Unskewing The Election Results

During the course of the campaign you may have heard about the movement to argue that the polls were “skewed” by including too many self-identified Democrats. The polls showed Mitt Romney ahead with self-identified Republicans, so clearly Romney’s winning.

And now you can see that the final results as measured by the exit polls are also skewed!

In this case we can easily enough re-do the calculation based on the assumption that there are equal numbers of Democrats and Republicans in the electorate and what we see is that Mitt Romney won a narrow victory with 50 percent of the vote to just under 48 percent for Obama. Not bad!


The issue, as it turns out, is that the polls weren’t skewed at all. They just had a lot of self-identified Democrats in their sample because a lot of people were self-identifying as Democrats. The underlying issue is that party ID is an “attidunal” variable (comparable to wanting to vote for Mitt Romney) rather than a demographic variable (comparable to being a white man). What seems to have happened is that the Republican Party brand has been in the toilet, so a lot of people who you’d demographically expect to vote Republican aren’t identifying as Republican. In a great year, Obama could persuade some of those folks to vote for him (that’s how he carried Indiana in 2008) but normally these are still conservative people and they vote for Republicans. This outflow of conservative-minded people from the bloc of self-identified Republicans and into the bloc of self-identified independents is interesting but ultimately doesn’t make any difference to the outcome.