Thursday, October 27, 2022

The Nonresponse Problem: What else are we missing?

The shy Trump voter hypothesis does not, by itself, do a good job explaining what we've been seeing in polling data for the past six years. If you are looking for a single theory that will explain large polling biases going in opposite directions in Tennessee and California, you will have to look elsewhere. On the other hand, it is entirely possible that under-response by Trump supporters is a part of the story, and if that is the case, political analysts and data journalists have completely missed perhaps the most disturbing implication.

If we start with the standard assumption of the shy Trump voter hypothesis, that these people tend to distrust and dislike interacting with mainstream media, then it's not just electoral polls that are being affected. All recent polling of Republican attitudes will not just be biased but will be far more biased than the electoral polls (Trump Voters/Registered Voters < Trump Voters/GOP voters).

We would expect the shy (or more specifically mainstream media averse) Trump voter to be, for starters, more likely to be a Trump loyalist rather than just a Republican, and if they are driven by distrust of and hostility toward non-right wing media, we would also expect them to be more likely to be election deniers, more likely to support the January 6th insurrection, more likely to be MAGA, more likely to be Qanon, and more likely to espouse ideas like white nationalism. This would mean that polling has seriously underestimated the support for all of these things within the GOP.  

Of course. We don’t know that the under-responding Republicans are driven by feelings about the MSM. It is perhaps just as likely that the self-selection bias is driven by embarrassment, traditional Republicans who don’t approve of Trump or the QAnon movement but who would not consider supporting a Democrat. 

This second theory might make us feel better but its implications are also troubling. Under the first, we underestimate the danger to the country from these extremists; under the second,  we exaggerate their numbers which helps them entrench their hold over the GOP.

We don’t know if we’re seeing the first, the second, both, or something else, but we can be reasonably confident that we are seeing self-selection biases in recent political polls and there is absolutely no reason to assume those errors stop with the “who are you voting for” section of the polls.

No comments:

Post a Comment