Tuesday, February 5, 2008
Interesting Use of Poll Statistics
I have been amazed at how much the RealClearPolitics (RCP) poll average is quoted. This number is basically meaningless. What they do is compile all of the polls and then take an average of the most recent several polls. In some cases I have seen the polls that are averaged being weeks apart. So for instance say McCain was up by 20 points three weeks ago in one poll. However more recent polls show only a 3-4 point spread between McCain, Romney and Huckabee. Because of their averaging method, the RCP average may still have McCain up by 5-8 points. This could be the case even if Romney or Huckabee is ahead in the more recent polls. Additionally, each poll has a different methodology of selecting their sample and asking the questions. So trying to compare what the Rasmussen poll means in relation to the SurveyUSA poll is not a simple task of looking at the numbers. A better way to look at the polling information is to look at each poll over a period of time. This is especially effective for those tracking polls that do a rolling sample over a few days. This way one can see if Obama has "momentum" or if Clinton's support is remaining steady. In the end, a poll doesn't tell the story like election returns.