Explanation of Involuntary Unemployment," Mimeo, Boston University, May 1982. Stiglitz, J.E. "Wage Determination and Unemployment in L.D.C. 's: The Labor Turn- over Model." Quarterly Journal of Economics 88 (May 1974): 194—227. "Prices and Queues as Screening Devices in Competitive Markets." IMSS Technical Report No. 212, Stanford University, August 1976. "Theories of Wage Rigidity." NBER Working Paper #1442, September1984. "Equilibrium Wage Distributions." Economic Journal 95 (September 1985): 595—618. Strand, J. "Unemployment As a Discipline Device Under Business Fluctuations." Mimeo, University of Oslo, 1985. Taylor, J. "The Role of Expectations in the Choice of Monetary Policy." NBER Working Paper #1044, December 1982. Topel, R. and Laurence Weiss. "Sectoral Uncertainty and Unemployment." UCLA De- partment of Economics Working Paper #384, September 1985. Vroman, W. "Cyclical Earnings Changes of Low Wage Workers." in R. Ehrenberg, ed. Research in Labor Economics, vol. 2. Greenwich, Ct.: JAI Press, 1978. References 59 Weiss, Andrew, "Job Queues and Layoffs in Labor Markets with Flexible Wages." Journal of Politica' Economy 88 (June 1980): 526—38. Weiss, Leonard. "Concentration and Labor Earnings." American Economic Review 56 (March 1966): 96—117. Wing, R. "Achieving Internal Equity Through Job Measurement." in M. Rock, ed., Handbook of Wage and Salary Administration. New York: McGraw—Hill, 1984. Yellen, J. "Efficiency Wage Models of Unemp'oyment." American Economic Review Proceedings 74 (May 1984): 200—205. References 60