No. If you want to predict the day before the election, you could come fairly close (but he corrects for polls that aren’t as reliable or which tend to shade to one side or the other).
But I think what is often missed is that his model predicted the results months in advance, and if you look at the graph on his site of what the model was predicting in terms of the popular vote, it didn’t fluctuate nearly as much as the RCP average of polls. Note too that even at the end, the RCP average only had Obama up 0.7, while Silver predicted a 2.5 point win–and the latter looks a lot closer to what happened.
This only makes sense if you believe that nothing that happened in the past few months actually had any influence on the outcome. Which seems highly unlikely to me.
If you accept that later events actually had an impact, then the fact that his pre-events predictions were correct would be coincidental.
I’m pretty sure gamerunknown is being sarcastic and you’re not seeing it. However, gamerunknown, I will point out that Stephen Colbert’s job is taken and you may want to adjust your style if nobody is getting your message.
The graph certainly seems to account for such things to me. You can see the small bounce in early September from the enthusiasm of the DNC, the HUGE crash in early October after the first debate, and the gradual rise as the second and third debates and Sandy came into play.
It could be coincidental that Nate’s initial prediction closely matches the final outcome, or it could be that his model sufficiently accounts for unexpected developments.
A look at the two states Senate races Nate seems to be losing are interesting. Nate missed on North Dakota (in whatever sense a probability is a miss), but there the spolls hardly seemed to contradict his prediction. I wonder what exactly it was that caused Sam Wang to predict a D victory in North Dakota.
In Montana, it is more interesting, because the poll average shows a narrow Tester win, but the state fundamentals are heavily skewed republican, which i s why he “missed” it. Now he did have a 1 in 3 chance for Tester, so it isn’t like he was well off.
But more at issue is how much predictive value is added by adding in the state fundamentals overall.
To be fair, Silver will be spending a lot of time trying to see if there is something that can be improved. And Nate himself has said that a mix of quantitative and qualitative analysis will provide the best predictions. The key is knowing how much of each.
Still beating your wife? Must be tough on you these days, what with restraining orders and all …
Either I’ve misunderstood you or this makes no sense. There’s no way a model can get the final percentage correct, if that final percentage is based in part on unknown future events.
A model can (in theory) “sufficiently account for unexpected developements” in terms of the confidence level surrounding the prediction. But it can’t nail down the actual number if that number is to be affected by future events.
I don’t have time to do it now, but I’d be interested in seeing a Nate Silver vs Sam Wang scorecard. Looks like Wang missed on Florida (which he called a cointoss, anyway), but nailed the popular vote totals, and did a bit better than Nate in the Senate races from what I can see. And Sam’s methodology is exclusively from polls, right?
I’m predicting a 92.8% chance of a very, very large party at Nate Silver’s house. By the way I read yesterday that he was responsible for about 20 percent of the New York Times’ web traffic in the month before the election. He’s in very good shape.
The Libertarian vote I think is the big wildcard in the Montana races these days. I think it’s quite difficult to tease out in a poll (or casual conversation) whether a Libertarian-minded person out here is going to grudgingly vote Republican or not. They’re not going to vote “for” Obama by voting Libertarian, but they will for someone like Tester who is very pro-gun and very pro-civil liberties*. And they did in a big way-- right now it looks like 6.5% went Libertarian, which even if it wasn’t decisive is definitely responsible for his large margin. The Lib vote was also decisive in the governor’s race.
(*And yes, I agree that thus far Obama hasn’t been particularly anti-gun and is by far the lesser of two evils on civil liberties, but that’s not the popular perception.)
Well now I feel bad. I’ve been assuming the guy was just a crank espousing an ideology, but his comments now make it plausible that he was a man with a theory which he tested and found wanting.
