MRP at NYT

Coincidentally, or maybe not, on the same day that our New York Times contact informed me about their new job opening, Yair points us to this new feature at the NYT website, “Times Voter Trends” at the NYT. Yair writes:

It’s a voter-file-based estimate of how every demographic group voted in every state, with survey models poststratified to the full file and calibrated to precinct results. It’s basically the approach we’ve been running at Catalist for years, which of course grew out of the MRP work you and I did together. Here’s a description of the methodology. They use gradient boosting instead of a multilevel model [what we’ve called Regularized Prediction and Poststratification — AG], but the overall structure is very similar. It’s great to see these methods going increasingly mainstream.

Here are the relevant research papers:

[2013] Deep interactions with MRP: Election turnout and voting patterns among small electoral subgroups. American Journal of Political Science 57, 762-776. (Yair Ghitza and Andrew Gelman)

[2020] Voter registration databases and MRP: Toward the use of [2020] Voter registration databases and MRP: Toward the use of large scale databases in public opinion research. databases in public opinion research. Political Analysis 28, 507-531. (Yair Ghitza and Andrew Gelman)

My Ph.D. adviser once told us that the true sign of success is when people use your methods without even citing you. That’s when you know the ideas are fully in circulation.