Sunday, November 04, 2007

Comparing the Polls

Reading the latest Ipsos-Reid poll, I was struck by the wide lead the firm gives to the Conservatives in Ontario and British Columbia, results which seem at odds with other polls. In trying to seperate the conflicting signals, I thought it might be interesting to put all the latest regional polls side by side to get a better handle on what is real and what is misleading:
ONTARIO:

Ipsos- Cons 43%...Libs 34%...NDP 14%...Greens 8%

Decima- Libs 38% Cons 32% NDP 17% Greens 12%

Strategic Counsel- Libs 40% Cons 33% NDP 14% Greens 14%

Environics- Libs 38% Cons 31% NDP 18% Greens 12%

Margin of errors are comparable, with the Environics poll having the biggest sample size. Three of the four polls show the Liberals leading by an average of 6.7%, while Ipsos has the Conservatives out front by 9%. Ipsos also has the Conservative support a full 10% above anyone else. Conclusion, fairly obvious which polling stands out as suspect.
BRITISH COLUMBIA:

Ipsos- Cons 42%...Libs 28%...NDP 16%

Decima- Cons 31%...Libs 26%...NDP 26%...Greens 12%

Environics- Libs 29%... Cons 27%... NDP 25%... Greens 17%

Strategic Counsel(west) Cons 41%...Libs 27%...NDP...21%

The SC poll is for the "west" as a whole, but a little common sense with Alberta eliminated demands a tight race in British Columbia. Again, Environics with the largest sample size, margins of error comparable, although higher than Ontario. Conclusion, Ipsos vastly overstates Conservative support, while underscoring the NDP support. In general, British Columbia is understood as a three-way race, with no one enjoying a solid lead. All the polls, except Ipsos, demonstrate that fact, which again makes its results suspect at best.
QUEBEC:

Ipsos- Bloc 50%...Cons 22%...Libs 17%...NDP 7%

Decima- Bloc 36%...Cons 21%...Libs 18%...NDP 13%

Strategic Counsel- Bloc 37%...Cons 26%...Libs 17%

Environics- Bloc 32%...Cons...25%...Libs 17%...NDP 17%

Crop(exclusive Quebec poll)- Bloc 31%... Cons 31%... Libs 17%... NDP 14%

Fairly good agreement on the Bloc support, apart from Ipsos, same for the NDP, apart from Ipsos. The Liberal numbers are very consistent, with all the polling, it would appear the teens are real. The Crop poll is the only one to show a close race, but it is noteworthy that they have a very good reputation for accuracy in Quebec. Apart from that, Ipsos once again produces a strange result.

When you look at the national numbers, it is the Ipsos results that look suspicious, with all the other polling outfits showing a far closer spread, and Harper nowhere near majority terrority. I suppose the Ipsos poll gives Conservatives great comfort, as they love to tout the results. However, if you are truly interested in accuracy, the above findings speak loud and clear, Ipsos results should be viewed with a sceptical eye. The Conservatives are not way out in front in Ontario, British Columbia is very competitive and the Bloc is not running away in Quebec.

29 comments:

Anonymous said...

Hey Steve.

Ipsos definitely is trending higher than anyone else. They do have a good reputation for Ontario numbers (I don;t have the exact results, but their last Ontario election poll was almost bang on with the results if I recall correctly...feel free to correct me!)

I don't put too much validity in the Environics results. They may have had a large sample size, but the time in which they were in the field was several weeks (5 I think?) and that can skew a lot as well.

The CROP poll in Quebec is the weird one, in that they had a 1000 people in Quebec only for this poll, leading to have much higher validity than any other regional poll.

The other weird thing is that Ipsos has now conducted 4 polls over 4 weeks, with all of them showing 39-40% for the Tories, and 28-29% for the Libs. To be able to replicate that result 4 times in a row is significant, but at the same time, they still seem very high.

My best guess - Tories are probably around 36-37%, Libs around 29-30% - pretty much exactly where they were in the election. They each have strong core support that probably won't vanish in the long run.

Steve V said...

rick

Fair points.

"The other weird thing is that Ipsos has now conducted 4 polls over 4 weeks, with all of them showing 39-40% for the Tories, and 28-29% for the Libs. To be able to replicate that result 4 times in a row is significant"

See, that means nothing to me, because if the methodology is suspect, then you can replicate bad results. Ipsos always seems to show the Tory support higher, and I can't buy the Ontario results.

Borges said...

Thanks for the post, it's very informative. Very interesting. Did you have to subscribe to some service to access this data or is it public on the web?

Anonymous said...

If those CROP results are right for Quebec (as you seem to believe), you ARE looking at pretty much a Conservative majority.

31% support for Cons in Quebec, with 17% Libs support and 31% Bloc support would add about 20-30 seats for the Cons in Quebec.
Yes they may lose some seats elsewhere in the country but 28 more and you have a majority. So perhaps you should be worried if you buy the CROP results.

Anonymous said...

Steve,

Sorry, I actually forgot to put my other point in about methodology, I agree with you, 4 weeks in a row does mean nothing if the methodology is off.

Of course polls are fun to track and review, but they actually mean nothing until you look at where the support is coming from. Libs should have a lead in Ontario, but a lot of their support is in Toronto, whereas the Tories have more rural support.

That is one reason why I trust the CROP poll more than others - typical national sample size, but concentrated in one area, and with breakouts of different regions. it is also why I always love the final SES poll before elections. They typically do a rolling 1000 people over 3 nights, but for the last 3 days of the campaign, they bump that up to 2000 people. Much larger sample size means better validity(not to mention they pol right until the end pretty much) No wonder they have the closest results on the past 2 elections.

Anonymous said...

I guess a lot depends on how the questions are asked, but I wonder - is Ipsos-Reid playing the game that people will vote for what they perceive could be the winner?

I'd like to see a SES poll.

ottlib said...

You really cannot compare the actual estimates of the different polls because they all used different methodologies and they all had different samples.

Hell, I am certain they did not use the same frames to draw their samples from so that adds to their non-comparability.

To get a vaguely real idea of what is happening you have to compare polls by the same polling company over time. However, even that is an inexact indicator because these polls use different samples each time so they are not really longitudinal in nature. As well, you have to assume they use the same methodology over time for each poll. As soon as they change that they break the string.

Decima and SES are the polling companies that have the most consistant polling methodologies for their polls. Ipsos and Strategic Council change theirs periodically and often without notice. I cannot speak to either CROP or Environics.

There are many statistically valid techniques a polling/survey company can use to get a poll result they are shooting for and they do not hesitate to use them. This is particularly true if the news organization writing the cheques wants a particular result to advance their current narrative.

I believe that explains the discrepancies between polling companies much better than actual changes in public opinion.

liberazzi said...

This is what I had to say about your last Ipsos post:
"Why all the hangwringing about Quebec? The numbers are not that far off what happened in the last election. The real problem is that the Libs cannot count on Ont as they once did and the Cons still have their trump card in Alta. The rest of the country is status quo. The Libs needs to concentrate their efforts in Ont and hope to make small gains elsewhere. The Cons at the level of support they are in Que can only hope to regain there current level of seats. In fact, if you look at this seat tracker, if the Libs actually raise their support to 32 to 33% they actually have a shot of winning the election.
http://www.nodice.ca/elections/canada/projections.php
(Take no notice of that bogus Ipsos poll). The last SES(the only credible polling firm out there) in Aug had the Libs at 33%. Of course, there is no chance of the Libs winning a majority at this point, but it is not a foregone conclusion that the Cons can win another minority and they need to get up to about 42% to gain a majority. The election at this point will not be won or lost in Que, it will be won or lost in Ont and we are still in minority territory here one way or another. Geesh, Libs need to relax, if they fall further, ok then start the handwringing, but at this point all is not lost."
Furthermore, I am getting quite frustrated with the media focusing on these bogus Ipsos polls. Jane Taber just had an orgasm on QP "what about this latest Ipsos poll!!, the Cons are creeping towards a majority!!" I am getting quite frustrated with the media not focusing on the fact that despite these bogus Ipsos polls, the Cons are stalled. The Cons are stalled! Everyone wants to focus on Dion's slips, but Harper has had 22 months to improve his numbers and they are still at the same levels give or take! Dion may not have what it takes to win a majority, but I dont think Harper does either! All the Sun and Posty writer can dream about a majority, but I dont believe it is going to happen, because fundamentally the majority of Canadians do not share the values of the Cons and that is the bottom line!

Steve V said...

"Did you have to subscribe to some service to access this data or is it public on the web?"

All public :)

rick

I take the CROP poll too, for the reasons you state.

ottlib

"You really cannot compare the actual estimates of the different polls because they all used different methodologies and they all had different samples.

Hell, I am certain they did not use the same frames to draw their samples from so that adds to their non-comparability."

Ottlib, I don't understand the logic here. Yes they may use different methodologies, but they are all in the business of gauging public opinion, so the hope is to reach the right conclusion. I agree, about trends with the same companies, but you can also see themes with the various firms, outlier's aside.

liber

Don't expect that number in the new SES poll, that was taken an eternity ago.

I agree with about the media and the Ipsos poll. In fact, TVO had a roundtable, and told people to check out the new Ipsos poll on their website, despite the fact there were 4 other polls published that week, no of which mirrored the results.

To be fair, I think the Cons are stalled, but you could argue the internals have improved. People have become somewhat more "comfortable" with Harper, not quite as worried about a hidden agenda, coupled with the solid showing on matters of leadership and direction of the country. Heading into a campaign, I would have to give the Cons the advantage, because some of the groundwork is in place, even though the polls still show hesitation. The leader becomes paramount in a campaign, I see some optimism there for Cons.

Steve V said...

If anyone watched Question Period today, they will have noticed the roundtable at the end, that discussed the Ipsos poll. Taber talks about the Tories on the cusp of majority, Duffy actually mentions the Ontario results, as evidence of a real Con uptick. I guess the real question isn't what poll to believe, but why the media seems obsessed with Ipsos, at the exclusions of other? Particularly strange, when you go to the SC site and see the CTV banner on the frontpage.

liberazzi said...

Steve, the Cons organization maybe strong, but I believe at the end of the day, Canadians will back off. The Cons demonstrated once again where they really want to lead this country and that should turn off swing voters. Furthermore, you can only give out so many bribes. Eventually, you need to start providing Canadians with good govt. I keep stating this, but what has this govt really accomplished in 22 months, besides polarizing this country? Furthermore, polls are one thing, but the seat projection is another and all is not lost for the Libs. Anyways, SES will be the true current tale of the tape.

Anonymous said...

One has to wonder who answers 1 800 numbered calls at night? The polls all together are suspect as so many as myself have call display and I never answer one of these calls in the evening. The other factor is the number of young people without a land line but only cell phone which don't get polled.

Could one assume anything from polling only those that actually answer a 1 800 number at night??

Any thoughts?

Steve V said...

"Furthermore, polls are one thing, but the seat projection is another and all is not lost for the Libs.

True, however I would argue that Quebec will not rebound, or not rebound enough to make any thoughts of a Liberal minority realistic. I'm operating within the Tory minority or majority proposition, barring some miracle or drastically changed landscape.

anon

"The polls all together are suspect as so many as myself have call display and I never answer one of these calls in the evening. The other factor is the number of young people without a land line but only cell phone which don't get polled."

Fair enough, but I still trust the polling, for the most part ;) The last election in Ontario was polled accurately, even the Angus Reid online survey was bang on in Quebec and Manitoba, so there are filters to deal with challenges you mention.

Anonymous said...

Here's a big sample BC poll.

http://www.mustelgroup.com/pdf/20071031.pdf

On the whole, Ipsos is always on the fringes. Good post.

liberazzi said...

One thing that Duffy keeps repeating that there might be some sort of coup brewing that could come to a head around Christmas. In any event, I hope these Ipsos polls are bogus, because I am not just a Lib supporter, but because I believe a Harper majority would lead this country down the wrong path.

Steve V said...

anon

Thanks for the link. What struck me, the Cons approval numbers on the environment, would have assumed it was lower.

Tomm said...

Steve,

I really don't like commenting on polls, outside of election campaigns, as you can probably tell since I tend to avoid those type of posts.

I have a question though, I recall that one of these firms got the last Federal vote almost bang on. Which one was it?

That may inform this debate. I personally don't recall and don't imagine I will bother going back to check.

Tomm

Steve V said...

SES :)

liberazzi said...

Steve, even 40% does not give a surefire majority to the Cons. I think you are letting the media's negativity get to you. In any event, I believe the Libs should defeat the Cons at the next opportunity and take their chances. I believe the Libs can hold the Cons to a minority, possibly even the status quo. The question if that happens is not only what happens to Dion, but what happens to Harper if he cannot get his conveted majority? Also, do the Libs have an automatic leadership review after the election? Furthermore, if someone like Prentice was running the Cons, would they be in solid majority territory?

ottlib said...

steve:

With regard to your question on my comment. You are assuming that these polling companies all want to come up with an "objective" result for their polls.

That is simply not the case alot of the time.

During the 2004 election Ipsos-Reid was the polling company used by CTV and the G&M. After that election they were fired and replaced by Strategic Council. Why? Because they were inaccurate perhaps? No that cannot be it because SC had a worse record than Ipsos in 2004. The problem is Ipsos overstated the Conservative numbers then and that went against the narrative CTV was flogging at the time.

For polling companies the key is to get a steady gig with one of the news organizations and to keep it they have to keep that news organization happy. It has been proven more times than I can count that most of the MSM cares more about advancing a narrative than actually report news. That is one of the biggest problems with modern corporate news.

So it is no stretch to assume that polling companies will use polling methods to advance a particular narrative. And as I stated in my first comment those methods are statistically sound so they are not doing anything unethical or dishonest.

It all comes down to money Steve, which is why all polls should be taken with a truck load of salt.

With regard to the conversation on QP, I would point you to my comments on one of your earlier posts about journalists cherry picking a particular poll to "prove" their opinions. It is no surprise that Duffy et al would do just that today.

liberazzi said...

When is the next SES due out anyways?

Steve V said...

liber

"I believe the Libs can hold the Cons to a minority, possibly even the status quo."

Don't get me wrong, I believe that too. I've just resigned myself to the fact the Libs are a real stretch to win.

ottlib

I hear what you are saying, I understand motivations. That said, I honestly believe the Libs are in the teens in Quebec, because every single outlet demonstrates that. The Bloc and Cons fluctuate, but the Lib numbers are completely identical.

RuralSandi said...

Mike Duffy said on QP, CTV today that Nanos, SES is working on a poll right now which should be out in a few days.

Walks With Coffee said...

liberazzi said...
"... you can only give out so many bribes. Eventually, you need to start providing Canadians with good govt.... what has this govt really accomplished in 22 months, besides polarizing this country? "

excatly, and why Dion is patient. He wants the public to see CPoC corruption, bribery, and scams. Eventually, the MSM has to start reporting facts as their businesses depend on the perception they are telling the truth and reporting the facts accurately... the MSM cannot afford - literally - to continue to print spin direct from the PMO. It will bankrupt more than just the Western Standard.

Be patient and carry a lot of truth in what you say.

Aside: The Election Canada scandel that I set in motion will bring down the government. At that time, be prepared with a positive message that Canadians can support.

As an aside to the aside and example, I put forward an interest rate policy change that Flaherty read. He flatly spoke against it to the MSM (who also read the piece). The CPoC answer to the collapsing manufacturing sector in Canada is to say "no help for you". Such actions and the obvious corruption are roughing pillars under the CPoC government. If patient, that rought will collapse the CPoC government.

Cheers,

Walkswithcoffee

burlivespipe said...

I'm thinking the Ottawa mayor issue could be a bigger albatross around the CON neck than the election spending issue, because it has the catchy kind of words that stand out in headlines and are very easy to understand. Shuttling funds around from one body to another during an election, it is likely very imoral but some how I get the sense that the cONs are happy to ride that out and just say 'More liberal favouritism by their stooges'... Add the whiff of Harper's strongarming the grassroots, and you might get a number of supporters sitting on their hands pleased to see another minority and a leadership review...

Lept said...

But imagine the numbers with another leader! We can examine the entrails of any number of polls and pollsters while the fact remained that the official opposition with it's pathetic leader is unable to funtion effectively.
I found this article by Gil Courtemanche (author of "Un dimanche à la piscine à Kigali")both moving and extremely pertinent.

Anonymous said...

Steve reliability is very different than validity. To produce the same result 4 times over means it is reliable. The fact that numbers are wrong means it has no validity.

With no validity the polls of Ipsos Reid are meaningless. If a scientist can produce the same results over and over but the methodology or premise of the experiment is flawed it is not published or considered valid.

The experimental results must be able to be reproduced by another independent experimentor. In this case the poll results would have to be reproduced by other polling firms to be valid. The polls of Ipsos Reid do not meet simple sientific criteria for being pulished or believed. No one should quote them or report them. So much so the Liberal Party could persue them if they wanted to.

If Liberal Support were a product and a competing product hired a polling firm to produce results showing that no one liked that product, but the methodolgy was deliberately skewed,and these reslts were reported in the press, in an attempt to sway consumers away for the Liberal Product, The Liberal Product company could sue.

Ipsos Reid is that bad. It is not scientific or valid. Its just propaganda.

Steve V said...

"But imagine the numbers with another leader!"

I have imagined ;)

Anonymous said...

Holy cow folks,

you keep talking about the Ipsos Reid polling methodolgoy being flawed. What would make you to believe that? The firm has quotas based on region, gender, age, etc, and then weights the data to reflect the most current census data. How is that methodolgy flawed? I've also answered an Ipsos poll and noticed that they put their vote intenion question right at the beginning of the survey to unbias it. Sounds like a pretty sound methodology to me...

I think the big difference between Ipsos' numbers and the others is where the Green Party is at. Look at where everybody else has the Green Party. Don't you think that is a bit high? Do you honestly think the Green party will get 10, 11, 12% support? In what world?

Whoever mentioned earlier that the Ipsos numbers in Ontario were bang on during the last number is correct. you are all going to be eating your words...