Reverse Bradley Effect and Undecided Voters

Charles Franklin, who taught me statistics at the University of Wisconsin, has a great post addressing the so-called Bradley Effect and undecided voters. Read it here.

Franklin uses empirical data to suggest that the Bradley Effect, if it exists at all, will have a marginal effect in the aggregate. Second, he notes that it appears that undecided voters will break evenly for McCain and Obama.

There's been much speculation about both of these issues. Franklin's analysis actually looks at data to answer the question and is a great example of fine social science.

On the question of undecided voters, I've heard the following:

a) Undecided voters are likely to break for McCain, who is the known quantity.
b) Undecided voters are likely to break for Obama, because they want change.
c) Undecided voters are likely to not vote.

So, which is it?

Context matters in figuring out undecided voters. In the case of a presidential election, voters have enormous amounts of information available to them about both candidates. In this instance, it is hard to see--especially given the amount of money the Obama campaign has spent and McCain's long presence on the national scene--that an undecided voter does not have enough information to make a decision about Obama or McCain.

This is different from your run of the mill congressional election. For example, a congressional election generally features lots of information about the incumbent and relatively little about the challenger. Undecideds, in this situation, are likely to break for who they know best ABSENT mitigating factors. If, however, national events and the challenger's campaign have raised enough questions sufficient to put doubts in the minds of the voter about the incumbent, it is more likely that undecided voters will break for the challenger to express those doubts and concerns.

Undecided voters in a high information contest without an incumbent such as this year's presidential race are likely cross-pressured voters, to use the classic term from one of the earliest studies on voting behavior (Berelson et al 1954). In short, there are issues that pull them in both directions simultaneously. A good example might be a voter who is concerned about the economy and national security. Economically, perhaps they like Obama's plans better, but McCain's experience on defense matters pulls them in the opposite direction.

Franklin's data suggests that undecided voters will split their vote on Election Day between the two candidates. That is probably the case. But also important to remember is those who are cross-pressured face a difficult decision which is cognitively unpleasant. A portion of those voters are just as likely to stay at home than make the difficult choice.

To summarize, I do not anticipate McCain picking up undecideds by two to one, as his campaign suggests. If he does, then he's got a shot on Tuesday.