For The Record: 2004 Federal Election Polling Post Mortem

Toronto, ON - A post election review of the available evidence indicates that while the final polls released in the last days of the 2004 Canadian Federal Election Campaign indicated a race that was "too close to call", and the seat models suggested an outcome that was different from the actual results, the explanation for this discrepancy is straight-forward: timing.

The survey methods of the various polling firms, and their subsequent seat models, were sound and accurate. However, the results of the 2004 election campaign were greatly influenced by a shift in voter support in seat rich Ontario, and to a lesser extent in Quebec, in the last 24 hours of the campaign.

Evidence shows that while the intensity of the shift in Ontario happened the night before the vote, it also appears that there was a net switch from the final Ipsos-Reid poll concluded on the night of Wednesday, June 23 to the end of voting on the evening of Monday, June 28 of 6.7 points - it would appear that approximately 340,000* voters in the province moved to the Liberals over that timeframe and changed the national outcome from what was recorded on Wednesday night. In Quebec, the Liberals garnered an extra 4.9 points or approximately 168,000** votes in the same timeframe.

In all, of the 13,489,713 voters who turned out on June 28 to cast their ballots, a mere 506,851 voters (or 3.78 percent of all votes cast) determined the outcome of the election in the dying hours of the campaign. It's also apparent that this shift cost the Federal Conservative Party in Ontario a significant share of seats and enabled the Federal Liberals to secure their victory. This, when combined with a boost in the Province of Quebec, handed the Liberal Party a solid plurality of seats even though five days earlier they were staring possible defeat in the face.

While no pollster in Canada predicted the magnitude of the Liberal's victory, this does not mean that the polling was "wrong". Rather, a review of the evidence indicates that the polling and seat models presented at the end of the campaign were on target, but that the campaign had yet to have its full impact on voters when the polls were undertaken and released.

Further, because Canada's election laws prohibit the release of new polls on voting day, the final shift would not have been reported in time to correct the record until after the votes were counted and the outcome known. In fact, lost in the coverage of the Monday night tally was a lone national exit poll (by COMPAS) that accurately predicted the outcome and underscored the importance of vote shifting in the last 24 hours of the campaign.

Please open the attached PDF to view the full release.

-30-

For more information on this news release, please contact:

Darrell Bricker
President & COO
Ipsos-Reid Public Affairs
(416) 324-2900