Tuesday, October 05, 2010

Rasmussen Poll shocks everyone who has no clue how Rasmussen polls

Various commentators seem surprised at the sudden change in the Rasmussen Generic Congressional Ballot poll. All through September Rasmussen reported a near constant wide Republican lead while other pollsters reported a much smaller and shrinking lead.

Now suddenly, as soon as we enter October, Rasmussen reports a Republican lead of only 3%. What could possibly have suddenly changed on October 1st ?

Rasmussen weights it's samples so that a constant fraction of weighted respondents self identify as Republicans, Democrats and Independents. The weights are based on the average response in the last month of polling. This might make some sense in polling on specific races, but there really isn't much difference between partisan identification and generic ballot voting intentions. I'm sure there are some people sho say that they are Democrats but voting for Republicans and vice versa.

Therefore the Rasmussen poll is roughly last month's self identified Republicans plus a constant near one times the fraction in the latest poll of independents who plan to vote Republican for the Republicans and a similar calculation for the Democrats. This means that the poll basically is only partially updated within a month.

No comments: