Julian Sanchez explains:
Ideally, professional pollsters have no particular agenda beyond accurately forecasting the outcome of a race. But pundits are trying to influence outcomes, and forecasts don’t just predict outcomes, but at least partially help to determine them. There’s plenty of social psychology literature showing bandwagon effects in elections: Voters on the fence often pick the candidate they expect to triumph anyway, because it’s nice to be on the winning side. Campaign workers become demoralized if they think they’re laboring those long hours for a hopeless cause. A 20 percent chance of victory is still a chance, after all, and you don’t want people throwing in the towel prematurely.
Here as in many areas of life, when the odds are heavily against you, being a perfectly accurate assessor of your chances can actually make the odds worse. If you are rational, you will want to have some irrational beliefs. So I don’t expect supporters of a candidate who’s unlikely to win on election eve to acknowledge this, any more than I expect the coach of an underdog team to deliver out an honest read of the stats as a pre-game pep talk. We don’t make fun of coaches for this, because we all understand they’re engaged in a bit of socially appropriate bullshitting.