Wednesday, September 14, 2011

McGuinty drops to minority

A new poll was released last night from Abacus Data, for Sun News. This poll will undoubtedly raise a few eyebrows, considering that it pegs Progressive Conservative support at 41%, compared to 32% for the Liberals and 20% for the New Democrats.

Obviously, that differs from the two most recent polls from Ipsos-Reid and Nanos Research, and is even at odds with the spate of pre-campaign polls released at the end of August and the beginning of September. The results aren't wildly outside the margin of error of other recent polls, but we're probably looking at either an outlier or a shift in voting intentions (this poll was taken up to September 12th, after all, while Ipsos-Reid and Nanos ran up to September 11th).

But another issue with the poll is its methodology. Polls generally just ask respondents for which party they intend to vote, and if they aren't decided towards which party they are leaning. There are some variations on how this question is asked (i.e. Nanos asks voters to rank their top two local preferences, Ipsos-Reid doesn't prompt for Greens, etc.), but it is generally the same.

Abacus is taking a completely different approach, at least at this stage of the campaign. They are asking respondents to rate their likelihood of voting for each party on a scale of 0 to 10, with the party receiving the highest score being considered the party that the respondent will vote for. In case of a tie, the respondent is considered an undecided.

This does a good job of gauging enthusiasm for each party, but does have some limitations. For example, someone could conceivably respond with a 9 for the PCs and an 8 for the Liberals and NDP, and be marked down as a decided PC voter in this poll. Another respondent could put the PCs at 10 and the others at zero, and also be considered a decided PC voter. But those two voters are very, very different.

But what this poll may be doing is recording what other polls cannot. With standard polls, it is up to the respondent to gauge whether they are decided or not, and whether they are leaning to one party or not. The decided voter in a standard poll might be the same as that voter who answers 10 for the PCs, while the leaning voter may be the same as the person who answers 9 for the PCs and 8 for the other parties. Or this method may somehow negatively influence how someone answers, compared to the standard method.

Whether this method has any merit or not is difficult to say. I think it is a worthwhile experiment, and I hope Abacus will continue it through to election day. Since it is not up to me to decide which methodologies are valid and which methodologies are not, I see no reason not to include this poll. And remember that ThreeHundredEight's weighting system does include the track record of each polling firm, so not all polls are considered equal.

Another poll out this morning is from Forum Research on the city of Toronto only. It finds that the Liberals lead with 39%, followed by the New Democrats at 30% and the Progressive Conservatives at 24%. Compared to their last poll conducted at the end of August, that is not much of a difference. If we look at Abacus's poll, they have the Liberals leading in Toronto with 38% to 35% for the PCs and 21% for the NDP. At the very least, we can say that the Liberals are leading in the city.

Prince Edward Island

Before moving on to the Ontario projection, let's briefly look at the results of a Corporate Research Associates poll for The Guardian, Charlottetown's major newspaper. This is data derived from CRA's quarterly poll conducted in August.

It finds that the Liberals lead the PCs on every issue such as having a vision for the future of the province (40% to 18%), being able to handle economic problems (42% to 19%), dealing with the health care system (34% to 23%), and keeping a balanced budget (32% to 21%). What is interesting in this poll, however, is that on the vision question the PCs and Liberals are tied at 35% apiece in Kings County, the easternmost part of the island and the only part of the island that elected PC MLAs in 2007. Perhaps the party still stands a shot at more than one or two wins in the region.

On to the newly updated Ontario projection. With this Abacus poll included, the Progressive Conservatives are back on top, gaining 1.1 points to lead with 37.2%. The Liberals have dropped one point to 35.2%.

The New Democrats are down 0.8 points to 22.5%, while the Greens are up 0.7 points to 3.8%.

In terms of seats, the Liberals have dropped two and the NDP one, with all three going to the Tories. That puts the Liberals at 53, one short of a majority, the PCs at 35, and the NDP back down to 19 seats.

The PCs have picked up two seats in Greater Toronto and one seat in Toronto itself, with the Liberals losing one seat in each region and the NDP one seat in the GTA. The Liberals are still projected to win a majority of seats in both regions, however. In and around Toronto, the Liberals dominate with 28 seats to six apiece for the PCs and NDP.

This is the first time that the PCs have gained in the projection, but as you can see in the chart above they are still well below where they were only last week.

The ranges have become much more favourable for the Tories, who lead in five close races and trail in 16 others. The Liberals lead in 16 close races and trail in five, while the NDP leads in one close race and trails in two.

That puts the Liberal range at between 37 and 58 seats, compared to a range of 30 to 51 seats for the PCs and 18 to 21 seats for the NDP. The Liberals can still win a majority, while the PCs are now able to beat the Liberals - but only enough to secure a minority.
The overlap for the two main parties is apparent in the chart above.

Unfortunately, the rumour of another new poll this morning did not pan out. For Ontario, I suspect we will hear from Angus-Reid this week and a poll out of Manitoba will have to appear at some point. The two Atlantic provinces are relatively quiet for the time-being, but I'm sure we will see something from both provinces before the vote.


  1. Also, its from the Toronto Sun, which might as well be the news wing of the Conservative party. (I'm not complaining, same goes for the Star and the Liberals)

  2. I guess my problem with the methodology on the Abacus is does not represent what happens in the ballot box. There is no "Well, I like them a bit and them a bit" in there. We get one choice. That and arbitrarily picking 9 and 10 on a scale as a choice which indicates people are definately going to vote that way assumes people actually LIKE who they vote for. Many of us are looking at ranking our choices at a level around 5 or 6 but holding our noses and doing so anyways. :-)

    What this poll does do is track loyalty, and thinking, and that is VERY intriguing.

    The spin from the Toronto Sun and the various paid conservative online contributors started within minutes of this coming out, BTW.

  3. You mentioned that track record plays a role in the weighting. Since Abacus Data has no track record for political polling, what was its weight (compared to a Nanos poll, for example)?

  4. eberndl,

    Abacus was active in the 2010 New Brunswick provincial campaign as well as the 2011 Canadian federal campaign, so they do indeed have a track record.

  5. I think the problem with the Abacus approach is that Tory voters will always appear more firm because if you are a rightwing person - there is no where else to do. The Liberal/NDP/Green vote will always appear "mushier" because you have more voters floating around between those parties.

    BTW: Its worth noting that in the 2007 election in the City of Toronto the Liberals took 45%, PCs 24% and NDP 22% - the Forum poll suggests the Liberals have lost 6% and the NDP has gained 8% and the PCs are being seriously dragged down (I suspect) by the extreme unpopularity of Ford and by Hudak committing such a gaffe by talking about "foreign workers" and coming across as a racist.

  6. all polling aside, ontario voters may very well restore mcguinty back to office in the name of political moderation...i am not a fan of a conservative trifecta...too much conservatism is not necessarily a good thing...we'll see what happens...hopefully no minority situations...

  7. You forgot to mention that the Abacus poll was done online, while Nanos was done by telephone. Who knows how Abacus got its online sample, and how representative it is. Add that to the non-standard question, and I'd suggest removing it from your projection.

  8. Online polling is not a problem - some of the best pollsters of the 2011 campaign (and earlier) used online methods.

  9. You are so blatanly liberal leaning your analysis is embarrassing.

    Need I remind you Abacus came the closest to calling the federal election correctly:

    They had
    CPC - 41
    NDP - 32
    LIB - 17

    CPC - 40
    NDP - 31
    Lib - 18

  10. I notice that the riding in Toronto that flipped was Don Valley West. Can you explain how your model seems to think that the chance of a PC takeover is so high for this riding? Recall that this is the riding in which Kathleen Wynne defeated John Tory by over 10% last time and in which the federal Liberal MP lost by only 500 votes despite the massive Conservative wave federally.

  11. Anonymous 12:26,

    That isn't even close to true. Abacus's last result was 37% for the Conservatives, 32% for the NDP, and 18% for the Liberals. The actual result was 40% CPC, 31% NDP, and 19% LPC.

    Abacus was fifth best of the pollsters, and were out-performed by Nanos and Ipsos-Reid, among others.

  12. Scratch that, they were sixth best.

  13. Andrew,

    Methodology is explained in full in the links in the right hand column.

  14. Eric:

    I'll be honest, i haven't read your analysis all the way through.

    Could you give a coles nots version as to how, if the Ontario PC's have a greater percentage of the popular vote you have them winning 20 fewer seats

  15. hi Eric:
    thanks for the update but it shows the flaw in the methodology.
    the Abacus poll will project a Tory majority.
    you should project on each individual poll and let the reader to average them out.
    the weighting is akin to mixing food in a blender best wishes

  16. You mention questions about Abacus regarding the way they asked the questions like if someone gave the PC's a 9 and the Liberals an 8 was that a vote counted for the PC's.

    But the Nanos poll as far as I can tell asked respondants for their top 2 preferences. So if someone said Liberal and Conservative, did Nanos count that as a vote for the Liberals?

  17. Matt,

    It's just how the vote splits for the Tories. Note how I said they are trailing by 5% or less in 16 races.

    Look at the seats they are winning in the riding breakdown. Most of them are with majorities of 50% or more. Look at the seats the Liberals are winning. Almost all of them are with less than 50% of the vote.

    That's the problem - the PCs win a lot of seats easily, but lose a lot of close races. The Liberal vote is more efficiently concentrated, and they can squeak out wins.

  18. Matt,

    We're of one mind. I thought of the same thing, and added it to the post before your comment was submitted.

    Nanos asks them to RANK their preferences locally, so it depends on which party is placed first in the preference, and shouldn't be a problem in case of strategic voting.

  19. On line polling has a great track record when done by reputable, long established pollsters who have very large on line sample bases that are well defined by geography, demographics, socio economic factors and political philosophy. They can therefore come up with a much more representative sample than a telephone survey can. They also don't have the weekday / weekend / holiday problems, nor the land line / cell phone problems that telephone surveys have.

    Angus Reid's final online polls were the most accurate in the past two federal elections, as well as many provincial elections.

    YouGov's online polling of the last UK election was also very good. Don't know much about Abacus but this result of 41% for the Tories with a 9 point lead on the Liberals is not much different from Angus Reid's last Ontario poll which had the Tories at 38% with a 7 point lead on the Liberals. It will be interesting to see if the next Angus Reid poll supports the Abacus results.

  20. Eric -- this Abacus poll really shouldn't be in the mash-up. Not only does Abacus have hardly any track record as an election pollster, but the methodology is entirely new, as far as I can tell. Has Abacus tested this approach by comparing it to a standard preference item in a split-sample design? Is there any reason to believe this would be any more accurate than the standard approach?

    Seems pretty silly to just treat every poll as being as valid as any other, particularly if you're giving increased weight to recent polls since the federal election.

  21. RELAX!!! First off Eric's results don't influence the outcome of the election. Second the election is just over three weeks away and the leader's debate has yet to be held.

    Eric interprets what the polls have to say. He includes all reputable polls. Those of you wanting this or that poll pulled because of bias are yourselves trying to get a preordained outcome. It will do you no good at all, as the voters on voting day will determine the outcome.

    This entire election is so close and volatile that it could turn on one wrong word uttered by any of the party leaders or a particularly effective ad. There is without doubt a strong desire for change, tempered in large part by Hudak's failure to put a friendly face on that change to date. The Liberals have in my mind raised serious questions over both the platform put forth by Hudak and his approach to governing. Despite Hudak's attempts to come across as a nice family man both his demeanor and actions so far have a hard edge to them, bordering on nasty.

    McGuinty faced with the albatross of incumbency and the broken promises that 8 years of governing will produce has tried to project himself as a man making hard choices for the future. His fiscal record reeks. However Hudak is not going to eliminate the deficit any sooner and somehow will find ways to spend more and tax less. Nor will he commit to eliminating the HST. To many Ontarians this is reminiscent of Mike Harris.

    It is with great reluctance that I may have to side with Dalton McGuinty on this one. Like many others I view his green energy plans as an enormous boondoggle. Yet my alternative is someone who can't tell me how he will pay for his spending increases and tax cuts. The most obvious sources are cuts to education and healthcare. I don't want a mini Mike Harris, and that is where Hudak is missing the boat. He thinks we want the second coming of Harris. I say we don't.

  22. The Abacus poll is a valid entry. Stop "bashing" Eric - his model is built around the methodology of including all relevant data. I have my own reservations about the Abacus poll methodology, but they deserve to be in this (with the appropriate weighting).

    I have to agree with Andrew that the Don Valley West call is questionable and Wynne probably has more staying power. Also, on the flip side, the Mississauga East Cooksville does not "compensate" for the retirement of Fonseca and should look much better for the PC. Eric, are you expecting any riding-specific polls to help out with these "personality" issues?

  23. Jolly Red Ontarian14 September, 2011 16:23

    This may seem kind of random and/or weird to mention but I was looking at the monthly provincial political polling trends and the Ontario trend especially. The liberals were leading until last August when they tied with the PC and from then until this September the PC were in the lead. For this month of September the liberals and PC are almost tied again, and what is really interesting is that they are almost tied at the exact same place and at around the same time of year as before.

    I'm not implying anything, all I'm saying is that it's just something to note

  24. Anonymous 12:26

    I think it will be fair to wait for the election results and draw conclusions afterwords.
    The reality is that pollsters and this blog's projection were far from stellar in the federal election. If some more projections will end up looking as having a leftist slant on election day (and I'm mostly looking at MB and ON) then, I guess, the credibility of this blog will be greatly diminished. And that will be very unfortunate, since I was looking for a long time for a Canadian alternative to Nate Silver's 538.
    In the end, you are out on third strike, not on first one.

  25. Thanks Pete. As for Mississauga East Cooksville, the retirement of Fonseca is taken into account, but the effect of not having an incumbent is, of course, not always the same everywhere.

    I don't know if there will be any riding-specific polls. If the 2011 federal campaign is any indication, the answer is that there will be none.

  26. Eric,
    Yes, I can see your methodology. I am wondering if it is flawed. You assign a "star candidate" factor to the PC candidate in the riding, even though she is really a little-known broadcaster. Somehow that factor is larger than the incumbency factor and cabinet minister factor. Being on the ground in Don Valley West, I can assure you that while it may flip if there is a PC tide, it will most assuredly not be one of the first to fall.

    I realize that without polling from individual ridings, it is difficult to predict riding results. I also realize that there are limitations to any model. Nevertheless, the contradiction between the model's prediction and what is happening on the ground causes me to question the methodology of the entire model.

  27. Andrew,

    No star candidate factor was applied in Don Valley West.

    I don't expect the model to be accurate in 100% of ridings, and people would be foolish to expect that. More important is the larger conclusions that can be drawn. The individual riding projections are the ingredients to the provincial projection. They aren't a meal on their own.

    Anonymous 16:34,

    If the projections end up appearing to have a "leftist slant", it would merely be an appearance. The numbers are the numbers, and as operator of this blog I care about being right, not about who wins (particularly in provinces in which I do not live!).

    My inclusion of the Abacus poll makes me a Conservative in some eyes, my questioning of its unusual methodology apparently makes me a Liberal in other eyes. I do not understand this need people have.

  28. Unfortunately, this site is referenced as a horse race site - watch the ponies as they go around the track. People need to have their "side" be winning so they can feel like the world is going to be what they want it to be.

    I must admit I saw it through that lens, as I had been reading fivethirtyeight before I came here; that site was very horse race oriented.

    I enjoy the attempt you are doing Eric to work out the science of this and its application to local races. Please keep going.

  29. I think it is unfair to claim that this blog has a "leftist slant". The final federal projection underestimated the success of BOTH the Tories and NDP. The 2011 federal election was hard to predict, and up to the end it seemed plausible that the Tories would win only a minority.

    When looking at the GTA results, the NDP surge allowed the Tories to squeeze through some traditionally Liberal strongholds with small margins. The NDP vote were stronger in certain parts of the GTA like Brampton, North York and Scarborough than other areas. A federal or provincial poll would be unlikely to catch that projection.

    When looking at the provincial election, the Tories have been projected in the lead until the last few days. I do not understand how people can cry foul now that the Liberals are in the lead (as per three of the last four polls).

    - Maple

  30. Anonymous said...
    Eric -- this Abacus poll really shouldn't be in the mash-up. Not only does Abacus have hardly any track record as an election pollster, but the methodology is entirely new, as far as I can tell.

    Or do you just want it removed from the "mash up" because it showed the PC's leading?

  31. You should have realized by now, Eric - for those that call you out as a leftist or a conservative or whatever, it's not about the science of what you do. It's about what the results can do for them.

    The only results that matter are the ones they want to see. The actual numbers are secondary.

  32. Well said Volkov. I may be an NDP partisan, but I am not going to criticize Eric because the results don't say what I want. I understand (at least on a surface level) the process behind the numbers, and understand that Eric isn't trying to put a partisan spin on the numbers. The need for others to see partisanship in the numbers is something I don't understand.

    I opposed the inclusion of the Abacus poll, but that was because it had an unprecedented and unproven methodology. Not because it had a particularly good result for the PCs or a relatively low number for the NDP. If that was the criteria I was using, I would have opposed including Nanos' poll before their current one that had the NDP at 16%.

  33. Heck, I'm a Green and won't complain even though I hate the numbers. The polls are done by other people, and are reporting what they see as the mood. Like it or not.

    The Ontario race has begun, but for most people it still isn't on the radar. I've only heard two people talk about it (one likely NDP one who is an angry PC who hates McGuinty), and in both cases I brought it up. I suspect most are too busy with back to school and the like to pay attention until the last 2 weeks, if then. That is why advertising is vital and the Liberal war chest will help them a lot.

    At the end of 2010 the surplus/deficit of each party...
    Liberal: +$3,193,549
    PC: -$380,327
    NDP: -$2,158,218
    Green: +$153,859

    Scary eh? The PC's and NDP in debt (NDP deep in it) while the Greens are a bit to the positive and the Liberals...well...lets just say they can afford to advertise.

    Wonder how those figures, if they were widely reported, would affect the election? Most here know that those in power have an easier time staying to the financial good side, but that sure doesn't explain the Green Party.

  34. There were a number of riding-specific polls in the fedelxn.

  35. Most of them were in Quebec, however, and were conducted by Quebec-based firms. Outside of Quebec, there were almost none.


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.