Archived Seat Projections

Ontario Opinion Polls Diverge
As we enter Week 3 of the Ontario election campaign, one of the most noteworthy surprises is how inconsistent the polls have been over the past two weeks. It isn't just that the polls have ranged from an 11 percentage-point Conservative lead to a seven-point Liberal lead (even wider if one screens for likelihood to vote), but the pattern has actually been bimodal. The two polls from Ipsos-Reid and one from Oracle have averaged a Conservative lead of close to 9 points, while those from Ekos and two from Innovative Research have a Liberal lead of over 6 points. Ironically, the pollster in the middle is Forum Research, which uses the controversial "robocall" (interactive voice response) approach to interviewing. The pollsters might not all be wrong, but they certainly can't all be right. A possible explanation for these diverging results could be the range of methodologies used. As the traditional assumptions of probability sampling are being replaced by variants of online sampling frames, necessitated by a declining response rates to phone interviews, such polling inconsistencies might become a bigger issue in the future. The seat projection below is based upon an aggregate sample of approximately 6000 respondents collected between May 2-15, in which the PC lead was just below one percentage point. The 2011 results are in parentheses. 

Projected distribution of seats by party, released May 20, 2014

Ontario PC Logo 2010.jpg
Ontario Liberal Party
Seat Projection
2011 Election
0 (0)
16 (17)
6 (5)
4 (4)
13 (13)
1 (1)
4 (3)
3 (4)
4 (4)
17 (14)
5 (8)
0 (0)
18 (15)
3 (7)
3 (2)
1 (1)
3 (4)
6 (5)


The 2011 election results are in brackets. The "regional swing model" is more fully explained in a paper originally prepared and presented by Dr. Barry Kay to the 1990 annual meeting of the Canadian Political Science Association, entitled "Improving Upon the Cube Law: A Regional Swing Model for Converting Canadian Popular Vote into Parliamentary Seats". It should be noted that the application of the model above does not make use of the "incumbency effect" described in that paper. In tests for past elections, using late campaign polls to project electoral outcomes, the model has proved to be accurate within an average of four seats per party since 1963. Readers interested in post-dictions for past federal elections dating back to 1963, for projections using pre-election polls dating back to the 1980 federal election and for three Ontario provincial elections, may contact me at bkay@wlu.ca.