The Nobel Prize in Economics was announced this week. As usual, Alex Tabarrok has a great description of the contributions of the winners. But I have seen a number of commentators, mostly on “the right,” question this award, especially for David Card. Mostly they have focused on the highlighting of Card’s paper (with the late Alan Krueger) on minimum wages, saying that this paper has been heavily criticized and debunked, or as evidence that “economics has degenerated into socialist propaganda.”
Yikes! If true, these are serious charges against a profession in decline.
But hang on. What’s really going on with the award for David Card? Again, Tabarrok sums it up nicely: “what really made the paper great was the clarity of the methods that Card and Krueger used to study the problem.” What was this clarity?
To understand their 1994 paper, we need to travel back to 1982. That was the year a comprehensive survey of the existing research on the minimum wage was published in the prestigious Journal of Economic Literature. If you know anything about modern methods for studying the minimum wage (or basically anything else in economics), reading this article is like stepping into another dimension. Up to that point, almost of the research confirmed the basic supply-and-demand intuition that higher minimum wages reduce employment. But importantly, almost all of the studies used time-series analysis, almost exclusively of changes in the federal minimum wage, focusing primarily on teenagers. In other words, if the minimum wage increases federally in, say, 1970, does teenage unemployment increase in 1970 and the few years that follow?
This is certainly an approach. But it’s an extremely weak approach. Lots of other things are changing. Teenage unemployment, like all unemployment, is cyclical with the business cycle. You can try to control for these, but it doesn’t actually achieve what you want to achieve. What you want is a natural experiment.
The sum total of the time-series papers was summarized as “All studies find a negative employment effect for all teenagers together and the signs are almost exclusively negative for the various age-sex-race subgroups.”
Well, that seems to settle it! All papers agree. A consensus! Leading to, somewhat famously, even the New York Times writing an editorial calling for abolishing (or at least not raising) the minimum wage. Why did they say this? Because there was “a virtual consensus among economists that the minimum wage.” Journalism follows the science, great! The trouble though, is that the empirical science just wasn’t very well done, and anyway even if it was everyone was basically looking at the same data (changes to the federal minimum wage), just slicing the data slightly differently.
There were a small handful of papers, it appears just three papers, that used state-variation in minimum wages to do cross-sectional analysis. In other words, do states with higher minimum wages have higher teenage unemployment rates than states that don’t? Even these papers were extremely weak by modern standards. For example, one of the three papers is Kalachek (1969), which looked at teenage unemployment across 75 MSAs, and asked whether the state had its own minimum wage above the federal level. 45 MSAs did have higher minimum wages. Sounds like a good start. But here’s how he approached the issue: he just created a dummy variable for whether the state had a higher minimum wage! And guess what? Even this very weak study found no effect of the minimum wage on employment!
A follow-up paper by Katz (1973) tries to look at the same data in another way, but again just using a dummy variable for whether a state had a higher minimum wage than the federal level. So many other factors are not controlled for, and that’s the point. In a simple cross-sectional, there are only so many things you can control for, but even controlling for everything doesn’t give you a true test! What we want to know is “if state X did (or didn’t) have a minimum wage, would employment be impacted?” You need a counterfactual, but just “controlling for stuff” isn’t a counterfactual.
This is the empirical void that Card and Krueger stepped into in their 1994 paper. Working in their favor, several states had been increasing their minimum wages, despite the consensus among economists. This provided an excellent opportunity to study the issue more carefully. Card and Krueger chose to focus on a change in the minimum wage in New Jersey and compared it to nearby counties in Pennsylvania (where the wage wasn’t raised). Did employment in fast food industries change after NJ raised the minimum wage? This state-by-state comparison was an attempt to identify and study a natural experiment.
Of course, it’s just two states. The data they used was questionable. Etc., etc. Their critics today are still focus on these issues. But the core new insight was a new way to study the minimum wage. Neumark and Wascher’s response to Card and Krueger is probably the most famous, but notice that they don’t critique the method, because the method was very good! They critique the data.
The famous NJ/PA paper from 1994 was not even Card and Krueger’s first attempt to use state-level changes to study minimum wage impacts. A 1992 symposium in the ILR Review contains a paper by Card looking at a change in California (comparing with other states that didn’t raise their minimum wage), as well as another paper by Card, a paper co-authored by Krueger, and even a paper by Neumark and Wascher, the constant critics of Card and Krueger. This research was referred to as the “New Minimum Wage Research” in the journal at the time, a term still used by scholars today.
To show the influence of Card (and the late Krueger), I would simply point to a recent attempt to survey the literature on the minimum wage co-authored by Neumark once again (this time with Peter Shirley). You will learn a lot by reading it, and I encourage you to do so even if you don’t agree with Neumark’s particular reading of the literature. What you will notice, if you look at the 66 papers since 1992 that are reviewed, is the enormous influence of Card and Krueger on this research. The approach they pioneered is now standard in the literature. You will find lots of different approaches and different areas of study: city-level minimum wages, comparisons of other states (Illinois vs Indiana), attempts to study all minimum wage differences near state borders, rather than just one state-pair in one year! Very ambitious stuff. But importantly, gone were the days when economists would run time-series analysis of federal changes with slightly different subgroups or different ways of measuring employment.
All of that later research is influenced by Card and Krueger’s initial research. And they were also constant contributors to the research, at least for a time.
And this is typically what economists win Nobel Prizes for: not for a particular result or a particular paper, but for a new way of doing economics, which later researchers not only utilize, but come to see as standard in the field. Of course, the methods get better over time. I would be shocked if the AER today would publish a simple two-state comparison of minimum wages, especially if it only used the basic discontinuity design that Card and Krueger used. But the AER also wouldn’t republish Hayek’s famous 1945 paper today either. Not because the papers are wrong, but because the papers are now part of the way we do and think about economics.