FRBSF Economic Letter
1996-29 | October 11, 1996
The Minimum Wage
- Evidence on minimum wage effects
- Criticisms of Card and Krueger
- Reconciling these results with economic theory>
- Likely impacts of the latest increase
August 20, 1996, President Clinton signed a bill passed by Congress that raises the federal hourly minimum wage from $4.25 to $5.15; this is scheduled to occur through increases of 50 in October 1996 and 40 in September 1997. Leading up to this increase has been a heated debate within the economics profession over the past several years regarding the employment and other economic effects of the minimum wage. This debate has focused largely on work by David Card and Alan Krueger of Princeton University (1995). Card and Krueger (henceforth CK) argue that economists’ traditional interpretation of the minimum wage–based on downward-sloping demand curves and consequent employment reductions–may be mistaken, and they cite quasi-experimental studies to support this and related controversial claims. Their ideas have generated substantial resistance from others in the profession. In this Economic Letter, I discuss the empirical and theoretical issues surrounding this debate, and also the likely effects of the new minimum wage.
One of the basic tenets of economics is that, with very limited exceptions, demand curves slope downwards–i.e., a decrease in the price of a commodity leads to an increase in the total quantity that market participants wish to purchase. For many years, there has been no clear reason to disbelieve the applicability of this “law” to the low-wage market, where the “commodity” (unskilled workers) is presumed to be relatively undifferentiated and the corresponding product market is highly competitive. Consistent with this view, much of the early evidence on the employment effects of minimum wage increases produced the expected negative effect on employment (albeit a small one). However, this evidence was based primarily on time-series regressions of aggregate U.S. employment outcomes on federal minimum wage changes; given problematic features of these data, it is surprising that any consistent effect was isolated.
In their book, CK summarize a variety of studies that update this work using alternative methodologies and that fail to confirm the conventional wisdom on the subject. CK base their arguments on two key sets of studies. One set uses published government figures and public use data sources such as the Current Population Survey (the federal government’s monthly demographic and employment survey) and focus on individuals for whom the minimum wage is most important (teenagers and low-wage adults). The other set uses specialized surveys of fast-food restaurants in New Jersey and Pennsylvania in one case, and Texas in another.
A key difficulty in devising a “quasi-experiment” to assess the effect of the minimum wage on employment is finding a control group that is relatively unaffected by the minimum wage, for which employment growth provides a suitable baseline for comparison. The CK studies do this in several different ways. For example, they compare employment outcomes for low-wage workers in states that increased their minimum wage (such as New Jersey in 1992 and California in 1988) to outcomes for comparable workers in other states that did not change their minimum wage. Alternatively, CK identify the effect of the 1990-1991 nationwide increase in the minimum wage by comparing its employment impact across states that vary in their share of affected workers, or by comparing the effects of the minimum wage across high-wage and low-wage establishments (which are differentially affected by the minimum wage increase). CK make these comparisons over a short period (1-3 years in general).
Using a variety of such experiments, CK find that the minimum wage increased wages earned by affected groups, which provides partial validation of their quasi-experimental design. However, they find no evidence that minimum wages reduce employment, and they even find evidence suggesting employment increases. In addition to evidence for the U.S. states, they cite similar results for Puerto Rico–a locale where the minimum wage is binding for a large portion of the labor force–and several foreign countries. In further analysis of U.S. data, CK also find essentially no effects of minimum wage increases on firms’ stock prices, and they find that minimum wages slightly reduce inequality and poverty in family earnings.
Critics have pointed to a number of potential problems with CK’s data and quasi-experimental design. For example, various researchers have expressed reservations about the comparison group being used, whether the estimated effects are biased by employer anticipation of the minimum wage change prior to the “experimental” period, and also by the possibility that the long-run employment impact may be larger than the short-run employment impact. The latter two criticisms–anticipation of the law’s impact, and a larger long-run effect–apply with equal or greater force to CK’s finding that stock prices of low-wage firms were little affected by news of the 1990-91 increase. Also, the survey data collected by CK have some anomalous features that are highly suggestive of measurement error, including large employment changes for some establishments. Such measurement error is likely to bias estimated effects of the minimum wage toward zero, and in some extreme cases may impart a positive bias to the estimated effects (particularly given the small sample sizes).
Furthermore, other recent work reports results that directly conflict with CK. For example, Neumark and Wascher (1992) exploited variation across states in the timing of state minimum wage changes and the degree to which the state minimum exceeded the federal minimum, for the period 1973-1989. By comparing the employment impacts of such changes over time, Neumark and Wascher (henceforth NW) implemented a test of minimum wage effects that is similar in design to the CK estimates based on state-specific changes in minimum wages, but for a longer time period and a larger number of states than in any of the CK tests. NW conclude that there are small negative employment effects of the minimum wage, consistent with the previous generations of time-series evidence. Among other recent work, Deere, et al. (1995), find particularly large disemployment effects of the 1991-91 increases in the federal minimum wage.
Of course, such results also are open to criticism. For example, the NW results hinge critically on whether school enrollment is accounted for in the estimation. In fact, NW find results consistent with CK when the school enrollment variable is included. Because this variable is related to teenage employment and minimum wages in a complicated fashion, deciding how to treat it in the estimation is a challenging problem. Also, although NW’s analysis of a newly collected Pennsylvania/New Jersey data set produced results that contradict CK’s findings, the NW data and findings have not yet been subject to widespread scrutiny, due to pre-publication restrictions on the use of these data. Among examples of other studies that conflict with CK, the Deere, et al., results may be affected by the time period studied, which encompasses the 1990-91 recession. Because data from other recessions are not included in their sample, their results may be biased by failure to account for business cycle effects on employment probabilities that differ across low-wage and high-wage workers.
Many economists do not believe the CK results because they fly so directly in the face of conventional theories of low-wage labor market operations. On the other hand, some economists are more willing to accept the assumptions needed to produce the CK results. The class of models that produce this result are “monopsony” models. In the standard competitive setting, individual firms can hire additional qualified low-wage workers at a constant wage rate which is equal to the wage rate paid to similar workers at other firms. In contrast, a firm has monopsony power in the low-wage labor market if, as CK put it, it must “pay a higher wage in order to maintain and motivate a larger [low-wage] work force.” In economic parlance, monopsony implies that labor supply curves to individual firms slope upwards, rather than being completely flat–i.e., firms are price-setters, rather than price-takers, in the labor market.
Many economists regard this model as a curiosity that applied historically to “company towns,” in which a single firm faced the entire local market supply curve of labor. Such economists feel that the model certainly does not characterize the constraints facing the small retail and other establishments that dominate low-wage labor markets today. However, the “dynamic monopsony” model described by CK embodies monopsony power in the relationship between the wage and turnover rates. In particular, CK note that as long as either the quit rate or the hiring/application rate varies with the wage, a positive wage/employment relationship exists for individual firms, at least in the short run. This reasoning may be particularly applicable to low-wage labor markets, in which high rates of employee turnover are a particular concern. Of course, the assumption that turnover rates respond to wages in ways that individual firms can control cannot be accepted a priori, and the empirical evidence that CK cite on this point is weak. They cite statistically imprecise results from an earlier study and results using their specialized data sets that do not confirm the precise predictions of dynamic monopsony models.
Several existing models with a more complex structure, in which monopsony power arises from informational imperfections in labor markets, also predict ambiguous employment effects of minimum wages. Some of these models account for the complexities of the job search and matching process for firms and workers, which produce equilibrium wage variation (akin to monopsony power) across firms. Other models account for the possibility that worker effort varies with the wage, and that firms’ costs of monitoring their workers decrease as wages rise but increase as employment rises. At this point, however, empirical testing of such models is limited at best, and their applicability to estimated minimum wage effects on employment and wages is speculative.
Given the theoretical ambiguities and the unavailability of high quality experimental data, disagreements over the effects of minimum wages on the level of employment are unlikely to be resolved soon. However, economists (including CK) can at least agree that such increases raise wages for low-wage workers who remain employed, and that the wage effects “spill over” to workers above and below the new minimum wage. This raises concern about the impact of the forthcoming minimum wage increase on wage and ultimately price inflation.
The estimated effect of the higher minimum wage on product prices depends on observable factors, such as the number of minimum wage workers and labor’s share of output, and also on the size of the “spillover effect.” Using the best available estimate of the law’s coverage (90 percent of the work force), data from the 1995 Current Population Surveys reveal that the $5.15 minimum wage will directly raise the hourly wage of approximately 9 million workers currently earning between $4.25 and $5.15 per hour, out of 110 million employed (excluding the self-employed). Assuming that the wages of those currently earning less than $4.25 are raised to preserve their relative distance from the minimum and that wages of those currently in the $5.15-$6.00 range also increase somewhat, this implies an increase in average economy-wide wage earnings of 0.40 percent. Assuming that labor’s share of output and costs is 0.70, the wage increase implies a price increase over the implementation period of 0.28 percentage points. Although this should lead to an approximately equivalent increase in inflation over the implementation period, the longer-run effect on inflation depends on whether the price increase is incorporated into workers’ and firms’ price expectations.
As for the employment effects of the new minimum wage, it is likely that they will be sufficiently small as to be difficult or impossible to observe. First, although the impending minimum wage increase is large (21 percent), in historical terms the new minimum wage will not be high relative to average wages in the economy. Second, despite disagreement over the exact employment effects of the minimum wage, evidence suggests that this effect always has been relatively small in the aggregate and that it may have declined over time. Whether the economy stays strong or weakens, any disemployment effects arising from the minimum wage will be difficult to disentangle from the prevailing employment trends. Ironically, the impact of the minimum wage increase may not be clear enough to help us measure the effects of the decision.
Card, David, and Alan B. Krueger. 1995. Myth and Measurement: The New Economics of the Minimum Wage. Princeton, NJ: Princeton University Press.
Deere, Donald, Kevin M. Murphy, and Finis Welch. 1995. “Employment and the 1990-91 Minimum-Wage Hike.” American Economic Review 85 (2, May), pp. 232-237.
Neumark, David, and William Wascher. 1992. “Employment Effects of Minimum and Subminimum Wages: Panel Data on State Minimum Wage Laws.” Industrial and Labor Relations Review 46 (October), pp. 55-81.
Opinions expressed in FRBSF Economic Letter do not necessarily reflect the views of the management of the Federal Reserve Bank of San Francisco or of the Board of Governors of the Federal Reserve System. This publication is edited by Sam Zuckerman and Anita Todd. Permission to reprint must be obtained in writing.
Please send editorial comments and requests for reprint permission to
Attn: Research publications, MS 1140
Federal Reserve Bank of San Francisco
P.O. Box 7702
San Francisco, CA 94120