The Shotgun Blog
Wednesday, June 03, 2009
Ontario PC leadership poll analysis
Ipsos Reid has come out with a poll for the Ontario PC leadership race. The numbers they give are: Elliott (35%), Hudak (24%), Hillier (22%) and Klees (20%). Perez Hudak has a further breakdown of these numbers. Assuming Perez Hudak is correct, here are my calculation of the how the leadership balloting will look.
1st Ballot: The 1st place votes are as follows - Elliott (19%), Hudak (13%), Hillier (14%) and Klees (11%). Since people that don't vote will not be counted, I adjust these numbers out of 57% to get the first ballot results - Elliott (33%), Hudak (23%), Hillier (25%) and Klees (19%). Klees has the lowest total and would be out after the 1st ballot.
2nd Ballot: The 2nd choice vote totals for the remaining three are - Elliott (28%), Hudak (18%) and Hillier (14%). Since we are not given transition matrices of the preferences (ie. we don't know who people that picked Klees first support 2nd), I assume that all the candidates have the Klees votes distributed based on their 2nd choice totals. The 19% from Klees would be broken up as follows - Elliott (9.0%), Hudak (5.8%) and Hillier (4.5%). The second ballot results would be - Elliott (42.3%), Hudak (28.6%) and Hillier (29.1%). Hudak would be out after the 2nd ballot.
3rd Ballot: Assuming the the Hudak votes split at the 2-1 ratio for Elliott (since her 2nd choice was twice as much as Hillier's), she would win easily with 61% of the final ballot.
There are some obvious criticisms of this poll. The most common one that I'm seeing is that they didn't have access to party lists, so they were just using people that self identify as PC supporters, whether or not they are party members. While this is completely true, it is partially mitigated by the 43% of respondents that did not answer. I'm sure some of the 43% could be party members that have yet to make up their minds, it seems more likely that these would not be members. Of the remaining 57% that did answer, there probably is a higher proportion of party members. While using PC supporters instead of exclusively using members is a problem, I don't think it is the biggest problem with the poll. If we break up the PC supporters into two categories: non-members and members. Do we have any evidence that the subset of members is different than the subset of non-members? If Ipsos Reid did ask if the respondent was a member, even with the smaller sample size, I would be curious to see if this subset was much different from the non-members.
I think the biggest issue with the poll is that the Ontario PC Party does not vote by a pure one-member one-vote system. Each riding is given 100 Electoral Votes (unless there are less than 100 votes cast in a riding, in which case the EV's would equal the votes cast). They way to combat this would be to weight the results by ridings and I don't see it indicated that this was done.
Even with the above criticisms, I don't see an obviously better (and cost effective) polling strategy. The only people with the membership lists are the campaigns, and I wouldn't trust a poll out of any of their camps. Ipsos Reid presumably has no motive to sway the results of the race and actually has incentives to be correct.
I am not confident that Christine Elliott really has this big of a lead, but it certainly should raise some questions about the conventional wisdom that Hudak is in the lead. I think it will come down to whichever candidate actually did sell the most memberships to their suppporters.
I heard the poll was conducted over a month, and began before the May 14th cut-off date. Is that true?
Posted by: Christian | 2009-06-03 11:19:44 AM
Frank Klees is in 4th place. Okay! Excellent analysis there. Not knowing how the poll was conducted (oh, sorry, "assuming Perez Hudak is correct") this tells us absolutely nothing.
Posted by: FreakyStyley | 2009-06-03 11:42:41 AM
Since people that don't vote will not be counted,
Posted by William Joseph on June 3, 2009
I'm glad you clarified that.
Posted by: The Stig | 2009-06-03 11:58:07 AM
Christian, I am not sure of the exact timing of the poll. The cut-off date seems to be irrelevent though since they weren't only polling members.
FreakyStyley, do you have any reason to believe Klees isn't in 4th?
Also, the breakdown of 1st and 2nd place support was done by Ipsos Reid. It isn't that unbelievable to think that Perez Hudak was able to get these numbers.
Posted by: William Joseph | 2009-06-03 11:59:47 AM
What is the point of releasing "statistics" with a 100% chance of being wrong?
Posted by: Wow, Invent a numbers | 2009-06-03 1:53:01 PM
Stronach has appeal, poll finds; Candidate scores high in popularity with Canadians, campaign workers say
4 February 2004
The Globe and Mail
Campaign workers for Belinda Stronach have buttressed their argument that opponent Stephen Harper won't sell to the general public, citing a poll saying their candidate is more popular.
The survey, conducted by the firm of Ms. Stronach's campaign director, shows that Canadians who didn't identify their political preference are more willing to give Ms. Stronach a try than they would Mr. Harper or Tony Clement.
Respondents were asked whom they would vote for if they were members of the Conservative Party, with the results that 41 would opt for Ms. Stronach, 28 per cent for Mr. Harper and 19 per cent for Mr. Clement.
Posted by: Christine Is The New Belinda | 2009-06-03 8:29:16 PM
What of waste of time and Internet space. Detailed predictions based on a meaningless and unrepresentative poll. Out of "nearly 500 respondents" as indicated the National Post's review of the poll, do you really think that's representative of anything. They don't give the margin of error!
"Even with the above criticisms, I don't see an obviously better (and cost effective) polling strategy."
I do, abstinence. This poll was only useful to tell the trend between the Liberals and PCs in Ontario. The rest is hot air.
Posted by: Nathalie | 2009-06-04 5:08:13 AM
William -- I am glad there is nothing else going on in Ontario like an eHealth scandal or a GM bailout or a tax hike or a border closure or a recession for you to write about. What will you do when the leadership race is over and you have to analyze real news instead of meaningless polls that tell us nothing about how this race will turn out? Wait a second: ARe you Peres Hudak???
Posted by: Charlie knows best | 2009-06-04 5:55:00 AM
It's amazing that people find this so hard to understand: the leadership will be decided by party members. Hudak and Klees signed up the most members (14000 and 9800 respectively), yet they are the first two out? Hillier who signed up only 3200 makes it onto the final ballot, along with Elliott, who signed up only 8100? Not gonna happen!
Ipsos is, at best, a six week old test of name recognition. It was in the field in April, long before the membership cutoff. 56% of people who said they support the Conservatives couldn't identify two candidates they would support, even when prompted with a list. Maybe Elliott and Hillier had gotten the most press by that point (Elliott because she's married to Jim Flaherty, Hillier because he released some controversial policies at the beginning). But this has absolutely nothing whatsoever to do with who is going to win.
Posted by: thinktwice | 2009-06-04 8:31:00 AM
An interesting analysis, but you fail to take into account the 100 points per riding. You can't use simple percentages. The race will now be determined by each campaigns ground game in all 107 ridings. Some ridings only 1 or 2 votes will equal a point where others will require 5 or 10, depending on the total number of members in the riding.
Posted by: BBS | 2009-06-04 5:32:06 PM
BBS, I actually explicitly mentioned this flaw. Read the 3rd last paragraph. I did my best based on the limited nature of the poll. Give me better numbers and I'll do more with it.
Charlie knows best, I write about what I feel like. I'm interested in the leadership race so most of my posts are about it. There are plenty of other bloggers here that write about other issues.
Nathalie, what would a margin of error accomplish? The margin of error is one of the most misused and misunderstood measures in the reporting of polls. All it does it say if you repeated a poll with the same procedure, you'd expect to get within x%, y times out of z. If you have a problem with the actual study, and it seems you do, that is completely unmeasurable by any confidence interval. But yes, I do think 500 can be a statistically ok poll, it just depends on the how it is done.
Posted by: William Joseph | 2009-06-04 5:58:10 PM
Nathalie, you look about college age. Why do you not know basic statistics?
z/2 (two tailed test) = 1.96
p +or- z sqrt(pq/n)
.35 +or- 1.96sqrt(.35x.65/500)
.35 =or- 0.0414
.308 to .392 19 times out of 20
There is a 95% chance, the true population is between 30.8% and 39.2%.
In other words your margin of error for Mrs. Elliott is 4.1% 19 times out of 20.
Posted by: 아레스 | 2009-06-04 10:40:24 PM
Phew! As a huge Hudak supporter and PC party member, I am extremely glad to know that this poll is unreliable. I still don't see how Hudak can be this far behind. I mean he's 2nd in fundraising, he's got the most endorsements, he was the front-runner immediately, AND as stated in the article it will come down to who signed up the most members, which was done by Tim Hudak (one report has Klees signing up the most, but most hav him in 2nd). I don't get how Hudak and Klees combine to sign up over 23,000 of the 35,000 members and end up in 2nd and 4th. Just doesn't make sense. I was at the Markham debate, and it seemed clear to me that Mr. Hillier (God bless his true Conservative values, he will be 2nd on my ballot) is out of this race. I mean he's from the smallest riding of them all, he has signed up the least new members, he has raised the least amount of money. I just don't see how he can even receive 15% of the vote.
Posted by: Jon | 2009-06-08 4:39:26 PM
If Christine Elliot wins, i won't bother voting in the general election just as I didnt bother voting for John Tory. I want to vote for a conservative not a liberal. If i don't have that option I don't vote at all.
Posted by: John | 2009-06-14 6:41:11 PM
The comments to this entry are closed.