To wrap up our election coverage, we have compared the predicted vs. actual results for the five major parties in every riding. Overall, the model correctly allocated 94% of the votes (overall accuracy) and correctly predicted 92% of the seats.

DOWNLOAD PREDICTED VS. ACTUAL PDF (800kb)

The following tables summarize the predicted vs. actual results for riding-by-riding seat predictions, overal seat counts and share of vote. An example of how vote accuracy is calculated is given at the bottom.

RIDING-BY-RIDING SEAT PREDICTIONS

SEAT COUNTS

REGION
NATIONAL	Actual	124	103	29	51	1
	Predicted	128	94	29	56	1
	Difference	-4	+9	0	-5	0
ONTARIO	Actual	40	54	12	-	0
	Predicted	44	49	13	-	0
	Difference	-4	+5	-1	-	0
QUEBEC	Actual	10	13	0	51	1
	Predicted	6	12	0	56	1
	Difference	+4	+1	0	-5	0
BRITISH COLUMBIA	Actual	17	9	10	-	0
	Predicted	19	9	8	-	0
	Difference	-2	0	+2	-	0
ALBERTA	Actual	28	0	0	-	0
	Predicted	28	0	0	-	0
	Difference	0	0	0	-	0
PRAIRIES	Actual	20	5	3	-	0
	Predicted	20	4	4	-	0
	Difference	0	+1	-1	-	0
ATLANTIC CANADA	Actual	9	20	3	-	0
	Predicted	11	18	3	-	0
	Difference	-2	+2	0	-	0
NORTH	Actual	0	2	1	-	0
	Predicted	0	2	1	-	0
	Difference	0	0	0	-	0

SHARE OF VOTE

REGION
NATIONAL	Actual	36.3%	30.2%	17.5%	10.5%	5.5%
	Predicted	36.8%	29.5%	17.3%	10.7%	5.7%
	Difference	-0.5%	+0.7%	-0.2%	-0.2%	+0.2%
ONTARIO	Actual	35.1%	39.9%	19.4%	-	5.6%
	Predicted	36.1%	37.9%	20.4%	-	5.9%
	Difference	-1.0%	+2.0%	-0.7%	-	-0.3%
QUEBEC	Actual	24.6%	20.7%	7.5%	42.1%	5.1%
	Predicted	25.6%	16.7%	8.2%	44.9%	4.6%
	Difference	-1.0%	+4.0%	-0.7%	-2.8%	+0.5%
BRITISH COLUMBIA	Actual	37.3%	27.6%	28.6%	-	6.5%
	Predicted	36.4%	30.3%	26.7%	-	6.6%
	Difference	+0.9%	-2.7%	+1.9%	-	-0.1%
ALBERTA	Actual	65.0%	15.3%	11.7%	-	8.0%
	Predicted	62.5%	15.8%	13.6%	-	8.1%
	Difference	+2.5%	-0.7%	-1.9%	-	-0.1%
PRAIRIES	Actual	45.6%	24.4%	24.8%	-	5.2%
	Predicted	48.2%	24.8%	22.3%	-	4.7%
	Difference	-2.6%	-0.4%	+2.5%	-	+0.5%
ATLANTIC CANADA	Actual	34.8%	40.0%	22.4%	-	2.8%
	Predicted	36.4%	40.8%	19.7%	-	3.1%
	Difference	-1.6%	-0.8%	2.7%	-	-0.3%
NORTH	Actual	23.7%	40.2%	29.6%	-	6.5%
	Predicted	21.0%	37.5%	32.5%	-	9.0%
	Difference	+2.7%	+2.7%	-2.9%	-	-2.5%

Example of how accuracy is calculated:

Nipissing-Timiskaming (ON)

RESULT
Actual	45%	34%	17%	6%
Predicted	45%	36%	17%	4%
Difference	0%	-2%	0%	+2%

Therefore, in Nipissing-Timiskaming, the model projected the Conservatives 2% too high and the Greens 2% too low. Thus, the model allocated 2% of the vote incorrectly. Therefore the accuracy in this riding is 98%.

DemocraticSPACE would like to thank everyone who visited us during the 2006 Election Campaign. The site attract considerable interest nationwide. FYI, the site statistics on election day (23 Jan 2006) were:

Hits: 2,624,240
Page Hits: 202,709
Unique Visitors: 40,917

DemocraticSPACE once again produced the most accurate published election prediction in the country, edging out (by just a hair) the equally astute Election Prediction Project, run by Milton Chan and based solely on gauging comments by voters.

DemocraticSPACE correctly predicted the winner in 92% of Canada’s ridings (283 ridings). Of the 25 ridings we incorrectly predicted, 17 were in our ‘too close to call’ category (i.e. margin of victory of less than 5%), leaving just 8 seats out of 308 that were off the mark. The Election Prediction Project’s call of C118/L104/B56/N29/O1, however, was closer to the mark (i.e. more ridings were wrong, but they cancelled each other out), being off by a net of 6 seats (vs. a net of 9 seats for DemocraticSPACE). Congrats to Milton.

What is equally clear is that the major polling companies, in particular Strategic Counsel (Allan Gregg) and Ipsos-Reid (Darrell Bricker) need to seriously adjust their seat projection models, or, better yet, hire DemocraticSPACE to do it for them! Ipsos-Reid projected 148-152 (150 average) for the Conservatives, while Strategic Counsel put the Conservatives inteh 140-145 range — both well off the 124 seats they received and nowhere near the 128 seats projected by DemocraticSPACE. What say you, Allan? Darrell? Next time, give us a call.

Edit: as people rightfully note in the comments, the UBC stock market also came very close to the final seat tally (off my a net of 7 seats). This prediction is a slightly different monster, as it (like many others) are not riding-by-riding predictions, but rather a more intuitive guess of the final tally (in particular, using a market approach). The above post is meant only to compare against others that also did riding-by-riding predictions.

DemocraticSPACE Predicts Conservative Minority Government
By Gregory D. Morrow

DemocraticSPACE is projecting that a Conservative Minority Government will most likely result from Monday’s election in Canada. The fate of the nation rests with 58 very close ridings across the country. With nearly half of these (28 ridings) located in Ontario, the province will once again be the deciding factor in the election. The democraticSPACE projection model is forecasting the most likely scenario will be the following (projected seats in parenthesis):

CONSERVATIVE			125-130 seats		(128 seats)
LIBERAL			90-95 seats		(94 seats)
BLOC QUEBECOIS			55-60 seats		(56 seats)
NDP			25-30 seats		(29 seats)
INDEPENDENT			1 seat		(1 seat)

DOWNLOAD GRAPH AS PDF

While a last-minute shift back to the Liberals would give the Liberals a narrow minority, the possibility of this occurance is low. The likelihood of a Conservative majority, barring a ballot-box flood of votes to the Tories, also remains low. The following chart summarizes the best- and worst-case scenarios for each of the major parties (worst case = lose all of very close ridings, best case = win all of very close ridings):

The best-case for the Conservatives would look something like this - this would require the Conservatives win all or most of the very close ridings (projected seats in parenthesis):

CONSERVATIVE			140-145 seats		(144 seats)
LIBERAL			75-80 seats		(76 seats)
BLOC QUEBECOIS			50-55 seats		(52 seats)
NDP			30-35 seats		(35 seats)
INDEPENDENT			1 seat		(1 seat)

The best-case for the Liberals would look something like this - this would require the Liberals win all or most of the very close ridings (projected seats in parenthesis):

LIBERAL			120-125 seats		(125 seats)
CONSERVATIVE			100-105 seats		(102 seats)
BLOC QUEBECOIS			50-55 seats		(52 seats)
NDP			25-30 seats		(28 seats)
INDEPENDENT			1 seat		(1 seat)

The FINAL (Jan 22) Election Update indicates that we are headed for a Conservative Minority Government, with both the Bloc Quebecois and NDP (barely) able to combine to reach the required 155 seats needed to pass legislation.

	FINAL 2006 PROJECTED RESULTS - 22 Jan
	CPC	LPC	BQ	NDP	IND	GRN
TOTAL SEATS	128	94	56	29	1	0
% OF SEATS	42%	31%	18%	9%	0%	0%

Note:
CPC+BQ=184
CPC+NDP=157
LPC+BQ=150
LPC+NDP=123

The following is the FINAL support levels, as generated by the democraticSPACE projection model.

REGION
NATIONAL	36.8%	29.5%	17.3%	10.7%	5.0%	0.7%
ONTARIO	36.1%	37.9%	20.1%	-	5.5%	0.4%
QUEBEC	25.6%	16.7%	8.2%	44.9%	3.3%	1.3%
BRITISH COLUMBIA	36.4%	30.3%	26.7%	-	6.0%	0.6%
ALBERTA	62.5%	15.8%	13.6%	-	7.0%	1.1%
PRAIRIES	48.2%	24.2%	22.3%	-	4.0%	0.7%
ATLANTIC	36.4%	40.8%	19.7%	-	2.6%	0.5%

The following is the projected seat distribution by region:

REGION
NATIONAL	128	94	29	56	0	1
ONTARIO	44	49	13	-	0	0
QUEBEC	6	12	0	56	0	1
BRITISH COLUMBIA	19	9	8	-	0	0
ALBERTA	28	0	0	-	0	0
PRAIRIES	20	4	4	-	0	0
ATLANTIC	11	18	3	-	0	0

Regional Analysis: 905 Region
By Gregory D. Morrow

The suburban 905 belt around Toronto should prove to be a bellweather region in Monday’s election, according to democraticSPACE projections. Seven ridings are projected to change hands, as suburban voters respond to targeted Conservative policies aimed at the middle-income earners who largely comprise the 905. One of the most significant and projected handily defeats will be Human Resources Minister Belinda Stronach, who jumped to the Liberals after being elected in 2004 as a Conservative. With much bitterness and cynicism over this move, a determined Conservative effort stands to reap its rewards on Monday, with Lois Brown taking Newmarket-Aurora. International Cooperation Minister Aileen Carroll appears set to lose her seat to Conservative Patrick Brown in Barrie. Former Provincial PC Finance Minister Jim Flaherty appears poised to claim Whitby-Oshawa from Liberal Judi Longfield. Four Mississauga-area seats are projected to swing to the Conservatives: Mississauga-Erindale (Bob Dechert projected to win over newcomer Omar Alghabra, who replaces maverick Carolyn Parrish on the Liberal ticket), Mississauga-Streetsville (Raminder Gill edging Wadij Khan), Brampton West (Baljit Gosal edging Colleen Beaumier) and Bramalea-Gore-Malton (John Sprovieri edging Gurmax Balhi). Each of these seats are very close, however, and could swing either way.

PROJECTIONS
Conservative: 12 (+7)
Liberal: 9 (-7)
NDP: 0

Regional Analysis: Toronto (416 Region)
By Gregory D. Morrow

Toronto should remain a Liberal stronghold, despite some gains by the Conservatives in Etobicoke and Scarborough. One of the most hotly-contested races is in Etobicoke-Lakeshore where all-star (parachute) candidate Michael Ignatieff is projected to edge out Conservative John Capobianco, however with just a 2.9% projected margin of victory, this riding could go either way. The NDP stands to make solid gains in 4 downtown ridings, and are projected to take 2 of them, with Peggy Nash edging out Sam Bulte in Parkdale-High Park and Olivia Chow edging out Tony Ianno in Trinity-Spadina. With margins of less than 5%, however, the incumbents could quite easily hold these seats, especially if NDP strategic voting is a factor. Despite a strong NDP candidate in Marilyn Churley, democraticSPACE is projecting incumbent Maria Minna to hold Beaches-East York, as the presence of Green Party leader Jim Harris in the riding could take a few hundred votes need to win. Likewise, NDP activist Gord Perks should make strong gains in Davenport, a long-time Liberal riding, but incumbent Mario Silva is projected to hold on. Like the previous two, these latter ridings are also too close to call. The remaining ridings in Toronto proper project to hold for the Liberals

PROJECTIONS
Liberal: 21 (-2)
NDP: 3 (+2)
Conservative: 0

Regional Analysis: Golden Horseshoe
By Gregory D. Morrow

Significant change and extremely tight races make the Golden Horseshoe a region to watch on Monday, with 7 of 12 ridings projected to change hands and half with projected margins of victory of less than 5%. If the democraticSPACE projections prove correct, this region could represent a significant loss for the Liberal Party, dropping from 9 seats to 2. The most significant change is the projected loss of Liberal House Leader Tony Valeri to NDP Wayne Marston in Hamilton East-Stoney Creek. The retirement of Liberal incumbent Beth Phinney opens the door for a tight 3-way race in Hamilton Mountain, projected to go to NDP Chris Charlton (given the narrow 4.5% margin, it should be considered too close to call). Liberals seats in Brant, Ancaster, Halton, St. Catharines, and Burlington all project to flip to the Conservatives, but with margins of victory of less than 5%, these ridings are up for grabs for both Liberals and Conservatives. Notable is the projection of Former MP Garth Turner to narrowly wining over Gary Carr in Halton and long-time Liberal Walt Lastewka narrowly losing his seat to Rick Dykstra.

PROJECTIONS
Conservative: 7 (+5)
NDP: 3 (+2)
Liberal: 2 (-7)

Regional Analysis: Southwestern Ontario
By Gregory D. Morrow

With three Liberal incumbents retiring, the Conservatives appear poised to pick up two, with the NDP potentially gaining the third. The departure of Rose-Marie Ur (Lambton-Kent-Essex) and Jerry Pickard (Chatham-Kent-Essex) almost certainly assures Conservatives Bev Shipley and David Van Kesteren comfortable wins. All three London ridings are hotly contested, with Labour Minister Joe Fontana and Sue Barnes projected to hold their seats. London-Fanshawe, formerly the riding of Liberal-turned-Independent MP Pat O’Brien, is one of a handful of tight 3-way races in the country, winnable for all three major parties (democraticSPACE projects a narrow win for NDP Irene Mattysson, although given the 2.9% margin, it must be considered too close to call). Also too close to call is Sarnia-Lambton, where long-time Liberal incumbent Roger Gallaway is projected to hold his seat by just 2.7% (thus, it is too close to call). Windsor is projected to be much closer than previous campaigns, especially in Windsor-Tecumseh. Both NDP incumbents (Brian Masse and Joe Comartin) are projected to hold their seats. The remaining rural ridings should be solidly Conservative.

PROJECTIONS
Conservative: 6 (+2)
Liberal: 3 (-3)
NDP: 3 (+1)

Regional Analysis: Central Ontario
By Gregory D. Morrow

Just one seat is projected to change hands in Central Ontario, according to democraticSPACE projections, with former pastor Harold Albrecht projected to take Kitchener-Conestoga from Liberal incumbent Lynn Myers (although with just a 3/10ths of a percent margin, this riding should be considered too close to call). Despite strong challenges from their Conservative opponents, Brenda Chamberlain (Guelph) and Paul Steckle (Huron-Bruce) are projected to hold their seats for the Liberals. Gary Goodyear’s razor-thin victory in 2004 should be more comfortable this time around.

PROJECTIONS
Conservative: 5 (+1)
Liberal: 4 (-1)
NDP: 0