President Obama is currently leading by as much as 4-5 points in the polls if attention were restricted to registered voters. Yet, polls based on the subsample of “likely” voters show a different story, with Mitt Romney equal or slightly in the lead. How sure are we that these “likely voter” corrections adequately assess who will turn out on November 6?
One problem is that people often misrepresent whether they voted. The people who feel the need to give the socially desirable answer in surveys may differ systematically from those who don’t. In a new article in Political Analysis (ungated version here) Stephen Ansolabehere (Harvard) and Eitan Hersh (Yale) validate what survey respondents in all 5o states said they did against public records from the 2008 elections. Their key conclusion is that:
[..] standard predictors of participation, like demographics and measures of partisanship and political engagement, explain a third to a half as much about voting participation as one would find from analyzing behavior reported by survey respondents.
This finding occurs because:
Well-educated, high-income partisans who are engaged in public affairs, attend church regularly, and have lived in the community for a while are the kinds of people who misreport their vote experience.
So: surveys exaggerate the differences between the characteristics of non-voters and voters. Yes, non-voters are less educated than voters but education is not nearly as important a correlate of actual turnout decisions as it is of respondents’ answers to the question whether they voted (and presumably whether they intend to vote).
This may cause trouble for “likely voter” corrections. If I understand it correctly, “likely voter” corrections are usually based on what Ansolabehere and Hersh call the “standard predictors of participation” and/or self-reported voting intentions or past behavior. If Ansohalabehere and Hersh’s findings generalize, then the difference between likely voters and registered voters may be smaller that currently reported, which could be good news for Obama. This conclusion relies on some additional assumptions and admittedly a cursory understanding of the actual likely voter corrections that pollsters use. Am I wrong?