Do polls overestimate Republican peformance?

While I was on Twitter I read a Tweet by Nate Cohn saying how the polls are looking good for Republicans. One of the Democratic replies had two reasons to discount that. I’ll get into one of them here and save the other one for a later post. The person said that Democrats beat the RCP average in every 2010 and 2012 poll. Was that true?

I figured I’d check, starting with 2012 Senate and Gubernatorial polls. I decided not to look at Presidential polls because 1) I wanted to go apples to apples running for the same position 2) Those polls all have the same two candidates and may be heavily influenced by that 3) We know Barack Obama beat the polls in 2012.

I only used states that had three polls in the last 10 days of the campaign, excluding those with one or two. This dropped a few states but I didn’t feel one or two polls provided enough data. I only used polls in the last 10 days. We know that in 2010 Republicans gained in the polls and that they beat most every poll taken more than two months before the election. So a Republican up 3 in the poll now might be up 7 later and could still not beat the poll and win.

I found 15 states that fell into the criteria in 2012. Eleven had Senate elections and 4 had gubernatorial elections. Democrats beat the polling averages in every contest, by an average of 4.5 points. That isn’t too surprising, since we know the polls showed Republicans stronger than they actually were based on Obama. While Republicans can come up with reasons it won’t be representative of 2014, (e.g. turnout, Obama not running), we can’t dismiss these ominous results. If pollsters have learned nothing from 2012, then the actual results might be better for Democrats than the polls show.

What about 2010? That’d seem to be better representative with similar turnout, no Presidential election, and the polls skewing Republican. Democrats also did better than polls indicated in 2010, but this time it was only in 21 of the 32 contests. I say “only” because it wasn’t the 2012 sweep. The percentage is still impressive. It includes three races Democrats won where they polled behind (IL-Gov, NV-Sen, Co-Sen) and one the GOP won where they polled behind (FL-Gov).

There’s nothing definitive about these results. Some of these pollsters aren’t active this year and Rasmussen was included in most of the races. So that one pollster skewed the polling average toward the GOP. We should still keep this in mind when looking at polls. The GOP might, once again, fall short.

Edit: Nate Silver tackles this very issue on 538.