Check Your Turnout Model

By Adam Probolsky

Knowing what questions to ask in an election is the easiest part of political opinion research; it’s modeling for turnout that’s hard.

Pollsters strive to develop a turnout model that accurately reflects the profile of who will actually cast a ballot in an upcoming election. If it sounds like common sense, it is.  But developing a turnout model requires careful examination of voter history in the political sub-division you are polling, and every pollster has their own approach.

Some pollsters simply ask voters how likely they are to vote in a coming election. While this seems like the obvious way to determine likelihood of participation, it does not capture true turnout. The reason: voters don’t like to admit failing in their duty to fulfill the social contract of voting.  Respondents claim they are more likely to participate than they actually are – especially if they are younger.

The same issues apply to pollsters who model their turnout predictions using other self-report data, such as ethnicity. Ethnicity is fast becoming a subjective metric. Is the voter whose mother is black and father is Hispanic, black or Hispanic? How about the voter who has an Asian father and white mother? Past exit polling results are one of the worst predictors of ethnic turnout and should be avoided entirely. Stereo-types that drove campaigning to ethnic voters in the past are largely obsolete, as is relying on this data for developing turnout models.

Our approach models turnout predictions based on as much fact as possible. Looking at voter registration rolls (maintained – in California –  by county election officials) for historical turnout is a uniquely effective approach to predict future turnout. These files not only show the percentage of voter turnout for past elections that may be similar to an upcoming election, but they also include profiles of the demographic composition of election turnouts, including age, party, gender and more that can be used to develop an accurate prediction of who will return their vote-by-mail ballot or show up at the polls on Election Day.

By getting it right, a campaign knows the relative importance of key messages, targetable voting blocs (and how receptive they are to alternative messaging) and can direct scarce dollars accordingly.

Adam Probolsky is president of Probolsky Research