Casey Mulligan's Strawman and the Numbers/Words Divide

It's nice to know that being a big-time University of Chicago economist writing in the New York Times doesn't preclude you from acting like a marginal blogger and beating the holy hell out of a strawman.

So we see, today, in Casey Mulligan's piece in the Times' Economix column, where he takes cheap shots at what he calls "New Keynesianism". Trying to take down a school of thought in a few paragraphs was a dubious mission at best to begin with, but I immediately noticed that his only actual source on what New Keynesianism is was the electric goop between his ears.

Professor Paul Krugman, Keynesian extraordinaire, agrees:

I’ve been asked for reactions to Casey Mulligan’s piece about the failure of New Keynesian economics.The short answer is, he should try reading a bit of Keynesian economics — old or new, it doesn’t matter — before “explaining” what’s wrong with it. For the doctrine he’s attacking bears no resemblance to anything Keynesians are saying.

This is fairly typical of freshwater economists. They know that what the other side is saying is obviously stupid, so there’s no need to read it; they picked up enough about it talking to some guy in a bar, or whatever, to criticize it...

...he presents as “the New Keynesian position” something that is just what he imagines, on casual reflection (or, again, maybe after talking to some guy in a bar) to be the New Keynesian position.

OK, so from now on I’ll assert that the Chicago position on unemployment is that we can cure it by sacrificing goats. Hey, I heard that somewhere — no need to actually read anything they say, right?

So, yeah, strawman. Krugman goes into greater detail about why the Mulligan piece is wrong, but you could pretty much assume it was wrong going in, so no worries there.

What surprises me, though, is just how much this differs from other social sciences. Citing is sacrosanct there. You simply don't refer to an idea or a position or a theory without pointing to someone, somewhere, that defines exactly what that theory is. (Something beyond Mulligan's half-assed links to "Investopedia".) It just goes to show what Noah Smith was talking about when he decried his graduate instruction in economics as little more than getting force-fed DSGE modelling with little instruction in real-world economics and even less instruction in competing economic theories. If you think that the only thing that matters is plugging numbers into a model, why on earth would you care about silly things like citing sources? Numbers are objective! Their only source is Divine Providence!

It points to a deep, deep problem that we are starting to have as a civilization. We know numbers. We know how to manipulate numbers. We have fantastic devices of unimaginable power in manipulating numbers. Humans, though, aren't fantastic at manipulating numbers. We're good at it. Some of us are really good at it.

No, what humans are really good at is manipulating words. Even the least of us has a greater faculty with words than all but the best of us do with numbers. We know words. We're familiar with words. We're not just good at it, we're fantastic at it. Best on the planet, maybe even best in the universe. Go, us.

Since we're familiar with words, we know not to trust words. That's why we focus on things like citation. People lie and misrepresent and omit and use rhetoric and all the rest to manipulate words to their advantage. You've seen it. You may have even used it. That makes us skeptical about words. We demand sources, and citations, and literature reviews, and all that other lovely academic folderol. We know better.

So when we delve into the world of numbers, many of us—maybe even most of us—set aside that healthy skepticism. We presume that "numbers can't lie". We presume, in turn, that quantitative, formal modelling can't be a lie. It can be wrong, but even then, we presume that it's the numbers that will prove the model wrong. If the numbers fit, then it must be true, right?

It isn't true. A model is just an opinion. It's the quantitative equivalent of saying "Colonel Mustard, in the Library, with the Pipe" in a game of Clue. It might be true. It might not. Certain choices of numbers might show it to be true. Others might show it to be false. Others might mean nothing at all. It's a theory. Words, numbers, whatever, that's how theories work.

That means you can indeed lie or misinform or misdirect with numbers. Choosing one set of numbers and ignoring another set might "prove" the model you want. The process of turning qualitative factors into quantitative variables ("operationalization") might also "prove" the model you want. And, hell, if the numbers prove multiple sets of models—which happens—picking one model and ignoring the rest can also "prove" the model you want. All of these are lying with numbers. They can happen. They do happen. Constantly.

I don't think that Casey's lying with numbers. I do think that Casey doesn't really realize what he's doing. He (and many others) forget that the same sort of care that we take with theories in words also needs to be taken with theories in numbers. You have to teach and learn and cite these theories, so that you aren't vulnerable to those who attempt to lie with numbers, and you aren't susceptible to making mistakes with your number-theories that amount to unwitting lies.

Most importantly, you have to be very careful to present the theories as they are, instead of how you think they are. If you don't, you end up like Casey Mulligan: beating the holy hell out of a strawman made of numbers.