Last week, the Trafalgar Group released a poll of Minnesota voters, which found Donald Trump and Joe Biden tied at 47-47% (1141 Likely Voters, margin of error 3%). This poll showed a remarkable shift from a Fox News poll taken in July that found Joe Biden up by 13 percentage points 51% – 38%. Is this new poll a reflection of a dramatic shift in voter sentiment or the difference of two different polls’ methodologies? Trafalgar’s public comments suggest their unique methodology is driving the dramatic shift.
Trafalgar weights its polls to account for “social desirability bias.” This bias is based on the theory that there are thousands of so-called “shy Trump voters” who are too scared or embarrassed to tell a pollster how they feel about Donald Trump. As a result, a portion of the respondents who say they are undecided, will support Joe Biden, or will support a third-party candidate, are in fact Trump voters. To account for this, Trafalgar weights respondents who stated they plan to vote for Donald Trump by more to account for the shy-Trump voters that falsely claimed they won’t vote for him.
As discussed in our course, weighting is a commonly used technique by pollsters to make sure the composition of the respondents surveyed resembles the electorate as much as possible. How a poll is weighted can make a major difference in the results it yields, which in turn can shape how voters and campaigns view an election. The most common factors pollsters weight on are age, race, gender, education, and geographic region. It is less common to weight a poll based on theories about how certain members of the electorate might respond to a poll. There is no clear academic consensus that shy-Trump voters exist in large numbers. The American Association for Public Opinion Research does agree that while some may exist, there are likely not enough to substantively bias a poll.
How a pollster weights a poll can change the reported results, which in turn can dramatically shape how voters and campaigns view an election. The state polls in the Midwest during the 2016 election were wrong because of a weighting error: the polls didn’t weight by educational attainment. While it is hard to examine the merits of certain methodological decisions made by a pollster without being an expert in the field, it is important to understand choices made in constructing a poll and how broadly accepted those choices are by the polling community.
Introduction to Polling
A Guide to the 2020 Election Season—our introductory to polling course is designed to teach you a framework to use so that you can confidently follow the polls through this election season and beyond.View Course