Who Says No? A Study of Non-Cooperation in 2016 Pre-Election Polls
Citation: Khanna, K., Salvanto, A., Ben-Poreth, E., & Willsie, B. (2018, May). Who Says No? A Study of Non-Cooperation in 2016 Pre-Election Polls. Paper presented at the American Association for Public Opinion Research Annual Conference, Denver, CO.
Abstract: Although national pre-election polls were, by and large, accurate in 2016, many have speculated that polls suffer from differential refusal rates between partisans. Did Republicans refuse to answer polls at a greater rate than Democrats did, once they knew it was a political pollster calling them? In this study, we offer a straightforward test using a series of random-digit-dial CBS News polls conducted prior to the 2016 presidential election. Taking the RDD-generated phone numbers of households that received the request to take the survey but refused to participate, we match these numbers to those that appear in voter registration lists. We then compare the demographic and political characteristics of households that refused and those that completed surveys. Our preliminary results confirm our expectations that, at least among registered voters, completes are older, whiter, and more politically engaged (as measured by voting history) than the refusals. However, we do not find significant partisan bias beyond the factors that typically contribute to nonresponse. These results may help to mitigate concerns that people of one political stripe systematically opt out of political polls. We also offer a discussion of which types of voters are more likely to be able to matched to voter lists, which types of households simply do not answer, as well as the implications for this and similar studies. Our findings have broad implications for survey practitioners and methodologists alike.
