The attack on election forecasting straw men

I agree (again) with Brendan Nyhan here.

As I wrote recently, I’m thrilled that, in the words of Michael Tomasky, “the conventional wisdom in the political press is largely driven by the political-science theory of presidential elections and economic determinism: that is, that the results of presidential elections are pretty much strictly a function of economic conditions, and if those are bad (defined by various measures, chiefly the jobless and growth rates), the incumbent will lose.”

When Gary King and I were working on our project that resulted in our 1993 article, “Why are American Presidential election campaign polls so variable when votes are so predictable?” the conventional wisdom in the political press was that elections were driven by campaign events, debates, charisma, etc. The new conventional wisdom is much better!

By “pretty much strictly,” I assume Tomasky is referring to the prediction error from a regression model. The prediction can be off by a few percentage points in either direction.

We don’t know what economic conditions will be next year, hence the prediction model can be just fine and there is still uncertainty about the election.

As Gary and I discussed in our paper, approximate economic determinism is consistent with the campaign mattering. The efforts of the two campaigns help to push voters toward the choices that would be expected based on the fundamentals. This argument works because the general election has two major candidates who clearly differ in ideology.

P.S. No, I’m not talking about dumb models like this one.

2 Responses to The attack on election forecasting straw men

  1. Alan T November 17, 2011 at 4:34 am #


    I would love to read your thoughts on Nate Silver’s post, “A Radical Centrist View of Election Forecasting.” Do you have anything to add to Brendan Nyhan’s comments?

    • Andrew Gelman November 17, 2011 at 9:13 am #


      I think Brendan said it well.

      I don’t know what to think about Nate’s piece. He writes that “political scientists as a group badly overestimate how accurately they can forecast elections from economic variables alone,” but I have no idea how to evaluate what “political scientists as a group” are thinking. I don’t think that Brendan or I or my co-bloggers here do this. I agree that overfitting can be a problem in forecasting and in statistical estimation more generally (see here, for example), but I don’t think that invalidates the work of Rosenstone, Hibbs, etc.

      Most of what Nate says is consistent with Rosenstone’s 1983 book on forecasting presidential elections and with my paper with King on why campaign polls are so variable. As the election comes up, there will be a temptation for journalists to jump at every movement in the polls. Such movements can tell us a lot during primary election season, but once the general election comes, I think most of the polling is a waste of time.

      P.S. I was amused by Nate’s expression, “debating the number of angels that can dance on a pinhead.” I’ve always heard the phrase, “head of a pin.” “Pinhead” makes me think of this guy. I wonder if the change from “head of a pin” to “pinhead” was made by an overactive copy editor?