The minimum wage pushback
Some valid worries, some less-valid worries
A couple weeks ago I wrote a long post about why a federal $15 minimum wage would be a pretty safe policy. Since then, there has been a vigorous pushback from minimum wage skeptics, including several of my own colleagues! These objections deserve to be taken seriously instead of shouted down. So let’s go through the objections one by one. But first, I want to make a general point about costs, benefits, and risks of policies.
Costs, benefits, and risks
Most of the discussion about the minimum wage centers on the potential costs of the policy — especially whether it will cause unemployment. This seems like a huge mistake. In general, every policy has costs and benefits, and to ignore the benefits introduces bias to the discussion.
The whole point of minimum wage is to reduce poverty. Many papers find that it does do this. For example, a 2019 paper by Arindrajit Dube finds that doubling the minimum wage would result in somewhere between a 22% and a 45% drop in poverty. That would mean a decrease of somewhere between 2.6 and 5.4 percentage points, since the overall poverty rate is around 12%. A 2018 paper by Kevin Rinz and John Voorheis of the Census Bureau finds that minimum wage hikes substantially increase income at the bottom of the distribution over a 5-year period. These effects, of course, take into account any employment loss that may or may not be happening — in other words, these papers find that minimum wages reduce poverty overall, not just for those who have jobs.
This is not to say that “minimum wage reduces poverty” is settled science. Undoubtedly there are some papers out there that disagree. But by refusing to even talk about the poverty reduction aspect of minimum wage, we’re doing the debate a disservice.
Note the huge difference from the discussion on trade. Advocates of free trade will admit that a minority of Americans will lose out from the policy. But they argue that because the benefits of free trade outweigh the cost overall, we should implement free trade and “compensate the losers”.
Now, minimum wage is not quite like that, because unless you think minimum wage increases productivity (which is not out of the question), the benefit to the poor comes out of the pockets of the rich and the middle class, in terms of higher prices and reduced profits. So to “compensate the losers” from minimum wage, if indeed there are any losers, would require taxing the rich even more in order to pay welfare benefits or provide government employment to those few who lose their jobs. But…OK. Sounds good to me, why not do that?
Of course, there’s also inherent risk in a policy change — we know what failing to raise the wage will look like, but we have only educated guesses as to what a world of $15 minimum wage will look like. In general, policymakers should be risk-averse. That weighs against making dramatic moves. Advocates of tax cuts, free trade, deregulation, and other laissez-faire policies likewise were likewise urging us to take risks in the hope of improving things — and those gambles didn’t always pan out.
But if government were so risk-averse that it never made any policy changes, policy would never improve. Sometimes, as Deng Xiaoping put it, you have to “cross the river by feeling the stones” — make a series of calculated policy gambles using your best guess, and be ready to change course if necessary.
And that means taking both costs AND benefits into account.
The strongest argument against the federal minimum wage is that it’s a one-size-fits-all policy that fails to take regional variations into account. Some states and cities have low costs and low wages; others have high costs and high wages. A minimum wage that is too high for rural Mississippi is too low for New York City.
And in fact, this argument is correct! The optimal minimum wage policy would be a local living wage policy that sets minimum wage to be a certain percent of median wage. Remember that existing evidence shows that minimum wage tends to have small employment effects up to about 59% of local median wage.
So it makes sense to worry about places for which costs and wages are both low, and so $15 is pretty high. In a recent Bloomberg post, my colleague Michael Strain writes:
According to data from the Bureau of Labor Statistics for 2019 (before the pandemic), in 47 states, at least one-quarter of all workers earn less than $15 per hour. In 20 states, half of all workers earn less than $18 per hour, and in 30 states, the median hourly wage is less than $19.
And in a blog post follow-up, my colleague Tyler Cowen adds:
No matter what you think about the recent literature on the minimum wage, all economic theories imply that minimum wages should be decided at the state and local level, given the economic heterogeneity of the United States. That is the message that you as an economist should be carrying forward.
Do you think Puerto Rico should be a state? Should they have a $15 minimum wage too? Come on. Yes, it is easy enough to make an exception for them, and by the way the median manufacturing wage in Mississippi is below $15 as well. Rinse and repeat.
Now, remember that because places with lower costs and lower wages also tend to have fewer employers, they tend to have more monopsony power. Remember that the more monopsony power a labor market has, the less dangerous a minimum wage is to employment. So a federal minimum wage is less dangerous to places like Mississippi than you might think, just from looking at local wage levels. BUT, “less dangerous” doesn’t mean “not dangerous at all”.
So why don’t we just set minimum wage at 60% of local median wage in every county in the country? In a word: Politics. “Fight for $15” makes a good slogan, and it’s a simple number that’s easy to understand. “Fight for 60% of local median wage” does not make a good slogan, and is not easy to understand.
Lots of policy discussions end up coming down to arguments about what is and isn’t politically possible, and this is no exception. There is strong popular support for a $15 federal minimum wage, and it looks like a thing that might actually get done. So I’m kind of just assuming it’s $15 federal minimum wage or nuthin’. Remember to never let the perfect be the enemy of the good!
In fact, the Democrats’ proposal already contains something along these lines — it would index future minimum growth to local median wages. And the proposal already includes a long phase-in — the minimum wage wouldn’t reach $15 til 2025. And Seattle already implemented differential phase-ins for different kinds of businesses, when it raised its minimum wage to $15; small businesses were allowed to take up to 7 years to phase it in.
So based on these existing plans and policies, it seems to me there’s a good possibility that we could retain the rhetoric of “Fight for $15” while not demanding that Mississippi or Puerto Rico immediately implement the full $15 level. In my opinion, people who are worried about regional variation should fight for that modification, not for the scrapping of the policy as a whole!
(Oh, and as for the danger of too-high minimum wage in recessions, you could allow a special exemption during recessions, but I think a wage subsidy automatic stabilizer would actually be far better for this purpose.)
What does the evidence really say?
(Warning: This section gets nerdy.)
It’s fine to talk about monopsony power and all, but ultimately the case that the minimum wage is a pretty-safe policy rests on the weight of empirical evidence. And that evidence has been powerful enough to change economists’ minds over a fairly short space of time. Remember that in 2013, 34% of top economists surveyed by the University of Chicago agreed that a $9 federal minimum wage would “make it noticeably harder for low-skilled workers to find employment”. Just two years, later, in 2015, the same survey found that only 26% of economists agreed that a $15 federal minimum wage would mean that “the employment rate for low-wage US workers will be substantially lower than it would be under the status quo.”
The actual evidence didn’t change in two years; it had been building up for a long time. But economists’ awareness of the evidence probably did change. In 2014, the Upjohn Institute published a book by Dale Belman and Paul Wolfson called What Does the Minimum Wage Do?, summarizing the evidence thus far. In 2015 they put out a paper reviewing the recent literature. They found that since the turn of the century, newer analyses using better data and more sophisticated analysis had increasingly found little or no effect on employment from minimum wage hikes. Their data is where this graph comes from:
A 2019 review by Dube found the same. Here is a chart from that review, showing the various effects:
You can see that although most estimates aren’t statistically significantly different from 0, the estimates do tend to be negative — minimum wage probably has had a slight negative effect on employment for low-wage workers!
Now, you might think publication bias could be biasing these numbers toward 0, since “minimum wage doesn’t actually kill jobs!” might be an exciting result. But a 2009 analysis by Doucouliagos and Stanley found that while there is a spike at 0, the distribution of results also skews toward the negative side. If the variation in these results were random, you’d expect to see a symmetrical distribution around the true value, even if the true value was negative.
They apply a correction for that asymmetry, and find that it greatly reduces the overall estimate of the negative impact of minimum wage on employment. In other words, people are probably throwing away results that show a positive employment effect of minimum wage, and the true effect on employment is probably even smaller than what Dube, Belman and Wolfson report. Maybe around -0.05 or so.
OK, so far so good. But economist David Neumark of the University of California, Irvine has done two literature reviews — one in 2007 with William Wascher and the other in 2021 with Peter Shirley — that find substantial job-killing effects of minimum wage! The most recent one has been shared widely around the internet in the last few days, as it seems to come just in the nick of time for critics of the proposed $15 hike.
So let’s focus on that one. Neumark and Shirley argue that existing literature reviews will tend to include a bunch of results that aren’t studies’ “preferred” results. Instead, Neumark and Shirley want to focus on the main or most interesting results from each paper. So basically they do two things:
They read each paper and decide what they think is the most important result for each.
They write to the authors of the papers asking them what they think their papers’ most important number is.
They then collect these “preferred” results in a graph, and it looks like this:
The estimates obviously lean toward the negative side.
OK, so. First of all, I see three problems in this methodology, right off the bat.
Problem 1: Subjectivity. Neumark and Shirley are obviously going to be biased in their own choices of estimates, as most researchers probably would be. For the papers whose authors didn’t write back — who you’d think might tend to be the ones who aren’t friends with Neumark and Shirley — they use their own judgment in determining what the most important result is.
Problem 2: Selection. Neumark and Shirley miss some big studies, including the one by Rinz and Voorheis (who, since they are the Census Bureau economists, obviously have the best data), as well as some others. How did they pick which ones to include and which to exclude?
Problem 3: Significance bias. If I know empirical economists, I know they’re going to want to choose the estimates with the three magic asterisks of statistical significance. But if so, this is bad statistical practice. You’re not supposed to do multiple comparisons, and then select just one that comes out significant! If you do this, your statistical significance immediately becomes meaningless.
OK, so this paper obviously has lots of issues. But what’s interesting to note is that its result probably isn’t that different from Dube’s!
Neumark and Shirley’s median “preferred” estimate is -0.116. Dube’s graph above notes that the median estimate for “any group” is -0.17. In other words, the studies Dube reviews tend to find some subgroup of workers — teenage dropouts, or whatever — who experience pretty significant job losses from minimum wage hikes.
In other words, if Neumark and Shirley are simply selecting the most negative result — or even just the most statistically significant result — from each paper, they’re probably doing something similar to Dube’s “any group” calculation. And both methodologies find numbers in a similar range. (Neumark and Shirley’s estimates tend to be more statistically significant than the ones in Dube’s graph, but as mentioned above, this is meaningless because of Neumark and Shirley’s methodology.)
In other words, these researchers may not actually be disagreeing very much. It may be true that a high minimum wage is very bad for certain small groups of workers, while not very harmful for low-wage workers overall.
Which brings us back to costs and benefits. Suppose we can lower the poverty rate in America by just 1 percentage point with a $15 minimum wage. That’s literally millions of people brought out of poverty! And income boosts for millions more who need it the most. It’s that benefit that we must weigh against the possibility of concentrated job losses among teenagers or other narrowly defined groups of workers.
Hours worked, RCTs, automation, etc.
There have been a few other criticisms of the $15 minimum wage, and I’ll go through some of these briefly.
Some people have noted that studies find more of an impact on hours worked by low-wage workers than on employment. But that’s fine. Remember that the Rinz and Voorheis study and other studies find that minimum wage makes incomes for low-income people go up — that includes any reduction in hours. Remember that leisure time has value — it lets people spend time with their families, or just enjoy life. So if hours go down and incomes go up, that’s a win.
Of course that’s just an average, and a few workers might experience hours reductions so severe that they actually lose income. But remember, all policies have “winners” and “losers”, and we have other policies that can help those who fall through the cracks.
Some researchers did an RCT in an online marketplace (similar to Mechanical Turk) where some employers were forced to pay higher wages. They ended up hiring less labor. But note that this is a very competitive marketplace with very low switching costs and many employers, where you’d expect minimum wage to have disemployment effects. Also you wouldn’t expect general equilibrium effects to be in play here; in the real world, people spend their minimum wage raises and thus raise total local labor demand, but that’s not going to happen in this case. So while this RCT is nice to confirm our intuition that labor demand curves slope down, it doesn’t have much relevance to the real-world policy debate.
Some people warn that if we raise the minimum wage, it will force companies to replace human workers with machines. All I can say is, if our only hope for maintaining full employment was to cut our wages to below-subsistence levels, then it was only going to be a few years before the machines replaced us anyway, and it’s time for UBI.
Minimum wage is an effective policy for bringing lots of people out of poverty. It probably has negative effects for some small groups of workers like teenagers, but we have other policies to help these people. Meanwhile, we can use keyhole solutions to deal with issues like local variation in costs and wages. A federal $15 minimum wage isn’t the optimal perfect policy by any means, but on balance it seems like a good thing to do, and the massive public enthusiasm means that it might be the only really substantial poverty-fighting policy we can get through Congress right now.
If it were me, I’d say go for it.