Thursday, May 15, 2014

Ipsos shows sustained Ontario PC lead

And the pendulum swings back again (again), with the new poll by Ipsos Reid for CTV News/CP24. The poll itself shows a few interesting trends, not all mutually supporting. The margin enjoyed by the Tories over the Liberals has expanded among eligible voters, but decreased among likely voters. The PCs have been making steady gains since November 2013, while the Liberals have also been steadily progressing among likely voters - while dropping among eligible voters. Let's try to make some sense of this.

The projection has returned to putting the Tories firmly in control, with 40% support (or between 38% and 44%) against 33.1% for the Liberals (or between 32% and 36%). The New Democrats remain in third at 22%, or between 20% and 24%.

In terms of seats, the PCs are just short of a majority with 53, with the Liberals projected to take 35 seats and the New Democrats 19. The seat ranges for the PCs and Liberals just about overlap, with 44 to 59 seats for the PCs and 28 to 45 seats for the Liberals being the most likely outcome at this stage. The NDP range is from 16 to 22 seats.

It has been some time since the projection has put the Liberals ahead in voting intentions, and I think this is an important point to make. The polls have been in stark disagreement - this is undeniable. But even in this cloud of noise there is a faint signal. In the polls that have put the Liberals ahead, their margin was smaller than the one enjoyed by the Tories in the polls where they were ahead. Over the seven polls conducted in this campaign so far, the Liberals have averaged a 2.4-point lead in the four polls in which they have led. In the three polls in which the Tories have led, that margin has averaged 8.6 points. This means that a completely unweighted average of all the polls conducted in the campaign so far lean PC by just over two points.

Ipsos was last in the field just a week ago, between May 6-9. Since that poll, the PCs have picked up two points to lead with 39%, while the Liberals dropped one point to 30%. The NDP was down four points to 24%, while support for other parties (including the Greens) was up three points to 7%.

The number of undecideds increased by four points to 20%.

Aside from that jump in undecideds, none of these shifts appear statistically significant. But they are part of some wider trends.

The Progressive Conservatives have been either holding or gaining support over the last four Ipsos polls, stretching to November 2013 when the party was at just 31%. That grew to 34% in February, held at 37% in April and early May, and is now at 39%. The Liberals, meanwhile, have dropped from 32% in April to 31% earlier this month and 30% this week.

However, it is a different story among likely voters (the tally used by the projection model). Here, the PCs gained one point to 43%, while the Liberals were up three points to 31% and the NDP down five points to 22%. None of these shifts were outside of the margin of error of a probabilistic sample of comparable size, but the Liberals have been making gains among likely voters: 27% in April to 28% earlier this month and 31% this week.

This could potentially be a trend that is aligning with the one recorded by Forum, which does apply a sort of turnout model to its numbers. It could be that the Liberals are falling among the total population, but among those who are actually going to vote they are improving their position.

Regionally, the Tories led in central Ontario with 55%, followed by the NDP at 19% and the Liberals at 15% (down 16 points). They were also ahead in the 905 region of the GTA with 48% to 34% for the Liberals and 16% for the NDP. In eastern Ontario, the PCs led with 44% to 30% for the Liberals and 23% for the NDP.

The New Democrats were narrowly in front in southwestern Ontario with 33% to 30% for the PCs and 27% for the Liberals, while they were more comfortably ahead in northern Ontario with 40% to 29% for the Liberals and 28% for the PCs.

The Liberals only led in Toronto, with 37% to 34% for the PCs and 21% for the NDP. But something unusual may be happening in this supposed Liberal fortress. Take a look at the projection tracker for the city:


That is a rather sustained gain for the PCs at the expense of both the Liberals and the NDP. Could Tim Hudak really be making inroads in Toronto?

This is one area where the polls are in actual agreement. Unlike the other regions of the province, Toronto is defined in the same way by both Forum and Ipsos. So we can compare the trends they are recording.

Forum had the Liberals up over the PCs in Toronto with 44% to 27% on April 7 and 44% to 28% on May 2-3. In their last poll, on May 12, the margin decreased to just seven points, with the Liberals at 39% and the PCs at 32%.

Ipsos, meanwhile, has also been recording Tory gains in the provincial capital. The margin was 40% to 24% for the Liberals in its April 15-17 poll, and then 45% to 26% on May 6-9. The PCs have since bounded eight points to 34%, against just 37% for the Liberals.

It could be a fluke, but it is odd that both firms are recording similar gains for the PCs in Toronto - something would normally be unexpected. This could be a developing trend to keep an eye on.

For the rest of it, we still don't have much of an idea of where things stand in Ontario. But the evidence seems to point to a PC advantage of some sort, with the Liberals possibly making gains. This is as good as it gets in terms of discerning what is going on in this campaign right now.

51 comments:

  1. Ok, so let's get a projection-galore going on! Using the reported numbres, it gives me:

    49 PC
    38 OLP
    20 NDP

    Using the likely-voters, I get:

    56 PC
    35 OLP
    16 NDP

    And using the average numbers from the site, I get:

    50 PC
    41 OLP
    16 NDP

    I have a feeling people may just be plain lying to the polls right now, either by being fed up with an other election or by wishing to outright confuse the undecideds. Such disparities can't make any sense from poll to poll, even with different methodologies.

    I will repeat what I said yesterday, I think come election night, there will be a major surge for one party and the result will end up being a majority, one which would have been hard to predict just looking at the polls.

    ReplyDelete
  2. Thanks for the run down, Eric. I wonder if you might comment on any enduring bias issues for the various polling companies. Can one make the claim that certain companies are more Liberal friendly? Or even that the media outlet sponsor (CP24, Toronto Star (!)...) is the more determining bias factor. I know in the US Rasmussen polling has had a small but significant Republican bias in the last couple of election cycles.

    ReplyDelete
    Replies
    1. I don't think claims to bias are very solid.

      Delete
  3. And the drums just roll on

    ReplyDelete
  4. The continual swinging of these polls is leading me to believe that people are only beginning to tune in to the election. It's hard for me to see Hudak winning when he's running what is an indefensible platform that pisses off absolutely everyone (except the corporate world, I guess). I think that when the first leader debate rolls around, his support is going to crumble to the benefit of Wynne. Instead of capitalizing on the gas plant scandal, debate is focused on his platform, and the election is now firmly a referendum on his leadership and not Wynne's.

    I would like to add, however, that the PC gains in Toronto are definitely real. Take a drive anywhere in the suburbs (Etobicoke, Scarborough, York) and you'll be shocked by the number of PC lawn signs, something unimaginable three years ago. They definitely have a decent chance in some of those ridings, especially in holding Etobicoke Lakeshore and gaining Etobicoke Centre

    ReplyDelete
    Replies
    1. No one here is qualified to state that Hudak's platform pisses 'everyone' off. These most recent polling numbers and even your admission about the inroads in Toronto would indicate that there is a bit of partisan hyperbole in that statement.

      Delete
    2. so far the Tory strategy is to keep the spotlight on them and their policies---this looks like a 5 week plan--wait for the last 2 weeks when it moves to flattening the Libs on all their misdeeds--Hudak has yet to bring any of this up---I see a landslide myself if they are this strong and have basically rolled out their bad news

      Delete
    3. Were it not for his disastrous performance in 2011 I'd consider this a foregone conclusion. We'll have to wait and see. I will admit that my gut feeling is the same as yours, Tony.

      Delete
  5. I could be totally wrong (I'm an American who simply follows Canadian politics), but my perception is that the electorate is pretty evenly divided between PC and Liberal, and the deciding factors to the voters will not be the overall provincial issues. It will come down to local, riding-centric issues, and who performs best in the leaders' debates (excuse my ignorance, I'm not sure if they happened yet). June 12 will be a night to get the popcorn for sure.

    ReplyDelete
    Replies
    1. This comment has been removed by the author.

      Delete
    2. ... and then Hudak will reply that he makes no apologies for being part of a government that lead job creation in North America, which is what he already said when someone brought up that exact same question. There are still a *lot* of people out there who admire Mike Harris' government. Always remember Peter that one person's devil is another person's hero before you make blanket statements about what people want or what could be devastating.

      Delete
    3. I really don't think that line of attack will work Peter. Ontario's economy grew 1.2% last year-less than the rate of inflation. That is anemic.

      If Liberasls and NDPers continue to use Mike Harris as a scare tactics voters may remember the economy did well under the Tories the last time then, stuttered, stalled and went into deep recession under the Liberals!

      Delete
    4. Joseph,

      As Tip O'Neill once pointed out - all politics is local.

      Delete
    5. Well, Tip's comment is so simplistic and - at best - only partially true. But I guess it makes for a bon mot.

      Delete
  6. I get the perception that the electorate is rather evenly divided between Hudak and Wynne, and it won't be the provincial issues that decide this election. It will come down to whoever comes out strongest from the leader's debates, and local issues as to where the pendulum will swing.

    ReplyDelete
  7. To what extent has polling methodology evolved to include voters (mainly younger) who use only mobile devices?

    ReplyDelete
    Replies
    1. This should not be a problem for an online panel, as mobile devices can be used to take part in the survey. For telephone polls, most pollsters, if not all, call cell phones.

      Delete
  8. What these polls seem to show is that polls are rather poor indicators at this stage in the game. I think there is a palpable feeling of exhaustion at the repeated Liberal problems in the last years of McGuinty, which the PCs did a good job of hammering away at even a year after McGuinty left. But it's hard to square that with a renewed love for Tim Hudak based either on his performance or announced policies. It's especially hard to see this playing well in the GTA.
    I guess what it points to is that different methodologies have seemingly resulted in very different samples. The other thing it might point to is that expectations of Tim Hudak were so low to begin with that the fact he has not spectacularly imploded is regarded as beating expectations.

    ReplyDelete
    Replies
    1. The Liberals demonstrated they can not be trusted except to further the cause of the Liberal party. People have not started to "love" Hudak they have lost faith in the Liberals and Hudak is the most reasonable and competent alternative.

      The GTA is the business centre of Canada. Reducing corporate taxes give many hope that they will keep their job in a tight and under-performing economy. You may think corporate tax cuts only appeal to stock holders but, they appeal to everyone who works in the business world from the CEO to the secretary and gardeners-what is good for the goose is good for the gander.

      Delete
  9. Fun with numbers

    I went back to the 2011 Ontario election to find out how Biased IR and Forum were. I got quite of a surprise

    Both firms published 2 polls in the campaign period.

    Forum polls averaged Liberal 36, PC 35.5, NDP 25, Green 4

    They had the PC vote pretty well bang on and under-estimated the Liberal vote by 1.65. The had the Green vote 1.08 ... they were basically right on with part of the Green vote voting Liberal.

    On the other hand IR, who I thought might be biased towards the PC had these averages:

    Lib 39.5, CP 34, NDP 24.5 and Green 2

    They had the Liberals and the NDP over estimated by just shy of 2% and the PC under estimated by 1.5%.

    maybe both firms are over compensating in their methodology to try to get closer to the actual results??

    ReplyDelete
    Replies
    1. In 2011 Ipsos was still doing all its election polling by phone...this time they are using an online panel - and online panels can be and often are heavily skewed.

      Delete
    2. and phone lists have become self-selecting by the enormous amount of refusals to participate.

      The sample is messed up by the fact that someone will actually do a poll....

      Same applies for IVR. What sort of person will give their time to an IVR survey (and if they do how much do they mess with it)

      Delete
  10. Éric, is there any way a pollster historical accuracy can be combined into the model? This site ranks them:
    http://www.pauljorgenson.com/pollster-rankings/
    and you can see that companies like Forum, Angus Reid and Léger do well, Compas, Innovative are bunk and Ipsos-Reid, Harris/Decima are very mediocre. With this, shouldn't the Forum result rank higher in a projection model than Ipsos? Does the range capture this?

    ReplyDelete
    Replies
    1. Historical accuracy is included in the model. I'll quote from my methodology page:

      An analysis of a polling firm's past experience in a province or at the federal level has suggested that polling firms that were not active in a jurisdiction's previous election have a total error 1.27 times that of firms that were active in the previous election. Accordingly, polling firms with prior experience in a jurisdiction are weighted more heavily than those that have none.

      Polling firms are also weighted by their track record of accuracy over the last 10 years. Their accuracy rating is determined by three factors: 1) the last poll the firm released in an election campaign, 2) their average error for all parties that earned 3% or more of the popular vote, and 3) the amount of time that has passed since the election. In order to take into account changes of methodology or improvements made over time, the performance of a polling firm in a recent election is weighted more heavily than their performance in an older election. The difficulty of each election is also taken into account: elections where the average error was lower are weighted more heavily than elections in which the error was higher. This is meant to take into consideration elections in which there were particular factors contributing to pollster error that were outside of the pollster's control. Conversely, in elections where the consensus was close to the mark a pollster has fewer excuses for higher error levels.

      The accuracy rating is determined by comparing the average error, weighted by how recent the election is, of the best performing polling firm to others. For example, if the best performing firm had an average error of 1.5 points per party, a firm with an average error of three points per party would be given half the weight.

      Delete
    2. "The difficulty of each election is also taken into account: elections where the average error was lower are weighted more heavily than elections in which the error was higher."

      This would seem to penalize Forum though, as they are most notable for coming closest in BC and Alberta when other pollsters were wrong. I think Forum was the only pollster to actually project a Liberal victory in BC.

      Delete
    3. Forum had the BC NDP two points in the lead six days before the election. That was their last poll. With their unexplained seat projection model, they did give the Liberals a handful of more seats, but pollsters are graded on their polls because they are pollsters.

      My thought process is the following: if an election is polled well, a pollster has no excuse for a mistake because others were able to do it. If an election is polled badly, there is no way to know that the one pollster who got it right was doing something right or merely lucky.

      Delete
    4. "
      My thought process is the following: if an election is polled well, a pollster has no excuse for a mistake because others were able to do it."

      What about the reverse - if the pollster has a track record of being right when everyone else is wrong, we should pay extra attention to that pollster.

      Delete
    5. Except Forum was statistical wrong on the BC election

      They had the Liberals under by 3.1% and the NDP over by 3.3%.

      On the relese of their last poll.... THAT WAS WRONG they stated:

      "Results based on the total sample are considered accurate +/- 3 %, 19 times out
      of 20."

      They were the 1 poll out of 20 that was the abberation yet again.

      They were not flirting with the edges of having a outlier they were beyond the statistical range that their poll should have been.

      So if a pollster is wrong when everyone else is wrong we should give them more credence?

      Delete
    6. "Except Forum was statistical wrong on the BC election"

      Yes BCVOR. Ipsos was more wrong though.

      Statistically, all the pollsters were wrong federally (though Nanos was pretty close).

      "So if a pollster is wrong when everyone else is wrong we should give them more credence?"

      By your standard all the polls are wrong all the time, and we should ignore all polls. Which frankly, I don't entirely disagree with.

      Delete
  11. Aren't the online panels basically self-selected? (i.e. a bunch of self selected people join the panel and receive polls on various issues every so often). This would mean that the online polls are a lot less accurate than telephone polls. It seems that it is mostly the online polls which are giving the PCs large wins, the telephone polls tend to favour the Liberals more often.

    ReplyDelete
    Replies
    1. IPSOS use Bayesian statistics to validate their results. Considering that telephone samples face their own sets of biases they seem to have developed a methodology that should yield pretty good results. One of their white papers summarizes part of the approach: "conditional ignorability can be achieved by combining multiple opt-in online panel and non-panel sources, which we refer to as blended online samples. In essence, by combining multiple sample sources, the “holes” in any one sample source can theoretically be filled out by any one of the others"

      http://www.ipsos-na.com/dl/pdf/knowledge-ideas/public-affairs/IpsosPA_POV_BayesianCredibilityIntervals.pdf

      It seems this survey was online only, but I'm assuming IPSOS have enough previous experience with blended survey's they can use Bayesian approaches to come up with a credible result now using only online panels.

      Delete
  12. Ugh at this point I hope the polls keep showing what they're showing and that on election night the NDP wins a landslide majority just to make things more hilarious.

    ReplyDelete
  13. It sure is interesting when the PCs can are projected to take 7 seats in Toronto and still come short of a majority. Liberals are really holding steady in some parts of the 905.

    Ridings such as Oakville, Oak Ridges Markham, Brampton West and Ajax-Pickering are going Liberal, while the likes of Eglinton-Lawrence, Willowdale, Scarborough Agincourt and Scarborough Guildwood are going PC. Bizarre.

    ReplyDelete
  14. The increase in PC support in the GTA is noteworthy. People are hurting even whilst unemployment remains relatively low.

    I don't see how the Liberals will be able to hang-on. Although the polls don't quite show it yet I predict a PC majority. I expect NDP numbers to decline further as this becomes more evident.

    ReplyDelete
    Replies
    1. The GTA results are certainly telling, when Liberals lose "Fortress Toronto" they tend to have very bad elections. We see this in the polls, the only region where the Liberals lead is the GTA but only by 3 points, so really they are tied with the Tories in Toronto.

      I don't think there much doubt we'll see a Conservative majority. It is clear from the Tim Hudak=Mike Harris scare tactics-Liberals themselves know they are on course to lose this election!

      Delete
  15. The oscillation in the polls is odd.

    The Ipsos polls refers to their on-line panel.

    Hang on ... how different is the membership of this panel, to the one from the previous poll. They don't keep polling a lot of the same people do they?

    As for pollsters calling mobiles. I can't recall a pollster EVER calling my mobile. And yet pollsters calling my Bell land-line are a weekly occurrence. It was pushing daily during the end of the last municipal election in Toronto.

    ReplyDelete
    Replies
    1. I've had pollsters call my cell on numerous occasions.

      Delete
  16. Pollsters routinely call mobile phones nowadays when its a national or provincial survey. They cannot include mobile numbers when doing a riding poll because while you can get lists of cell numbers that are in Ontario or even in Toronto (e.g. starts with 416 or 647) you cannot geolocate them to specific riding

    ReplyDelete
  17. After re-reading the comments on this article it really indicates to me that the polling science in Canada (at least) has broken down.

    Pollsters have basically no way of getting a representative sample and what sample they do get has to have major adjustments to account for over/under sampling a demographic group. Then they have to run various algorithms to account for incumbency and try to factor in common sense.

    In the for media publication polls there does not seem to be any BS filters in the questions or internal validation of the answers given. I have seen no polls with simple civic questions to establish the validity of the opinion.

    When is the next federal election? Who is the Leader of the Official opposition?

    There are far too many 1 out of 20 results of polls taken near to the measurable event. The "people changed their vote on election day" excuse is wearing very thin. It is calling into question the political fickleness of the actual voters.

    There is no publication of how many attempts were made to get the sample, how many people quit the poll before it was done.

    Once the pollsters and media paying for the polls and publishing them recognize that their product is no longer scientifically valid it is a small step to just making the polls results and headlines come out to whatever the best result is for them.

    I personally have a high degree of skepticism that public polls have become tools for political parties the same as mailers and rallies and platforms to influence the voter rather than the unbiased measurement of current voter intent that they are advertised to be.

    ReplyDelete
  18. I'm calling for a PC majority government right now. Hudak has dominated the news for the first two weeks of the campaign. Wynne has been nowhere to be seen other than pooping up in Walkerton to make some good points. Horwath refuses to release her platform.

    Hudak is winning the campaign. Maybe things change next week when political advertising begins, but at this point Hudak is clobbering his opponents.

    ReplyDelete
  19. What the polls and the comments in these threads show beyond a doubt is that the polling has become a major player in the campaign, right or wrong, for good or bad. As a partisan, this may be terrific news, but as a pollster, this should be the last thing one wants.

    ReplyDelete
    Replies
    1. It's more than just the polling. Whether one likes it or not, Tim Hudak has been squarely in control of the message and has been forcing Wynne and Horwarth to respond to his platform rather than the other way around, which in turn draws attention away from their own policies and instead focuses it on that of Hudak and the Progressive Conservatives. That in of itself has a big impact on campaign momentum and public response.

      Delete
    2. Horse race trumps content every time.

      Delete
    3. Should for the media, but not the pollster. It would render Eric's work meaningless. @Nick, stop drinking the kool-aid and try to analyze things a little more objectively. If the polls are wrong (and some MUST be at this point), then your description of how Hudak is driving the media messaging is equally wrong. Of course, you could be right, but just anecdotally there are a lot of counterfactuals out there.

      Delete
  20. Another poll is now out showing a 7 point lead for the Liberals. It seems quite literally a swinging pendulum back and forth as each poll is released. Is there any sense to be made out of any of this? If there is, lord knows I can't see it.

    ReplyDelete
  21. New poll from Ekos:

    http://www.ipolitics.ca/2014/05/16/the-ekos-poll-advantage-wynne/

    I don't trust Ekos as a pollster but I sure hope Ekos is right!

    ReplyDelete
  22. New EKSOS poll conducted May 13-15 gives the Liberals 7 point lead.
    http://www.ipolitics.ca/2014/05/16/the-ekos-poll-advantage-wynne/

    ReplyDelete
  23. The polls are weird this time. I cannot recall another case where we'd see one party with the lead, then another, then back again and so on for this long. I think the leaders debate will be very, very big as will the media's perception on who wins/loses it. If Hudak does well then the NDP could see support plummet as fears about a PC win cause many to shift due to the Liberals giving an NDP budget and a desire to keep the PC's out (fears of it being like Harper or ..brr.. Rob Ford). If Wynne does well and climbs in the polls then things could be good for the NDP, strangely enough, as they'd feel safer voting NDP.

    Turnout will be a big factor this time too. If both Wynne and Hudak flop (or just are 'ok') then we could see both parties supporters lose interest to some degree (or get a 'pox on both houses'). Then the NDP has a really, really big opportunity.

    A reminder, I am a candidate for the Greens (in Thunder Bay-Atikokan) just as full disclosure. I obviously want to see Mike Schreiner win in Guelph and maybe a 2nd or 3rd Green make it (there are a few ridings we are competitive in) but being shut out of the debates, as always, means it is a steep uphill battle.

    ReplyDelete
  24. Bonjour Eric: le nouveau sondage de EKOS renverse la tendance à la hausse des conservateurs à Toronto. J'imagine qu'avec 52% (et le NPD à 15%), les libéraux remporteront 21/22 sieges à la capitale provinciale.

    ReplyDelete
  25. CBC Website

    "The leaders of Ontario's three main political parties will square off in one 90-minute televised debate on Tuesday, June 3.

    NDP Leader Andrea Horwarth, Progressive Conservative Leader Tim Hudak and Liberal Leader Kathleen Wynne will participate in the live debate hosted by TVO's Steve Paikin at the CBC Broadcast Centre in Toronto."

    ReplyDelete
    Replies
    1. The debate will get underway at 6:30 p.m. ET and will be broadcast live on CBC, CTV, CHCH, CPAC, Global, Sun News Network and TVO. Viewers are advised to consult local listing for specific broadcast details.

      More details on the format of the debate are expected to be released in the coming days.

      And the link:

      http://www.cbc.ca/news/canada/toronto/ontario-votes-2014/ontario-votes-leaders-debate-set-for-june-3-1.2645644

      Note: No Green Party member !!

      Delete

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.