Latest CPAC/SES poll

[daleo]

CPAC-SES Nightly Tracking (1,200 Canadians, 1,010 Decided Voters)Federal Ballot:
CPC 37%,
Lib 32%,
NDP 18%,
Bloc 10%,
Green 4%

The Globe reported 42 to 24 today (18 points), this one has a 5 point difference.

Note that this adds to 101, so someone is getting a slight bump from rounding.

Edited to add link.
http://www.sesresearch.com/main.asp

[ClusterFreak]

1. I believe the gap is closing.....

....certainly not that it's an 18 point lead. Globe and Mail/CTV poll is typically skewed to favour their right wing views. I think the gap will close enough to only hand Harpo a minority.

[V. Kid]

2. No actually...

...the methadology is simply suspect. The poll asks leading questions before, such as "who has the momentum", that tends to lead people to think gee if X, has the mometum, and I'm leading towards X, I should then vote for X. It has nothing to do with them being biased, they're just not very smart, just think me defending the media.

[daleo]

3. That's one way you skew polls on purpose

Anyone running a polling firm will know better than to do that if he/she wants an objective result. It is pretty basic social science methodology.

Potential sources of bias:
- non-response bias (e.g. phone polls only get people with land lines who answer pollsters).
- question order.
- leading questions/wording.
- post-analysis sifting of data ("likely voters, very likely voters", etc).
- over and under sampling in certain regions and/or demographics that can be trusted to give the responses you want.
- post-analysis re-weighting in intended directions
- not reporting polls that don't give the answers you want (depends how much money you have to spend).
- selectively reporting data that is favorable to one side, ignoring that which is not favorable.

Some of these are unavoidable, some can be done on purpose.

The best guess on how valid a poll will be is simply who commissioned it, in my opinion. Probably the best polls are those done by the political parties themselves, but are not reported.

On edit - Post-analysis is bad wording. I really mean post-survey, at the data analysis stage.

[V. Kid]

4. BC results, reputible pollster:

http://www.mustelgroup.com/pr/20060119.htm

A recent Mustel Group poll taken among a random sample of 1,002 BC adults indicates that the Conservative Party has taken a strong lead in BC, garnering 38% of the decided vote. The NDP and Liberal Party are essentially tied in second place, with 29% intending to vote for the NDP and 27% planning to support the Liberal Party in the up-coming federal election.

These levels are similar to those posted by the parties in the June 2004 federal election. However, given that one-third of BC’s seats were decided by less than 5% margins in the last election, there could be changes in the seat allocation by party.

The Conservatives have the lead in the suburbs of Greater Vancouver and in the interior. The Liberals lead in the City of Vancouver, whereas the NDP have a dominant position on Vancouver Island (including Victoria) and in the Burnaby/New Westminster area.

In addition, the Liberals are the party of choice among those of Chinese descent.

[daleo]

5. We will know in a few days

I don't have any frame of reference for this polling group. Do they say who commissioned the poll?

I saw their "record of accuracy" page. It is nice that they are sticking their necks out, but they could be cherry picking favorable results.

[Telly Savalas]

8. Yeah. The Elections Canada poll is pretty accurate.

:)

[daleo]

9. I believe so

I don't know if every country can make that claim though.

[Telly Savalas]

10. It'd be nice if the U.S. sent teams up here on Monday,

to take notes and observe how it should be done.

[V. Kid]

11. I think they did themselves.

They're a BC based company that mostly does samples for the BC scene, provincially and federally. They were only one point off for the last BC election. They seem to have solid methadology, and before elections don't really seem to deliver stinkers. And while the other regional results from the bigger polling companies are all over the map, they seem to match things well enough.

Don't get me wrong, I hope the Conservatives fall somewhat. But unfortunatley there's a lot of voters here who believe that simply voting the bums out, with whomever is most likely to kick the bums out, is the best idea, regardless of ideolgy.

[daleo]

13. Thanks for that info

I analyze survey data (not specifically political data though) on my job sometimes, so I don't dispute the scientific basis of polling. I also have a number of courses on survey theory and practice, so I know that a well done, objective poll does give good information. I also know that there are lots ways to bias a poll, so that is why I sometimes express some skepticism.

I guess my preferred result would be a Liberal-NDP minority, but a more solid one than the shifting alliances of the last election.

I suppose I don't trust the current bunch of Conservatives, coming from Alberta as I do. They can't fool me with feigned moderation - I know this lot too well. These are not your Joe Clark types, nor even your Brian Mulroney types.

I don't expect everyone here to cheer-lead for the Liberals or NDP, though. Honest, intelligent debate is fine by me.

[Harper_is_Bush]

6. The Globe uses "strategic council", which is being ridiculed in

the media today.

They are owned by a long-time conservative pollster, and it IS CTV and Globe&Mail, after all.

Angus Reid dropped them from thier tracking list after yesterdays result, which was the 18 point difference one. Juuuuust a bit out of step!

The real difference is 5/6%, as is confirmed by two other polls reflecting the same thing.

After tonights appearance on "your turn" Harper should be finnished.

[daleo]

7. Strategic Council

The name says it all.

[AX10]

12. The media is ridculing that poll.

Wonderful!

[daleo]

14. They are saying 37-29 today

I guess the Globe and Strategic Council decided that they had to come back to reality some, or they would be laughed out of the joint. I expect a "big push" could still be in store from them, though.