03 December 2012

Why Was Colorado Presidential Polling So Off?

In the polling data provided to [The New Republic] by Mitt Romney's chief internal pollster Neil Newhouse, Romney has a 2.5 point lead on President Barack Obama in Colorado over the weekend before Election Day. As you know, the President won Colorado by almost 5.5 points. The range of explanations offered by Newhouse in this story vary from Latino voters (which while significant, don't fully cover the spread), over-reliance on self-identified "highly likely voters," and the perils of polling on a Sunday. Each one of these, the story goes, unintentionally helped contribute to the false sense of optimism projected by the GOP going into Election Day.
From Colorado Pols.

Needless to say, an 8 percentage point gap is far outside the margin of error for any decent poll.  Romney's internal polling was similarly biased in other swing states.

President Obama's internal polling was widely credited with being far more accurate nationwide in a data driven, technically flawlessly executed campaign was criticized for being too numbers driven.

Nate Silver at the New York Times sponsored 538 blog, who maintained a comprehensive running list of Presidential race surveys prior to the 2012 election did better, predicting an Obama win in the end in Colorado, but by a much smaller margin than the 5.5 percentage points by which Obama ultimately won Colorado.  Since he relied upon multiple polling operations rather than just one, his underestimate is even more notable as it should be robust to poll specific methodology mistakes.

The bottom line is that the pollsters don't have good explanations for why they screwed up, although problems with their estimates of likely voter demographics based on off year turnout in 2010 and ignoring the impact of President' Obama's get out the vote effort (particularly in key swing states like Colorado) seem to be among the most likely culprits.

An inability to accurately poll Colorado is a particularly troubling thing for someone trying to predict electoral college winners as Colorado is a bellwhether state.  It was the marginal state that pushed President Obama over the top to an electoral college majority in both 2008 and 2012.  Indeed, one could accurately have determined the winner in both elections simply by looking at Denver's first ring suburbs like Araphahoe and Jefferson counties, which is where the median voters in Colorado live.

No comments: