A Preference Isn't a Decision

My Pollster.com colleague David Moore is right that the standard vote choice question typically used by pollsters deliberately pushes voters to make a decision, when their level of certainty about how they will actually vote is somewhat less than complete. From George Gallup on down, pollsters have learned that pushing harder, especially in the last few days before an election, usually produces a more accurate forecast.
Unfortunately, media accounts often lose sight of the fact that the preference expressed on surveys is not set in stone, and may still change. A preference is often not a final decision. We all learned this lesson the hard way earlier this year in New Hampshire. When David and his colleagues at the University of New Hampshire did their final survey before the primary, they had Barack Obama leading by nine percentage points (39% to 30%). Nearly all of the Democratic primary voters (94%) expressed a preference for a candidate (i.e only 6% were ("undecided"), but only 53% described themselves as "completely decided" and one in five (21%) said they had "considered some candidates but are still trying to decide.
We all know what happened next. Most of us remember that Obama led on virtually every poll conducted just before the New Hampshire primary, but how many of us knew about all that uncertainty there in the final days before the election?
I used to be a campaign pollster -- someone hired by Democratic candidates to conduct polls for their campaign. One point I'd to add from that perspective is that virtually all campaign pollsters do something analogous to what David is proposing. They ask the standard vote question that presses respondents hard to make a choice. Then they follow-up with a question asking how certain they are to support that candidate on election day.
Campaign pollsters focus much more on the measure of certainty than on "undecided" in thinking about targeting messages and organization resources, because they have learned though experience that voters with a preference can sometimes change their minds.
















how do pollsters take early voting into account?
the results in nh on the day of voting may very well have reflected the polling acurately. how many people in nh voted before obama won in iowa?
October 21, 2008 12:15 AM | Reply | Permalink