Friday, May 8, 2009

Dewey Beats Truman

The Angus Reid Strategies poll is out and here are the headline numbers

  • Liberals - 44%
  • NDP - 42%
  • Green - 10%
  • Others - 4%

This the poll I was part of the other day.

In general I am certain that the methodology ARS is using is what all polling companies will be using from now on. I was impressed the sort of questions they asked and the attempt to drill down to get detailed results. But I can not shake the feeling that this is something similar to the 1948 US election in which polling was part of the reason for the famous Tribune headline.

I believe there is an issue with who is in the sample for ARS.

Every polling company wants to correct in predicting the popular vote in the election because it is free advertising for them. Mustel, Ipsos and Angus Reid Strategies all are working to make sure their results are an accurate reflection of what the public is thinking. No one wants to be wrong. The problem is that in this election someone is going to be wrong.

Adding ARS into my mix gives me the following numbers:
  • Liberals - 45.2%(-1.4)
  • NDP - 39.2%(+2.0)
  • Greens - 11.7%(-0.7)
  • Others - 3.9%(-0.1)
I will be updating my numbers on the seats later today if I get enough of a break from work. What I will quickly note is that the numbers I am getting to is still showing the NDP down enough from 2005 to lose a significant number of seats.

On the ARS numbers alone we would still see a Liberal victory on May 12th. Defending a seat you are holding is easier than winning one and since the Liberals are defending more seats than the NDP, the are naturally set to win more seats. This effect is core to the cause the 1996 NDP "wrong winner" election.

The ARS poll may have the result of boosting the Liberal vote on election day. As long as it was looking like the NDP had no hope of winning a lot of right wing people who really can not stomach voting for the Liberals could stay home and not vote. The threat of an NDP government could boost the Liberal vote by one to two percentage points on election day.
Post a Comment