Leiberman: Just another science doubter?

Leiberman says he doesn't believe the polls that say that McCain is going to lose.

Sen. Joe Lieberman (I-Conn.), one of GOP presidential nominee John McCain’s top supporters, said Friday that the Arizona senator’s poll numbers are “so extreme that I don’t believe them.”

While I can understand the sinking feeling that Leiberman has in his gut (something I've felt it in every national election--save two--since 1980), is his disbelief grounded in rational thought or wishful thinking?

Polling is a pretty scientific process in the political arena these days. But the difference in models can make a big difference in results. But when you look at the national polls only a week and a half before the election and see that Obama is up in every single one (from +3 to +14), it is hard to deny that things are looking up for the Democrats.

As any scientist can attest, the analysis is only as good as the model. Or the sample. Or the way the questions are posed in the poll itself. I'm the first to admit that election polls lack the rigor of more conservative statistical analyses. But in many ways, they mirror the same problems that all scientists face in analyzing our data. The accuracy of the analysis is dependent upon a number of important factors:

• That our sample is representative of our population
  
• The individual data points represent reality (i.e., respondents don't lie)
• That our assumptions (which vary by model) reflect reality

We can't know these things with certainty, which is why we allow for uncertainty in our models. That level is generally 5% in biostats anyway. My problem with polls is that they don't give their level of uncertainty. I have no idea whether there is any significant difference in the poll that results in a +3 for Obama or a +14 for Obama. It all depends on my margin of error.

We know for a fact that telephone polls that depend on landlines are not reflective of the population. Young people are vastly underrepresented among landline users. Time of day of a poll can also favor one age group over another. Disqualifying respondents based on past voting behavior also leaves out newly registered voters and voters energized by the novelty (?) of this presidential election. Interestingly enough, each of these widely reported problems with polls seems (by a common sense approach) to me to under represent Obama's likely voters (i.e., Obama may lead by even GREATER numbers). For this and many other reaons, I tend to think of polls as just slightly more reliable than my horoscope. Polls should be taken with a grain of salt.

But, in the grand scheme of things that is the last week of this election cycle, I'll take my grain of salt and weigh it against Leiberman's wishful thinking any day.

In a week, all those pollsters can figure out what was right and what was wrong in their models and improve a bit for the next time. While they do that, and no matter who is elected, I will certainly be celebrating the lame duck status of dear old 43.