What Pro-Minimum Wage Economists Say

I recently came across this excellent document by Arindrajit Dube. I was curious about what pro-minimum wage economists actually think. How do they sell a policy that seems so at odds with a straightforward application of economic theory? The document is Dube's written testimony to the U.S. Senate Committee On Health, Education, Labor, and Pensions. For me, it serves a dual purpose. It is 1) a brief literature review on the empirical effects of the minimum wage and 2) the arguments that pro-minimum wage economists present to politicians. Number 2) is important because presumably this is what politicians are referring to when they say "minimum wages don't cost any jobs" or that it puts money into the hands of poor people. When politicians say such things, I like to know where they are getting it from so I can somehow respond. (It's hard to respond to a claim that's made up on the spot, pulled out of thin air.) In the future, I'll refer to this document as the presumed justification for such claims.

[For example, Illinois recently passed a minimum wage bill, and state representative Will Guzzardi said in justification: “Raising the minimum wage has no net effect on employment, it doesn’t drive jobs out of the state...All it does is put money in people’s pockets who need it." We will see that this is not a fair representation of the research, and it's not even what pro-minimum wage economists say.]

There's no substitute for reading the Dube document in full, but you should know that there are meta-studies that reach the opposite conclusion. I recommend the book Minimum Wages by Neumark and Wascher. Here is a long paper by the same pair of authors, if you don't want to shell out for the actual book. And here is a summary of their conclusions in a final chapter of Minimum Wages:

This is a contentious field. There seems to always be someone reading the same set of papers and coming to the opposite conclusion. Read the Dube document and the Neumark/Wascher meta-analyses with some skepticism.

What follows are my reactions to the Dube testimony.

Section I.A: Rising Inequality

The first section past the introduction starts with: "The minimum wage has failed to keep pace with productivity, while top pay and corporate profitability have grown rapidly." This is a statement about how inflation has lowered the real minimum wage; keeping the real minimum wage high would require indexing it to inflation. It strikes me as incredibly loaded language. The word "failed" implies that there is some natural law or some inherent justice in keeping the real minimum wage fixed. Dube attempts to argue this further along in the document, but to open with this line is to admit to the author's bias. And the comparison to top pay and corporate profitability is irrelevant. The appropriate prevailing wage in the low-skilled labor market shouldn't necessarily move in tandem with rising pay of top-executives. The former could be relatively flat, while the latter could be increasing because of increasing firm size or the increasing importance of top executives.

Dube admits as much a couple pages down (page 5):

It is quite remarkable that had the minimum wage kept up with overall productivity, it would have been $22 per hour in 2011. Had it kept up with the growth in income going to the top 1 percent, it would have been even higher, at $24 per hour; and the wage would have exceeded $33/hour at its peak in 2007.

This evidence does not suggest that the minimum wage should be increased to $22 or $24 per hour.

Emphasis mine. The first paragraph is unfortunate, because it suggests that the minimum wage should be raised very high. Many readers will read this without the important qualifiers in the next paragraph. Politicians will use the first part as intellectual firepower to implement a higher minimum wage without necessarily disclosing the second part.

There is some discussion of the minimum wage "decreasing inequality." This seems quite implausible given the small number of people for whom the minimum wage is actually binding. The Bureau of Labor Statistics reports that it's only about 2.3% of workers. Notably, this is down from the previous year, and down from 13.4% in 1979; this jives with Figure 3 in Dube's testimony, which shows the real minimum wage declining over this period. Presumably, the real minimum wage becomes less and less binding as it decreases. A minimum wage that is binding on more workers means wages should be higher for those workers than otherwise, but it also means that those workers will work fewer hours (assuming they keep their jobs), and the loss of hours could completely offset the wage increase (as happened in Seattle recently). It's not clear a priori what effect this would have on the income distribution. Dube cites some literature that attempts to measure effects on income inequality. There is lower tail inequality, measuring the ratio of the 50th percentile to the 10th, and upper-tail inequality, measuring the ratio of the 90th percentile to the 50th. Dube is citing literature that shows that the 50/10 inequality "improves" with a higher minimum wage. He concludes this section by saying,

To sum up, while there is some scholarly disagreement about the exact magnitudes of the impact of minimum wages on inequality, we know that the decline in the real minimum has played an important role in increasing inequality in the bottom half of the wage distribution, especially for women.

Um, no. This is an open question, and Dube is overstating the consensus. See the table above from Minimum Wages, particularly the part about the income distribution. Neumark and Wascher reach the opposite conclusion. Apparently many studies in this vein fail a falsification exercise:

That is, if theory predicts an effect of x (the minimum wage) on y (lower-tail inequality), and such evidence is found, researchers often also explore whether there is an effect of x on another variable, z (upper-tail inequality), which is conceptually related to y but for which theory does not predict an effect on z.

In other words,

"My statistical model says that minimum wages decrease inequality!"
"Yeah. You're statistical model says all kinds of crazy things. Shall we take them all together? Or shall we pick and choose?" 

In fairness to Dube, Minimum Wages was published in 2008 and Dube cites a study from 2011 in that section. I am always very conscious about dates, because it seems that most of the compelling papers were written in very recent years.

Section I.B: Minimum Wages Have Not Kept Up with Cost of Living

The next section is about the minimum wage and the cost of living.

Over the last three decades, the minimum wage has failed to keep up with cost of living.

This is a whopping non sequitur. The minimum wage is a very poorly targeted anti-poverty program. Most minimum wage earners are not the fabled "single parent with a full-time minimum wage job raising a family." These are mostly secondary or tertiary earners in a household, and most of them are working part time. I think Dube is trying to build on the earlier section, where he thinks he's demonstrated that minimum wages help poor people (by decreasing inequality). Insofar as he hasn't actually done so convincingly, nothing from Section B of the document actually follows. He ends this section saying, of indexing the minimum wage to inflation:

This is why even some economists who are skeptical about minimum wage policies nonetheless support indexation.

Notice that he says "some" and not "most". In fact, I think most economists who are skeptical about or hostile to the minimum wage think that inflation is an excellent natural sunset clause for a bad policy. It means that if we made a mistake, we don't get stuck with it forever. In fact, I remember Walter Block describing his disappointment with Gary Becker's moderation, because Block wanted to abolish the minimum wage outright while Becker wanted to see inflation gradually kill it off. I think many business owners, while doing long-term planning, recon that they can "ride out" a minimum wage increase because inflation will gradually take the bit out of it after a few years. Indexing to inflation locks a bad policy in place. It also takes away a source of variation by which to judge the impact of the minimum wage. If the real minimum wage were the same in all times and all places (as a federally imposed minimum wage indexed to inflation would do), then there is no "control group" by which to judge the policy. If we're going to have social planners impose economic policy on us, let's at least have an escape hatch and local experimentation so we can know which policies are harmful and quantify the harm.

Section I.C: Minimum Wages Have Fallen Behind Median Wages

This next section is interesting, and I think it catches the pro-minimum wage economists in a contradiction.

When analyzing the strength of minimum wage policies, economists typically use the ratio of the minimum to the median wage, also known as the Kaitz index. There are three reasons to pay attention to this measure. First, a comparison of the minimum wage to the median offers us a guide to how binding a particular minimum wage increase is likely to be, and what type of wage the labor market can bear.

Emphasis mine. What does it mean that the "labor market can bear" a minimum wage? Presumably Dube thinks that the labor market cannot bear a minimum wage that is too high, because he implicitly knows that demand curves slope downwards. Raise the price, and people buy less.  He should be more clear to his readers that this effect is real. He just doesn't know at what wage level it starts to be "unbearable." Also presumably, though Dube might argue with this, those same economic forces are present at lower levels of the minimum wage, they are just weaker and harder to measure. It's easier for employers to make adjustments other than actually laying off workers. I think Dube should admit that there are giant question-marks about the actual effects of the minimum wage, and that some of the harm is not obviously captured in government statistics.

Second, a comparison also provides us with a natural benchmark for judging how high or low a minimum wage is across time periods or across countries that vary in terms of their labor markets and wage distributions.

Like I said, wouldn't indexing to inflation destroy this variation and make it hard to form a comparison group?

Third, the median wage also provides a natural reference group for judging how reasonable a minimum wage level is: most people would not think fairness concerns dictate that the minimum wage should be set equal to the median wage, but they may find it objectionable if it is much lower (say a fourth or a fifth as large). 

What do people's feelings of "fairness" have to do with anything? Is this an economist's summary of the literature, a "positive analysis", or are we pandering to vague feelings of fairness, a "normative analysis"? The document is a mix of both kinds. I think people have very bad intuitions about "fairness". Economics often points out that popular policies have terrible negative consequences. It might seem "unfair" to let landlords increase rents to whatever they like, but the alternative (rent control) leads to horrific decay of the housing stock. Oh yeah, and "unfair" behavior by landlords, like discrimination and failure to do basic maintenance. It's the job of an economist to fight this kind of populism. Maybe I'm being unfair here and Dube is simply acknowledging that this is a reality; there is a realpolitik constraint on how low the minimum wage might be set. But given the thrust of the document, it seems like he's willing to indulge the public's bad intuition.

A natural target is to set the minimum wage to half of the median wage.

This. Let's have the politicians and activists calling for $15/hour read this sentence, over and over again. Dube, who is perhaps the most esteemed pro-minimum wage economist, is telling us that there is a limit to how high we can set it. He's not saying "The empirical literature shows that the minimum wage has no effect whatsoever, no matter how high we raise it," which is how some politicians and even some economists have summarized the literature. Dube is picking a finite value, which is relatively low compared to what some agitators are asking for, because he realizes that Econ 101 will start to bite at a higher wage. No state has a median wage that would justify a $15/hour minimum wage by Dube's criteria, and in fact some states have median wages less than $15/hour.

What follows in that section is a comparison of the United States to the rest of OECD, with us being near the low end.  I don't care for this kind of argument, because it implies that the average of a bunch of other countries' policies must be right. Maybe they're all victims of the same kind of voter populism that gives us so much bad economic policy, and we're the only one's actually getting it sort of right. The international consensus is often wrong.

Section II of the document is titled "How are Increases in the Minimum Wage Absorbed?"

Section II.A: Employment Effects

This section starts with this very subjective summary of the literature:

For the range of minimum wage increases we have seen in the U.S. over the past two decades, recent evidence based on credible methodologies do not find job losses of any sizable magnitude.

 Again, read the Neumark and Wascher meta-analyses. They come to a very different conclusion. Don't let Dube convince you he's speaking for the entire profession here. This is still very much an open question. And new groundbreaking papers written since Dube's testimony (given in 2013) have found large disemployment effects.

Studies comparing similar neighboring areas right across the border account for these problems and find no impact on jobs either for sectors like restaurant and retail or groups like teens.

Dube obviously has a fetish for these cross-borders comparisons. I've seen it in many of his other public statements and critiques of various papers. I think the best evidence we have here is from Seattle, and this favors the use of synthetic controls rather than cross-border comparisons.  (My write-up here, and the working paper here.) Dube has this graphic in his testimony, which suggests using border counties as natural treatment-control groups for comparison:

The first paper by Jardim et. al. from the Seattle studies has this figure:

Jardim et. al. argue that using the surrounding region as the control group (represented by the two panels on the left) fails a falsification test. Look at the lower left graph. If this is a proper control group, it sure looks like the minimum wage is "causing" the increases in earnings above, say, $27/hour, where no reasonable person would expect to find an effect. The panels on the right are the synthetic controls, which don't show any obvious effect of the minimum wage at these much higher wages, where you wouldn't expect any effect. (Interestingly, the Economic Policy Institute takes Jardim et. al. to task saying that their synthetic controls actually do not pass the falsification test. If you pool the effects at higher wages instead of looking at them one-by-one, it becomes "statistically significant." I wrote about that here. Dube and the EPI folks should get in a room and decide whether they like falsification exercises or not.)

Maybe Seattle is a special unique snowflake, but my take-away from this (contra Dube) is that the synthetic controls are the better methodology. Dube's written testimony is from 2013, so obviously he didn't have access to the papers by Jardim et. al. which were only published in the past couple of years. If he's revised his opinion in light of this new information I haven't seen statements by him to that effect (but good on him if he has!). (There is a footnote at the bottom of page 13 of Dube's testimony that I think neatly summarizes the academic disagreement over synthetic vs. cross-border controls. Read that if you're curious about the details.)

This was an interesting passage:

Over the past two decades, the variation in minimum wages has been highly regionally selective: the states that have seen greater increases in the minimum wage— typically in the northeast and the west—have tended to be those with lower underlying growth in demand for low-wage workers.

Here is my take on this. Politicians intuit that a minimum wage has real disemployment effects, so they only implement an increase if they think the economy can handle it. Unscientific and self-confirming maybe, but it's hard to deny that something like this happens. To take an extreme example, the first minimum wages were only a few cents an hour; nobody was asking for $15/hour. Everyone rightly intuited that there was such a thing as "too high"; and everyone waited until it became feasible (not to say sensible) to start asking for $15. If this political will is idiosyncratic enough, not necessarily correlating with measurable economic variables, that would mean that studies of minimum wages in places that implement them understate what the effects would be in places that decline to implement them.

There is a section about the changing attitudes an opinions of the profession. Of surveys of economists asking whether minimum wages led to unemployment, Dube says,

Surveys of AEA members in 2000 found 46% agreeing with a similar proposition, while surveys concluded in 1992 and 1978 revealed 79% and 90% of economists agreeing with similar statements.

He makes much use of an IGM Forum that found no clear majority on either side. (See Don Boudreaux's post about it here, particularly the part at the bottom about one economist's answer with the comment that perhaps the Fed would accommodate. (In the IGM Forum, economists can leave comments explaining their answer, and these are sometimes revealing.) This is amusing. It's kind of admitting that the minimum wage has real disemployment effects, and that the Fed would bring about some inflation to accommodate it if this effect was taking too much of a bite. This basically assumes that we're in Econ 101 world, and furthermore that other economists all mutually understand that we are in Econ 101 world, and that this understanding is deep enough that we can fiddle with various policy levers to undue a minimum wage increase that is too large. If we all share these Econ 101 intuitions, let's say so.)

Dube tries to make the point that "weighted by confidence" the IGM forum comes out much more strongly in favor of minimum wage increases. I actually take it the other way around. Economists who are appropriately humble about their ability to engineer society are naturally more willing to admit uncertainty. Interventionists will have a bias toward overconfidence. This is evidence of hubris, not consensus.

The third of the panel that expected job losses were split on their support for the policy, while the third that were sure that there would not be job losses were unanimous in their support.

Again, this is more a sign of overconfidence and crassly politicized thinking than of academic consensus.

Section II.B: Turnover and Job Flows

This section discusses how job turnover falls when the minimum wage increases. I don't think there's much disagreement on the fact of this point, but there is some disagreement about whether it's a good thing or not. Pro-minimum wage economists tend to emphasize the cost savings of low turnover, a very real effect for sure. If you don't have to keep re-training new staff, that saves your business some money and headache. But perhaps the lower turnover is actually a bad sign. Low-wage workers know that jobs have become more scarce, so they desperately cling to any job they can get. This isn't necessarily a good thing. There is much wage mobility, particularly for low-wage workers who have a lot more space to move upward than downward. If these workers are clinging too long to a job rather than taking other opportunities, it could be a sign of a sclerotic labor market. Besides, employers know about their turnover costs. If the savings are truly valuable, they would pay higher wages without being forced to do so. This is a common theme in pro-minimum wage arguments. Some "benefit" is proposed without any explanation as to why the business owners aren't already trying to capture it. By all means, let's count this in our ledger of total costs and benefits to society, but let's not pretend we're doing business owners a favor by forcing them to do something that they could do without our "help".

Models with search frictions in the labor market— which have become increasingly popular—can help explain this pattern of small effect on employment coupled with larger effect on turnover. Of course this cannot be true at all levels of the minimum wage—with a sufficiently large increase, employment levels will most likely fall as well.

Emphasis mine. Once again, Dube is acknowledging that the disemployment effects are very real. It's just that we don't know where they start to kick in. I wish Dube would place more emphasis on this uncertainty. The "benefits" of minimum wage are highly questionable and poorly targeted, and the costs, which even he acknowledges must exist, are highly uncertain.

Finally, there are other channels through which minimum wages may positively impact employment. A higher minimum wage can spur those who are unemployed to search more intensely for jobs, as the value of a job rises.

 This passage was very jarring for me. In my economics education, this effect is counted as a cost, while Dube seems to be counting it as a benefit (as implied by the language "positively impact employment.") See the diagram from this post.

This is a standard treatment of the economics of minimum wages, as it was first presented to me by Steven Landsburg's excellent Price Theory text. The intense searching by minimum wage workers (represented by the rectangle in pink) is a cost, not a benefit. If employees are investing in more intense signalling (credentials and other resume boosters that have little to do with actual productivity), or if they're competing to curry favor with employers in hopes of obtaining a now-scarce job, this activity represents a cost imposed on the worker that does not get recouped by society as a whole. Bizarrely, Dube is treating this as some kind of offsetting benefit.

Dube goes on:

Generally speaking, workers’ bargaining power may be insufficiently low for the purposes of efficiency. By increasing workers’ pay, a minimum wage policy can improve the functioning of the low wage labor market.

Emphasis mine. Bargaining power? This seems backwards. Increasing the minimum wage decreases their bargaining power. For most low-skilled workers, the only bargaining chip they have is "Okay. If you won't hire me at that price, you can pay me less. Deal?" Once they've bumped up against a minimum wage, there's nowhere else to go. Granted, there are times when tying your own hands improves your bargaining position. Stephen Pinker gives the example in his book How the Mind Works of two people playing the game of chicken, in which one of them locks and removes the steering column, waving it for the other player to see. "Sorry, I can't possibly budge." The other player has to swerve. It is far less clear if this kind of hand-tying benefits workers. As even Dube acknowledges, there is a wage that's high enough that employers no longer wish to pay it. ("Oops, turns out I'd already locked my own steering column. Guess you'll have to unlock yours in a hurry.")

Dube, Freeman and Reich (2010) finds that replacement costs are around 8% of annual salaries, and are sizable even for blue collar and service workers. Reduced turnover can, therefore, increase productivity through reducing recruitment and training expenses.
These additional channels of adjustment can help explain why moderate increases in minimum wage seem to have small employment effects.

Once again, companies know about turnover costs. Presumably they're already setting wages to optimize the trade-off between turnover costs and payroll. (It reminds me of this post by Alex Tabarrok on efficiency wages.) This entire section strikes me as an attempt to recast costs as benefits. Carelessly and clumsily describing these topics in plain English allows for this kind of casuistry.  A little bit of Econ 101 clarifies what is happening, what is a cost, what is a benefit, and how these things play out for society at large.

Section II.C: Prices, Inflation, and Indexation

I have no disagreements with Dube here. He makes the point that minimum wages do not cause general inflation. There are cost pass-throughs (acknowledged in this written testimony), but they are unlikely to cause a general increase in price levels. This lay-person argument against the minimum wage, that the increasing prices will "inflate away" the benefit, is a fallacy. No argument there.

Section III. The Minimum Wage, Poverty, and the EITC

This section starts with:

Minimum wages tend to increase income going to working class and poor families. However, the anti-poverty aspect of minimum wage is limited by the fact that many families under the poverty line do not have substantial attachment to the labor force.

The first sentence seems to conflict with the table at the top of this post from Minimum Wages. And the second sentence seems to contradict the first. If most families below the poverty line don't have a strong attachment to the labor market, then presumably they aren't working a lot of hours. Raising the minimum wage is a very poor way of helping these people, compared to (for example) giving them direct transfers of cash or other benefits. To put it into simple mathematical terms: the fewer hours these households work, the less the wage level affects their income. Also, the more binding the minimum wage, the more likely these folks are to see their hours reduced. Once again, see the Seattle studies, where a reduction of hours-worked offset the wage increase. Finally, many minimum wage earners aren't in poverty-line households at all. If we're trying to target poverty, minimum wages are the wrong policy tool.

I'm scratching my head about some parts of this document, because Dube seems to be saying that the minimum wage complements the earned income tax credit. The thinking goes, if you raise the earned income tax credit, employers can afford to pay their employees less, because they know those employees are getting a subsidy. (Imagine if your employer knew that some philanthropist were paying you $1 million to stay employed at your current job; they'd be slightly less inclined to cut you a bonus. The argument about the EITC is this same intuition on a smaller scale.) I thought that Steven Landsburg debunked this argument, here and here. Granted, these posts were written after Dube's written testimony, but they're employing basic Econ 101. Any other thoughtful economist could have worked this out prior to Landsburg's blog posts. This gets back to the big pink rectangle in the diagram above. If there is intense, wasteful competition for those now-scarce minimum wage jobs, there will be a big deadweight loss as low skilled workers attempt to out-do each other. The benefits of the EITC are competed away, more so than if there wasn't a minimum wage. If the competition for jobs isn't a big deal after all, then it's possible Dube is right.

This was a useful exercise. I'm left with a feeling that I understand the pro-minimum wage economists arguments much better, assuming Dube is speaking for them as a group. I think they are too cavalier about dismissing the negative effects of the minimum wage. They take "p > 0.05" to mean "there is no effect" rather than (more appropriately) "there is too much noise to signal". I think that the politicians and political agitators who "cite" them are missing the important caveats and qualifiers, and these economists have fallen down on the job of correcting them when they misstep. I also think that they are just wrong about what the literature says. They tend to understate or explain away or leave undisclosed those studies that do find large disemployment effects. I don't know any good way to settle this other than sitting down with the original studies and deciding for yourself which ones are convincing. But, just to be clear, pro-minimum wage economists tend to admit that the disemployment effects are real and will show up if we raise the statutory minimum high enough. If Dube is serious about targeting half the median as the appropriate minimum wage level, then he's a lot closer to Neumark and Wascher than he is to the "Fight for Fifteen" crowd. I wish he'd say so more clearly.