For someone whose vaunted statistical model was forecasting an 85 percent chance of an Obama victory next month, Nate Silver is really beginning to sound as if he’s attempting to retreat from his current predictions:
I feel as though it’s my duty to tell you when my subjective estimate of the odds differs by a material amount from the ones that our model produces. On Friday and Saturday, I wrote that I thought the model was underestimating Mr. Romney’s chances.
The model is designed to distinguish essentially random changes in the polls from more permanent reversals in the state of play. But it takes a one-size-fits-all approach to do this. Had there been no major developments in the news cycle over the past several days, there would be reason to be skeptical that the shift toward Mr. Romney had been quite as clear as the polls had seemed to imply. There have been other points in the election cycle when the polls appeared to show a shift in the race but without much news to drive it; the model has been fairly “smart” about avoiding being taken by these false alarms.
The trade-off, however, is that the model may be too conservative about accounting for a shift when there is real news behind it. The model is able to account for changes caused by some types of economic reports, since those are incorporated directly into the forecast; we also have special procedures to handle polling around the party conventions. Other types of news events, however — like the debates, major foreign-policy developments, or the vice presidential selections — may not be handled very adroitly by the model.
Those who like to get on my case about Obama defeating Hillary for the Democratic nomination tend to forget that I make a single prediction about the election long before the nominees are even known. In 2008, I predicted that the Republicans would serve up a sacrificial lamb and the Democrats would win nearly 18 months before the election. I was wrong about the specific individuals, but the general theme was correct. In 2010, I predicted that the Republicans would win in a landslide. In 2012, I predicted that the Republican candidate would win easily unless it was Mitt Romney, in which case I still expected him to win.
Meanwhile, the “professionals” like Silver are not only making new model-based predictions on a weekly or even daily basis, but are attempting to separate themselves from their predictions. I leave it to you to determine which process is more useful, even if the success rate of those predictions made months in advance is lower than those “predictions” made in real time.
Yes, I was wrong about Clinton winning the nomination, about Pataki being the sacrificial lamb, and about Obama actually dropping out, but based on all the articles that have suddenly begun appearing about whether he is even interested in seriously contesting the election – and articles written by some of his most fervent supporters – it is clear that in the case of the latter, I was on the right track even though events did not turn out exactly as I thought they would.
It may be worth noting that despite what Silver and the mainstream polls say, I’m not the only one anticipating a Romney victory in November:
An update to an election forecasting model announced by two University of Colorado professors in August continues to project that Mitt Romney will win the 2012 presidential election. According to their updated analysis, Romney is projected to receive 330 of the total 538 Electoral College votes. President Barack Obama is expected to receive 208 votes — down five votes from their initial prediction — and short of the 270 needed to win.