Monday, August 11, 2008

Reverse Bradely Effect

538 points out that Barack Obama actually outperforms polls on election day. Typically, political campiagns expect minority candidates to under-preform polls because there is a sliver of voters who harbor some racists sentiments but are not willing to admit those sentiments to pollsters. The phenomenon is know as the Bradley effect in reference to Tom Bradely, a black Los Angeles mayor, who lost the race for Governor of California despite a lead in the polls going into election night, and projections by media outlets that he had won.

Obama on the other hand seems to do better than his polls would suggest. My guess is that both the orginal Bradley effect and the new Obama effect don't have as much to do with lieing but with emotionality at the voting booth.

In Obama's case he outperforms in areas with high black populations.

In those areas there is likely some fraction of the black population that on election night was seriously considering voting for Hillary Clinton. However, once inside the voting booth, the emotiona force of a potential black president overwhelmed their issue based calculation and they cast a vote for Obama.

Remember that voting itself is not rational unless there is some emotional reward for doing so. It is highly unlikely that any single voter will swing the election. Therefore, voters vote more for emotional reasons: civic duty, patriotism, boredom, the right to through it non-voters faces, etc.

An interesting study, I believe, would be testing whether or not there is a general Obama Effect, that the candidate who has more emotional connection outperforms his polls.

A simple test might be whether or not likability ratings predict overperformance. We know that more likable canididate, get more votes. However, are they more likely to get even more votes than they are forecast to get?