[Posted by Karl]
To quote CBS News political director John Dickerson, right now there are more polls than in a Warsaw bar. For political news junkies, Nate Silver offers helpful advice on poll-watching. I don’t necessarily agree with every point (his attacks on forecasting models are not only misguided, but downright funny coming from Silver, who built a model that is worse than those he criticizes), but overall I recommend reading the whole thing.
The main point Silver misses — perhaps because poll analysis is his blogging bread and butter — is that head-to-head polls at this point in election cycle explain less than 50% of eventual results.
This early in the cycle, I would prefer to follow Pres. Obama’s job approval number. Sean Trende notes the longstanding correlation between the incumbent’s job approval rating and the vote share ultimately received on Election Day. He also notes the close correlation in the 2012 campaign so far:
Since January of this year, the president’s share of the vote against Romney has been, on average, within .55 points of his job approval in the RCP Average on any given day (the median is .5 points). There has only been one day, back on Jan. 10, where he ran more than two points ahead of his job approval. This tendency translates to individual polls as well. ***
On average, Obama runs .93 percent ahead of his job approval. We also might note that there seems to be some systemic bias in Pew that has the president running unusually well (vis a vis other pollsters) compared to his job approval. If you assume that something in Pew’s methodology renders it an outlier, the president would run .42 percent ahead of his job approval.
Thus, this early in the cycle, rather than fret over each poll that comes in, a good number to consider would be Obama’s average job approval +.5 percent.
However, it’s also possible that number could be a ceiling at any given moment. Harry J. Enten has a very nice piece extending Silver’s final point about the small number of elections usually studied (16) makes the notion of “rules” about elections a risky enterprise. Read the whole thing, because there’s plenty of good stuff there beyond what he says about presidential job approval:
Approval ratings are great at predicting winners, but they are inexact. The perceived ideology of an opponent may not play the biggest role in determining the winner, but it does play some role. Also, these historic approval polls are of adults generally, not actual voters – who tend to be somewhat more Republican.
In this particular cycle, Mitt Romney’s moderate image probably helps him by a percentage point or two (which is why Team Obama is trying to paint him as the most right-wing candidate since Barry Goldwater). And the fact that the electorate trends more GOP than the general population also helps Romney. Sean Trende is correct in claiming Obama can win if he boosts his job approval by a couple of points, which is entirely possible. However, Jay Cost correctly notes Obama has not managed this consistently in over two years.