Let’s look at bit more at the CBO’s job loss estimate of 500,000 from their minimum wage report out last week (I’m focusing largely on their analysis of the increase to $10.10, then indexed to inflation).
My point last week was that even if the budget office is right about this employment loss guesstimate, we’re still talking about a policy that lifts the pay of 24.5 million low-wage workers (16.5 million directly, 8 million indirectly), meaning 98% of the target audience reap the benefit of the policy.
This post is about whether their employment loss estimate is defensible. To telegraph the result: yes, but it’s from the upper end of the range, and that’s an unfortunate choice given the political intensity around this issue. Just a few years ago, the budget office was careful not to weigh in on this question, pointing out that the range of estimates was wide. If anything, the range has widened since then, making their current choice even more puzzling.
Recall that the CBO did not do original research on the impact of higher minimum wages on employment. They just went into their “demand elasticity supply closet” and took a number that many of us who follow this work thought was from a pretty high shelf.
That number was -0.1 in the case of teenage workers for the $10.10 increase, implying a 10% increase in the minimum wage would lead to a 1% loss of jobs held by teenagers. The CBO suggests—though it’s not clear how they landed on this number—that they took the midpoint between a range of estimates.
Two problems here. First, I’m pretty sure that’s not the midpoint. Summarizing a vast literature (more on that exercise in a moment), the CBO says that the midpoint lies between zero and -.15, or -0.075 (there are credible estimates of positive impacts of minimum wage increases on jobs, but let’s stick with their range).
The CBO did choose this lower midpoint for their simulations of a $9 minimum wage, but kicked it up to -0.1 for the $10.10 version. They offer a number of rationales for this, and while they’re not unreasonable (e.g., unlike the $9.00 simulated policy, the $10.10 increase is indexed to inflation), they’re really just hunches and given the sensitivity of the estimate, why apply hunches? And why do all their hunches just go in one direction (toward larger job losses)?
Second, as economist Michael Reich points out in a recent piece on the current state of research on this question of the minimum wage and job loss, if the budget office had weighted the various estimates by the quality of the studies, they would have given a heavier weight to those which find no significant impact on employment. Reich discusses the methodological advances that the CBO discounted, generating results like those summarized here in Table 2 from testimony by Arin Dube, a co-author of Reich’s on various careful analyses of this question.
Had they done so, they might well have landed on a smaller negative, like -0.05, which would have cut their projected job loss by half, to 250,000. In fact, the most accurate assessment would have been something like a range from zero to 500K, with an acknowledgement that zero is a plausible outcome supported by high-quality work (this was the White House’s position in their critique of the CBO report), as is a negative number in the low hundreds of thousands.
In fact, one part of this that I found puzzling was the fact that in a 2007 CBO analysis of a minimum wage increase, the budget office said this about job losses:
For this tabulation, CBO assumed that no changes in employment or hours would have resulted from the higher minimum wage rate .
And here’s the footnote:
 The economics literature includes numerous studies on the employment effects of increases in the minimum wage, which indicate a wide range of potential impacts.
So, here’s the question: since 2007, have econometricians learned that much more now about this question that we can now make an accurate assessment when we couldn’t do so, at least in CBOs judgment, back then?
I don’t see it. If anything, the research since 2007 has widened the range—e.g., it now includes zero (and remember, we’re talking just about teenagers here; even the early research wasn’t sure if older workers’ employment was affected at all by moderate increases).
I suspect the core of the problem is that most economists simply haven’t digested the newer work. Though CBO cites it, they don’t give it heavier weight, as Reich argues they should.
I predict that this will change in the next few years in no small part due to a book about to be released by Dale Belman and Paul Wolfson: The New Minimum Wage Research, an extremely detailed analysis and review of the best work over the past few decades. I’m just looking at a pre-publication manuscript, and man, if you go for that sort of thing, it’s a real page-turner.
A key chapter provides a meta-analysis of the many studies they review, wherein they summarize hundreds of estimates, carefully accounting for quality, without, as far as I could tell, an ideological thumb on the scale. In fact, about two-thirds of the employment elasticities in the chapter are negative, and a fair number are statistically significant. But after what has to be the most exhaustive existing analysis of the question of the impact of minimum wage increases on employment, here’s what Belman and Wolfson conclude:
Bearing in mind that the estimates for the United States reflect a historic experience of moderate increases in the minimum wage, it appears that if negative effects on employment are present, they are too small to be statistically detectable. Such effects would be too modest to have meaningful consequences in the dynamically changing labor markets of the United States.
This begs the question as to why moderate increases in the price of low-wage labor do not consistently generate “statistically detectable” effects, and I’ll get into that some other time (Reich does a bit in the link above and I’ve done so here as well). The point for now is that CBO clearly highballed the job loss estimate. They’re not “wrong” but in an organization that strives for balance, this is not their most balanced work. And given the importance of the issue and the divisiveness of the politics, that’s just very unfortunate.
UPDATES: First, Paul Wolfson himself reminds me that an earlier meta-analysis by Stanley and Doucouliagos of employment elasticities from minimum wage research lands in a similar place: “with 64 studies containing approximately 1,500 estimates, we have reason to believe that if there is some adverse employment effect from minimum-wage raises, it must be of a small and policy irrelevant magnitude.”
Also, I stumbled on this excellent discussion of many of the above points by John Schmitt, one of the premiere synthesizers of minimum wage research.
Next: why did the CBO deflate the minimum wage with the PCE deflator? (I know, you can’t hardly wait!)