Call It 'The Obama Effect'
Why undecided voters will swing to McCain.
by Arnon A. Mishkin
10/27/2008 12:00:00 AM
http://www.weeklystandard.com/As Election Day draws near, people are wondering if the presidential race will tighten. Will the undecideds swing to McCain, or will Obama continue to maintain his 4 to 11 point lead?
Some point to a "Bradley effect" suggesting that voters are hiding their true feelings from pollsters because of Obama's race, while others say the Bradley effect either never existed or no longer exists. People who think there is a Bradley effect believe that the substantial majority of undecideds are likely to vote for McCain, enabling him to close some of the gap.
McCain should win a larger share of undecided voters than Obama, but it has little to do with race.
With Obama outspending McCain by upwards of 4 to 1, getting enormous traction with newspaper editorial boards, generating the enthusiasm to bring out crowds measured in the tens of thousands, and with Palin treated as more of a punch line than a candidate by the press--it seems likely that if voters are not ready to tell a pollster that they are with Obama, they are unlikely to get there.
But the phenomenon of undecided voters' breaking for McCain need not be called the "Bradley effect." Call it the "Bloomberg effect"--where after $100 million of spending, his mayoral challenger was able to capture essentially all of the 10 point undecided vote. Or call it the "Clinton effect"--where almost all the undecided vote swung away from the popular incumbent and went to Bob Dole. Or call it the "Reagan effect"--where even during the Republican 1980 primaries, voters were apparently reluctant to say they were going to vote for the "elderly washed up actor" and he got the preponderance of the undecided vote.
They all amount to essentially the same pattern. Call it "the Social Effect." Where there is a perception that there is a "socially acceptable" choice, respondents who do not articulate it, are likely not to agree with it. Are they lying? Or just genuinely torn about taking that route or another? I am not going to psychoanalyze what is going on in their heads, but in the end, the pattern tends to be that those undecided voters vote against that "socially acceptable" choice.
In fact, we saw a preview of this during the Democratic primaries this year. Typically, Hillary Clinton won substantial majorities of all late deciders (those who decided in the last three days of the primary)--i.e. Obama tended to lose the "undecided vote."
At the same time, there were examples where Obama outperformed his final poll numbers--even though Clinton was winning the late deciders. What seems to have happened were two effects that had opposite impacts on polling accuracy. It seems they amount to "The Obama Effect:"
1. The "Social Effect" where "undecided" voters were not going to vote for Obama, and
2. Lopsided Enthusiasm, or an Enthusiasm Effect, leading pollsters to underestimate the turnout of Obama supporters--e.g. African Americans, the young, and the more independent voters.
To test this, I correlated the difference between pre-primary poll estimates of Obama and his actual vote against the relative size of African-American share of Democratic primary vote.
I am using the African-American share as a proxy for the size of the Obama-enthusiast group. Obviously, there are other groups that tended to be very enthusiastic for Obama, notably the young, but the population of young people is much more evenly distributed than the population of African Americans, and thus the African-American share is a fairly good proxy for the relative size of the total number of pro-Obama enthusiasts. I used the Real Clear Politics final average going into the primary as the pre-primary poll estimate.
What is surprising is how accurately the percent of African Americans correlates with the error on the polls (the actual vote minus the RCP average of polls). In a regression, I found that fully 89 percent of the difference between Obama's actual vote and his final poll average could be explained by percent of African Americans among the state primary voters.
The analysis suggests that both effects were at work during the primaries. When one controls for the size of the Obama-enthusiast population, Obama basically gets what the pre-primary polls said he would get ("The Social Effect"), but when one looks at the actual result, it is clear that it is highly dependent on the size of the Obama-enthusiast population.
For pollsters, it suggests the need to get a better handle on measuring likely turnout among groups of highly enthusiastic voters, especially if they come from groups that have not turned out in high numbers in the past. Historical "likely voter" screens which focus on a respondent's voting history/knowledge, may be very ineffective this year. It suggests that pollsters should be more open about what kinds of voters--especially first time voters--are in their likely voter screens.
It also suggests that as the election winds down, one should look less at the difference between Obama and McCain, and more at the actual number that Obama is getting in the polls.
Arnon Mishkin is a management consultant and partner with Mitchell Madison Group. He is on the Fox News Decision Team.
© Copyright 2008, News Corporation, Weekly Standard, All Rights Reserved.
