Nate Silver vs. the world, ctd
My post on Nate Silver generated some thoughtful email responses. I thought I’d share a few.
This one came from a political science professor:
I think your column on Nate Silver today, while perceptive, missed three points.
1. The probability of an Obama victory and his predicted margin of victory are not the same. Silver predicts a very close outcome at least in the popular vote.
2. When the polls were pointing to a massive Obama landslide, SIlver (sic) was more cautious, pointint (sic) to fundamentals that suggested a closer race.
3. Most importantly, he is right about Gallup being an outlier. Today’s tracking polls all shifted in Obama’s direction, save Gallup. Conversely, Gallup was still showing Obama above 50 well after the second debate. Some of this is because of their long time lag (seven day rolling average!) but not all. I suspect a lot of it is because of their extremely restrictive likely voter model which I think will overstate the predictive power of enthusiasm on turnout.
Admittedly, I am also one of those Democrats to whom Silver is providing a bulwark against panic, but I also know a thing or two about reading polls. And the story about polls in this election is ultimately going to be about the likely voter models and how well or badly they forecast turnout. I wouldnt’ bet money on any of them. But especially not on Gallup, for the reasons Silver analyzed so cogently.
Another poly-sci professor was less enamored with Silver:
I briefly looked at Nate Silver’s numbers about a month ago and saw that he was sometimes including two polls from the same polling organization — usually pro-Obama outliers — in a state average, and was often including polls well over a month old — again usually pro-Obama outliers — in a state average. Both seem like questionable practices. Unfortunately I don’t have any more time to spend on him, since this is not really what I do.
Lastly, I think this reader sums it up both our arguments fairly well:
I read Nate daily and I am for Romney. I also know something about gambling. Unfortunately, if you look at all the “prediction markets” (i.e. gambling markets) Obama is a 2-1 favorite. This includes the compendium of London bookmakers that you can find on the net. Unfortunately, I think they are right, based on the state of the polling of the electoral college vote.
Now, if Gallup is right, or even close to right, and Romney pulls off a 3-4 point popular vote victory, history says there is no way the electoral college comes out differently. Even Nate has said that this can almost never happen if the margin in more than about 1.5 points.
I suspect more emails on this topic will arrive in my inbox over the next couple of days…