"Mr. Gregg, who did not publish seat projections in 2004, said the main reason pollsters got it wrong last time was because the main premise of projections is that previous elections are the strongest indicator of how the next election will turn out.
Unfortunately for pollsters in 2004, there was no previous election that involved the newly formed Conservative Party.
Trying to blend results from the Canadian Alliance and Progressive Conservative parties threw off the formulas."He is wrong. The principal problem for seat predictions in 2004 was that the last published polls in Ontario did not capture a last second and very substantial movement to the Liberals. Combining the Alliance and Conservative vote to create a base for the new party was at worst a minor issue, certainly for me.
My model performed no better or worse that it ever had when I inputted the actual vote shares after the election.
Back to the campaign, I still am not seeing the consensus on poll numbers that I expect to see at the end of a campaign, so it is still somewhat fluid. That is reflected in varying seat forecasts. I will do one with the Globe's numbers when the Strategic Counsel posts them to the web. They have not yet done so. However, one point is clear: the polls don't yet show a Conservative majority.