Could This Paper Be a Paradigm Shift for Minimum Wage Thinking?
According to a new working paper by Texas A&M economists Jonathan Meer and Jeremy West, raising the minimum wage may have little or no effect on the level of employment, but it does hurt growth in employment for years after the increase goes into effect. In a recent column in the Washington Post, Robert Samuelson put it this way: “In the short run, even sizable increases in mandated wages may have moderate effects on employment, because businesses won’t abandon their investments in existing operations. But companies that think themselves condemned to losses or meager profits won’t expand.”
The new paper will, I am sure, be cited by opponents of current proposals to increase the federal and various state minimum wages. But, there are important reasons to be skeptical about the findings. (Apologies in advance for the long post!)
Two Red Flags
The first red flag is that the negative effect of the minimum wage estimated by Meer and West applies to the entire workforce (not just workers affected by the minimum wage) and appears to lie well outside the range of almost all earlier research on the minimum wage. Meer and West find that “a real minimum wage increase of 10 percent reduces job growth in the state by around 0.53 percentage points (during these years, the average state employment growth rate was 2.0 percent annually).” (p. 16) So, according to Meer and West, a 10 percent increase in the minimum wage would reduce the growth rate in total state employment by about 25 percent, roughly from an average of 2 percent in the absence of the minimum wage to 1.5 percent after a 10 percent increase. This is a very large effect. Economists typically treat policies that might raise or lower economic growth by 0.1 or 0.2 percentage points as a very big deal.
What is especially surprising is that their estimated effect does not refer to the employment growth for groups that typically earn low wages, such as teenagers and restaurant workers, who have been the main focus of almost all minimum-wage research. These estimated effects apply to the entire workforce, not just the 5 percent or so of workers directly affected by the typical increase in the minimum wage. (The last seven increases in the federal minimum wage, for example, directly affected between 1 and 6 percent of workers, and, indirectly, perhaps a few percentage points more. See Table 1 in this document.)
Even researchers who are most critical of the minimum wage typically find that a 10-percent increase in the mandated wage floor would reduce current employment among teens (or, in some analyses, fast-food restaurant workers, or in a few studies, adults with less than a high school degree) by 1 to 2 percent. (For a summary of employment effects written from a point of view that is generally critical of the minimum wage, see David Neumark and William Wascher’s 2008 book Minimum Wages. My reading of the evidence is that employment effects of moderate increases hover near zero.) It is hard to see how a policy that has that small an impact on directly affected workers could have such a large effect on the overall population.
But, it is difficult to compare the size of the effect that Meer and West estimate, which is defined in terms of future growth rates, with the much more standard measures of employment effects, which are defined relative to pre-minimum wage employment levels. In fact, in the most recent version of their working paper, Meer and West go to some lengths to explain that what appear to be large employment effects are smaller than they look. Meer and West emphasize that the 25-percent decline is the estimated impact of a permanent increase in a state’s minimum wage. In practice, they argue, minimum-wage increases are eroded by inflation (and whenever other nearby states increase their own minimum wage, reducing the differences in minimum-wage levels across states).
Even so, when Meer and West do the arithmetic to try to convert their finding into the more standard way of representing the employment impact of the minimum wage, they conclude that “each 10 percent increase in a state’s real minimum wage, relative to its regional neighbors, causes a 1.2 percent reduction in total employment relative to the counterfactual by the end of five years.” If I am reading that correctly, this suggests that a 10 percent increase in the minimum wage reduces total employment (after five years) about 1.2 percent relative to where it would have been in the absence of the minimum-wage increase.
But, as Meer and West note, “not all workers are affected by increases in the minimum wage.” The estimated effects on the total workforce are likely to fall overwhelmingly on workers at or near the minimum wage. Since this group typically makes up less than 10 percent of the workforce, these estimates of the total effect imply an impact that is roughly 10 times higher on the groups that will actually be affected by the minimum-wage increase. If for the workforce as a whole, a 10-percent increase would reduce employment by 1.2 percent, this suggests that for teens, say, a 10-percent increase would reduce employment by about 12 percent. These implied estimates for affected groups are six to 12 times higher than the range that critics of the minimum wage identify as the likely impact of the policy for teens (and, I repeat, I think even that range is too high).
The second red flag is that Meer and West’s key result — that minimum wages reduce employment growth — does not hold for teenagers or fast-food restaurant employees, the two groups that researchers overwhelmingly use to analyze the impact of the minimum wage.
In a new working paper, Sylvia Allegretto, Arindrajit Dube, Michael Reich, and Ben Zipperer (ADRZ) use two different data sets to test Meer and West’s conjecture. One data set covers teen employment; the other, fast-food workers. In both cases, following procedures that are similar in spirit to Meer and West, they find “no evidence that minimum wages reduce employment growth.” For both groups, the estimated effects on employment growth are “close to zero” (p. 28) ADRZ conclude that these results “from low-wage groups actually affected by the policy suggests that the negative association found by Meer and West is unlikely to be causal.” ADRZ believe that the Meer and West results may reflect “their lack of controls for the overall state of the labor market.” (p. 28)
Another working paper, by Dube, William Lester, and Reich, looks at exactly the kinds of employment dynamics that interest Meer and West (separations and hires in the wake of a minimum-wage increase), but focuses specifically on teens and fast-food workers. DLR “find striking evidence that separations, hires, and turnover rates for teens and restaurant workers fall substantially following a minimum wage increase” — but with no significant net effect on employment. (p. 2)
Given that Meer and West find implausibly large effects of the minimum wage on the total workforce and given that applying their insights to workers that are most affected by the minimum wage (teens and fast-food workers) reveals no impact of the minimum wage on growth rates for these low-wage workers, what most likely is happening is that something other than the minimum wage (but correlated with it) is moving overall state employment.
Meer and West’s specifications appear to be suffering from omitted-variable bias. Some important factor (or factors) is missing from their model. That factor — not the minimum wage — is actually driving their results. The failure to incorporate that factor into the economic modeling leads the model to give the false impression that it is the minimum wage that is influencing state employment growth.
The very large effects on total employment and the absence of an effect on low-wage workers, however, suggest strongly that whatever is going on, it isn’t the minimum wage.
As I mentioned at the outset, the Meer and West paper will likely be cited widely by opponents of the minimum wage. But, many of those who deploy the paper to suggest that the minimum wage reduces job growth (by reducing future rates of job creation) will feel uncomfortable with many of the paper’s other findings and the economic theory used to justify the results.
On the empirical side, Meer and West actually corroborate the findings of many other researchers going back at least as far as David Card and Alan Krueger “of no measured effect of the minimum wage on the level of employment” (Meer and West, pp. 17-18; see also their Table 2, row 2). Meer and West brush aside this finding by noting that “this is unsurprising in light of our [earlier] discussion” (p. 18) that the true effects of the minimum wage are on the future growth of employment, not the current level of employment. But, if their theoretical arguments are correct, Meer and West are also dismissing all earlier research on the minimum wage that looks for effects on the level of employment, including all of the work traditionally cited by opponents of the minimum wage.
If true, the clear implication of Meer and West’s paper is that we should not find the impact of the minimum wage on the level of employment, only on the growth rate. To the extent that earlier researchers did find negative employment effects, these results are, in Meer and West’s framework, spurious. Meer and West cannot negate the findings that the minimum wage does not reduce the level of employment without simultaneously negating the earlier findings that the minimum wage did reduce employment levels.
With respect to economic theory, Meer and West are also likely to cause traditional opponents of the minimum wage some discomfort. In academic circles the skirmishes around the minimum wage are as much about the best way to model the low-wage labor market as they are about the exact employment impact of the minimum wage. Until Meer and West, opponents of the minimum wage have staunchly defended some version of the standard “competitive” model. In the competitive model, employers and workers all have good information about the labor market, including the number and location of vacancies, the going pay rate, each worker’s productivity levels in every possible job, and any other variable needed to participate effectively in the labor market. Together, these and a few related assumptions ensure that employers pay workers exactly what they’re worth to the production process — no more and no less. In this framework, a minimum wage that pushes the wage above this market level must reduce employment. With minor variations, this is the model that has motivated academic opponents of the minimum wage and it is — almost without exception — the model of the labor market that nonacademic opponents have in mind when they oppose the minimum wage.
Meanwhile, several alternative models emphasize that the low-wage labor market diverges from the competitive ideal in a host of important ways. These deviations from the assumptions of the competitive model mean that the unequivocal prediction that a binding minimum wage lowers employment need not always hold. This alternative view is, in fact, the starting point of Meer and West’s analysis, which highlights these complexities and acknowledges the implication that the employment effects of a minimum wage are an open empirical question:
“In this class of models, the minimum wage has opposing effects on job creation. Although it reduces demand for labor by raising the marginal cost of employing a new worker, a higher minimum wage increases the gap between the expected returns to employment relative to unemployment, inducing additional search effort from unemployed workers…. The theory thus has ambiguous predictions for the effect of a minimum wage on job creation.” (p. 6)
Meer and West’s paper includes multiple references to assumptions and observations that constitute serious violations of the competitive model. For example, their focus on future rates of job creation, rather than immediate instances of job destruction — the core idea in the paper — is motivated by concerns around “non-trivial fixed costs associated with hiring a new employee” and “[p]sychological factors” including “the negative feelings managers have when terminating employees, sometimes called ‘firing aversion.’” The standard competitive model ignores transaction costs. The standard competitive model assumes that any employer who allows psychological factors to affect their judgment would quickly go out of business. Yet, Meer and West build their theoretical and empirical analysis around these (and other) assumptions that violate the standard model.
For Meer and West, economic theory is ambiguous on the employment impact of the minimum wage. This is exactly the position articulated early on by Card and Krueger (and many other researchers, then and since), a position that was widely ridiculed as bad economics by conservatives in both academic and non-academic settings. That serious, young researchers such as Meer and West take these theoretical ambiguities seriously — and use them to motivate their own work — is a sign that the paradigm may be shifting.
Originally written for the blog of The Center for Economic and Policy Research, which was established in 1999 to promote democratic debate on the most important economic and social issues that affect people’s lives. CEPR is committed to presenting issues in an accurate and understandable manner, so that the public is better prepared to choose among various policy options. Follow CEPR on Twitter @ceprdc.