Friday, January 27, 2012

Liberals up, NDP steady, Tories down

Because the next election is more than three years away, four federal polls were released this week amid the heightened tension that the upcoming vote in 45 months has caused.

In all seriousness, though, this spate of new polls does give us a good idea of where Canadians stand as the 2012 political year starts to get going. The implications of these polls are, of course, relatively minor. But getting a bead on the mindset of Canadians is never a meaningless exercise.

Take a look at my article today on The Huffington Post Canada website here for a shorter overview of these polls, what trends they are showing, and what that might mean.

As we rarely get the chance to look at a series of polls taken over so short a time outside of an election campaign, let's examine these four federal surveys (by Abacus Data, Angus-Reid, Harris-Decima, and CROP) side-by-side.
The three Canada-wide polls were taken between January 12 and 22, surveying a total of more than 4,000 Canadians. The dates of the three polls all overlap with one another, except Abacus and Angus. As they both use online panels, that is perhaps a good thing.

The three polls run the gamut of scenarios. Angus-Reid gives the Conservatives an 11-point lead, with 39% to the NDP's 28%, while Harris-Decima gives them only a three-point lead, with 32% for the Conservatives and 29% for the New Democrats.

Since Abacus's last poll in early December, the Conservatives have dropped three points to 37%. The NDP has also dropped three points to 28%, while the Liberals are up three points to 21%. Angus-Reid was last in the field in September, and since then the Conservatives are unchanged, the NDP is down one, and the Liberals are up one to 22%.

Harris-Decima has the Tories down two since their early December report, while the NDP is up one and the Liberals are up three.

Though the Liberals don't exactly have the momentum of a runaway freight train ("Why are you so popular?"), this is a positive trend across the board in their favour. The New Democrats appear to be steady, while the Conservatives appear to be slipping.
British Columbia, however, is far less clear. The Conservatives ranged between 30% and 47% in the three polls, putting them either 12 points behind the NDP or 26 points ahead. This is what smaller samples can do.

The New Democrats ranged between 21% and 42%, while the Liberals ranged between 13% and 18%. Angus-Reid has the NDP up 10 points and the Conservatives down nine since August, while Abacus has the Conservatives up one, the NDP down 18, and the Liberals up 11 since December. Harris-Decima also shows wide variation, with the NDP up 11 points and the Conservatives down seven. That is a little too wild to conclude anything definitive, but we can certainly say that the race in British Columbia appears to be solely between the Conservatives and the New Democrats.

Alberta is a little more cut-and-dry, as usual. The Conservatives lead by between 44 and 57 points, with between 61% and 72% support. The New Democrats stand between 10% and 17% while the Liberals have between 11% and 17% support. This is where it gets a little muddier. Most polls have shown the NDP solidly in second in Alberta, but Harris-Decima shows the Liberals in second. Compared to everything else we've seen, that makes it a small outlier.
The Prairies pose another problem, but this is not new. Abacus Data shows the usual Conservative lead and relatively strong NDP showing, while Angus-Reid gives the Prairies to the Tories in a walk and Harris-Decima has a neck-and-neck race. Most recent polls have been between the results of Abacus and Harris-Decima, so we have to consider Angus-Reid to be the outsider on this one.

These are important differences. Angus-Reid would give most of Saskatchewan and Manitoba to the Conservatives, while Harris-Decima would actually split the provinces between the NDP and the Tories. This is one of the reasons why I expect Manitoba and particularly Saskatchewan to be more important in 2015 than they have been in a long time.

Ontario is slightly more in line. The Conservatives range between 35% and 42%, the Liberals stand between 29% and 34%, and the NDP between 24% and 26%. This generally jives with what other polls have been showing: the New Democrats are holding on to their gains while the Liberals have eaten into the Conservative lead. This is the major reason why the Liberals have been stronger in recent months.

Since August, Angus-Reid has the Conservatives unchanged at 42%, while the Liberals are up two and the NDP is down two. Since December, Abacus has the Conservatives down five while Harris-Decima has them down one. The Liberals are up four points according to Abacus and three points according to Harris-Decima, while Abacus has the NDP down three and Harris-Decima has them down one. In other words, the trends point to the Conservatives and (to a lesser extent) the NDP slipping in Ontario to the benefit of the Liberal Party.
We have a richer set of data in Quebec as CROP also reported this week with a survey of 1,000 Quebecers, at least twice as many as any other poll in the field. CROP also has the most up-to-date data, as it was in the field between January 19 and 23, overlapping with all three other polls but also stretching later than all of them.

Across the board, the New Democrats are leading with between 29% and 37%. Optimists might point to Abacus's 37% result, but the three other polls are far more in line with what the general trend has been since December.

The Bloc Québécois holds second in three of the four polls with between 21% and 23% support, a very tight grouping. The Conservatives have between 17% and 24% support, while the Liberals are also tightly grouped at between 17% and 19%.

In what direction the NDP is heading is difficult to say. Since August, Angus-Reid has the party down seven points. CROP also has them down seven points since their mid-December survey. But Abacus Data has them up one point since December while Harris-Decima has them up six. It would appear that the most likely situation is that the NDP is holding generally steady in the mid-to-low 30s.

The Bloc, however, seems to be slipping. Though Angus-Reid has them up two points since August, the three others that last reported in December have them either holding steady (CROP) or dropping (four points according to Harris-Decima, five according to Abacus). In any case, it does not seem that the Bloc is making any new gains, contrary to what seemed to have been the case in December.

The Conservatives seem to be holding, with no changes larger than two points since December (positive in the case of Harris-Decima and CROP, negative according to Abacus), while the Liberals have made three to four point gains in CROP and Abacus's polling. Harris-Decima has them down one, however.

This all seems to point to a general status quo in Quebec. That is good news for everyone but the Bloc, as the NDP slide seems to have stopped and both the Conservatives and Liberals are polling above the last election's result.

Finally, in Atlantic Canada the New Democrats lead by good margins in two of the three polls, and overall average between 27% and 42%. The Conservatives range between 26% and 30%, while the Liberals stand somewhere between 21% and 34%. With its small sample sizes and tricky three-way race, it is difficult to discern any real trend here. But the NDP has being doing well in the area recently.

Both Abacus Data and Angus-Reid see generally similar situations - a large Conservative minority. Abacus Data's polling would result in 147 Conservatives, 96 New Democrats, 60 Liberals, four Bloc seats, and one Green, using the current 308-seat boundaries. Angus-Reid's numbers would give 150 Conservatives, 101 New Democrats, 50 Liberals, six Bloc MPs, and one Green.

Harris-Decima, on the other hand, shows an extremely weak Conservative minority: 118 Conservative seats, 106 New Democrats, 77 Liberals, six Bloc, and one Green. Undoubtedly, this Conservative government would not last very long.

Broken down regionally, these three polls would give the Conservatives between 11 and 26 seats in British Columbia, with the New Democrats winning between four and 20 and the Liberals between four and five. The Greens win one seat in each of these three polls.

In Alberta, the Conservatives range between 27 and 28 seats, while the New Democrats could win one or none.

The Conservatives win between 12 and 26 seats in the Prairies, with the New Democrats winning between two and 11 and the Liberals between none and five.

In Ontario, the Conservatives range between 48 and 64 seats, the Liberals between 20 and 37 seats, and the NDP between 21 and 22 seats.

In Quebec, the New Democrats range between 39 and 52 seats, the Conservatives between eight and 16 seats, the Liberals between 11 and 13 seats, and the Bloc between four and seven.

And in Atlantic Canada, the Liberals range between nine and 17 seats, the New Democrats between five and 14 seats, and the Conservatives between nine and ten seats.
Taken altogether, the Conservatives range between 116 and 171 seats. This means they could win a majority government with these polls, or could even be replaced by the New Democrats, who range between 72 and 121 seats. I only see a 9% chance of an NDP victory, however.

The Liberals range between 45 and 78 seats, meaning they could potentially form the Official Opposition, while the Bloc ranges between four and seven seats. That keeps them out of official party status.

So what do all of these polls tell us? Generally speaking, Canadians haven't moved too much from where they were in May 2011. If they have moved, outside of Quebec it has been from the Conservatives to the Liberals while inside Quebec it has been from the New Democrats to either the Liberals or the Conservatives. Though it is somewhat more complicated than that, as we seem to have the NDP making gains in British Columbia and Atlantic Canada as well, this is what we're seeing.

But overall, the Liberals and NDP can take the most from these polls. The Liberals are showing signs of life, indicating that a future recovery is possible. The New Democrats are showing staying power, indicating that the next leader will not have to play catch-up, at least outside of Quebec. Of course, the Conservatives are still in control. But they have a majority government and are looking less towards 2015 than the other parties. There is nothing to worry them just yet, but the situations in British Columbia, the Prairies, and Ontario point to the potential for problems when Canadians next cast their ballots.


  1. That 42% for the NDP in BC is probably the highest it has ever been, and the fact that the NDP has the potential to tie in the Prairies means the Conservatives stand to lose a lot in both regions (especially in Saskatchewan) given the NDP has only 14 seats, while the Conservatives have 34 seats in those two regions. There is quite a lot of variance between the Conservatives and the NDP so it's anybody's game at this point.

    All three polls point to a Liberal recovery (and it's the first time the Liberals have led in Atlantic Canada since the election so that means that support may be coming back) so at this point it will be the Conservative government's election to lose in 2015.

    The most probable outcome would be a repeat of 2008 but with the NDP having the majority of Québec seats meaning it would be possible for the Conservative government to be defeated by a Liberal-NDP coalition (without the BQ).

    1. An NDP-Liberal coalition is what I am betting on next time around. The real question is who will be in control? If the NDP can hold on to Quebec my money is on them.

  2. seriously, why is H-D given any credence still? They perpetually under-estimate conservative support and over-estimate the liberals. It happens during elections and between. They were pretty much the worst pollster in the election, other than hitting the NDP number accurately.

    What would the seat ranges be without this outlier?

    1. You should really bring up your issues with Harris Decima.

    2. Agreed on the Conservative side, though less on the over-estimating the Liberals. If you look at H-D's last result before the federal election, they were dead on for the Liberals, NDP and Bloc. Where they were off was underestimating the Conservatives and overestimating the Greens. This could just be statistical error, but I agree that it's happened enough times that there's probably an underlying bias.

  3. Have to agree with Bryan about HD. I too would like to see the seat ranges without HD or at least the Conservative seat range if it is possible Eric.

  4. With just Abacus and Angus-Reid, the ranges are 135-170 CPC, 79-108 NDP, 46-65 LPC. But Harris-Decima's numbers look most plausible for Atlantic Canada and Quebec, so I don't think it is worth dismissing them so easily.

    1. What I find frustrating is that while H-D has a good sample size, it's history of showing some bias makes it hard to draw conclusions from it (especially at a national level).

      At the same time, the sample sizes of Abacus and Angus-Reid are so small at the regional level and their results so divergent (case and point the spread between the Conservatives and NDP in BC) that their regional results are practically useless.

      It'd be nice if H-D would develop some sort of likely voter screen like EKOS has to correct for any bias in a fair and transparent way. It'd also be nice if Angus Reid and Abacus would survey more people in the smaller regions and down-weight them where necessary. :(

  5. Harris-Decima and EKOS do appear to have a significant problem identifying Consrevative support. EKOS seems to think this is the result of not filtering for likely voters. I wouldn't be surprised if this is the case for some others, as well.

  6. Ira,

    Even accounting for "likely voters" wouldn't move Ekos' numbers that much... they'd still likely fall below 35%. It's just-outside-the-margin-of-error movement, so it's really inconsequential.

  7. Volkov,

    I thought when EKOS went back and looked at their numbers with a likely voter screen after the election, they found they would have been bang on? I know their likely voter numbers were within the MOE for Ontario at least.

  8. Rob Silver makes a good point about the Liberals. Historically their pre-writ numbers are good. They are by virtue of history and being in the mushy middle the "default" option between elections.

    However, once the election is called and election day actually happens the Liberals rarely do so well.

  9. Some reflection:

  10. Ryan,

    I'm not saying it can't change the results to make it closer to the actual number, but just using that Ontario example - EKOS' numbers in Ontario in their May 1st poll were 39.6 Con, 28.2 Lib, 26.5 NDP, 5.5 Green with a 3.1% MOE. What actually happened is basically within the MOE. So likely voter models versus what they actually had wouldn't have made a tremendous amount of difference, at least not in Ontario's case.

    The thing about a "likely voter" model versus just the registered voter model is that the former is based off of what pollsters believe are likely voters. In the US, there was controversy over how pollsters like Rasmussen chose their criteria for a "likely voter." The best way to calculate likely voters is to do a post-election survey to determine which demographics turned out to vote and which didn't, and determine your model based off of that. Even then that makes the assumption that those demographic groups will behave the same way they did last election...

    There's a whole science behind this that I'd be happy to go into one time. However, in my opinion and the opinion of some pollsters as well, likely voter models based on the last elections just bias the data, meaning that you could be shifting a party's numbers in the wrong direction in a pre-writ period. That's a no-no. Applying it retroactively, as EKOS did, is just benefit of hindsight. It's not necessarily worth looking at.

    And you could argue that pollsters should do it in the pre-writ period, like now. Poll for voting intentions as well as whether or not they're likely to vote. This could work but as I've said, it's not necessarily reliable either - besides, anyone that actually bothers to answer these pollsters on the phone or online are usually the engaged voters. Response rates are horrible for pollsters, and most of the ones that don't respond are going the vast bulk of the electorate who don't know whether or not they're likely to vote, they simply just don't want to deal with some stupid guy from EKOS calling them during dinner.

    The only time I'd recommend doing a likely voter poll is during the writ period. That's when it matters, that's when people are engaged. But pre-writ is unreliable, and a likely voter model based off the last election makes the results biased.

  11. Volkov the thing is that some of these pollsters break out their numbers into various demographic groups.

    So they are weighting their over all numbers against the census.

    That means their voting pool assumes that all voting age Canadians are equally likely to vote regardless of age, gender, education, income, language, etc etc.

    We know that is not true.

    Isn't it better to try and make an educated guess than to simply go with something you already know is wrong ?

    Even if you just did age alone it would probably help a great deal to bring down the artifically high greens numbers and shore up the CPC vote to be more accurate.

    No need to get too fancy, its been proven over countless cycles that young people vote less than older people.

  12. Crothers,

    I am 99% sure that pollsters already account for that in their "demographic weighting" - they would be folly to otherwise.

    "Likely voters" however is more complicated, as it concerns not just voter turnout, but demographic voting patterns, enthusiasm levels, partisan commitment, and many other factors. If you just want to fiddle with turnout levels, that's different.

  13. As far as I know Volkov they don't.

    Its why we frequently see write ups with top line numbers that are so so for the Conservatives and then an important rejoinder buried somewhere to the effect of:

    Conservatives may be at an advantage because their voters are more motivated or have a higher intensity.

    Or their supporters skew older and are more likely to show up at the polls.

    Usually this shows up in the analysis or in the interview and not in the top line numbers you see published that go into Eric's projections.

    Were that it was so.

  14. I'm talking about likely voter models where you ask people how likely they are to vote. Which is what EKOS is doing now and what a lot of pollsters in the US and the UK. I think you're mistaken in what you think I'm advocating for.

    Also, team red has a good poll from Nanos I see.

  15. Ryan,

    I covered that - they can produce interesting results, but it's like asking who they expect to vote for in 20 years. Half of voters don't know if they'll vote by the time next election comes around - it's a sample that doesn't exist yet. All you have is people who SAY they'll vote, and we know that at least 30% of them never do. It's not necessarily unreliable data, but it isn't a great predictor.

    And don't fall for that poll.


    Well, we don't have access to pollster's data or their weighting but I think some of them do. There's little reason for EKOS to otherwise ask for their respondent's ages, if they're not going to do anything with that data.


COMMENT MODERATION POLICY - Please be respectful when commenting. If choosing to remain anonymous, please sign your comment with some sort of pseudonym to avoid confusion. Please do not use any derogatory terms for fellow commenters, parties, or politicians. Inflammatory and overly partisan comments will not be posted. PLEASE KEEP DISCUSSION ON TOPIC.