Monday, October 18, 2010 at 11:42 pm by Darryl
Rasmussen released a new poll today in the race ‘twixt Sen. Patty Murray (D-WA) and real estate seminar motivational speaker circuit superstar Dino Rossi (R). The poll of 750 likely voters has Murray leading Rossi 49% to 46%. This is the sixth poll in a row in which Murray has led Rossi.
By itself, this poll suggests (based on a Monte Carlo analysis of a million simulated elections) that Murray would win with a 72.6% probability (719,456 wins to Rossi’s 271,740 wins). Here is the distribution of simulated election outcomes:
A more interesting analysis includes all seven of the October polls, and pools them into a single analysis. Doing so yields 4,906 “votes” of which 4,602 go to either Murray or Rossi. Murray gets 2,398 (48.9%) and Rossi takes 2,204 (44.9%). And 304 (6.2%) are “other” or “undecided,” The outcome of a million simulated elections from the seven polls is that Murray wins 978,721 times and Rossi wins 20,545 times. In other words, in an election held now, Murray would win with a 97.9% probability. Rossi would win with a 2.1% probability.
Last week, with six pooled polls, the probability of a Murray win was 97.6%. So the current poll improves Murray’s chances, even if the results of the new poll only favor Murray by +3%. This is because the sample size has grown by 750, adding considerable certainty to the statistical estimates.
Finally, I’ll point out that this poll is completely consistent with the robopoll/live poll/voter enthusiasm hypothesis. Recent robopolls (from Rasmussen, FOX News and SurveyUSA) seem to be systematically biased toward a tie. Live interview polls give a considerably higher percentage to Murray. My hunch is that the less enthusiastic Democrats are more likely to hang up on robopolls. I’m sure there are more polls to come to help us evaluate the hypothesis.