Saturday, April 30, 2011

COMPAS: Tories by 20

COMPAS has the Tories at 46%, the Dippers at 26%, and the Grits at 17%. Needless to say, this poll is grossly out of line with everything we've seen. Because the previous COMPAS poll was also pretty crazy, I will simply assume that COMPAS' methodology is flawed and ignore this survey.

Note that this decision will counteract a significant fraction of the Tory ballot box bounce included in the model. Of course, that bounce is relative to "normal" polls, which this one is not.

16 comments:

Anonymous said...

Quick question, is anyone factoring the fact that 10% of Canadians already voted during the advanced polling before a lot of this "surge" started getting so much attention.

In a lot of close races that could make a huge difference especially for the bloc and the Liberals. I know there were signs of it but it was just all talk

Anonymous said...

The 8-9% of voters who voted in the advance poll will constitute about 15-17% of the total vote given that only 58-65% of all registered voters will show up and vote.

Anonymous said...

And in answer to your question......no one among the pollsters if factoring in that as part of their weighting of poll responses.
This site, however, does make an adjustment to poll data for projection purposes that adds 2-3% to the Tory margin for a "Tory ballot box bump". The advance poll is one way of supporting that. The fact that the pool of actual voters tends to be older, wealthier and more conservative than the pool of ALL registered voters explains it as well.

Andy JS said...

Hello everyone,

I've put together a Conservative target list which I thought some people might find interesting. We use target lists a lot for UK elections although I'm not sure whether you have them in Canada. I'm intending to fill it in with the results on election night:

http://bit.ly/lPVBkj

Thanks,
Andy JS

Anonymous said...

The Compas poll could also have detected the possible swing due to the Masturgate scandal. My expectation was, however, that there would be a bigger bump for the Liberals. However, if the public has given up on the Liberals then maybe the result is a swing to the Conservatives. Remember - the NDP vote is a Layton vote. Damage to Layton is damage to the NDP moreso than with Harper and Count Iggy would be to their parties.

Bernard von Schulmann said...

The Compas poll is also a very small sample for national poll, it is as far as I can tell it is the smallest one in the whole election.

A 750 person poll is very small for a nationwide poll in Canada, that is an average of not even 2.5 people per riding.

The idea of a statistical margin of error makes some assumptions of enough homogeneity to all for a reasonable chance to be a complete picture. I am not at all convinced that a 750 person poll can capture the nature of the Canadian political scene.

Frankly, polls of less than 2000 people nationwide in Canada are not really very useful unless they are part of a long series.

Andy said...

Thanks Andy JS. That's useful and, no, we don't use them in Canada to nearly the extent one sees them in Britain. I suspect that Canadian voting patterns have traditionally been too regional for the idea of a nationwide swing (that is relatively uniform) to be of as much analytical value here.

Election Watcher said...

Anon 11:02p: I'll be looking at that in the final projection, but only for extremely close races. As pointed out by a commenter previously, if you're undecided, you probably don't vote early.

Anon 11:17p: Note that the adjustment adds 2-3% to the Tory-NDP margin. It adds 1+% to the Conservative vote, and subtracts 1+% from the NDP vote.

Anon 11:40p: No, the poll was taken 4/28-29, so the vast majority, if not all of the sample was before Sun broke the news.

Bernard: Absolutely, but the 46% for the Tories is still completely out of line. I mean, even if the truth were 39%, higher than what everybody else is showing, this would still be 4 standard deviations out! Moreover, the weird COMPAS poll mid-campaign had a 2000+ sample. So I'm almost sure it's more than just sample variation, and COMPAS must be using a very biased methodology.

Skoblin said...

EW,

Does the polling industry have its own regulatory body? Or an association of approved members based upon recognized methodologies?

Funny, I was just thinking yesterday about COMPAS and whether or not they would have another poll out showing a 20 point Tory lead...shazam.

Election Watcher said...

Skoblin: I don't think it does. But the free market seems to be working pretty well here: every observer is suspicious about COMPAS, and that writeup didn't exactly suggest much faith in the horserace numbers.

Plus, you know, you will get that 1 in 15,000 outlier once in a long while. (~50 national polls this campaign, counting only fully new polls, so once every 300 campaigns)

Skoblin said...

How much impact do you think this "massage scandal" will have on the electorate? I am only relying on internet sources so I have now idea how much coverage this is receiving.

Election Watcher said...

Same here, but I doubt much. Like we say in Québec, "Le jello est pris."

Skoblin said...

I guess the question now is - will it be blue jello or some sort of orange jello with a liberal amounts of grits in it?

Election Watcher said...

It'll be a blue jello, served with tangerine wedges and a liberal amount of red sprinkles mixed with some light blue ones. The question is whether the tangerine wedges and sprinkles overwhelm the jello.

willge said...

As a Conservative supporter, I love this poll of course :), but can't really put much stock into a poll with such a small sample size.

Anonymous said...

There is no way that the Liberals are at 11% in Atlantic Canada. I doubt they are below 10% in the West, so COMPAS must have really flawed methodology to bump up Conservative numbers by 10%. A more reasonable guess is 36 CON, 29 NDP, 23 LIB 6 GRN 6 BQ (I think Atlantic Canada is a tight three-way race with the Liberals ahead of the pack by about 3%)