It's an interesting fact of political life that campaigns and their staffers/volunteers are quite bad at predicting whether or not they will actually win a given race. To give one salient example, despite the fact that Romney's internal campaign polling told them they would likely lose, they still believed they would win. Indeed, reports were that the campaign team was "shell-shocked" after the loss.
A pair of political scientists explain, campaigns are generally over-confident and over-predict the final outcome in a loss.
The researchers also point out that the northeastern quadrant is filled with campaigns who though they would win, but lost.
Equally interesting -- but unsurprising if you think about it -- is that campaign volunteers are especially bad at predicting a "win." This isn't surprising because campaign volunteers are most likely to be very ideologically committed to the candidate's party and cannot brook the idea that they may have wasted a bunch of their free time working on a losing campaign. (N.b., the researchers compared presidential campaigns, House races and state elections races; the further down the food chain one goes, the worse, generally speaking, the predictions were).
As the authors note:
Consistent with our expectations, incumbent campaigns are both more accurate and more pessimistic at predicting their election outcome than open-race or challenger campaigns. We have also shown that races that are competitive and for higher offices are generally associated with better predictions. We have shown the added nuance that staffers are substantially better predictors than volunteers, but only in the Presidential campaign.
And why does any of this matter, you ask? Because if elections are close then competition among politicians makes it more likely they will be responsive voters' wants. If a race is not close -- a politician knows he's going to win -- why bother with the effort in determining voters' preferences.
Political scientist John Sides, commenting on this research, cogently states:
Here's why I think this finding is important.Typically speaking, I discount what campaigns say about their chances of winning because I assume that (a) they will selectively release polls to manipulate commentators and reporters or (b) they'll lie if the data don't look good.
The bad news: there is probably not a whole lot that can be done about this in the near future. Politicians are humans and they are suffer from cognitive dissonance and confirmation bias just as much as we do.