Tuesday, May 27, 2014

Ontario Liberals inch ahead in Abacus poll

The new poll by Abacus Data for the Sun News Network shows the Liberals narrowly edging out the Tories among both all eligible and likely voters, part of a trend that seems to be building in favour of the Liberals - or, perhaps more accurately, against the PCs.

As the results of the Abacus poll among likely voters did not differ greatly from the projection yesterday, the numbers have not moved dramatically today. The Liberals are still in front with 35.8% (or between 34% and 39%), followed by the Tories at 33.2% (32% to 36%) and the New Democrats at 22.7% (or between 21% and 24%).

The seat count is virtually identical to the standings at dissolution: 48 for the Liberals, 37 for the PCs, and 22 for the NDP. The ranges still overlap between the Liberals and PCs, however, at between 40 to 58 seats for the Liberals and between 30 to 47 seats for the PCs. The New Democrats sit at between 16 and 24 seats.

For once, we're spared the whiplash caused by a new poll contradicting the one that came before it. In fact, this Abacus poll is broadly similar to the most recent poll by EKOS Research. The Liberals have now led in three of the last four polls, and the methodological trend seems to have been broken. It was starting to emerge that the IVR polls favoured the Liberals and the online polls the PCs, suggesting that there was a methodological factor behind the discrepancy. But now that is no longer the case.

A bit of a trend is developing. The charts below are the same as the one posted yesterday, but highlighting only each party. The vertical lines represent each day a poll was in the field, with a rough approximation of the margin of error relating to the poll (assuming a probabilistic sample). The horizontal lines track the trends for each individual pollster. When looked at in this fashion, the discrepancies recorded so far in the campaign do not seem too large, and the trends seem a lot clearer.

Generally, the numbers have not been moving very much. But it is clear that there is a positive trend in favour of the Liberals and a negative one for the Progressive Conservatives. The New Democrats appear stable. The Liberals have gone from the low-30s to the mid-to-high 30s, while the PCs have gone from the high-30s to the low-30s. There is a bit of a signal in all this noise.

The Abacus poll shows a great deal of stability, with even the regional results not changing from Abacus's last poll by more than five points (keeping them well within the margin of error of a probabilistic sample of similar size).

Among all voters, the Liberals picked up one point to move into the lead with 34%, as the PCs dropped one point to 32%. The NDP was also down one point to 25%, while the Greens were steady at 6%. The number of undecideds increased by one point to 15%.

The interesting result was among likely voters, which has favoured the Tories in past polling by Ipsos Reid and Abacus. Here, however, the Liberals are increased to 36% (+3 from last week) while the PCs also increase, but just to 33% (-3 from last week). Have the Tories lost their turnout advantage?

At the regional level, the races are quite close. The Liberals led in Toronto and eastern Ontario, with 39% and 40%, respectively. The Tories were second with 31% and 34%, and the NDP in third with 22% and 21%.

There was a tie in the GTA/Hamilton-Niagara region at 33% between the two parties (the PCs led here last week), while the NDP was at 25%.

Elsewhere, the PCs led in southwestern Ontario with 34% to 28% for the NDP and 27% for the Liberals, while in the north the NDP was ahead with 34% to 32% for the Liberals and 22% for the PCs.

It will be interesting to see if other polls start to echo what Abacus is recording. They are all recording drops in support for the PCs, so that is a trend that seems to be consistent. Whether that drops them into a tie or further behind, however, seems dependent on the pollster. But for now, at least, the polls are making a bit more sense.


  1. Since all the pollsters seem to be showing a PC slide I think that by election time the situation will be distinctly clearer

  2. It's interesting to me that both Abacus polls during the campaign have been very close to your projection at the time they came out. I wonder if this is a sign of Abacus doing good polling this election.

  3. still think the deciding point will be the main debate. the leaders' performances there will decide who wins this one.

  4. Using the reported numbers, I get:

    46 OLP
    40 PC
    21 NDP

    With the likely-voters numbers, I get:

    50 OLP
    39 PC
    18 NDP

    And with the 308-average, I get:

    48 OLP
    41 PC
    18 NDP

  5. So much for the Green Surge of last week!

  6. I was genuinely surprised to see no whiplash for the first time between this update and the previous one! This campaign seems to be making a bit more sense now. :)

    Still curious on how the disgruntled dippers will vote this time. Both the Liberals and Greens are campaigning further left than then NDP. I wonder if we'll see Ontario's first Green seat, or Liberals picking up NDP-held areas?

    1. Toronto looks like the most likely site of losses for the NDP, judging by the regional results in the polls and because the most progressive (therefore most discouraged) NDPers seem to be in Toronto.

    2. But so far it looks like the NDP may hold on in Trinity-Spadina & Davenport in the latest projection, so perhaps they won't lose any seats in the 416, afterall. I don't think the Greens will be winning a seat, though.

    3. Yes, the NDP does seem to have steadied in the last few days...

  7. Short of an overwhelming performance in the debate by Hudak I don't think you will see any major shift Bryan.

    The distrust and dislike of Hudak won't disappear overnight no mater how somebody thinks it will.

  8. Does anyone know if the polling firms are adjusting their methodologies?

    Also, Abacus did some great polling to identify swing voters in their previous poll, something Eric alluded to in the previous post on Abacus. Are these numbers holding up in the new poll?

  9. Interesting, the projection is now only one seat off of the legislature at dissolution, that one seat difference being the one that was vacant

  10. Sorry to be a bother you Eric, is it possible for you to put in some more detail on how you weight the polls? Ideally a number showing the weight each one carries and how you "decay" them and any weighting for the comapny

    1. I have it explained here: http://www.threehundredeight.com/p/forecasting-methodology.html

      To put it briefly, polls are reduced by 35% each day. Sample size weighting is based on the margin of error (or theoretical margin), with 3.1%=1 and the weight given for a poll's margin of error being relative to that. The weighting given for each polling firm based on their track record is not something I release, because it is just trouble if I do.

      Support for others is estimated based on the performance of fringe parties and independents in 2011, and the number of candidates in 2014.

      The chart on the projection page shows the relative weight of each individual poll, but if you're interested in the actual current weighting, it is:

      Abacus (May 21-24): 45%
      EKOS (May 16-23): 31%
      Ipsos (May 20-21): 11%
      Forum (May 20): 10%
      Abacus (May 14-16): 2%
      EKOS (May 13-15): 1%
      Ipsos (May 12-14): 1%

      And a number of other older polls weighted below 0.5%.

    2. Does the total not need to add up to 100%? If so, how can you have a uniform reduction in the value of polls over time?

      Next, is the "date" for the Abacus poll May 22.5?

      Part of the reason I am asking for more details is because I am not sure why the Ipsos and Forum polls would have such a lower weight than the EKOS given the relative median date is almost the same for all three.

    3. Yes, the total adds up to 100%. I was rounding. I'm not sure I understand your question. The total 'value' of all the polls is currently 0.965 in the model, the current 'value' of the Abacus poll is 0.430. Tomorrow, all the polls will be reduced by 35%. If no new polls are added, the total 'value' drops to 0.627, and Abacus to 0.280. But it will still represent about 45% of the weight.

      I used to use median dates, but I found that this rewarded firms like Forum, who poll on one day, and penalized other firms. It did not seem appropriate to me that a Forum poll done at the end of the election should be valued more than a poll done over the last two days of the election or the last three days, so I moved to just counting the last day of the poll.

      I realize that I did not update this on the methodology page, though it has been used in the last two elections. I will update.

    4. I think a better understand what I asked in the first part and my only comment on it would that is strikes me at that if there is a gap of a number of days before a new poll it would have a very strong impact on the model. A poll coming out tomorrow with an end date of May 28th could end up being more than well over half the weight?

      So the change in the date model gives Forum less weight but then gives EKOS more weight, right? In the median model Forum had a date of May 20th and EKOS of May 19/20th. In the final date model Forum remains May 20th but EKOS is three days newer and has three times the weight of Forum.

      How much a difference does it make to your model does using median date versus final date?

    5. The difference can be significant, but polls like EKOS's are rare - most are done in three or four days now (even EKOS's previous poll was over three days). So, I think the real world effect is minimal.

      Also, note I have a hard cap of 66.7% for the weight of any one poll within the model.

    6. There seem to be fewer companies offering polls this time than in 2011, Nanos, Angus Reid and Leger seem to be missing this time. Can you see any difference on your model this time compared to 2011 when there a lot more polls coming out? Were your ranges narrower in 2011?

    7. The model didn't calculate ranges back then, so I can't compare. The ranges are also not based on the number of polls, but rather the past performance of polls.

    8. So more companies and more polls will not narrow the range?

    9. No. That was the system I used in BC. It assumed that the more data we had, the better. But BC showed that more data doesn't make for a better projection if the data is flawed to begin with.

      The ranges instead assume the polls will be wrong rather than right, and estimate the over- and under-estimation of the aggregate that is likely to occur.

    10. Thanks for the explanation

  11. I wonder who shall replace Tim Hudak when he resigns? John Baird would be one possibility. If he were smart, Chris Alexander might also make the leap, rather than facing either a decade in opposition or a decade in Jason Kenney's shadow.

    1. Chris Alexander is a career diplomat. His policy strength and principal interest are both in foreign policy -- I expect, if the Conservatives stay in government, that he'll soon be Minister of External Affairs. A switch to provincial politics would be a poor fit for him.

    2. The PCs need to ignore the hard-right rump that has been steering their ship onto the shoals ever since the Liberals took power. Only then will they be able to nominate a leader who will appeal enough to the general electorate to win enough swing voters to return to power.

    3. eg what you are describing is the remains of the Mike Harris group. As such the best approach for the PC party in Ontario is to split. The old "Red Tories" of Bill Davis era could still prevail. This current alignment probably never will.

      Now the BIG question is if Hudak loses again can the party dump him ??

    4. If they cannot be rid of him, further wandering in the wilderness awaits them.

    5. Mike Harris was not "far-right" he was a fiscal conservative whose priority was to get Ontario's books in Order after nearly a decade of NDP-Liberal spendthrift ways. The deficit decreased every year Harris-Eves were in power culminating in four consecutive surplus budgets. It is a record the big spending McGuinty-Horvarth Liberals have been unable to emulate not because they lack the political skills but, the political will. A re-elected Liberal government may very well be forced to cut since, Ontario's finances are rushing full speed towards a fiscal crisis. The Liberals have tried to hide the mess by claiming the deficit is only 11.7 billion but, that is only the operating deficit- the real deficit when infrastructure is included is $36 billion!

      Ontario can no longer afford not to turn right!

  12. Looks like the Abacus poll leads to the Liberals doing better in their Eastern Ontario seats (eg. Ottawa-Orleans, Glengarry-Prescott-Russell, Ottawa West-Nepean, etc.) but possibly losing some ground in a few other places.

    Quite a few of the seats are close between the Liberals, PC's & NDP since I've noticed about half a dozen of them switch back & forth in each projection, which may make the final outcome difficult to predict, yes?

  13. Hey Eric I love your site . How about putting % changed from update to update on main score board ?

  14. Hi Eric, thanks for the update! It's nice to see your numbers are jiving with polls and that the whiplash is dying down. Just wanted to bring our attention to some riding polls! I think the Green numbers are left out though: http://metronews.ca/news/canada/1047411/poll-suggests-tories-far-behind-in-northern-ridings-after-tim-hudak-skips-debate/

  15. The complete abacus poll is actually a good read. Page 13 shows which way swing voters are leaning. The biggest group (NDP/Liberal or 20% of the electrorate) is leaning Liberal by a ratio of 40% to 20%. It looks like the old Liberal adage of campaigning from the left and governing from the right (lets face it, some big spending cuts are coming in Ontario) is holding.

    Kudos to Abacus for doing such intelligent and detailed breakdowns.

    Could someone from Ontario comment on how this reflects voter profiles in federal elections? My impression is that unlike in Quebec, BC, or Alberta the symbiosis between federal and provincial wings of the big 3 parties is pretty strong.

    1. Just elaborating on the Abacus data:

      The latest Abacus poll shows that the main reason for the Liberal trend is that Liberal/NDP swing voters are leaning towards the Liberals. The four-way swing vote is also leaning heavily Liberal.

      20% of voters: Lib54%-NDP35%
      13% of voters: Lib45%-PC20%-NDP20%-Gr15%

      On the other hand, the PC seems to have some appeal:
      13% of voters PC65%-NDP30%
      10% of voters PC50%-Lib44%;

      The PC/Lib difference seems to be statistically insignificant.

      I'm really surprised that PC-NDP swing voters outnumber PC-Lib swing voters in Ontario. It defies conventional wisdom.

      Abacus I (page 13)
      Abacus II (page 13)

    2. More analysis of Abacus:

      The Abacus poll shows that the Liberal ground game appears to be the strongest and gaining momentum, while the NDP ground game is fading away slowly. The Liberals are leading in the number of voters contacted:

      Voters contacted by party (comparison to previous week)
      Lib: 24% (up 9%)
      PC: 22% (up 7%)
      NDP: 14% (up 4%)

      Interestingly, the Libs and PCs seem to be duking it out in Liberal ridings, with the Liberals winning large by taking the NDP vote.
      Contact in Liberal Ridings Lib:40% PC:29% NDP:15%

      Conservative ridings are being neglected. The Conservatives have contacted more people in Liberal ridings than in their own. They are obviously going for victory.
      Contact in Conservative ridings: PC:25% Lib:14% NDP:07%

      The NDP is fighting strong to hold onto what it already has.
      Contact in NDP ridings: NDP:33% PC:21% Lib:15

      The Liberals have obviously hooked onto a winning campaign strategy, and are successfully holding onto their ridings by appealing to the NDP swing vote. Their path to a majority is to go after the soft NDP vote but in Conservative ridings in the GTA. That means a shift in the ground game from Liberal ridings to selected Conservative ridings in the GTA and in small cities (Peterborough, London. Guelph. etc.). Given the surprising strength of their grass roots campaign, that could win them a small majority.

      The Conservatives will need to refocus on holding onto what they have in the GTA. They might not do it though because Hudak needs a plurality of votes to survive and might just continue to do so by aggressively canvassing Liberal ridings. Unless things change though, that will guarantee a Liberal majority.

      Source: Abacus Poll 2 (page 15-16) http://abacusinsider.com/abacusinsider/wp-content/uploads/2014/05/OntarioElection_Ballot_Wave2FINAL.pdf

      I'm going to go on a limb here and predict a Liberal majority barring any self-sabotage by Wynn. She doesn't seem the type, though.

    3. At this point, after the PC math issue (counting hours of work instead of jobs to get to their million figure...so it is the million hour program ... ouch) I have to think this election could turn into a cake walk for the Liberals if they don't pull a John Tory (finding defeat in the jaws of victory).

    4. Gus,

      You are being selective in the data you report. 60% of eligible voters have not been contacted by any party; of the 40% who were 24% were contacted by the Liberals, 22% were contacted by the PCs and 14% were contacted by the NDP.

      The reason why many PC voters have not been contacted is as Abacus shows the PCs hold a significant lead in rural areas: 36% PC, Lib 27%, NDP 24% and a strong PC lead in PC held ridings 51% PC, 23% OLP, 20% ONDP.

      The NDP is not "fighting to hold on to what it already has" Abacus clearly shows a wide and significant NDP lead in NDP held ridings 53%NDP, 23% PC, 19% OLP

      There is good reason why the Tories concentrate their efforts on swing ridings-rural Ontario will vote strongly PC this time around, the PC majority will be won on the back of Liberal constituencies in the GTA, Eastern Ontario and the South West.

      The Liberals will lose, look at the polls! One need only bring up Ornge or gas plants or questionable parliamentary spending to see why the PCs are leading. Also in the Abacus poll 51% of eligible voters believe it time for a change in government but, only 25% believe the OLP should be re-elected. That and your notable omissions of the Abacus poll say more than your selective inclusions that are meant to show the OLP in a stronger position than the data demonstrates. I think it speaks to a desperation among some and a intuitive knowledge the OLP will soon lose government.


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.