Surveys always ask respondents who they're going to vote for, and sometimes surveys ask who respondents think is going to win. I'm big on that latter question. I've written about it far too often.
What's new? An AP-GFK poll just out puts the race as a statistical dead heat. But when respondents were asked whether Obama will get re-elected -- the prediction question -- we see a very different result. Below, the results by date, earliest to most recent:
Will Barack Obama be ...
Dec 2011 June 2012 August 2012
re-elected 49% 56% 58%
voted out 48% 35% 32%
don't know 4% 9% 11%
no answer * * *
The * for "no answer" means it's too small to round to 1. There are some rounding errors in the table above. Ignore all that. Pay attention to the steady improvement (if you're an Obama fan) in the predictions of him getting re-elected and the steady decrease in those predicting he will be "voted out of office." That last category seems to be people shifting from thinking Obama won't be re-elected to "don't know" in the survey. That's kinda interesting and I don't have a good explanation for that, other than it suggests doubts about Mitt Romney.
People tend to believe their own preferred candidate will win. That's called wishful thinking in the social science literature, and yes I've published research on the topic and even delivered a paper this summer that examined the connection between who people are for and who they think will win from 1952 to 2008. I really should sit and write a version of it for a site better than this blog. Yo, Politico, got space to fill? In some years people are more wishful, and therefore less accurate, than others. I'll save the details for another time, another post, or elsewhere.