Hang Right Politics

Posted by COgirl on November 2nd, 2006 in Polls

Hat tip to Lorie Byrd at Wizbang Politics for pointing us in the direction of this link and this statement.

“The fact that the online population is more heavily composed of Republicans than Democrats is principally a function of the Republican party’s higher composition within the overall electorate,” said Nielsen//NetRatings analyst Ken Cassar. “This is exacerbated by the fact that online penetration continues to be deeper among affluent households, which have historically skewed Republican,” he continued.

I’ve written about the inherent bias in the polls we’ve been seeing and I’ve referenced DJ Drummond many times. One of his biggest gripes with the polling has been how they weight the responses among Democrats, Republicans and Independents. DJ has argued that it’s too heavily weighted to Democrats given recent election results. This statement from Ken Cassar supports DJ’s claim.

We’ll know for certain on November 7, but I’m betting we’re going to discover there were a lot of polling problems this year and no one will be more pissed off about it than Nancy Pelosi’s interior decorator in Washington. (Pssst. . . cancel the order for the new drapes.)

UPDATE As I tossed and turned in the middle of the night, it dawned on me that perhaps I should give an example of the impact of overweighting the Democrat vote. First of all, when the polls are taken, they don’t know how many Republicans, Democrats or Independents they will get until the data is in. Often, they find that the sample they have taken does not reflect the actual (or assumed) makeup of the population. So the statisticians adjust.

Suppose that 100% of the Republicans said they would vote for Mr. R, while 100% of the Democrats said they would vote for Ms. D. If the pollster believes that 52% of the population is Democrat, then the poll would say Ms. D wins with 52% of the vote. The point that Ken Cassar makes is that there more Republicans around. In my simple example here, that would mean that Mr. R should really get over 50% of the vote and win the election. In reality, the number of considerations are far more complicated (e.g., turnout) and each factor has its own assumptions.

DJ Drummond believes, as do others, that polllsters are using a traditional allocation that has not been updated in a number of years to reflect actual voting patterns. When the raw data is provided, you can judge the assumptions the pollsters made, but often times the data is not released. And that is probably because of the poltical agenda of whoever is paying for the poll. So take the polls with a grain of salt and be prepared for many surprises on Wednesday morning.