WOW that was a good speech

The problem with using polls as a way to predict the outcome of an election is that there are several micro-biases which can skew the results ... In most cases these biases tend to (when combined) balance out to being (roughly) equal but I have had a feeling that they won't in this election.

First off, it is typically assumed that non-respondents hold views that are (roughly) in proportion to respondents so you can generalize the views of the respondents over the entire population. Now, I could be wrong but, I think that when you compared the non-respondents to the respondents you would find that there was greater support for McCain in those who did not respond to polls. The reason for this assumption is that Obama supporters want to be heard because of what supporting Obama implies about them and what an Obama presidency implies about the country, while McCain supporters are much more likely to be supporting him as the better of a bad choice.

Another issue is that, if a person detects that you have a preference, they will typically answer the way that you want them to when asking a question rather than the way they truly feel. They really don’t even need to detect this preference on a conscience level in order to act upon it. Most polling companies hire college students part time in order to conduct their polls because college students tend to be more professional on the phone than other potential employees. In most elections college students tend to be (generally) pretty disinterested in the election but in this election they are very interested and heavily support Obama. Even if they try to sound objective it is likely that some of the pollsters will telegraph their preference some of the time and this will have an effect on some of the people they polled.

Now, I’m not expecting McCain to win given the polls we have seen, but I am certainly expecting him to lose by less than 5% (which is about the high end of the polls for him).