How election models factor in high turnout

Pollsters' likely voter screens can be flexible enough to pick up changing turnout patterns

A month or so ago, a reader named Paul wrote to me:

Hi Elliott,

Recently I've noticed differences in polls of the same state, and I have a quick question about likely voter polling: Do you know if anyone has compared the LV screens and methods used by different public pollsters?

- Paul

This is a really good question! And answering it also has the bonus of providing me with the opportunity to explain how election forecasting models can factor in the high turnout we’re seeing in early voting data. But let’s take this one step at a time.

First, let me briefly touch on how a poll of likely voters even works. There are two main ways that pollsters get samples that are representative of the likely electorate. The primary method is to simply ask people if they’re going to turn out and then subset the poll to respondents who say yes. But this is, of course, only a crude approximation. Because we know that people are prone to overestimate their likelihood of eventually casting a ballot, many polls…