Wednesday, October 5, 2011

Final Ontario Projection: McGuinty's Liberals win majority

Prince Edward Island and Manitoba were appetizers to what is the main course of this election season: the provincial election in Ontario. ThreeHundredEight projects that the Liberals under Dalton McGuinty will win their third consecutive majority government, with Tim Hudak's Progressive Conservatives forming the Official Opposition. The New Democrats under Andrea Horwath will be the third party in the Ontario legislature.

The projection has changed since the last update, as Angus-Reid and Forum Research released some new data. They made it a consensus - the Liberals lead in every final poll of the campaign, but the size of that lead stands at anywhere from one point to as much as 10. Either way, the Liberals should be re-elected with a majority. Even Forum's numbers point to a slim two-seat majority for them.

UPDATE: EKOS ruined my night by posting their final numbers a little before 11 PM. I've updated everything (again) with these new numbers, but it changes little.

Before the successful projection of the Manitoba election, the volatility and uncertainty indicated by the polls would have left me ill at ease going into the Ontario vote. I don't know how the polls will perform tomorrow, but I am confident that if the polls do well the projection model should also do well. Of course, the 2011 Ontario might be one of those enigmatic elections, like the one in 1990 that left everyone surprised.
ThreeHundredEight's aggregation of the polls gives the Liberals a 3.3-point lead over the Tories. The Liberals are projected to take 36.6% of the vote, with the Progressive Conservatives taking 33.3% and the New Democrats 24.7%. The Greens come in fourth with 4.1% of the vote.

The Liberals are projected to win 58 seats, with 29 being won by the Tories and 20 by the New Democrats.

The Liberals are projected to win 15 seats in Toronto, 15 seats in the Greater Toronto region, 14 seats in southwestern Ontario, seven seats in Ottawa and eastern Ontario, three seats in the Hamilton/Niagara region, and two seats in central and northern Ontario each.

The Progressive Conservatives are projected to win nine seats in central Ontario, six in Ottawa and eastern Ontario, six in southwestern Ontario, three in the Hamilton/Niagara region, two in Greater Toronto, two in the north, and one in Toronto itself.

The New Democrats win seven seats in northern Ontario, six in Toronto, four in the Hamilton/Niagara region, and one each in Ottawa, Greater Toronto, and southwestern Ontario.

When the campaign began, there were about 30 races which were projected to be 'close'. That has dropped to about a dozen.

The Liberals lead in seven close races and trail in four, putting their likely seat range at between 51 and 62 seats. In other words, the possibility for a Liberal minority still exists.

The Progressive Conservatives lead in four close races and trail in six, putting their likely seat range at between 25 and 35 seats.

The New Democrats are leading in one close race and trailing in two, putting their likely seat range at between 19 and 22 seats. Note that the NDP won 10 seats in the 2007 election.

But with the last month's worth of polling, one can't help but feel like anything could happen on election night. Ontarians have been some of the most difficult voters to call in the last three federal elections. Will they fall in line as they did in 2007, when the polls were within 1% of the PC and Liberal results?

A lot of polls were released in this campaign. Twelve were taken in the five first days of October alone. In the 26 polls released since September 1, the gap between the Tories and the Liberals has been three points or less in 12. It has been an amazingly close race, and only in the final week has it really become clearer who is ahead.
And that party is, of course, the Liberals. They took the lead on September 29, two days after the debate, and have never looked back. The gap has narrowed in the last few days, but the Liberal lead is larger than any the Tories had during the campaign, except on one day.

Generally speaking, though, the campaign has moved very little. The New Democrats have been steady at around 23% to 26% since Day 1. The Greens have faltered, though, dropping to about 3% or 4% about the same time as the leaders' debate took place. One wonders if the presence of Mike Schreiner in the debate would have changed anything.

As with the PEI and Manitoba elections, though in neither case did the same kind of error repeat itself, let's take a look at some other possible scenarios if the polls are inaccurate in the same ways as they were in the 2007 provincial and 2011 federal elections.
In the 2007 provincial race, the polls were quite accurate, and there would be little change in the projection. The only difference is that the New Democrats take a little less of the vote, giving the PCs one more seat.

If the polls are wrong in the same way as they were at the national federal level in the May election, the Liberals would win the bare minimum of a majority government.

But what if the polls repeat the same error that happened in the federal election in Ontario? Here, the Conservative vote was greatly underestimated. However, the Liberal vote was about right. So what we have in that situation is the Liberals and Tories tied at a little more than 36% and the NDP below 23%. This results, again, in a Liberal majority. Indeed, the Liberals are the beneficiary of a greater drop in NDP support.

The PEI and Manitoba projection each had very different levels of accuracy. As I called 98.2% of races correctly in Manitoba, that gives the Ontario projection a two seat window - in other words, 56-60 seats for the Liberals, 27-31 seats for the Progressive Conservatives, and 18-22 seats for the New Democrats (and 0 to 2 seats for the Greens, if you wish). But with the 85.2% accuracy rating of PEI, that extends the margin to 16 seats, or 42-74 seats for the Liberals, 13-45 seats for the Tories, and 4-36 seats for the New Democrats. That seems to be a margin too wide to be worth anything, but if the Greens win 16 seats you'll have heard it here first.

More so than probably any other province, Ontario has a great degree of political regionalization. Unfortunately, the model was unable to take that into account due to the lack of uniformity in how the pollsters break down their polls. This could be a problem when the votes are counted, but I don't think it will be. If the polls have this one right, and even if they are off by a little, Ontario will elect a slim Liberal majority government on Thursday night. But if this campaign has taught us anything, it is to expect nothing.


  1. Thank you Eric for your great work! You have been very busy the last couple of weeks, but note your efforts have been greatly appreciated!

    Thank you!

  2. Andrew Drummond06 October, 2011 06:54

    I also agree this is fantastic work. Though for greater accuracy, you should consider some application of the effect of retiring MPPs.

    In particular, from other sites, it seems like Kenora-Rainy River is a close fight because of Hampton's retirement. There needs to be a way for the model to reflect this phenomenon. (Welland will also be an example of this)

  3. Retiring incumbents are factored into the model, they lose a little more than 10% of their support compared to where the swing would expect them to be.

  4. Eric another NFLD poll is out.


  5. And it looks like one of the Anonymous group and myself have been right since day one?

    All I would like to see is the seat numbers for the NDP and PC's reversed. After this campaign it's definitely time to replace the PC leader.

    Gee it seems that happened the last election to ??

  6. Robert,

    Thanks. The next Newfoundland update may have to wait until Monday, however, due to scheduling issues.

  7. Since the 1920's in Ontario the lowest a party has been in popular vote and still won a minority is 35.7% by the Conservatives in 1943. For a majority it is the 1990 NDP at 37.6%.

    FYI: 1919 was an oddity with a party getting a minority with just 21.7% of the vote (United Farmers) while the party with the highest vote % (Conservative at 34.9%) came in 3rd for seats. The UF joined with another party that had just over 11% to have a majority of seats despite the two having fewer votes than the Conservatives. Hmm...wonder if a 'Landowners Party' could do the same today, ruling rural Ontario while the NDP/Liberals/PC's split the urban zones - much like the Bloc did in Quebec. It'd be interesting to see.

  8. When you have three competitive parties (or, perhaps more accurately, three uncompetitive parties) this sort of thing is going to happen.

  9. Yeah, the closer we get to 33.3% for a winner the more it is just a random draw, suggesting there is nothing making people choose one over the other. That is, unless the Greens suddenly climb up to 10%+ in which case 30% or less could start winning elections much like in 1919 where 4 parties had 11%+ plus 4 others won seats, and the 2 (of the big 4) with the lowest vote totals of the 4 were able to gain majority power working together.

    Funny - if a few parties decided to start running in just small areas of the province (say, Northern Party, Eastern, South-West, and Toronto ones) then we could see a royal mess thanks to a system that rewards regional strength over general strength.

  10. With the polls all in, I just wanted to endorse EKOS' new practice of explicitly distinguishing in its headline numbers party support amongst 'eligible voters' from support amongst 'likely voters'. While I know most pollsters make adjustments for likelihood of voting, it is not my impression that most are explicit about the difference it makes. And the more that public opinion continues to diverge from voter opinion, the more difference that difference is going to make!

  11. Eric,

    Guaranteed that a significant amount of the error in your final projections can be attributed to the "undecided" vote.

    There's an implicit assumption in your model that the undecideds will split in proportion to the decided vote.

    In reality, the undecided vote is most likely to lean towards what they see as the "least of the evils".

    That's vague and hard to quantify, but in the federal election it was probably the Conservatives who benefited. Was it because the Conservatives are the incumbents? Or was it because right-leaning Liberals were trying to stop an orange wave?

    In the provincial election, I'm predicting that the undecided vote breaks towards McGuinty as the "least of the evils". He has the advantage of incumbency, and the dynamics make Hudak into a bigger risk. But who knows. Maybe the undecideds will break against McGuinty as a "change" election. Maybe they'll even go NDP

    The undecideds are the main source of volatility. If you had data about the undecideds and leaners, you'd probably be able to reduce the last couple percentage points in your Margin of Error.

  12. @John_Northey, I don't know how likely the possibility of a province-wide three-way race really is. Even local "three-way races" are usually oversold and represent a temporary quirk of weak party with strong candidate or vice versa. I have a hard time believing that the ONDP would be into the numbers it is now if Jack Layton hadn't died-- it's a bit of a fluke.

    Regarding the overselling of three-way races, note the CBC's story regarding a "three way race" in Ottawa Centre, based on the allegedly strong showing of a Conservative candidate in the federal election. Ottawa Centre was an absolute steamroller victory for Paul Dewar, who got over 52 per cent. The Conservative candidate did finish second... but with under 22 per cent, and he actually lost almost 2 per cent relative to 2008.

    As well, as the Conservatives will probably find out tonight, and as the Manitoba Conservatives can tell you, regional strength isn't all it's cracked up to be--after a certain point it's just wasted votes.


  13. I think it won't be a stretch to say that the final Liberal and PC numbers will be higher than the final NDP and Green numbers. I suspect NDP numbers may be around the 22-23% range while the Green support may drop to 3%.

    What a bland campaign. I forgot what issues were discussed again. There was too much negativity from Tim Hudak, that it left the entire campaign sour. I guess the "Tax Man" will be winning a third straight majority.

    I wonder what direction the Ontario PC party will head now. Will Tim Hudak stick even around for another campaign? The party's last two elected premiers Bill Davis and Mike Harris are from different ideological camps. But most party members seem more comfortable with the latter and that makes it unlikely that someone like Christine Elliot will become leader. Then there is the big elephant in the room, Randy Hillier...

    - Maple

  14. If Randy Hillier becomes PC leader, they can kiss any chance of government goodbye. He is far too nakedly ideological. As both Bill Davis and Dalton McGuinty have proven (and as Bill Davis said) about Ontario politics, bland works. Bland isn't exciting, but it is, at least in Ontario, effective.

  15. It strikes me as implausible that the PC's will only win 2 seats in the GTA (ie. the suburbs, not in Toronto itself), given that every poll I've seen has them ahead there, and in some cases by a fair margin. I suspect this will be the main source of discrepancy between your prediction, and the actual results.

  16. Liberals - 61
    PC - 27
    NDP - 19

  17. @hosertohoosier - a lot of those polls include a far larger area as the "GTA" than Eric does. Abacus' region stretches all the way to Hamilton and Barrie for example.

  18. "Guaranteed that a significant amount of the error in your final projections can be attributed to the "undecided" vote. There's an implicit assumption in your model that the undecideds will split in proportion to the decided vote. In reality, the undecided vote is most likely to lean towards what they see as the "least of the evils".

    I have to disagree here. I think that the 10% of people who say they are still undecided the day before the election are almost all NON-VOTERS. In reality about 45% of the eligible population will not vote. The undecideds will not vote.

  19. Looks like you've underestimated the Conservatives for the 4th time and overestimated the NDP.

  20. Eric - there seems to be a lot of growth in advance polls over the last few years - how do you think they are impacting poll results - are people voting before late swings in opinion? If I was called after I voted early, but I had changed my mind, I'm not sure how I'd respond to the pollster. Or are early voters less likely to change their mind? Do you think it's impacting your projections?

  21. Anonymous 22:44,

    Yes, the polls have under-estimated Conservative support again. I think I need to develop a system for predicting how the vote will be different from the polls.

  22. I'm guessing not being able to properly put regional numbers into your model accounts for a lot of the error? Quick glance looks like the South-West is the only place where your numbers were really off.

  23. Just a question; Since there are fixed election dates in Ontario, what occurs if the governing minority party can't win over enough support to pass their budget? Essentially is it possible for there to be an election outside the fixed date if the governing minority party can't get enough votes in a vote of non-confidence such as the budget?

  24. Ryan,

    Yes, that was a problem. I think I will need to work on a regional model for Ontario.


    The fixed election dates only apply in a majority government. If the Liberals don't pass a budget and neither Hudak nor Horwath can form a government, we'd be back to an election.

  25. What I find remarkable is how close the results are to your prediction adjusted for the 2007 polling error.

    As of now, it's:

    Liberals - 37,5% vs 37,4% (adjusted prediction)
    Conservatives - 35,0% vs 34,8%
    NDP - 23,2% vs 23,0%

    Those are extremely close results. Nonetheless, the Conservatives seem to have gained votes in the right place, electing around 37 MLA vs your prediction of 30 for the adjusted results.

  26. well looks like forum research was the closest at the end of the campaign. interesting 1 point over/under lib/green and 2 point over/under for ndp/PC in your projection. very interesting that you under-estimated Lib support but they got 8 less than your projection. did the 2 point PC improvement really make that big a difference in your projection? I look forward to seeing how your model would perform with the actual vote %s.

  27. After three elections within a twelve month time frame, hope Ontario can be election free till 2014!

    Will be looking forward to the federal NDP and Liberal leadership races though.

    - Maple

  28. So, who is the early bet to get a cabinet seat by flipping from PC/NDP to Liberal? :)

  29. over/under on an NDP floor crossing happening within 2 weeks?

  30. @Bryan, good question. However, I wouldn't be surprised if one of the few urban Conservatives (coughPeterShurmancough) went first, because the post-Hudak leadership campaign is probably going to drag the party to the wacky rural right.

    And the right floor-crosser (coughPeterShurmancough)is gonna get to retire as a cabinet minister...

    Also, can we stop talking about the Greens like they're a real party now?


  31. "Also, can we stop talking about the Greens like they're a real party now?"

    And not a moment too soon either. As a political party/force they NEVER have existed in anything but the media.


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.

Note: Only a member of this blog may post a comment.