“Tough choices in election forecasting: All the things that can go wrong” (my webinar this Friday 11am with the Washington Statistical Society)

Friday, October 11th, 2024, 11:00 A.M. – 12:00 P.M. Eastern Time, Remote Only The seminar talk is live-streamed. Please register to receive the link.

Tough choices in election forecasting: All the things that can go wrong

Andrew Gelman, Department of Statistics and Department of Political Science, Columbia University

Different election forecasts use different models but yield similar predictions, which is no surprise as they’re using the same information: state polls, national polls, and what we call the “fundamentals”: economic conditions, presidential popularity, incumbency, and the relative positions of the states in previous elections. And when the forecast is close to 50/50, which it is at the time of this writing, it’s hard to be called out. Back in the day, it was considered impressive for a forecast to predict 45 or more states correctly, but now everyone knows which are the key swing states, and it’s not hard to come up with a reasonable state-by-state forecast with historically-calibrated uncertainty. Nonetheless, there are lots of ways a forecast can go wrong, and lots of choices involved in modeling, data inclusion, testing and validation of the forecasting procedure, and communication of results. We discuss this in the context of our efforts in election forecasts from 1992 to the present day.

At the beginning of the session, before the talk, I’ll be interviewed by Edward Wu, who has a Ph.D. in statistics and is currently the Data, Polling, and Election Analytics Senior Producer at CNN.

I was told that CNN doesn’t allow questions to be prepared in advance, which is actually fine with me because I prefer to be surprised!

If you in the audience want to be prepared, I recommend my recent article, Grappling with uncertainty in forecasting the 2024 U.S. presidential election (with Ben Goodrich and Geonhee Han), my article from 2020, Information, incentives, and goals in election forecasts (with Jessica Hullman, Chris Wlezien, and Elliott Morris), and my article from 1993, Why are American Presidential election campaign polls so variable when votes are so predictable? (with Gary King).