Social Desirability Trumping the Polls

• In Event: Poster Session Part II: Elections, Public Opinion, and Voting Behavior

Abstract

The election of Donald Trump as the 45th President of the United States took many by surprise. Even the day before the election many pundits, pollsters and political analysist were confident that Hillary Clinton would win since polling data throughout the election consistently showed Clinton ahead of Trump. How was this possible? This paper begins with the idea that Trump ran an unconventional campaign which violated established norms. Trump mocked a disabled reporter, encouraged violence among his supporters, engaged in numerous public feuds and much more. As a result, Trump’s norm violations made public support for him socially undesirable for some people. Social desirability can leave some respondents hesitant to admit their actual positions, leading them to hide behind more socially acceptable answers or don’t know (DK) responses (Berinsky, 2004; Klar & Krupnikov, 2016). This paper argues that social desirability concerns systematically biased 2016 election polls by encouraging some respondents to hide their support for Donald Trump. Using data from 179 polls from 22 organizations and 2 daily tracking polls this study demonstrates that higher rates of DK responses in election polls disproportionately cost Donald Trump support. Furthermore, DK responses were connected to social desirability factors of the polls and the electoral context. Taken together the evidence indicates that social desirability systematically biased polls in the 2016 Presidential election.

Author

• Tracy Goodwin, Stony Brook University