Do legal minimum wages create rents? A re-examination of the evidence.

I. Introduction

The effects of minimum wage laws have interested economists for many years. While much of the research regarding minimum wage laws has focused on its employment effects, several economists have been concerned with the laws' effects on the non-pecuniary aspects of jobs. There is disagreement in the literature as to whether minimum wage legislation creates rents in those jobs for which the law is a constraint. Several researchers have argued that firms will offset any rent created by the minimum wage by reducing the non-wage aspects of the job [4; 7; 14; 15]. In a recent paper Holzer, Katz, and Krueger [5] present evidence suggesting that employers do not completely offset the minimum wage. They use information on application rates for job openings found in the Employment Opportunities Pilot Project data set (EOPP) to test for the existence of rents on minimum wage jobs. Minimum wage jobs are found to have longer queues of applicants than other jobs, ceteris paribus, which suggests the existence of rents.

This paper presents additional tests of the existence of rents. Using the EOPP data, we find that workers who received starting wages equal to the minimum wage have a significantly higher probability of quitting than workers paid either less than or more than the minimum wage. This result seems to contradict the findings of Holzer, Katz, and Krueger regarding minimum wage jobs. As we see it, our findings support Wessels' [15] contention that since wages are more visible than non-pecuniary aspects of compensation, low wage workers mistakenly queue for minimum wage jobs, thinking them more attractive than they are, but then quit when reality sets in.

The paper is organized as follows. Section II discusses the existing theoretical and empirical research regarding minimum wages, section III presents our empirical procedures, section IV discusses the EOPP data set and in section V we discuss our results. Finally, in section VI we draw some tentative conclusions and discuss future research.

II. Literature Review

The textbook treatment of minimum wage legislation ignores the laws' effects on non-wage aspects of jobs. According to this analysis, when employers are forced to raise their wage offers due to minimum wage legislation, they simply move up along their demand curves for labor, reducing employment. Thus, while minimum wages create disemployment, those workers who remain employed in jobs covered by the minimum wage benefit. The theory has clear implications for labor market dynamics. Namely, those individuals covered by the law will be less likely to quit because the expected net return to search is reduced. Those not covered will queue for minimum wage jobs [9].

Others have suggested that employers will be able to offset the minimum wage by reducing the non-wage aspects of jobs. This hypothesis has been tested by Leighton and Mincer [7] and Hashimoto [4] who both found evidence of reduced on-the-job training in response to minimum wage legislation. Others have found that legal minimum wages may reduce fringe benefits such as paid vacation, sick leave, shift premiums and severance pay [14; 15; 1].

Holzer, Katz, and Krueger, question whether the firm fully offsets the minimum wage. They employ an efficiency wage model to demonstrate that the employer may not want to completely offset the minimum wage. Previous research, while finding evidence of at least some offset, has not been able to determine whether the offset is complete. To get at this issue Holzer, Katz, and Krueger consider application rates for jobs paying the minimum wage. They find that these jobs have a significantly larger number of applications than jobs paying either below the minimum wage or jobs paying above the minimum wage. They conclude that "overall, these results suggest that employers do not fully offset rents from the minimum wage by reducing fringe benefits or eroding working conditions" [5, 754]. Holzer, Katz, and Krueger note that researchers have used turnover rates to test for the existence of rents in high wage industries and large firms. They contend, however, that employee turnover is more appropriate for measuring ex post rents but that application rates will better measure ex ante rents.

These researchers conclude that minimum wage jobs offer economic rents because workers perceive rents, ex ante. Implicit in this conclusion is the assumption that the worker is fully informed about the non-wage aspects of the job prior to working since only in this case will the perceived ex ante rent accurately reflect a true rent. If, ex ante, the typical worker is misinformed about characteristics of the job, then a queue could form even if the job does not provide a rent. The literature on job search has long recognized that workers will not be fully informed about jobs until they are actually working on those jobs. This imperfect information results in "job-shopping" on the part of workers [6]. Wilde [16] for example, develops a model of equilibrium job-quitting in which jobs have both "general" characteristics - which are observed prior to starting a job - and "specific" characteristics - which cannot be observed until a worker is actually on the job. In this model, if the realized values of the specific characteristics are too low, workers will quit.

Of course, for large queues to form at minimum wage jobs even if the jobs offer no rents, there must be reasons why workers systematically over-estimate the non-pecuniary "specific" aspects these jobs. Wessels has offered such a reason by arguing that changes in the legal minimum wage may increase turnover in the short-run because workers will quit current jobs in order to pursue minimum wage jobs with apparently higher "full wages."(1) According to Wessels, workers are able to observe the offsets on their current jobs, but not on others. Hence, they believe those other jobs offer rents relative to their current jobs. This is a short-run theory since, presumably, workers will eventually learn of the offsets and develop more accurate predictions regarding the search characteristics of potential jobs [15, 70-71]. While Wessels does not state that a short-run queue of workers for jobs paying the minimum wage is a likely consequence of an increase in the minimum wage, such a finding is consistent with Wessels' theory. Of course, a true test of Wessels' theory would require knowledge of the applicant's wage at his/her most recent job.

The effects of minimum wages on quit rates have been examined previously. Mixon [10] tests the textbook theory by considering the effect of minimum wage legislation on industry quit rates. He finds a negative relationship between the level of the minimum wage and industry quit rates. His minimum wage variable is the ratio of the nominal minimum wage to the average wage in the industry. He regresses the quit rate in the industry on, inter alia, his minimum wage variable and finds that a one per cent change in the ratio will result in a change in the quit rate of "from just under one-half of a percent ... to almost 3 percent" depending on the industry.

Finally, if the firm responds to the minimum wage by reducing on-the-job training, and if at least a portion of the training is firm-specific training, then turnover will be higher since there is a smaller quasi-rent created [4; 7].

IIl. Empirical Procedure

We have argued that both Wessels and Holzer, Katz, and Krueger present theories that are consistent with applicants queuing for minimum wage jobs. The reasons for the queues, however, differ in the two cases. In order to discriminate between the two theories, we estimate quit propensity regressions for the workers in the EOPP. If Holzer, Katz, and Krueger are correct about minimum wage jobs offering rents, then the minimum wage should negatively affect the quit propensity, ceteris paribus. If Wessels is correct about rents being offset by reduced non-pecuniary aspects of minimum wage jobs, then, ceteris paribus, the minimum wage will increase the probability that the worker quits.

Our model of quit behavior is based on labor market search and human capital theories. We estimate a reduced-form quit equation similar to others in the literature [2; 12; 13]. An individual will quit his/her job if tie expected present discounted value of the current job is less than the expected present discounted value of the next best alternative, net of moving costs. The expected value of the best alternative is a function of the worker's alternative wage distribution and/or the value of non-market alternatives. Define:

V = present discounted value of the current job,

A = present discounted value of the next best alternative,

C = mobility costs,

e = a random error term, and

Q = 1 if the individual quits, zero otherwise.

Then:

Q = 1 if V - A + C < e. (1)

Assume that the value of the current job is a linear function of a vector of firm and job characteristics ([X.sub.v]), that the value of the best alternative is a linear function of a vector of the individual's general human capital and of labor market conditions ([X.sub.a]), and that mobility costs are a linear function of individual characteristics, and labor market conditions ([X.sub.c]):

  V = [v.sub.0] + [v.sub.1] [multiplied by] [X.sub.v]   (2a)
  A = [a.sub.0] + [a.sub.1] [multiplied by] [X.sub.a]   (2b)

C = [c.sub.0] + [c.sub.1] [multiplied by] [X.sub.c]. (2c)

Then

Q = 1 if ([v.sub.0] - [a.sub.0] + [c.sub.0] + [v.sub.1] [multiplied by] [X.sub.a] + [c.sub.1] [multiplied by] [X.sub.c] < e. (3)

If we assume that the error term is normally distributed, we have a probit regression model:

P{Q = 1} = [phi]([v.sub.0] - [a.sub.0] + [c.sub.0] + [v.sub.1] . [X.sub.v] - [a.sub.1] . [X.sub.a] + [c.sub.1] . [X.sub.c]) (4) where [phi](.) is the standard normal distribution function. In this model the minimum wage dummy is an element of [X.sub.v]. Under Holzer, Katz, and Krueger's interpretation it represents a rent accruing to the current job. Under Wessels' theory the dummy variable proxies a lower than expected level of non-pecuniary aspects of the job.

One might also argue that the minimum wage dummy should be considered an element of [X.sub.a], since holding a minimum wage job may influence a worker's alternative wage distribution. For example, if firms reduce rents by providing less on-the-job training in minimum wage jobs, then, to the extent this lost training would have entailed investments in general human capital, the value of the worker's alternatives would be reduced. On the other hand, firms may attempt to reduce rents by being more selective when hiring from their queues of applicants. A potential result of this is that prospective employers might interpret an applicant's holding a minimum wage job as a signal that the he/she is likely to be of higher than average productivity and this may enhance the applicant's chances of being hired into other minimum or above minimum wage jobs. Our reduced-form quit equation can accommodate either of these possibilities.

IV. Data

The data used in this study come from the second wave of the EOPP. The EOPP is a two-wave survey of firms conducted in 1980 and 1982. The 1982 survey directs the firm to consider "the last new employee your company hired prior to August 1981 " and asks several types of questions dealing with both "that person and the position he or she was hired to fill."

The dependent variable in our regressions, QUIT, is a dummy variable set equal to one if the employee is reported to have left the finn voluntarily. The starting wage for each employee is used to create MINIMUM WAGE, a dummy variable set equal to one if a worker's starting wage was equal to the minimum wage that prevailed at the time of hiring.

Firms also indicate whether, at the time of the hiring, the job was permanent, temporary or seasonal. To control for the type of job, we include PERMANENT, a dummy variable set equal to one if the job was reported to be permanent. Firm-wide data include the number of employees in all the firm's establishments within the geographic site defined in the EOPP. The log of this number, SIZE, is included as a control variable in our regressions, though we recognize that it is neither purely a measure of plant size nor of finn size. We control for local labor market characteristics by including the variable TIGHT, a dummy variable set equal to one if the employer indicates diat it is "somewhat difficult" or "very difficult" to find reliable unskilled workers at reasonable wages.

The major weakness of die EOPP is that some demographic variables for workers are missing. In particular, the data set has no information regarding race, marital status, or the size of the worker's family, if any. Nonetheless, it does allow us to control for the worker's age, education, sex, related experience, and tenure. RELATED EXPERIENCE is defined as the answer to the question: "How many months of experience in jobs that had some application to the position did N"E) have before (he/ she) started working for your company?" TENURE in months for those workers who are no longer with the firm is given in the EOPP. For those still employed by the firm, TENURE is calculated using the worker's starting date and the date of the survey.

Finally, to test for the possibility that MINIMUM WAGE is simply catching the effect of having a "low-paying" job, we include BELOW MINIMUM WAGE, a dummy variable equal to one if the starting wage was below the minimum wage. Note that there are 107 jobs paying below die minimum wage in the sample. As Table I shows, these are primarily with small employers in retail and service industries, which account for most of the uncovered portion of the labor market [11.]

Table 1. Characteristics of Below Minimum Wage Jobs (N = 107)
                            Percent in Industry     Mean SIZE

Construction                       1.87%               7.00
Manufacturing                      1.87               39.50
Retail                            57.01                8.72
Wholesale                          3.74               12.00
Service                           31.78               42.59
Finance, Real Estate,
 Insurance                         3.74               26.25

The firms surveyed were selected from across the country but were concentrated in the South and Midwest.1 The survey concentrated on relatively small and low-wage employers. Given the well known differences in pay and personnel policies between large and small employers, one should be circumspect in generalizing the results presented in this paper [3].

To make our results comparable to those of Holzer, Katz, and Krueger, we use the same criteria they employed to drop observations. Namely, we delete any observations in which the hire was prior to 1978 and for which there is not a complete set of the relevant variables. In addition, we keep only those observations in which die individual hired was between die ages of 16 and 65 and we drop any restaurant jobs in which the starting wage was less than die minimum wage, as well as any job which paid less than $1 to start.(3) Hires prior to 1978 are dropped since "retrospective information is likely to become less accurate with time" [5, 748]. Restaurant jobs are dropped because these workers receive an unknown portion of their pay in unreported paying no more than $5 per hour. We report results both with and without this last restriction. The restricted sample has 1374 observations while the unrestricted sample has 1920 observations. There were 278 workers in the sample earning the minimum wage when first employed. Table II presents descriptive statistics for the full and restricted samples.

[TABULAR DATA OMITTED]

V. Results

It is common practice to estimate separate quit equations for men and women since empirical evidence has often indicated that quit behavior differs by gender [2; 12; 8]. Likelihood ratio tests performed on both our full and restricted samples, however, fail to reject the null hypothesis of equal slope coefficients for men and women.(6) Thus, we report results in Table III for our full and restricted samples pooled by gender.

Our primary finding is that workers whose starting wages are equal to the minimum wage are more likely to quit. The coefficient for the labor market interaction term (MINIMUM WAGE x TIGHT) indicates that the minimum wage effect is enhanced in tight labor markets, although the interaction effect is statistically insignificant. These results hold for both the full data set and for the restricted sample and are stronger in the restricted sample. The BELOW MINIMUM WAGE

[TABULAR DATA OMITTED]

dummy variable has a statistically insignificant effect on the quit probability, indicating that it is the minimum wage per se, and not just low wages that cause higher quit rates.

We also find that TENURE has a significantly negative effect on the probability of quitting and that the tenure effect is greater for minimum wage jobs. In fact, the interaction effect quickly outweighs the independent MINIMUM WAGE effect. For example, an individual with one month of tenure in a minimum wage job is about 48 percent more likely to quit than the same person in a non-minimum wage job (P{QUIT = 1\MINIMUMWAGE = 1} = .455 vs. P{QUIT = I\MINIMUMWAGE = 0} = .308). The asymptotic t-statistic for the MINIMUM WAGE at one month is 2.21.(7) MINIMUM WAGE has a statistically insignificant effect for the same person with 1 year of tenure and by 2 years of tenure the person on the minimum wage job has a significantly lower probability of quitting.(8) These findings may be caused by heterogeneous preferences on the part of workers. Assume that a worker learns the nature of the non-wage aspects of the job soon after becoming employed. The subjective value of the lower-than-expected benefits will vary across workers. Those workers who are dissatisfied with the realized full wage of the job will quit relatively quickly. Those who place a low value on the non-pecuniary aspects of the job earn an economic rent due to the higher wage and thus are less likely to quit.(9)

A common interpretation of the negative relationship between tenure and quit probabilities relies on human capital theory. Those workers with longer tenure have more specific human capital and thus are less likely to quit. However, it is commonly argued that the firm will respond to minimum wage legislation by reducing on-the-job training [4; 7]. The negative coefficient on the tenure/minimum wage interaction variable contradicts this hypothesis. Wessels, however, is careful to point out that the firm can reduce the non-pecuniary aspects of the job in any of several ways [14, 304], and, therefore, not much should be made of the fact that our results are inconsistent with one particular possibility.

Overall these results are consistent with Wessels' theory of turnover. Workers appear to overestimate the full wage and to queue for minimum wage jobs, but once the true full wage is learned, they are disappointed and quit. The fact that this effect is somewhat larger in tight labor markets suggests that workers are more likely to respond to the disappointment when there is a high probability of finding alternative work.

VI. Conclusion

The estimates presented in this paper cast doubt on the conclusion that minimum wage jobs offer economic rents to all workers. They are consistent with Wessels' argument that firms will reduce the non-pecuniary aspects of jobs in the face of legal minimum wages and that workers may systematically overestimate the value of minimum wage jobs in the short run because they do not anticipate this reduction. We are not able to identify how the firm reduces the non-pecuniary aspects of the job to offset the minimum wage.

The EOPP data were collected at a time of frequent changes in the minimum wage. It is logical to assume that during periods of relative stability in the minimum wage, workers are likely to become better informed about the full wages on alternative jobs, and hence less likely to quit during the first several months of employment. Further research is needed to consider the law's effects on quit propensities during the 1980s when the minimum wage did not change. Since Wessels' theory is a short-run theory based on misinformation, we would expect to see the positive effect of minimum wages diminished during this time period. While it would also be informative to consider changes in application rates during the 1980s on minimum wage jobs, we are not aware of any data that would allow such a project.

References

[1.] Alpert, William. The Minimum Wage in the Restaurant Industry v. New York: Praeger Publishers, 1986. [2.] Blau, Francine D. and Lawrence M. Kahn, "Race and Sex Differences in Quits bv Young Workers." Industrial and Labor Relations Review, July 1981, 563-77. [3.] Brown, Charles and James Medoff, "The Employer Size Wage Effect." Journal of Political Economy. October 1989, 1027-59. [4.] Hashimoto, Masanori, "Minimum Wage Effects on Training on the Job." American Economic Review, December 1982, 1070-87. [5.] Holzer. Harry J., Lawrence F. Katz, and Alan B. Krueger, "Job Queues and Wages-" Quarterly Journal of Economics, August 1991, 739-68. [6.] Johnson, William R., "A Theory of Job Shopping." Quarterly Journal of Economics, May 1978, 261-77. [7.] Leighton, Linda and Jacob Mincer: "The Effects of Minimum Wages on Human Capital Formation," in The Economics of Legal Minimum Wages, edited by Simon Rottenberg. Washington: American Enterprise Institute, 1981. [8.] Meitzen, Mark, "Differences in Male and Female Job-quitting Behavior." Journal of Labor Economics, April 1986, 151-67. [9.] Mincer, Jacob, "Unemployment Effects of Minimum Wages." Journal of Political Economy August 1976, s87-s104. [10.] Mixon, J. Wilson, Jr., "The Minimum Wage and Voluntary Labor Mobility." Industrial and Labor Relations Review,, October 1978, 67-73. [11.] U.S. Department of Labor. Employment Standards Administration, Wage and Hour Division. The Fair Labor Standards Act of 1938, As Amended. WH Publication 1318, 1986. [12.] Viscusi, W. Kip, "Sex Differences in Worker Quitting." Review of Economics arid Statistics, August 1980, 388-98. [13.] Weiss, Andrew, "Determinants of Quit Behavior." Journal of labor Economics, July 1984, 371-87. [14.] Wessels, Walter J., "The Effect of Minimum Wages in the Presence of Fringe Benefits: An Expanded Model." Economic inquiry, April 1980, 293-313. [15.] _____ Minimum Wages, Fringe Benefits, and Working Conditions. Washington: American Enterprise Institute, 1980. [16.] Wilde, Louis L. "An Information-Theoretic Approach to Job Quits," in Studies in the Economics of Search, edited by S. A. Lippman and J. J. McCall. New York: North-Holland, 1979.

(*) The authors would like to thank Richard Gonce, Wayne Snyder and an anonymous referee for helpful comments and suggestions. The usual caveat applies. (1.) Firms compete for workers by offering a package consisting of a stream of wages and non-wage benefits - Wessels calls the value of this package to the worker the full wage" [14]. (2.) Survey sites included firms from Alabama, Colorado, Florida, Kentucky, Louisiana, Ohio, Texas, Washington, Wisconsin, and Virginia. (3.) Thus, the sample we construct consists of workers who started their jobs between 1978 and 1981. The minimum wage was increased in January in each of these years. It was raised to $2.65 from $2.30 in 1978, to $2.90 in 1979, to $3.10 in 1980 and to $3.35 in 1981. (4.) The period 1978-1981 was a time of rapid inflation. Nonetheless, Holzer, Katz, and Krueger used a nominal wage of $5.00 to restrict the sample. When we use a real wage of $5 to restrict the sample (with base year set at either 1978 or 1981), our results are qualitatively unchanged from those reported below. (5.) Henceforth, the sample restricted to jobs paying less than or equal to $5 will be referred to as the "restricted sample." (6.) x.sub.14.sup.2 = 16.418 for the full sample (p -value = .289) and x.sub.14.sup.2 = 19.202 for the restricted sample (p-value = .157). (7.) The reported probabilities and t-statistics are for the regression estimated on the full sample and are computed at the means of the independent variables. (8.) At 24 months of tenure, the probability of quitting is .00734 for minimum wage jobs and .063 for non-minimum wage jobs. The asymptotic t-statistic for MINIMUM WAGE is -2.18. (9.) The average tenure for those who quit non-minimum wage jobs is 9.51 months while the average tenure for those who quit minimum wage jobs is 6.58 months. The difference is significant (t = 3.22).