Member Newsletter
One thing we’ve talked about a lot this year in the Backchannel and the podcast is changes pollsters have made to their methodologies over recent years, in large part because of 2016 and 2020 polling errors tied to Trump. Kyle Kondick, of Larry Sabato’s Crystal Ball, posted two good links on this that I wanted to share with you. The first is this short interview with Professor Charles Franklin of Marquette Law School who runs what is generally considered the signature in-state poll in Wisconsin and one of the most reliable nationwide. (Some of you may remember that Franklin was our polling methodology advisor back in the days of TPMPollTracker.) Then there’s this short article which goes over the changes industry-wide.
Trump-related polling errors aren’t the only driver of this. There’s also a move toward hybrid use of phones and online questionnaires that is driven in part by changes in behavior and changes in technology. But the big thing is weighting samples for demographic features pollsters mostly didn’t use to weight for. The biggest one is education level which, as most of us know, has become a core element of political polarization in the country. But it’s not just that.
The key theory in Trump-driven polling errors isn’t “shy” Trump voters. The thinking is that Trump tends to draw support from people less plugged into politics and more suspicious of major institutions in the dominant culture. So things like the mainstream media and pollsters. In other words, Trumpers are a bit less likely to take polls. And even a small difference in those percentages can have a big impact on poll results. So the general effort is to find new things to weight for to remove this under-sampling from the results. There’s party identification. There’s who people say they voted for in the last election. In Franklin’s case, moving to partly online questionnaires meant working from a voter file. That gives you among other things more geographical clarity about where a voter lives. And, as we know, another key element of political polarization is geographical — rural areas vs cities. So that’s another part of the equation.
There are many more details. And the interview and article I linked above are good places to start.