NATURE DOESN’T RUN ENOUGH ELECTIONS. Arnold Kling’s meta-interpretation appeals to me because he uses an analogy to a problem that I have posted on before—that in economics—“Nature doesn’t run very good experiments.” (See this post from 2007, for example). There are almost always a number of possible explanations for an event, and it is difficult to disentangle which are the important ones.
Kling gives the example of the financial crisis: “After the financial crisis, it was remarkable how many economists found their world view confirmed by it.” Now, he predicts, everybody is going to use the 2016 election to prove their favorite theory.
Kling writes: “If we are looking at singular events, like the financial crisis, we have no chance of definitively sorting them out. When we look at macroeconomics in general, too many factors change to enable us to draw firm conclusions.” Similarly, we have only one election, far short of the right kind of elections to test which of the possible explanations of the 2016 election are correct.
There is a burgeoning industry of statistical “experts” attempting to use data to make predictions with presidential elections. You also hear talking heads recite stats about “X state hasn’t done Y since 1976” or something like that.
But oddly, none of these people talk about the fact that by the time you get a meaningful sample of presidential elections so much time has passed that it’s ludicrous to compare them. 40 presidential elections ago it was 1856 and the Civil War was looming.