Tuesday, January 17, 2006

Seat predictions

I think election prediction models like mine are going to be weaker in this election than previously. I see from pollsters stories of asymmetric shifts, for example unusually large gains for the Conservatives among the affluent. My guess is that we will see some ridings where there are very large swings that overwhelm the capacity of any formula to foresee them.

I am also guessing there is going to be increased strategic voting of all sorts – last minute switches from NDP to Liberal, some Liberal to NDP where appropriate, even Conservative to NDP to block Liberal wins. This too will be damaging to accurate seat forecasts.

It seems to me the same basic pattern of voting has existed since 1993 and we are witnessing a sea change, not necessarily as long-term as in 1993, but on a significant scale.

There is plenty of disagreement among the polls, and now among the seat predictors. Andrew Coyne has helpfully assembled most of them into one place here: http://andrewcoyne.com/2006/01/seven-sleeps.php

I crunched the numbers today on the latest Ipsos-Reid Online poll with its massive sample size. My numbers are very different from theirs.

Ipsos Reid Seats
C – 149 to 153
L – 64 to 68
NDP – 29 to 33
BQ – 57 to 61

TC Norris Seats
C – 137
L – 72
NDP – 37
BQ – 62

I don’t see how Ipsos get the Conservatives as high as 149 to 153 on the strength of a four point lead in Ontario and a performance in B.C. very close to the last election. Strategic Counsel has the Conservatives up seven points in Ontario and eleven in B.C., numbers that inevitably produce much better seat numbers for Harper. My intuition is that I may have the NDP too high.

However, my real hunch is that all the egocentric seat predictors like me could wind up looking foolish on the morning of the 24th.

No comments: