Why People Distrust Polls: Perceiving respondent dishonesty is a far more robust predictor of distrust in poll than motivated reasoning

Lees, Jeffrey M., Barry Lam, Sara Purinton, and Daniel Wodak. 2020. “Why People Distrust Polls: Meta-cognitive Reasoning, Not Motivated Reasoning.” OSF Preprints. November 23. doi:10.31219/osf.io/53hc7

Rolf Degen's take: https://twitter.com/DegenRolf/status/1331118327343820800

Abstract: Many pollsters and news organizations have expressed growing concerns over public distrust in the accuracy of polls, especially in light of widespread perceived polling failures during the 2016 U.S. Presidential Election. Scholarly investigations into distrust in polls have highlighted motivated reasoning as a source of distrust, where individuals reject or accept polls based on whether their findings conflict with personally held attitudes and desired outcomes. Yet a separate, domain-general explanation for distrust in polls exists that may outweigh the effects of motivated reasoning: meta-cognitive reasoning (i.e., attitudes about the thoughts, attitudes, and motives of others). Here we investigate two distinct domains of meta-cognitive reasoning which may better explain distrust in polls than motivated reasoning: beliefs about whether the pollsters who conducted a poll are biased, and about whether poll respondents are honest in their responses. To examine this hypothesis we utilized a repeated-measures survey with a nationally-representative sample of Americans (N[Observations]=3510, N[Participants]=351) conducted five days prior to the 2020 U.S. Presidential Election. Participants viewed a stimuli set of 56 polls, pretested to be widely distributed across liberal-conservative favorability in results while neutral in aggregate, and responded to measures of motivated reasoning, meta-cognitive reasoning, and political ideology/affiliation. Utilizing mixed-effects modeling to allow for generalizable inferences, we find that meta-cognitive concerns regarding whether pollsters are (un)biased and poll respondents are (dis)honest are much stronger predictors of (dis)trust in polls than motivated reasoning or political ideology/affiliation. Our results are the first to demonstrate that perceiving respondent dishonesty is a far more robust predictor of distrust in poll than motivated reasoning. In general, meta-cognitive reasoning about pollsters and the people polled is the best explanation for why individuals distrust polls.