Overconfident Pundits Predicting Victory for Their Side

Given the extreme closeness of both the national and battleground state polls, one would think that serious political commentators would avoid making bold predictions about the outcome of the presidential race. After all, an overconfident pundit who turns out to be wrong will have egg on their face in just a few days. This is especially true in a situation where state polls and national polls seem to be in tension with each other.

Yet one of the striking things about recent election commentary is that most conservative Republicans are confidently predicting a Romney victory, while liberal Democrats seem equally convinced that Obama is sure to win. Karl Rove, for example, is predicting a clear Romney win. Liberals such as Joan Walsh and Mark Mellman are just as confident that Romney is doomed.

What explains such seemingly irrational overconfidence? One possibility is that these people are simply engaging in biased wishful thinking. Like sports fans, committed political partisans tend to overvalue evidence that reflects favorably on their preferred “team” and ignore or downplay anything that cuts the other way. But another factor may be the desire to create a “bandwagon effect” by convincing as many people as possible that their candidate will win. As I explained here, a small number of swing voters will tend to gravitate to the side that looks like it’s going to win. In a close election, they could make a decisive difference. If the Roves and Mellmans of the world can persuade the public that their guy has the momentum and is likely to win, it could turn out to be a self-fulfilling prophecy.

A recent Gallup poll shows that 54 percent of the public believe that Obama will win, compared to only 34 percent who predict that Romney will prevail. That certainly does not prove that Obama really will win. It’s possible that Romney will amass enough support to offset the bandwagon effect. But the public’s perception that Obama is the likely winner does give him an edge.