Monday, September 26, 2011

Ontario election 2011 prediction (september 26)

Outline by S. Smith
This past weekend, Forum Research published the largest poll in Canadian history for the provincial election in Ontario. 40,000 people were polled, asking which party they were supporting. This poll was so large, individual riding results were also published within a certain degree of accuracy.

For this projection, I will only be looking at this particular poll, due to the fact that there is no way you can get more accurate than 40,000 respondents- the margin of error was just 0.5%. The results of the poll are quite interesting; it showed an exact tie between the Tories and the Liberals at 35%, with the NDP trailing not too far behind at 23%. The Greens meanwhile are at a respectable 6% province wide.

Looking at individual ridings, according to Forum Research these numbers translate to 47 seats each for the Liberals and Tories and just 13 for NDP. This is where the errors comes in. With a tie vote, the Liberals should have the edge in seats, but they don't. This can be explained by large rural swings against them, which I factored into my projection. But one statistic that doesn't sit right is the 13 seats for the NDP. That would mean a net increase of just three seats on a swing of 6%. This seams ridiculous considering the party was able to win 17 seats on just 18% of the vote in the 2008 federal election. But who knows, perhaps the party is doing well, but not in any concentrations.

The report published in the Toronto Star indicated that there were 28 seats within 5% of the vote, but didn't say who was leading or in second place in them. They were: Ancaster—Dundas—Flamborough--Westdale, Beaches--East York, Bramalea—Gore—Malton, Brampton—Springdale, Halton, Kitchener Centre, Kitchener—Conestoga, Kitchener—Waterloo, London North Centre, London West, Mississauga East, Mississauga—Erindale, Mississauga South, Niagara Falls, Oak Ridges—Markham, Ottawa South, Pickering--Scarborough East, Richmond Hill, St. Catharines, Thornhill, Thunder Bay—Atikokan, Thunder Bay--Superior North, Timmins--James Bay, Trinity—Spadina, Welland, Windsor—Tecumseh, Windsor West and York Centre.

Some of those seats make sense, but others are quite odd, like McGuinty in danger of losing his seat in Ottawa South, or the NDP at risk of losing Timmins—James Bay. Additionally, there are some notable absences from this list, like my own riding of Ottawa Centre. Some of the individual riding sample sizes are quite low however. Timmins—James Bay had just 54 respondents in the poll.

Sunday morning I looked for some individual riding numbers, and found some for the Greater Toronto Area. Looking at the map found that most ridings seemed to be accurate, but there are a few oddities. Specifically, how poor the NDP seems to be doing in Toronto, despite polls from earlier last week showing the party was doing better than they were at federally. Forum Research has the party losing Parkdale—High Park surprisingly, and in danger of losing the two strongholds of Trinity—Spadina and Beaches—East York. Also, the party is way behind in York South—Weston, despite some indications of a strong NDP campaign there. Also, the Tories being ahead in two Brampton ridings is a bit of a surprise, considering I would have expected other ridings to go first. I suppose many Indo-Canadians are warming up to the provincial Tories at the expense of the Liberals.

I also found some numbers from Bruce, Grey and Simcoe Counties. The big surprise here is the collapse of the Liberal vote in Huron—Bruce which is part of the reason why I suspect the Liberals are way behind in their rural seats.

And so, I am left to make a new projection from all of this. I didn't just copy and paste the poll numbers into the ridings, but I did make some adjustments where I could find the numbers. Some wonky ridings I nearly ignored completely, but then again, it's possible the polls are correct about those seats. I will be taking a close look at these seats in the next week and a half before election day.


Seat by seat projection

Riding
Lib
PC
NDP
Grn
Oth
Projected winner
Ajax—Pickering
43
37
13
6
1
Joe Dickson
Algoma—Manitoulin
37
15
41
5
1
Michael Mantha
Ancaster—Dundas—Flamborough—Westdale
36
41
16
6
1
Donna Skelly
Barrie
26
47
19
7
1
Rod Jackson
Beaches—East York
32
18
39
10
1
Michael Prue
Bramalea—Gore—Malton
36
32
27
4
1
Kuldip Kular
Brampton West
42
38
16
4
0
Vic Dhillon
Brampton—Springdale
38
39
17
5
1
Pam Hundal
Brant
40
34
20
4
2
Dave Levac
Bruce—Grey—Owen Sound
29
42
13
11
5
Bill Walker
Burlington
31
46
17
5
1
Jane McKenna
Cambridge
25
45
24
5
1
Rob Leone
Carleton—Mississippi Mills
35
47
12
5
1
Jack MacLaren
Chatham-Kent—Essex
27
45
23
5
-
Rick Nicholls
Davenport
35
11
47
5
2
Jonah Schein
Don Valley East
43
30
22
4
1
Michael Coteau
Don Valley West
47
38
8
6
1
Kathleen Wynne
Dufferin—Caledon
24
49
10
17
0
Sylvia Jones
Durham
27
47
19
6
1
John O'Toole
Eglinton—Lawrence
50
36
9
4
1
Mike Colle
Elgin—Middlesex—London
25
49
21
4
1
Jeff Yurek
Essex
24
40
32
4
-
Dave Brister
Etobicoke Centre
47
35
12
4
2
Donna Cansfield
Etobicoke North
48
25
21
4
2
Shafiq Qaadri
Etobicoke—Lakeshore
41
34
17
6
2
Laurel Broten
Glengarry—Prescott—Russell
37
44
13
5
1
Marlissa Gosselin
Guelph
45
28
15
11
1
Liz Sandals
Haldimand—Norfolk
20
60
15
4
1
Toby Barrett
Haliburton—Kawartha Lakes—Brock
34
47
13
6
0
Laurie Scott
Halton
34
46
14
4
1
Ted Chudleigh
Hamilton Centre
23
18
50
7
2
Andrea Horwath
Hamilton East—Stoney Creek
26
28
40
4
2
Paul Miller
Hamilton Mountain
30
27
37
4
2
Monique Taylor
Huron—Bruce
32
38
25
4
1
Lisa Thompson
Kenora—Rainy River
29
31
36
4
0
Sarah Campbell
Kingston and the Islands
44
29
20
5
2
John Gerretsen
Kitchener Centre
40
35
19
5
1
John Milloy
Kitchener—Conestoga
36
41
18
5
-
Michael Harris
Kitchener—Waterloo
37
43
14
5
1
Elizabeth Witmer
Lambton—Kent—Middlesex
35
40
19
4
2
Monte McNaughton
Lanark—Frontenac—Lennox and Addington
33
43
18
6
-
Randy Hiller
Leeds—Grenville
25
55
14
6
0
Steve Clark
London North Centre
39
32
21
7
1
Deb Matthews
London West
44
38
17
4
1
Chris Bentley
London—Fanshawe
35
29
30
4
2
Khalil Ramal
Markham—Unionville
48
29
18
4
1
Michael Chan
Mississauga East—Cooksville
42
38
15
4
1
Dipika Damerla
Mississauga South
44
40
11
4
1
Charles Sousa
Mississauga—Brampton South
43
36
16
4
1
Armit Mangat
Mississauga—Erindale
42
38
15
4
1
Harinder Takhar
Mississauga—Streetsville
44
36
13
7
-
Bob Delaney
Nepean—Carleton
30
49
13
7
1
Lisa MacLeod
Newmarket—Aurora
34
46
14
6
-
Frank Klees
Niagara Falls
42
36
16
5
2
Kim Craitor
Niagara West—Glanbrook
21
54
18
5
2
Tim Hudak
Nickel Belt
25
20
51
4
-
France Gelinas
Nipissing
34
46
16
4
-
Victor Fedeli
Northumberland—Quinte West
35
40
18
6
1
Rob Milligan
Oak Ridges—Markham
39
42
14
4
1
Farid Wassef
Oakville
43
39
11
5
2
Kevin Flynn
Oshawa
18
42
35
4
1
Jerry Ouellette
Ottawa Centre
32
20
39
8
1
Anil Naidoo
Ottawa South
42
35
16
6
1
Dalton McGuinty
Ottawa West—Nepean
37
41
16
5
1
Randall Denley
Ottawa—Orleans
43
41
11
4
1
Phil McNeely
Ottawa—Vanier
41
27
24
7
1
Madeleine Meilleur
Oxford
23
52
18
5
2
Ernie Hardeman
Parkdale—High Park
32
14
44
7
3
Cheri DiNovo
Parry Sound—Muskoka
19
48
22
10
1
Norm Miller
Perth—Wellington
37
42
13
6
2
Randy Pettapiece
Peterborough
42
29
23
5
1
Jeff Leal
Pickering—Scarborough East
42
36
16
5
1
Tracey MacCharles
Prince Edward—Hastings
34
39
21
4
2
Todd Smith
Renfrew—Nipissing—Pembroke
18
65
13
4
0
John Yakabuski
Richmond Hill
42
37
14
6
1
Reza Moridi
St. Catharines
40
35
18
5
2
Jim Bradley
St. Paul's
43
28
19
8
2
Eric Hoskins
Sarnia—Lambton
23
45
26
4
2
Bob Bailey
Sault Ste. Marie
51
16
27
4
2
David Orazietti
Scarborough Centre
43
28
24
4
1
Brad Duguid
Scarborough Southwest
39
24
31
5
1
Lorenzo Berardinetti
Scarborough—Agincourt
50
27
17
4
2
Soo Wong
Scarborough—Guildwood
41
30
23
4
2
Margarett Best
Scarborough—Rouge River
45
21
28
3
3
Bas Balkissoon
Simcoe North
24
52
17
7
-
Garfield Dunlop
Simcoe—Grey
23
54
13
10
-
Jim Wilson
Stormont—Dundas—South Glengarry
28
51
16
4
1
Jim McDonnell
Sudbury
49
14
32
4
1
Rick Bartolucci
Thornhill
41
45
9
4
1
Peter Shurman
Thunder Bay—Atikokan
32
23
41
4
0
Mary Kozorys
Thunder Bay—Superior North
42
12
42
4
0
Michael Gravelle or Steve Mantis
Timikaming—Cochrane
35
17
44
4
0
John Vanthof
Timmins—James Bay
29
22
45
4
0
Gilles Bisson
Toronto Centre
44
20
26
8
2
Glen Murray
Toronto—Danforth
26
11
52
9
2
Peter Tabuns
Trinity—Spadina
33
14
45
7
1
Rasario Marchese
Vaughan
51
33
11
4
1
Greg Sorbara
Welland
24
33
38
4
1
Cindy Forster
Wellington—Halton Hills
27
55
11
7
-
Ted Arnott
Whitby—Oshawa
26
48
19
6
1
Christine Elliott
Willowdale
43
36
15
5
1
David Zimmer
Windsor West
33
20
43
4
-
Helmi Charif
Windsor—Tecumseh
41
20
34
4
1
Dwight Duncan
York Centre
41
40
13
5
1
Monte Kwinter
York South—Weston
39
14
42
4
1
Paul Ferriera
York West
51
17
27
4
1
Mario Sergio
York—Simcoe
24
52
17
6
1
Julia Munro

Seat changes since last projection
The Tories were the main beneficiary of this projection, gaining six seats in my theory that they are winning all of the rural Liberal seats. These are Huron—Bruce, Lambton—Kent—Middlesex, Northumberland—Quinte West, Perth—Wellington and Prince Edward—Hastings. Also, the absence of Ottawa West—Nepean on that close race list means that I believe it is firmly in the Tory column as well, making it the sixth Tory gain. Three of those Tory gains were from projected toss up seats. The NDP also gained on seat in this projection, in Algoma—Manitoulin (from toss up). Brampton—Springdale and Ottawa—Orleans are new toss ups, as those previously Liberal projected seats are now close races.

Conclusion
While I didn't come up with a tie in seats like Forum Research did, I did get a close result. Either the Tories or the Liberals can win this, but it will be a minority. The Tories are presently two seats behind. Even though I shifted many Liberal seats into the Tory column, I still couldn't get to the number provided in the forum research poll. And, I have no idea where they have the NDP gaining seats, but I suspect their low sample sizes in the north have meant that their numbers weren't too NDP friendly. In any event, I have the NDP winning at least 17 seats, not 13. And they will be holding the balance of power. 



Methodology/disclaimer

I have looked at each riding individually and tried to figure out what its voters might do in the upcoming election. For this, I looked at the riding history, and tried to come to a conclusion as to why certain parties did as well they did, or as poor as they did compared to other years. For certain ridings, I assumed a uniform swing, because I felt its historical voting trends follow the provincial voting trends. For other ridings, I looked at the results of the recent federal election and adjusted the results to reflect the most recent poll. And for other ridings, I could only make estimates, again using past election data. It all depended on the character of the riding.

Of course, there is bound to be some oversights on my part. If you think I have a riding wrong, please let me know. I want to get this as right as possible. The key to making correct projections is not all about the math. Rather, it's about the local conditions. That's why I will be looking at qualitative data just as much as I will be looking at quantitative.

8 comments:

  1. How do you compensate for the IVR method? One can only claim +/-0.5% on 40,000 respondents if the sample is representative of the population?
    People who hang up immediately when they hear a recording instead of a real person bias the data. Is there any idea how much of an impact this has?

    ReplyDelete
  2. I think that those random errors are more likely on the riding by riding level. I don't know how they asked the questions, but they could have asked demographics. I know EKOS does. That would mean they could weight the results. But with 40,000 people, any skewing would probably be balanced out, probably.

    Also of interest, is the fact that application of respondents to ridings may be off. I doubt the respondents were asked which riding they lived in, so they were grouped by telephone exchange. That could mean massive errors for Toronto.

    ReplyDelete
  3. Any error associated with voter response or methodology is systematic not random. The size of the sample can't compensate for non-random sampling. This is the same argument as the mandatory long form census.

    Interesting point about phone exchanges - perhaps that partially explains the poor NDP results in down town Toronto. Highly concentrated support could be 'spread' into other ridings. This would preserve the overall party vote percentage but would ruin riding level predictions. If true that's one mother of a systematic error!

    ReplyDelete
  4. Im also starting to think some soft NDP-Liberal voters in Toronto will be voting Liberal to stop the Tories, because of vote splitting (as evidenced in the federal election). This may make sense in some ridings, and not in others. But, voters will vote strategically even it makes no sense. I can tell you that here in Ottawa Centre there are voters voting strategically for the Liberals to stop the Tories, when perhaps a vote for the NDP would make just as much if not more sense. (although the Tories have a next to nil chance of winning here).

    ReplyDelete
  5. I guess that all depends on if and how the dynamic shifts after last night. If the Tories are perceived to have some momentum that may very well happen. But if the Liberals seem to lack credibility they may shed soft support to both opposition parties or even experience a suppressed voter turn out on Eday.

    ReplyDelete
  6. Have you considered making a cartogram of Ontario proportional to population or ridings? I find that for a very unequally populated country like Canada, electoral maps are often very hard to do a visual scan of, since you end up focusing mostly on huge rural ridings. It would be great to have an Ontario template on Wikipedia so future map-makers to use.

    ReplyDelete
  7. I could do a cartogram, I guess. If I knew how to, I would make one of those stretchy cartograms based on population. It would be after the election(s) at any rate.

    ReplyDelete
  8. Oh we at the Green Party would have loved to have gotten some of those predicted results of 6-8% instead of 3%

    ReplyDelete