Canada - On the Reliability of Projections

2006 National Projected from 2004:
Con: 36.3 Lib: 30.3 Ndp: 17.4 Blc: 10.5
Con: 134 Lib: 93 Ndp: 27 Blc: 53
Con: 133.0 Lib: 92.6 Ndp: 28.0 Blc: 53.4

2006 Regional Projected from 2004:

	Con	Lib	Ndp	Blc
East	34.5	39.9	22.7	0.0
PQ	24.6	20.7	7.5	42.1
ON	35.1	39.9	19.5	0.0
MB/SK	45.7	24.3	24.7	0.0
AB	64.9	15.4	11.6	0.0
BC	37.3	27.9	28.3	0.0

	Con	Lib	Ndp	Blc	Con	Lib	Ndp	Blc	Con	Lib	Ndp	Blc
East	8	21	3	0	9.5	19.3	3.1	0.0	7	22	3	0
PQ	5	12	0	58	4.9	13.2	0.0	57.0	0	21	0	54
ON	39	56	11	0	37.6	56.8	11.6	0.0	24	75	7	0
MB/SK	20	4	4	0	19.7	3.6	4.7	0.0	20	4	4	0
AB	28	0	0	0	27.8	0.2	0.0	0.0	26	2	0	0
BC	21	10	7	0	19.9	9.7	8.4	0.0	22	11	5	0
All	121	103	25	58	119.4	102.7	27.9	57.0	99	135	19	54

Actual:

	Con	Lib	Ndp	Blc
East	9	20	3	0.0
PQ	10	13	0	51
ON	40	54	12	0.0
MB/SK	20	5	3	0
AB	28	0	0	0
BC	17	10	9	0
All	124	103	29	51

Clearly, the national projection wasn't very accurate. It was off by 10 seats. The regional projection was better nationally, although it had big problems in PQ and some problems and BC. It got the liberal seat count right on, while the national model was not close. I will get into the "lucky effect" in the next post.