It has been the custom for LISPOP to produce a post-mortem analyzing the final pre-election projection using the actual election vote splits rather than the aggregation of pre-election polls by region. As indicated in the table below, the first number by party in each region was the pre-election projection, the second number in brackets is the post-election projection using actual voting data, and the final number is the real number of seats won. Our pre-election projection was Libs-138, Cons-127, BQ-36, NDP-34 Greens -2 and Independent-1. The Independent was based upon a riding poll, not the use of the swing model. The LISPOP numbers were relatively close to the other sites doing projections, CBC Poll Tracker, 338 Canada and Calculated Politics.
On these numbers, the largest discrepancy from the actual results pertains to the Liberals who actually won 157 seats. However, our projection of 67 LIB seats in Ontario was based upon a 6% Lib advantage over the Cons in the province resulting from averaging the polls. In reality, the Lib advantage was an 8.3% lead which would have places nine additional seats in the Liberal column. For those who might wonder about the calculation, the ridings involved were Burlington, King-Vaughan, Oakville North, Richmond Hill, Aurora-Oak Ridges, Newmarket-Aurora, Northumberland, Whitby and York Centre, each of which had a Liberal winning margin over the Cons of between 2-4% in 2015. That was the greatest discrepancy, but there were others in the various regions as well.
Table 1: Federal Election 43 - Post-mortem
|Canada||138 (153) 157||127 (127) 121||34 (29) 24||36 (27) 32||2 (2) 3||-||1 (0) 1|
|Atlantic||23 (23) 26||7 (7) 4||2 (2) 1||-||0 (0) 1||-||-|
|Quebec||35 (38) 35||8 (12) 10||2 (1) 1||36 (27) 32||-||-||-|
|Ontario||67 (76) 78||41 (34) 37||13 (11) 6||-||-||-||-|
|Prairies / North||6 (7) 6||22 (23) 21||3 (1) 4||-||-||-||-|
|Alberta||0 (0) 0||33 (33) 33||1 (1) 1||-||-||-||-|
|British Columbia||10 (9) 11||16 (18) 17||13 (13) 11||-||2 (2) 2||-||1 (0) 1|
Note: The "regional swing model" is more fully explained in a paper presented by Dr. Barry Kay to the 2009 annual meeting of the American Political Science Association, entitled "A Regional Swing Model for Converting Canadian Popular Vote into Parliamentary Seats 1963-2008." 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 [email protected].