So as the minimum wage blog discussions drag on I've run across three memes (the original "meme", not the pictures) that I want to push back on a little (although one I need a little help with... it's more of an inkling right now).
1. Privileging theory.
The first is the tendency to say that when empirical evidence contradicts solid economic theory, you have to dismiss the empirical evidence. I have seen this put out there with varying degrees of strenuousness, of course. In any case, my view is that this way lies madness. You might be very curious about an empirical result for which there is no theoretical explanation (although in this case, of course, there is). You might want to think about how you might be getting that result. In my dissertation I'm coming up with a funny finding - a labor demand subsidy is resulting in reduced employment but increased earnings. Strange. My working hypothesis (that I'm going to test) is that part-time jobs converted to relatively fewer full time jobs without necessarily changing total hours worked (you have to offer a full time job to get the subsidy). I have no obligation to take my result at face value. But I don't just dismiss it because it doesn't at first blush match how I think labor markets work.
The moment we start privileging theory over empirics like that, our theories are going to start getting unmoored from reality and we're going to have doctrines and dogma rather than theory. Science is an interaction between theory and observation - and a constant revision and testing of theory to ensure that it is a story that best explains the world we experience. You figure that out by going and collecting data on the world we experience.
2. Empiricism as a popularity contest
I've also seen references to the fact that X number of empirical studies support the view that the minimum wage has disemployment effects. Who cares? You don't do meta-analyses by counting up studies. You organize and assess the studies by the quality of the methods they use. Any undergrad STATA monkey can run a state-level fixed effects model. That data is readily available and the model is very straightforward. So I don't care if there are a million variants of a state-level fixed effects model telling me that the minimum wage has disemployment effects. I mean, it's nice to have confirmation that that model works out that way I guess but when you use similar methods on similar data, you tend to get similar results. The important question is whether that's the right specification. The quasi-experimental studies may be outnumbered (I don't know this for sure, but I suspect it's true, particularly if you start trolling the lower tier journals [the original "trolling", not the thing Ryan Murphy does]), but the point is they are better. That's not to say you can't criticize them. Bob Murphy recently raised some important criticisms of how the quasi-experimental studies use time trends, which you may or may not be convinced by. But by almost any econometric standard the identification of the quasi-experimental studies is stronger than the fixed effects studies.
This is very similar to the fiscal multiplier debate or the immigration debate (in the empirical economics literature). These debates are not about counting studies - they are about dueling methods. If you run fiscal multipliers with certain identification techniques, you tend to get a certain magnitude of results. If you run immigration impact on native studies with Card's methods you tend to get different results from if you run it with Borjas's methods. Knowing this, nothing could matter less than how many of each type of study was run and published. What matters - once we establish the basic distribution of point estimates - is which specification we think is the best one. We might still disagree on that, but you don't argue that point by counting studies.
3. Racism and the minimum wage (this is the one I need a little help with).
I've heard a lot that the original proponents of the minimum wage wanted to push blacks out of the labor market. It's a little hard to parse... it requires the assumption that blacks are less productive than whites and that in the early twentieth century people needed the minimum wage as an excuse to discriminate against blacks. But it always seemed plausible to me simply in the sense that lots of progressives at the time were racists and progressives favored minimum wages. (And you can wave your hands over the fact that opponents of the minimum wage at the time were pretty racist too, I guess.)
The other day for completely unrelated reasons I was reading a portion of Bruce Schulman's book From Cotton Belt to Sunbelt, on the economic development of the South from the 30s to the 80s. He was briefly discussing the minimum wage legislation in the 30s and he noted that the proponents were advocates of black Southerners and that the opponents were the defenders of Jim Crow and racial terror in the South.
I doubt it's a clean story. You've probably got a mix. I am very concerned now that people who argue this point are cherry-picking cases and that there's no real solid correlation. But I do need help on this - does anyone know anymore details? My advice is to be cautious about this meme.
Friday Night Music: When I’m 64
15 hours ago