In the days following the first presidential debate between President Obama and Mitt Romney analysts have been asking what impact Romney’s apparent victory would have on the overall race. It is still early, but the polls generally have shown Romney gaining on Obama, or even taking the lead in some instances. As the new numbers come out, the debate over party identification has reignited. Some liberals are complaining that the new polls, favorable to Romney, include too few Democrats and too many Republicans. Conservatives complain that the polls still include too few Republicans, and that Romney should be winning by an even larger margin. In reality, the new polls demonstrate what has been known to pollsters for a long time. Party identification is not a simple split to be uniformly applied to every poll, as Unskewedpolls.com has done, but changing number which must be recalculated with every new poll.
As an example, lest us look at the newest poll from Public Policy Polling, which shows President Obama leading with 49 percent of likely voters compared to 47 percent for Mitt Romney in Wisconsin. While PPP does show Obama winning in the Badger State, Romney did make big gains from their last poll that showed Obama up by seven points in Wisconsin. PPP’s survey was conducted entirely after the first debate, and had a party identification split of +1 D (34 percent Democrats and 33 percent Republicans). In their last poll from Wisconsin, released in mid-September, PPP showed a party identification split of +3 (34 percent Democrats and 31 percent Republicans).
PPP’s numbers immediately drew criticism from both the left and the right. Liberals claimed that their poll did not include enough Democrats in a state that is traditionally more progressive. Conservatives claimed that PPP is still not including enough Republicans, and that Romney must actually be winning.
In reality, PPP’s survey shows the reality of party identification splits. Party identification is not a static number. Voter often switch party allegiances depending on what candidate they support or the news of the month, day, or week. PPP did show a stronger split for Democrats in their surveys before the debate, but that was probably because many Republicans identified themselves as “independents” before the debate. As Romney’s support increases, the number of people who identify themselves as Republicans in a sample will also increase. Conversely, when President Obama is trending down the number of people who identify themselves as a Democrat may also go down.
Party identification splits are not some way for pollsters to “fix the numbers” for either Romney or Obama. Instead, they are reflection of what the electorate is thinking. An electorate that votes for Obama is more likely to be Democrats, while an electorate that votes for Romney is more likely to be Republicans. When Obama was polling well in the last few weeks it is therefore no surprise that the polling sample showed more people identifying themselves as Republicans. As Romney gets a bounce from his debate performance, it should be no surprise to see samples with more Republicans than before.