A quick thought on the polls

Everyone is screaming at the top of their lungs about how wrong the polls were. However, what is interesting is that according to Real Clear Politics' average, Obama's average support was roughly 38% going into the New Hampshire Primary(http://www.realclearpolitics.com/epolls/2008/president/nh/new_hampshire_democratic_primary-194.html#polls) and he ended up getting 37%. There's certainly no scandal in that. Clinton's support, though, surged from 30 to 39% from the polling average. When one disects the numbers we find that the women's vote came out very differently than expected. American Research Group, based in NH, projected Obama to win the women's vote by 3 points and instead lost it by 13%. Obama's support among men was in line with what the polling showed and his support among women was actually largely in line with what he was polling going in. Clinton just happened to clean up among undecided women which is not entirely surprising.

With how dramatically polling supposedly shifted after Iowa, it seems possible to me that in the couple days after the debates that they could have shifted back. Clinton had a pretty good debate performance as the exit poll reflects: http://www.cnn.com/ELECTION/2008/primaries/results/epolls/index.html#NHDEM. Among those who said the debates were very important, 48% of the total vote, she won by 8%.

Also, the time when voters decided is very important. A huge chunk said they decided before a month ago and they broke for Clinton with 48% of the vote. Obama did lead substantially among those deciding between a month and a few days ago. This also possibly reflects the fact a lot of voters voted absentee and thus were not impacted by Iowa.

Then, we come to the raw strangeness of New Hampshire. In 2006, there were two incumbent Republican Representatives who seemed to be in pretty good shape. The western district, being challenged by Democrat Paul Hodes was a possible pick up as it is generally more liberal containing the Vermont border region. However, the eastern district was completely safe by nearly every measure and it too fell to the Democrats. This was a stunner. While I projected the bulk of House races correctly, I never saw that one coming and neither did anyone else except the most optimistic Democrats. This isn't limited to New Hampshire though. In Georgia in 2002, Governor Roy Barnes(D) was expected to coast to a 8-10 point win over challenger Sonny Purdue(R). The result: Purdue won the governorship by 5 points.

The lesson here: Polls are usually pretty good, but don't bet the farm on them. Some races are hard to predict due to various factors. Primaries are particularly notorious due to the fact they rely more on turnout operations than general elections.