Megan McArdle On Problems In Studying The Minimum Wage

Click here to read it. Excerpt:

"A lot of economists have tried, and I think it's fair to say that the consensus reading of all those papers is that the minimum-wage increases they studied probably caused some unemployment ... but not all that much.

Unfortunately, even that result isn't as useful as you might think in the current debates over the minimum wage. For one thing, a lot of those studies examined relatively small increases in the minimum wage. For example, the famous Card and Krueger study, which is probably the single paper most frequently cited by advocates of a higher minimum wage, involved a 1992 increase to $5.05 an hour from $4.25 -- the equivalent of raising the minimum to $8.50 from $7.15 today.

It wouldn't be all that surprising if a small hike in the minimum wage had little effect on unemployment. But that doesn't mean that you can extrapolate that result to very high minimums, like the Sea-Tac law, which hiked the local minimum wage by more than 50 percent from a level that was already well above the national average. To illustrate the problem, imagine raising the minimum wage by a penny. It's extremely doubtful that anyone would fire workers in order to save 40 cents a week. But you'd be foolish to conclude that it would therefore be safe to raise the minimum wage to $100 an hour. The size of the increase matters.

That's not the only problem, however. It's not even the biggest one. The hardest part about studying the minimum wage is trying to figure out what will happen over the long run.

Let's go back to Card and Krueger. The reason that that paper is so justly famous is that they exploited a neat little natural experiment: New Jersey raised its minimum wage, while Pennsylvania didn't. So the economists compared employment at fast-food establishments in New Jersey to nearby stores in Pennsylvania. What they found is that employment actually rose slightly in New Jersey compared with the control group across the state line.

Pretty neat, huh? Here's the problem: That study only covered what happened to 410 stores over a period of less than a year. I'm not quarreling with the design of their study, mind you; there are very good reasons to stick with a limited sample over a short period of time. Over longer time periods, more and more extraneous factors will start to swamp your results: changes in state labor law or tax policy, local recessions, a municipal ordinance banning cheap restaurants with lurid signs.

The problem is not that the study was bad; the problem is that the study was very limited in what it can actually tell us. Even if we assume that the results are correct, they don't "debunk" the possibility that the minimum wage costs jobs; they just tell us that a 20 percent increase in a relatively modest minimum wage will not produce massive unemployment in the short term.

Over the long term, however, things might be very different. For example, Card and Krueger kept careful track of store closings. But they had no way to track stores that might not have opened because the new minimum-wage law lowered the potential return on investment. Or the aspiring franchisees who decided to open a Subway, which employs only a few people, instead of a McDonald's, which employs many more. Even if Card and Krueger had tried to track those effects, most of those decisions would not have shown up during the time period they studied, because it takes longer than a few months to open a new restaurant.

Nor did their study track longer-term impacts on the establishments they did include in their sample. Did more restaurants run into financial trouble over the long run and decide to close down? Did they abandon expansion plans or decide to invest in labor-saving equipment, such as drink stations where customers can fill their own cups rather than having a drink filled by a server? These are the sorts of effects that take years to materialize; if you just stare hard at a short period, you'll miss them. But if you try to broaden the time period, you run the risk of finding spurious effects that were actually caused by some completely unrelated change in the local laws or economy. The shorter the time period of a study, the more I tend to trust its results -- but the less I'm willing to generalize those results to broad statements about the effects of increasing the minimum wage."