Reluctant Republicans, Eager Democrats? Partisan Nonresponse and the Accuracy of 2020 Presidential Pre-election Telephone Polls

Using the registration-based samples and disposition codes of state-level pre-election telephone polls conducted by the National Election Pool as part of the National Exit Poll in 12 states, we test whether likely Democrats were more likely to cooperate with the National Exit Poll than likely Republicans and independents. Using information about both respondents and nonrespondents, we find that Democrats are more likely to cooperate with telephone interviewers than Republicans and independents by 3 and 6 percentage points, respectively, even after controlling for individual and geographic features plausibly related to nonresponse (e.g., age, gender, race, urban/rural, community support for President Trump, and effects of COVID-19). Equalizing the partisan cooperation rate when post-stratifying to account for the partisan differences in cooperation decreases the average polling error on the margin of victory by 4 percentage points in the polls we examine, but sizable errors remain in critical swing states because of within-party differences in who responds and/or errors in the available partisanship measures in the voter file.