Monday, October 02, 2006
Steve and Microsoft Excel Team Up to Bring You the First Ballot Results
So, as we're all aware by now, 409 of the Liberals' 469 delegate selection meetings have reported their results. As we're also all aware, the delegates chosen by these meetings do not make up the totality of the delegates who will pick the party's next leader - there are also ex-officio delegates, as well as the delegates from the 60 delegate selection meetings yet to report. Besides that, 115 of the delegates who were selected on the weekend have yet to declare their intentions.
I've done a little bit of number-crunching to predict what the results will wind up being on the first ballot. I won't go too far into my methodology, because doing so will probably expose me to accusations of "ignoring biases inherent in the available data" or "being totally incompetent" - if anybody wants a copy of the spreadsheet, my e-mail address is in my profile and I'd be happy to send it to you.
In brief, the way I treated the uncommitted delegates was to assume that they'd break exactly the same way as the other delegates in their provinces. From there, we get the following first ballot results:
Michael Ignatieff - 30.7%
Bob Rae - 20.4%
Gerard Kennedy - 17.7%
Stéphane Dion - 17.3%
Ken Dryden - 4.8%
Joe Volpe - 4.7%
Scott Brison - 4.0%
Martha Hall Findlay - 1.0%
These results are similar, but not identical, to the ones being published as Super Weekend's "results" (the difference comes from the fact that I've allocated the uncommitted delegates by province rather than assuming that they'll break according to the national totals).
Next, I factored in the delegates from the 60 missing meetings. This was tough, because all I knew about these meetings was their provinces - I had no idea how many delegates were to be elected at each one (with a few exceptions, such as Nunavit's missing meeting which was surely for the 14 delegates allocated to the riding). I calculated the total number of delegates who were supposed to be elected this weekend as being 4865 (source: the Hill Times), and assumed that the missing 671 delegates were evenly-divided between the 60 seats (in fact, this isn't the case - besides the fact that the missing meetings are of different types, it's likely that some meetings failed to elect their full allotment of delegates - but it's the best assumption available under the circumstances). This means that each of the 60 missing meetings will elect 11.2 delegates. I then apportioned these, too, by province, to come up with a total elected delegate count:
Michael Ingatieff - 30.2%
Bob Rae - 20.8%
Gerard Kennedy - 17.7%
Stéphane Dion - 17.1%
Joe Volpe - 4.8%
Ken Dryden - 4.8%
Scott Brison - 3.7%
Martha Hall Findlay - 1.0%
Basically, this means that Ignatieff and Dion should slip slightly, Rae should gain a little, Volpe should pass Dryden (by all of 3 delegates - assuredly within this simulation's margin of error, which I estimate to be approximately a billion delegates), and Brison should do even worse than he already appears to be doing (which makes sense, since Nova Scotia's meetings have all reported). The only thing that is totally outlandish about my model's predictions here is that it gives Kennedy all 14 of Nunavit's delegates, on the basis that the only one we've yet seen is a Kennedy guy.
Of course, there are also the much ballyhooed ex-officio delegates - MPs, Senators, defeated/nominated candidates, riding association presidents, members of various Liberal boards, privy councillors, etc. According to Wikipedia, there should be about 867 such delegates, of whom 445 have gone on record as backing a specific candidate. At this point, all I did was project the intentions of these 445 across the whole 867, which gives the following results among ex-officios only:
Michael Ignatieff - 34.6%
Stéphane Dion - 18.2%
Gerard Kennedy - 15.5%
Bob Rae - 12.8%
Ken Dryden - 9.2%
Scott Brison - 7.2%
Joe Volpe - 1.6%
Martha Hall Findlay - 0.9%
It might be interesting to further divide/project the ex-officios on the basis of such things as geographic location and type of ex-officio (MP, Senator, etc.), but for the time being this is what I've got, and I saw little reason that the 445 who have declared should be seriously unrepresentative of the whole crop.
Combining the two categories (elected and ex-officio), we get the following results:
Michael Ignatieff - 30.9%
Bob Rae - 19.6%
Gerard Kennedy - 17.3%
Stéphane Dion - 17.2%
Ken Dryden - 5.4%
Joe Volpe - 4.3%
Scott Brison - 4.2%
Martha Hall Findlay - 1.0%
The ex-officios will allow Dryden to save a little face, but won't help Brison pull past anybody.
Is this an accurate prediction of what the first ballot will look like? I'm inclined to say yes, despite some factors suggesting the opposite, such as
1. the various sources of uncertainty in my model,
2. the fact that a high proportion of delegates won't actually show up to the convention, and the fact that the question of which delegates fail to show is not geographically neutral, and
3. the fact that those delegates who aren't committed one way or another on the first ballot may decide to back a perceived winner, or vote to stop Ignatieff, or what have you.
Predictions by such (cough) luminaries as Warren Kinsella and Jeffrey Simpson that Michael Ignatieff will hit 35% support on the first ballot seem unsupported by the data.
What does this mean about who will wind up winning the leadership race? I'll save those thoughts for a later post, but I'm no longer nearly as confident as I was in my Rae prediction.
|
So, as we're all aware by now, 409 of the Liberals' 469 delegate selection meetings have reported their results. As we're also all aware, the delegates chosen by these meetings do not make up the totality of the delegates who will pick the party's next leader - there are also ex-officio delegates, as well as the delegates from the 60 delegate selection meetings yet to report. Besides that, 115 of the delegates who were selected on the weekend have yet to declare their intentions.
I've done a little bit of number-crunching to predict what the results will wind up being on the first ballot. I won't go too far into my methodology, because doing so will probably expose me to accusations of "ignoring biases inherent in the available data" or "being totally incompetent" - if anybody wants a copy of the spreadsheet, my e-mail address is in my profile and I'd be happy to send it to you.
In brief, the way I treated the uncommitted delegates was to assume that they'd break exactly the same way as the other delegates in their provinces. From there, we get the following first ballot results:
Michael Ignatieff - 30.7%
Bob Rae - 20.4%
Gerard Kennedy - 17.7%
Stéphane Dion - 17.3%
Ken Dryden - 4.8%
Joe Volpe - 4.7%
Scott Brison - 4.0%
Martha Hall Findlay - 1.0%
These results are similar, but not identical, to the ones being published as Super Weekend's "results" (the difference comes from the fact that I've allocated the uncommitted delegates by province rather than assuming that they'll break according to the national totals).
Next, I factored in the delegates from the 60 missing meetings. This was tough, because all I knew about these meetings was their provinces - I had no idea how many delegates were to be elected at each one (with a few exceptions, such as Nunavit's missing meeting which was surely for the 14 delegates allocated to the riding). I calculated the total number of delegates who were supposed to be elected this weekend as being 4865 (source: the Hill Times), and assumed that the missing 671 delegates were evenly-divided between the 60 seats (in fact, this isn't the case - besides the fact that the missing meetings are of different types, it's likely that some meetings failed to elect their full allotment of delegates - but it's the best assumption available under the circumstances). This means that each of the 60 missing meetings will elect 11.2 delegates. I then apportioned these, too, by province, to come up with a total elected delegate count:
Michael Ingatieff - 30.2%
Bob Rae - 20.8%
Gerard Kennedy - 17.7%
Stéphane Dion - 17.1%
Joe Volpe - 4.8%
Ken Dryden - 4.8%
Scott Brison - 3.7%
Martha Hall Findlay - 1.0%
Basically, this means that Ignatieff and Dion should slip slightly, Rae should gain a little, Volpe should pass Dryden (by all of 3 delegates - assuredly within this simulation's margin of error, which I estimate to be approximately a billion delegates), and Brison should do even worse than he already appears to be doing (which makes sense, since Nova Scotia's meetings have all reported). The only thing that is totally outlandish about my model's predictions here is that it gives Kennedy all 14 of Nunavit's delegates, on the basis that the only one we've yet seen is a Kennedy guy.
Of course, there are also the much ballyhooed ex-officio delegates - MPs, Senators, defeated/nominated candidates, riding association presidents, members of various Liberal boards, privy councillors, etc. According to Wikipedia, there should be about 867 such delegates, of whom 445 have gone on record as backing a specific candidate. At this point, all I did was project the intentions of these 445 across the whole 867, which gives the following results among ex-officios only:
Michael Ignatieff - 34.6%
Stéphane Dion - 18.2%
Gerard Kennedy - 15.5%
Bob Rae - 12.8%
Ken Dryden - 9.2%
Scott Brison - 7.2%
Joe Volpe - 1.6%
Martha Hall Findlay - 0.9%
It might be interesting to further divide/project the ex-officios on the basis of such things as geographic location and type of ex-officio (MP, Senator, etc.), but for the time being this is what I've got, and I saw little reason that the 445 who have declared should be seriously unrepresentative of the whole crop.
Combining the two categories (elected and ex-officio), we get the following results:
Michael Ignatieff - 30.9%
Bob Rae - 19.6%
Gerard Kennedy - 17.3%
Stéphane Dion - 17.2%
Ken Dryden - 5.4%
Joe Volpe - 4.3%
Scott Brison - 4.2%
Martha Hall Findlay - 1.0%
The ex-officios will allow Dryden to save a little face, but won't help Brison pull past anybody.
Is this an accurate prediction of what the first ballot will look like? I'm inclined to say yes, despite some factors suggesting the opposite, such as
1. the various sources of uncertainty in my model,
2. the fact that a high proportion of delegates won't actually show up to the convention, and the fact that the question of which delegates fail to show is not geographically neutral, and
3. the fact that those delegates who aren't committed one way or another on the first ballot may decide to back a perceived winner, or vote to stop Ignatieff, or what have you.
Predictions by such (cough) luminaries as Warren Kinsella and Jeffrey Simpson that Michael Ignatieff will hit 35% support on the first ballot seem unsupported by the data.
What does this mean about who will wind up winning the leadership race? I'll save those thoughts for a later post, but I'm no longer nearly as confident as I was in my Rae prediction.
