Scott Sumner  

Paul Krugman on the minimum wage

Econlib in Top Three!... Sadie Hawkins Rant...

Paul Krugman recently had this to say on the minimum wage:

Until the Card-Krueger study, most economists, myself included, assumed that raising the minimum wage would have a clear negative effect on employment. But they found, if anything, a positive effect. Their result has since been confirmed using data from many episodes. There's just no evidence that raising the minimum wage costs jobs, at least when the starting point is as low as it is in modern America.
This struck me as very odd. I've done work that suggests that minimum wages probably cost jobs (although admittedly my research was on the Great Depression, and hence may not be applicable to today.) But it's widely known that there is lots of other research suggesting that minimum wages cost jobs. In addition, economic theory suggests that when you make something more expensive, people will buy less of it. Perhaps strangest of all, Krugman's claim is contradicted on the very first page of the study he links to as having "confirmed" his claim:
On balance, case studies have tended to find small or no disemployment effects. Traditional national level studies, however, have produced a more mixed verdict, with a greater propensity to find negative results.
Paul Krugman has said that he doesn't read conservative blogs, so obviously he may not be familiar with the literature on how the minimum wage costs jobs. But that's no excuse for not even reading the first page of the study he links to in support of his claim.

Krugman doesn't come right out as say that the Card-Krueger study provides support for the Democratic Party's recent attempt to raise the minimum wage, but that's surely the implication that most readers will draw from his post. And yet even the Card-Krueger study doesn't necessarily support the Obama administration's proposal for a $10.10 minimum wage, or the Congressional Democrats attempt to raise the minimum wage to $12/hour. The studies he cites look at the effects of small increases in the minimum wage.

There are conflicting empirical studies of the effect of minimum wages. When that occurs, it's probably safest to go back to the basic theory. That doesn't necessarily mean that minimum wages are bad policies, perhaps the gains in income outweigh the cost in unemployment. But it's disingenuous to claim that we can raise minimum wages without any disemployment effects.

Comments and Sharing

COMMENTS (37 to date)
Urstoff writes:

The minimum wage debate continues to baffle me. Given that the evidence is so mixed, and there are clearly superior alternatives (e.g, the EITC), why people (economists especially) support the minimum wage needs explanation. For the layperson, simple ignorance seems to be the reasonable explanation. For economists and policy wonks, though, I would guess a combination of confirmation bias and anti-corporate sentiment.

And surely even Krugman would agree that a minimum wage of, say, $75,000 a year, would cause massive unemployment, even if phased in gradually. So what's different about lower minimum wages?

Daniel Kuehn writes:

I was 100% in agreement with you until the very last sentence (and maybe the second to last two depending on how strongly it's read).

Why is it disingenuous?

Andrew writes:

@Daniel Kuehn --

It is disingenuous because if you say that there aren't any disemployment effects of the minimum wage then there isn't any reason to phase them in over time or even just stop at $15.

What you're reading into is that it's not disingenuous to suggest small increases over time do not cause disemployment effects.

This is drastically different than saying the minimum wage doesn't have any costs.

Just because I threw a rock at a glass house and didn't break anything this time doesn't mean it's a good idea to throw rocks at glass houses.

Don Boudreaux writes:

Coincidentally (or not?), I posted this piece on my and Russ Roberts's blog, Cafe Hayek, just this morning. A slice:

In the United States, national minimum-wage legislation has been around since 1938 (or, depending on how you assess such things, even earlier: 1931), and has been increased (by my count) 24 different times. It might well be that the disemployment effect of each individual minimum-wage hike was small (perhaps even undetectable), but it does not follow from this accession of small and undetectable effects – each one the result of a measurement conducted after each isolated rise in the minimum wage – that the disemployment effect of the minimum wage is small.
Nick writes:

There are two basic theories, though, one of labor markets in perfect competition, and one of labor markets where low-wage employers have market power. Identifying which one more accurately describes the labor market is an empirical question, not something where we can go back to 'the' basic theory.

The article Paul Krugman linked to references the literature that compares states, and that literature does indeed find a disemployment effect, but the article is about how that literature might not adequately control for unobserved differences between states.

The authors argue that the proper comparison is between counties that share a border but have different minimum wages, which, hopefully, controls for unobserved differences. They use national data and find no effect.

Perhaps Paul Krugman should have said "there's just no convincing evidence that raising the minimum wage costs jobs' since clearly papers have been published which do find a disemployment effect. But it's the nature of statistical research that there will almost always be studies on either side.

Daniel Kuehn writes:

Andrew - I see. Was that what you meant Scott? I had assumed the reference point here is minimum wages of the sort that the studies have covered. I don't think Krugman would say that of any increase would he? Would anyone walk away thinking that? (So I guess I didn't agree with Scott 100% up until the end because my read of the evidence is that the proposed $10.10 increase is within the range of historical increases - see Table 2 and Figure A here:

Agree on the glass houses point but I don't see how that's a good analogy unless someone already agrees on the conclusions (in which case why employ the analogy?).

Urstoff writes:

Nick, any potential statement like that needs to be more precise: "there's just no convincing evidence that raising the minimum wage by X costs jobs...." After all, even Krugman would agree that raising the minimum wage to 100k a year would cost jobs. And again, if the evidence is so murky, and there are obvious superior alternatives, why even support a minimum wage at all?

Don Boudreaux writes:

Commenter Nick writes:

There are two basic theories, though, one of labor markets in perfect competition, and one of labor markets where low-wage employers have market power. Identifying which one more accurately describes the labor market is an empirical question, not something where we can go back to 'the' basic theory.

This formulation misses the mark. First, the law of demand does not have as a prerequisite that competition be perfect (as perfect competition has been defined by economists for the past century). If it did have this prerequisite, then the law of demand would almost never apply in reality.

Second, monopsony power among buyers of good or input L is only a necessary condition for a coerced hike in the price of L not to cause fewer units of L to be purchased; such monopsony power is not a sufficient condition. If, for example, monopsony buyers of L face competition in the output market, then a coerced increase in the price of L might well still cause the quantity demand of L to fall (if only by causing some buyers of L to go out of business).

Third, if the question is the existence or not of monopsony power in the hands of employers of low-wage workers, then test for such power directly rather than - as is far too often the case - inferring it simply from the findings of empirical studies that detect no, or even positive, employment effects of the minimum wage.

Dave Smith writes:

Krugman, Paul (1994), ‘Past and Prospective Causes of High Unemployment’, Federal Reserve Bank of Kansas City, Economic Review, Fourth Quarter, 23-43.

Daniel Kuehn writes:

Don Boudreaux -
But both monopsony and perfect competition satisfy the law of demand. The law of demand is not the issue here. Nick correctly identifies the theoretical issue (of course "monopsony power" wears many potential guises with frictions and bargaining over surplus not always appearing the same way they do in an Econ 101 monopsony model - but it's still all market power, all the same fundamental argument).

Nobody here that I know of is disputing the law of demand.

john b writes:

it seems to me that if core economic theory does not apply to human labor, because humans are special; then we can't realistically make any assumptions about economic theories applicability to human labor.

I mean - Krugman seems to be saying,based on how I read it, that because humans are clever and want income, increases in the minimum wage don't change demand the same way other cost increases change demand. (As opposed to a box of cereal - it is not clever, so it can't "outsmart" cost increases).

But if humans are clever in this way, it seems reasonable to assume they are clever in other ways, and no standard economic theory applies until we are able to validate it in the context of humans.

Zeke writes:


"The authors argue that the proper comparison is between counties that share a border but have different minimum wages, which, hopefully, controls for unobserved differences. They use national data and find no effect"

I am flummoxed. Why would anyone assume the largest difference in labor markets between two countries was the minimum wage? Even if you believe it to be the dominat individual variable, the sum of all the other variables would make minimum wage insignificant. The proposed study seems worthless.

Jay writes:


Agreed, but I think EITC is seen as the "other guy's solution" to most of MW supporters no matter how much more superior or effective it is. It is also hard to explain in less than 10 seconds when you can just say that MW is "giving American workers a raise" no matter how many times you point out how absurdly low % of people earn it.

Dustin writes:

I don't think the "then why stop at $15 / hr" argument is sound because the function need not be linear.

Any chance that some level of minimum wage increase has a money creation effect that increases AD and so offsets the disemployment effect that would result from raising the price of labor?

Andrew_FL writes:
That doesn't necessarily mean that minimum wages are bad policies, perhaps the gains in income outweigh the cost in unemployment.

Not for the people who's reduced freedom of contract throws them out of employment entirely.

Oh the outrageous things people say when they believe in interpersonal utility comparisons!

@Jay-Many union worker's contracts explicitly tie their wages to the minimum wage. They would get an automatic "raise", too.

Nick writes:


I meant to write 'between counties', not 'countries'. The idea is that you compare counties which are close together, and so hopefully very similar (not different) to each other in all respects except for the minimum wage.

Ben Hughes writes:
I don't think the "then why stop at $15 / hr" argument is sound because the function need not be linear.

There is nothing about that argument that requires linearity. It implies that the function be monotonic. Not the same thing.

I think the burden is on detractors to explain why it would be non-monotonic.

Nick writes:

@Don Boudreaux

Putting aside your first two paragraphs, (a) which I agree with and (b) whose relevance I admit I don't entirely understand, I have a question about where you say

> then test for such power directly rather than - as is far too often the case - inferring it simply from the findings of empirical studies that detect no, or even positive, employment effects of the minimum wage.

I'm curious if you are thinking of specific ways to run these tests, or an existing literature which does this?

AlexR writes:

Given that EconLib was recently ranked highly as an economics education site, mostly for its Concise Encyclopedia of Economics I believe, it's appropriate to quote from Linda Gorman's Encyclopedia entry on minimum wages:

"In the early 1990s, after a telephone survey of 410 fast-food restaurants in New Jersey and Pennsylvania, economists David Card and Alan B. Krueger challenged the consensus view that higher minimum wages shrink employment opportunities. Their results appeared to demonstrate that a minimum wage increase resulted in increased employment.12 Because telephone survey data are notoriously prone to measurement error, Neumark and Wascher repeated Card and Krueger’s analysis using payroll records from a similar sample of restaurants over the same time period. The results from the payroll data showed that 'the minimum-wage increase led to a decline in employment in New Jersey fast food restaurants relative to the Pennsylvania control group.'13"

Now, one may ask why the Neumark-Wascher study was less influential than Card-Krueger's. The answer may not reflect well on the practice of our profession. Imagine physicists overturning their consensus view of the constancy of the speed of light based on an initial report about some neutrinos passing through a leaky pipe, then ignoring subsequent evidence to the contrary.

Don Geddis writes:

@Dustin: AD is controlled by the central bank (via control of the money supply). Acting as a negative-feedback "thermostat", the Fed is able to hit nominal targets (like AD) regardless of changes to aggregate supply factors (like minimum wages). So, no: due to monetary offset, there is no chance that MW increases could have a "money creation effect" that could increase AD.

Daniel Kuehn writes:

Thanks for pointing the article out, AlexR. I hadn't read it before - but her treatment of the 2000 AER articles on it is extremely problematic.

She goes over the N&W payroll data result which is fine, but then in characterizing C&K's paper in the same 2000 AER she says they retreated from their position!

What she DOESN'T share with readers in that discussion is that C&K's 2000 article found a lot of problems with the payroll data N&W were using and using administrative wage data they confirmed their earlier finding of no disemployment effects (as she writes, I believe the effect was somewhat smaller than the phone survey effect - which she characterizes as "retreating").

Don Boudreaux writes:

Nick asks, in response to my suggestion in an earlier comment that claims about monopsony power in the labor market be tested directly rather than merely inferred from the results of some minimum-wage studies,

I'm curious if you are thinking of specific ways to run these tests, or an existing literature which does this?

That's a good question. I don't have any bright ideas for how to conduct such studies - which of course doesn't mean that people smarter than me cannot design and actually conduct some.

But I can point to a large body of existing empirical evidence against the claim that policy-relevant monopsony power is rampant among employers of low-skilled workers. This evidence is of the Holmesian dog-not-barking sort - namely, the overwhelming failure of those who allege the existence of such monopsony power to enter the market with their own funds in order to profit from hiring all these underpaid workers away from their current employers.

Because in the United States entry barriers into retailing, food service, and other industries that employ lots of low-skilled workers are very low (entrepreneurs and new firms are incessantly entering and exiting these industries), the fact that nearly every academic and pundit who alleges the existence of monopsony power (as a reason to raise the minimum wage) steadfastly refuses to act on this alleged knowledge is very telling.

If Jones claims that $20 bills line the sidewalks of your town but, as he does so, refuses himself personally to bend down to profit from this alleged reality, Jones's behavior is relevant and powerful empirical evidence against the veracity of his claim. Ditto when Jones claims that profit opportunities in the form of legions of underpaid workers line the avenues of America but refuses to take action in the market to seize those alleged profits.

Dustin writes:

Don Geddis - the offset theory is clear enough, though my question was about the implication of claiming no disemployment effects. In other words, if a wage increase results in no disemployment effect, it would have to be because (and perhaps others):
1) Labor supply and demand are highly inelastic, or
2) AD is increased and offsets the supply gap

The plausibility of these alternative scenarios would be useful in evaluating the 'no disemployment' claim if the data is truly inconclusive.

AlexR writes:

You make fair points, Daniel Kuhn. The term "retreat" is freighted with negative connotations. I think the Card-Krueger result is best described as a non-result. (I'm thinking of their 1992 [?] paper; I don't think I read the 2000 one.) They fail to find a significant effect on employment from a small increase in the minimum wage, and frankly conclude that their findings aren't consistent with any theory (including monopsony and job search models). Now let's put on our Bayesian caps. Is this non-result particularly surprising, given the welter of confounding effects that might be in play (some of which Card-Krueger try to account for)? Even if surprising, is this non-result powerful enough to overturn a previous consensus on minimum wage effects?

Imagine that a scientist drops an ice cube into a pot of water. The mercury thermometer in the pot is a little hard to read, but seems to show no change in water temperature. In fact, the water might be getting warmer! The published study creates a sensation; one dazzled physicist notes that "before this experiment, we all blithely believed in the laws of thermodynamics, but facts trump theory..."

BTW, Neumark and Wascher have a 2006 NBER working paper surveying empirical studies of minimum wage effects on employment, available at:

Scott Sumner writes:

Everyone, Unfortunately I don't have time to answer each comment, but a few general observations:

1. There have been dozens of studies of the minimum wage, there's no point in obsessing over any single study.

2. If the minimum wage were raised to $10.10, it would be nearly double the 2006 level. There are no studies that I know of that suggest that sort of increase would not raise the unemployment rate. That rate is higher than the median income in Puerto Rico, for a full time worker. There are also calls for a $12 or even $15 minimum wage.

3. The market for unskilled labor is generally assumed to be relatively competitive, at least in the long run.

4. Krugman's claim is clearly false. It's pointless to say that if he had made some other claim it might have been true. He knew exactly what he was doing.

5. The evidence that the minimum wage has no disemployment effects is far weaker than the evidence for lots of conservative ideas that Krugman routinely dismisses, such as the claim that unemployment insurance increases unemployment. He is using a double standard for empirical evidence, disregarding studies that don't reach the policy conclusion that he prefers.

Alexandre Padilla writes:

I understand that the current post and most of the discussion on the minimum wage is often centered about whether or not increasing minimum wage has or doesn't have some measurable disemployment effects. But are we not forgetting that employment is more than just a wage? A job even for unskilled workers is more than just paying them a wage. There are many things that a job's compensation encompasses: fringe and/or perquisites benefits, training vs. no training, decision-making responsibilities, performance evaluation and how is your performance rewarded, what are your tasks, etc.

I am not saying we should ignore the question of potential disemployment's effects of increasing minimum wages but they are many other effects that just disemployment we can't ignore.

Arthur_500 writes:

Massachusetts has long had many unfriendly policies as well as California. I want to ignore California for the moment as there is a large undocumented population that may be working below the posted minimums as therefore skewing the documented results. so lets look at the East Coast.

This State continues to develop labor standards, business restrictions, etc. They were the first to implement mandatory auto insurance and maintained the highest rates in the country for decades.
They were the basis for ObamaCare and it was a money hole filled with US Taxpayer dollars. A minor detail ignored by the State residents and the supporters of ObamaCare.

through all this they survive. The eastern MA residents will tell you it is proof how great these policies are as they find ways to avoid them. It is the mentality of equality for everyone except me because I am special.

So we increase the minimum wage and it increases the prices of most everything so everything remains in equilibrium. Of course the poor remain the same just with larger numbers for income. Essentially you solve nothing but you feel good for a day.

So, I can understand how a liberal could be excused for not seeing that they have changed nothing and thinking that increasing the min wage does no harm. As to why an intelligent person who is supposed to think these things through misses the point, I regret that Mr. Krugman has some explaining to do:)

W. Peden writes:

[Comment removed pending confirmation of email address. Email the to request restoring this comment. We have tried to reach you previously and you have not responded. A valid email address is required to post comments on EconLog and EconTalk.--Econlib Ed.]

Dustin writes:

@Ben Hughes
Yes you are correct - monotonic

RPLong writes:

Has anyone (i.e. in the literature) ever considered the possibility that one of the reasons we don't observe strong disemployment effects from minimum wage increases is not because they don't exist, but because the average market wage in the United States is already above the minimum wage?

J.D. writes:

The chief flaw with C&K's study was that they counted FTEs instead of actual human beings, and then didn't control for hours worked. That creates big problems when trying to argue that it increased or had no effect on the number of jobs; if there are less discrete people who now have to work more hours to maintain a constant level of productivity, then it's possible to observe FTE numbers increasing even though the net amount of discrete jobs gained (i.e., actual people who are not satistical abstracts) is negative. Hard to argue one's policy preference has a benefit for real people when your dependent variable isn't even a real person.

Thomas writes:

If PK wanted to use his blog to educate "progrrssives" on the minimum wage he would point out that even though the studies he most credits find "little or no " harm to low wage earners,
a higher EITC is a better way to transfer income to them. The same could be said about the multiple posts on the subject on this blog.

Nick writes:


Most studies of the minimum wage focus on industries -- like the restaurant industry -- where the minimum wage actually does matter.

Floccina writes:

@Urstoff I think the reasons that knowledgeable people sometimes support minimum wage even though they know that it is inferior to a wage subsidy are:

1. They care for low wage works, especially those who might be too shy to ask for a raise and they think think that the a wage subsidy is too difficult get.
2. They want to be seen as caring for low wage works.
3. They resent and do not like employers who pay low wages.
3. They resent and do not like middle class and rich people who use businesses that pay low wages.

RPLong writes:

@ Nick - I realize that, but I'm still not sure that's "good enough."

Consider, for example, the fact that the average wage for a dishwasher in the United States is $8.62/hour, compared to the $7.25/hour minimum wage.

I haven't done a full search of all jobs or job categories, hence my question, but there is a lot of evidence out there that would suggest that most low-wage workers already make substantially more than the federal minimum wage.

Might that be one explanation for why we see mixed empirical evidence for disemployment effects?

Daniel Artz writes:

It never ceases to amaze me that whenever the Card-Krueger study is brought up, or any of its predecessor studies, like the Card or Katz-Krueger studies, the only thing mentioned is the conclusion. Never any detailed look at methodology.

In all scientific studies, but most especially in the social sciences (to the limited extent that such a phrase is not oxymoronic), methodology is EVERYTHING. A poorly designed study can reach any conclusion that its author sets out to find. And any empirical study in economics which doesn't address the ceteris paribus assumption, specifically identifying all extraneous variables which can affect the data and at least attempting to control for such variables, is deeply flawed.

Sure, there are several studies that can be used to support Krugman's conclusion that there are only minimal disemployment effects of small increases in the minimum wage. And every single one of them suffers from very serious flaws in methodology. Take the Card-Krueger study, for example. It purported to examine an increase in New Jersey's minimum wage that became effective on April 1, 1991. So the study examined the period from February and March, 1991 through December, 1991. But the increase in the minimum wage had been on the books since 1989, meaning that employers had 2 full years to adjust to the coming increase. The Card-Krueger study, however, assumes that all adjustment to the increase in the minimum wage occurred in the 30-60 days immediately prior to the effective date. Any adjustment in labor structure that employers made before that time is completely ignored. And, although the study purports to examine "the fast food industry", it ONLY looked at big chain restaurants - KFC, Burger King, Roy Rogers. But that is the best capitalized portion of the fast food industry, the segment best equipped to handle regulatory changes and best equipped to take advantage of other methods of controlling costs through bulk purchase and centralized distribution. The Card-Krueger study does nothing to examine the effects on the single store, mom and pop segment of the fast food industry; if a higher minimum wage simply creates additional competitive advantages for big chain restaurants at the expense of the low end mom and pop sandwich shops, the study will completely miss the disemployment effects of that change. Krugman, of course, completely ignores these flaws in the Card-Krueger study, just as he ignores all of the studies which come to a contrary conclusion. Why? Because he chooses to believe in the beneficence of state control of the economy, so why let facts stand in the way of that belief.

Comments for this entry have been closed
Return to top