The readers at fivethirtyeight.com don’t need to be told that the election results from 2008 matched Nate’s pre-election poll aggregations in almost every state. But it’s also interesting to look at some of the discrepancies:
The biggest discrepancy is in Hawaii, which Obama visited a few days before the election to go to his grandmother’s funeral. Then there’s Nevada–I don’t know what was going on there, but probably Nate does. Beyond that, there’s a systematic pattern that Obama did better than the polls in Deomocratic states and worse than the polls in Republican states. Does this represent a real pattern of voters–perhaps people reverting to their more predictable positions at the last minute, with Vermonters moving to the Democrats and residents of Wyoming going the other way? Or maybe it’s an artifact of the poll aggregation, with the predictions being pushed too close to 50%, on average? I don’t know. It would be interesting to do similar aggregations of state polls from 2000 and 2004, and also to look at national polls, to see if this pattern was occurring in earlier years.
The discrepancies had very little impact on the election forecast because most of the problems were in landslide states. If the poll aggregation predicted Obama would get 60% in Delaware and he actually got 62%, it doesn’t really matter anyway. But it could be interesting to study such systematic discrepancies with the goal of improving the methods for the future.