The Week In Review - Poll Noise

There has been a lot of talk in the last few weeks about national polls and their wide swings between companies. As an example, the current Gallup tracking poll shows a dead heat between John McCain and Barack Obama, while at the just days ago the Los Angeles Times reported Obama with a 12 point lead. How could this be? The truth is they are more statistically similar than you might think. The following is a list of all national polls that have been completed within the last two weeks, as well as the calculation I use to determine the outcome. (Please note: This is not my official estimate, only an interpretation of the national data)

The Gallup tracking poll has a sample of 2,605 voters over 3 days and has a margin of error of 2%. The Los Angeles Times poll was taken of 1,115 voters over 5 days and has a margin of error of 3%. Based on this if we were to assume that both polls are at the extremes of their margins of error, it is possible that The Gallup Poll is underestimating Obama by 2 points and overestimating McCain by 2 points. The new net result is Obama 46, McCain 42. If the opposite is true with the L.A. Times poll, it is possible that the results are Obama 46, McCain 40. Neither of these results are unacceptable. In fact, they appear to be in the range of most other polls in this time frame. Any single poll can have wide swings of this nature, which is why it is ultimately important to look at many polls over a similar time period to determine the actual results.