Mark Blumenthal | March 2, 2008
Topics: 2008 , ARG , Barack Obama , Divergent Polls , Hillary Clinton , Mason-Dixon , Mike McDonald , Pollsters , PPP , Quinnipiac , SurveyUSA , Washington Post
Last week, I emailed questions about poll demographics to all pollsters that had fielded recent surveys in Ohio. Fortunately, those requests were less necessary than in previous contests, as more and more pollsters have been including demographic composition data in their releases. A thank you is in order, however, to the pollsters at the Washington Post, the University of Cincinnati, Quinnipiac University and Public Policy Polling that shared data not already in the public domain.
The racial composition of the Ohio Democratic electorate is less of a puzzle than in Texas, if only because Ohio's Latino population is relatively small (amounting to 3% or less on the various polls that reported it). Still, the surveys show meaningful variation in their African American composition from a low of 12% on the University of Cincinnati "Ohio Poll" to a high of 22% on today's new poll from the Cleveland Plain Dealer and Mason-Dixon Polling and Research. With Barack Obama winning the overwhelming majority of black voters, differences of a few percentage points can have a significant affect on vote preference. If African Americans had been 16% of the respondents instead of 22% in the PD/Mason-Dixon poll, Clinton would lead by roughly 10 rather than 4 percentage points.
The variation in age has been bigger, (although again, age comparisons are more difficult because pollsters are inconsistent about the age breaks they use). The percentage of 18-to-45-year-olds varies from a low of 26% by the University of Cinninnati and 28% by Quinnipiac to a high of 46% by SurveyUSA. The percentage of 18-to-50-year-olds varies from a low of 44% in today's PD/Mason-Dixon poll to highs of 57% and 60% in the most recent surveys from ARG.
As with Texas, I have included comparable numbers from the 2004 exit poll (based on final data from the Roper Center Archives), although the "right" answer for this year will be unknowable until all the votes are cast and this year's exit poll is available.
Also, as in Texas, I asked pollsters to estimate the percentage of Ohio adults represented by their samples (unless they included the necessary data in their releases). This statistic is a rough measure of how tightly they screened for likely voters. Ohio's Democratic primary drew 12% of eligible adults in 2000 and 15% in 2004 (as per Michael McDonald's turnout page). How high will it go on Tuesday? The Ohio Secretary of State is predicting a total turnout (for Democrats and Republicans) of roughly four million voters, representing a 60-70% increase compared to the last two presidential elections.
With that in mind, consider the percentage of adults that some of these surveys represent. The number of polls included in the table below is smaller, because fewer pollsters were willing or able to provide an estimate. Obviously, the percentages of adults sampled are much higher than previous turnouts, higher than even the optimistic projection from the Ohio Secretary of state. They vary from a low of 27% in the PPP survey to a high of 40% in the most recent poll by SurveyUSA.
What all of this means is that polls are in disagreement about who will vote in Tuesday's primary, and that uncertain composition will likely determine the winner. The polls we have before us can tell us a great deal about how preferences differ across the key demographic and regional groups, but the tools of survey research are simply not powerful enough to predict who will vote with great precision. I'll have more thoughts on this issue after we see the final round of surveys tomorrow.