« ORSON SCOTT CARD ON MEDIA BIAS METHODS | Main | Selling Art to the Masses and Tracy Lawrence's "Paint »

July 14, 2004

The Effect of the Minimum Wage

Steven Landsburg usually makes more sense than he does in the Slate post in which he discusses the minimum wage. He makes three claims that seem to me wrong. The claims are (1) and (3) in my paraphrase):

1. Published studies of the effect of the minimum wage on employment cannot be trusted because of a selection effect: a study which found no effect would not be published, but a study which found a study due to a trick in the data *would* get published.

2. "It is almost impossible to maintain the old argument that minimum wages are bad for minimum-wage workers."

3. A minimum wage increase will hurt employers.

First, let's have some discussion of the theory. Why do economists think that an increase in the minimum wage reduces employment? Assume some employers are actually paying the minimum wage before the increase (that is, we don't have a minimum wage of $.25/hour in an economy where nobody works for less than $5.00/hour anyway.) Consider three types of employers:

1. Employer A does not change the number of hours of worker time he buys when the minimum wage goes up.

2. Employer B reduces the number of hours of worker time he buys when theminimum wage goes up.

3. Employer C *increases* the number of hours of worker time he buys when the minimum wage goes up.

How many employers of each type will there be? Lots of type A, and lots of Type B, I would think-- or, if you like, at least a *few* of Type B. But I would expect zero employers of type C. Why would any employer react to minimum wage increases by hiring more workers? If he is so generous as to like to give away money, he would have done that even before the minimum wage increase. Thus, there will be some employers who don't react, and some who reduce employment, so on average employment will fall.

The theory is therefore unequivocal: people will be working fewer hours if the minimum wage is increased.

But how big will the reduction in hours be? That is the real question, and it might be very small, especially in the short run. If we increase the minimum wage from $6.00 to $7.00 today, employment might well be unchanged tomorrow. Even over six months, it might not change much, if managers need time to ponder, for example, whether it is worth cutting back on the hours a fast-food restaurant is open. Much of the impact will occur in the long run-- over a period of several years-- as employers decide not to open new outlets or not to bother refurbishing old outlets whose profitability has been hurt by the higher wages. In the meantime, old outlets may keep on operating with the same shop hours even if the wage is higher, given that the other costs of the shop are sunk already.

Finally, we must keep in mind that the theory just says that employers will hire less labor, not that they will hire fewer workers. A fast food restaurant might go from 30 employees at 6 hours per employee down to 30 employees at 5 hours per employee. That keeps employment exactly the same, but labor hours have fallen from 180 hours to 150 hours, the equivalent of firing 5 of the 30 employees.

Indeed, an increase in the minimum wage could even *increase* employment. Our restaurant might go from 30 employees at 6 hours per employee to 40 employees at 4 hours per employee. That is a big increase in employment, but a reduction in hours worked from 180 hours to 160 hours.

This is important in evaluating studies such as that in the famous 1995 book by Card and Krueger. Their original study looked at employment in a clever comparison of New Jersey with neighboring Pennsylvania, two states with different minimum wage laws. They concluded that an increase in the minimum wage had no effect. Taken literally, their statistical results seem to show that the increase in the minimum wage in New Jersey *increased* employment, but they don't push that conclusion, since they don't have a theory for it. Indeed, the result is so odd as to cast doubt on their entire study, because it suggests that unknown to them, something else entirely different was happening in New Jersey that coincided with the minimum wage increase.

Or, it might be that restaurants went from 30 employees at 6 hours each to 40 employees at 4 hours each as I suggested above. Neumark and Wascher (American Economic Review, 2000) take a look at hours instead of employment in New Jersey and Pennsylvania and come to a different conclusion; Card and Krueger, (American Economic Review, 2000) reply and criticze Neumark and Wascher.

So it is hard to measure the effect of the minimum wage. Now back to Landsburg's three points.

1. Published studies of the effect of the minimum wage on employment cannot be trusted because of a selection effect: a study which found no effect would not be published, but a study which found a study due to a trick in the data *would* get published. Point (1) is nicely hit by a July 9 comment of Jim Glass on Brad DeLong's weblog.

Is Landsburg really saying 95% of all studies have found the minimum wage to have no effect on employment -- and by so finding were deemed too "uninteresting" to publish, like Card & Krueger? If so, that ought to be easy enough to verify. Calling all studies!

A second, bigger, problem with point (1) is the claim that a study which found no effect could not be published. The conventional wisdom is that the minimum wage does reduce employment, so we'd actually expect selection bias *the other way*. "The minimum wage reduces employment" is a "Dog bites man" story. "The minimum wage does not affect employment" is "Man bites dog". Card and Krueger got a lot of mileage out of their study precisely because the results were so counter to theory.

A caveat:studies which show no effect would often not get published because editors would rightly be concerned about lack of statistical power-- a concept I explain in a recent post of mine. Suppose the data is not good enough to pick up the effect of the minimum wage-- even if it is a large effect-- because too many other things are going on in the economy that affect employment. Then, a study which cannot reject the null hypothesis of no effect also would not be able to reject the null hypothesis of a large negative effect.

Jacob Levy, at the Volokh Conspiracy, writes about this selection theory. His post prompted me to write this, since the Card-Krueger result has come up in the Indiana Law and Econ Lunch before and since Levy wrote

The econo-bloggers all seem to think Landsburg is basically right about the consensus view among economists.

I'm an economist who dissents from that view. Tyler Cowen has an interesting angle too: just as product quality falls when a price maximum is imposed, so we would expect job quality (e.g., air conditioning) to fall when a wage minimum is imposed.

Also, Steve Bainbridge has a good post where he says,

Being curious as to whether there really was a new consensus to which folks like Sowell and Neumark are just outliers, I did a little digging and came across "Consensus Among Economists: Revisited" by Dan Fuller and Doris Geide-Stevenson, published in the Journal of Economic Education (Fall 2003). They find a decline in agreement among surveyed economists between 1990 and 2000 with respect to the minimum wage: "It is likely that the recent research and debate concerning the effect of a minimum wage increase on employment have shifted economists opinion toward less agreement." Yet, while there has been a shift, in 2000 a plurality of the surveyed economists (45.6%) still agreed with the statement "Minimum wages increase unemployment among young and unskilled workers." Another 27.9% agreed with provisos, while only 26.5% disagreed. So perhaps there is less of a consensus than some would have you believe.

2. "It is almost impossible to maintain the old argument that minimum wages are bad for minimum-wage workers."

Jim Glass points us to a Cleveland Fed survey by Neumark, Schweitzer and Wascher (two of them were the critics of the Card-Kreuger study I cited). But I would not rely on the many empirical studies that do find negative employment effects. Finding the long-run effect on hours worked of a 20% increase in the minimum wage is hard to do accurately if at the same time, (1) tax rates are changing, (2) the economy is rising and falling, (3) import competition is increasing or decreasing, (4) big companies that hire lots of low-paid workers are changing their policies in various ways, (5) the criminality of young unskilled workers is rising or falling, (6) schools and junior colleges are changing the quality of the workers they produce...

That is why Card and Krueger tried comparing just two regions in adjacent states-- to control for these other things. But with just two states, you end up with the problem that maybe something special about one of the states that is not included in the study is driving the results.

So I find the theory more believable. A standard example of why the theory is compelling is to ask whether you believe the effect of an increase would be small if the increase were from $5.00/hour to $50.00/hour. If you think that big an increase would have an effect, how about from $5.00 to $6.00? From $6.00 to $7.00? From $7.00 to $8.00? ... From $49.00 to $50.00? To quote Jim Glass again,

... if we keep changes small enough so we don't see ourselves doing any visible harm we will be free to imagine we are doing a lot of good!

3. A minimum wage increase will mainly hurt employers.

Brad DeLong caught what's wrong with this. Employers who hire minimum-wage labor are likely to be in highly competitive industries-- fast food, agriculture, production of low-quality goods, and so forth. Their profits are just a normal return on capital. If their costs rise, their prices will rise too. There will be some short-run loss of quasi-rents-- with higher prices, sales will fall and some of the employers will go out of business. But those employers were not earning more than a bare competitive return to their talents and capital anyway, so they haven't lost much. And, indeed, starting from Landsburg's premise of no employment loss, there won't be any sales loss either, and thus no exit from the industry (if sales fell, then employment would have to fall too, unless we believe employers are willing to pay workers to stand around idle). Instead, the losers are consumers of the products and services of minimum-wage workers.

Posted by erasmuse at July 14, 2004 03:51 PM

Trackback Pings

TrackBack URL for this entry: http://www.rasmusen.org/mt-new/mt-tb.cgi/21

Listed below are links to weblogs that reference The Effect of the Minimum Wage:

» Minimum Wage Update from Deinonychus antirrhopus
Eric Rasmusen has an interesting post looking at the Landsburg article. I particularly like this part, A second, bigger, problem with point (1) is the claim that a study which found no effect could not be published. The conventional wisdom is that the ... [Read More]

Tracked on July 19, 2004 06:04 PM


what would be interesting is if the minimum wage were different in different parts of a single state

(e.g., a lower minimum wage in counties with high unemployment over the past 5 years or so or counties with population loss over the past 20 years or so)

Posted by: mhw at July 15, 2004 11:26 AM

If the minimum wage is well below the market wage, raising it to a point that's still below the market wage won't make any difference.

Posted by: Steve Sailer at July 15, 2004 03:47 PM

The minimum wage can, per definition, not be below market wage if plenty of people are actually working for the minimum wage (today I think 2-3% of American labour earns this much).

Unless of course you believe the monopsony story, which I personally find quite unconvincing considering the kind of businesses that employ minimum wage earners. (it is however incorrect to state that no theory predicts the minimum wage will increase employment, this one does!).

For non-economists it basically means that if there is very little competition for poorly skilled workers in one place and if they refuse to move to earn more in other locations than a large employer can exploit workers by paying them less what they produce.

Regarding economic research I think this is similar to the Rodrik story. For ideological reasons plenty of economist simply do not WANT minimum wages to hurt employment! So there is a very large demand for anyone who can show otherwise. Most economists are pro-free markets, but in their harts of hart they are like most elite academics, subject to the social pressure to become civilized liberals and stop with all their anti- progressive logic.

What is crucial in all this is that it is very hard to show an empirical fact that is 100% robust to all statistical tests in the first place. We cant make controlled experiments, and have to take the data we are given.

The pro-miniumwage guys failed to show that the minimum wage was harmless, so they did the next best thing: used arguments based on statistical methodology to invalidate the results they didnt like.

Now the fragility of this sort of study makes the above easier that it might sound. We dont know all the parameters involved, as there is in practice a million different things happening at once. And even if we did the quality of the data is never perfect.

The most important thing to keep in mind is this: EVEN IF we accept that all studies showing that minimum wages matter are now invalid, that does NOT mean you have shown that minimum wages do not matter!

The burden of evidence is equally on the shoulders of those who make this claim. And they have to be ready to accept the same ultra-skeptical robustness check on their data.
So as the post above said: bring it on.

Posted by: Chicagoboy at July 15, 2004 07:28 PM

A reader emailed me this:

Hashimoto (AER, Dec. '82) did a piece on min. wage effects on training
which showed (theoretically) how a min. wage would LOWER the total
compensation received in equilibrium. He also tests this argument. Total
compensation---including things like how well employees are treated---is
virtually impossible to measure, so we don't see these negative effects
of the min. wage.

Posted by: Eric Rasmusen at July 16, 2004 02:50 AM

Of all the arguments against the minimum wage, the often-seen "If the $7.00/hr minimum wage is so great, wouldn't $70.00/hr minimum wage be even greater, huh, answer this? How about a minimum wage of $1,000/hr?" has always seemed the silliest to me. It might be that a minimum wage is a bad idea, but this argument certainly does not convince me about this. Let's use the exactly same logic for the Navy:

"If having ten aircraft carriers is good, wouldn't hundred be even better? How about a thousand aircraft carriers? Of course not! Do you see that having a thousand aircraft carriers would be totally absurd and destroy the economy? This clearly proves that aircraft carriers are altogether a bad thing."

Posted by: Ilkka Kokkarinen at July 16, 2004 06:52 AM

There is nothing silly about the argument, you are just missing the point.

People generally agree that there is a cost to having aircraft carriers; you have to take taxes to pay for them. You dont have people demanding that we increase the fleet but insisting that it is a free lunch.

But A LOT of people, usually those who support the minimum wage, simply do not accept that raising it will costs jobs. They see it as merely a transfer from rich companies (with endless pockets) to poor workers who are being exploited.

By asking them why we should not raise the minimum wage to say 20$/hours you can often force the person in question to at least acknowledge that the price of labour matters in terms of demand. I suppose one reason is as the blogger himselfs notes that it is easyer to ignore the effect of small increases.

Posted by: chicagoboy at July 16, 2004 05:47 PM

I've been blogging on the minimum wage for quite some time now. It's a pet hobby horse of mine. There's one story on the top page, and you can find others in the archive.

Posted by: Russ Nelson at July 18, 2004 12:14 AM

Post a comment

Remember Me?

(you may use HTML tags for style)