Signs of Desperation

The presidential polls are not going the way conservatives expected. Obama is up by about 4% in national polls and is polling even better in the battleground states.

One response to this phenomenon among conservatives is to criticize the polls themselves, arguing that they have a Democratic bias. A good example of this is found on the National Review web site in an article by Jim Geraghty and quoted repeatedly throughput the conservative blogosphere.

This really smacks of desperation. For example, one statistic cited repeatedly is that the last Gallup poll before the election showed Clinton up by 12%, when in fact he only beat Bush by 5.6%. A very strong Democratic bias, yes? Well, no. If you look at the actual data, that difference between the last pre-election polls and the final result had nothing to do with a pro-Democratic or anti-Republican bias. The change in the results is entirely accounted for by a large shift away from Clinton towards Perot. The polling for Bush was quite accurate. If anything, the polling in 1992 shows a bias against independent candidates.

I looked at the data published by Gallup for Presidential polling going all the way back to 1932 comparing the last poll before the election with the actual election results. I ignore exit polling, which is notoriously unreliable. I also assume that any final polling result that gets the actual election results +-1% shows no bias. The result? There is a measurable partisan bias towards Democrats in seven elections and a measurable partisan bias towards Republicans in eight elections! In other words, no consistent bias at all.

 

As recently as 2000, Gallup showed a bias against the Democratic candidate. The largest single polling error was in 1932, when FDR underpolled the actual result by 6.8%, a huge GOP bias. Interestingly, this anti-Democratic bias is even larger than the famous polling error of 1948, when Truman unexpectedly won against Dewey.

The problem is not the polling. It is the candidate.