Paul Krugman invokes a famous fallacy

Paul Krugman is an economist.  He is not a political scientist.  Mostly, though, economists manage to avoid making certain mathematical mistakes.  Yesterday, Krugman made a doozy, and it is a famous one in political science.

Here is yesterday's column.  I recommend that everyone read Krugman and John Cochrane, to get a sense of the competing views floating around, and while one can take issue with things Krugman says, he usually isn't this careless.  The gist of his argument, when he gets around to it, is that the downtrodden in states like Mississippi are tricked into voting against their economic interests.  He even mentions... Kansas!  This brings us back to a famous book, to which I will not link:  What's the Matter with Kansas?  (Hey, that title... sounds familiar...)  The argument was that voters in states like Kansas are duped into voting against their self-interest through the use of social issues.

Here's the problem.  It was total bullshit, and it was based on a mathematical error called "ecological inference."  That's some statistical jargon.  It is the term we use for looking at aggregate-level patterns, and trying to infer individual-level behavior.  States that are poorer tend to vote Republican.  Does that mean poor people vote Republican?  No.  That was the hole in the reasoning.

We've known about the "ecological inference problem" for a long time, and Andrew Gelman saw this shit, and wrote Red State, Blue State, Rich State, Poor State to debunk the shit out of this nonsense.  Here's the executive summary.  In rich states, like Connecticut, wealth and partisanship are unrelated.  In poor states, like Mississippi, they are related, and in the expected direction.  Wealthier people are more Republican within the state, but the state itself is shifted more GOP.  So, basically, there is no puzzle to resolve.  You just have to look at individual survey data rather than state-level election results.

The thing is, "ecological inference" is a known thing.  Krugman should recognize an ecological inference claim when he sees it, even if he shouldn't be expected to have read Gelman.

Ecological inference is off-limits.  Journalists?  OK, they don't get stats training.  What's Krugman's excuse?

Tweet

Subscribe to receive free email updates: