The State of the Polls

by Eyck Freymann

Like many politically-minded people, I keep abreast of the dozens of daily polls by looking at the graphs and charts on realclearpolitics.com. This site aggregates the polls and averages them. This initially seems effective. But when we think about it more closely, the validity of RealClear's numbers is called into question.

[Strongly Recommended that you click "Read More"]

Why is this? There are three main reasons.

1. Different pollsters tend to swing in favor of different parties. Rasmussen tends to tilt Republican (by that I mean usually has the Republican a point or two higher than reality) while CBS/NY Times tends to favor the Democrat. Because the average only consists of the last five polls, one of these can throw off the average.
2. Intermediary events: most polls are taken over the course of several days and then released a day or two after that. Events, gaffes, and media moments taking place after the surveys were taken do not make it into the numbers. In other words, there is a lag time.
3. Outliers: polling is statistically flawed. One can never guarantee the voting pattern of millions of voters based on a survey of a few hundred people. Various methods of cross-analyzing can improve the numbers (by making sure that women, men, blacks, hispanics, whites, churchgoers, etc. are all proportionate to their actual numbers in the population) but the result is imperfect. Occasionally a poll will appear that clearly demonstrates a flaw. If we have five straight polls showing Bush's approval at 30% and one showing it at 40%, one of them is clearly not correct. It demonstrates a major error in sampling and should be discarded.

So when we see this latest batch of Quinnipiac swing state polls, bells should be ringing upstairs. Although Quinnipiac generally provides some of the most reliable data available, we should be able to tell immediately that Obama is not up by 8 points in Florida and Ohio and 15 in Pennsylvania. Much as it would be convenient for Democrats to believe this numbers, it should be clear that this entire poll was sampled incorrectly. The first two are dead heats. Obama has a five point lead in Pennsylvania. These are the facts. Rather than selecting the truth we want, we should deal with reality (especially if it's not that bad anyway).
Obama '08