Thursday, December 27, 2012

Measuring Liberal polling volatility

A few weeks ago, I was having a debate with Darrell Bricker of Ipsos-Reid on Twitter concerning the high degree of variation between poll results for the federal Liberals. Bricker felt that we were mostly looking at the fluctuations caused by using different polling methodologies and the difference between good and bad polls. I felt that there was something going on beyond the methodological differences that was causing such great volatility.

I investigated this question in today's article for The Globe and Mail, which can be found here. Please read it, as it goes into greater detail than I will here, where I'm mostly just going to show some of my work.

I identified the current polling period for the Liberals, stretching back to Justin Trudeau's leadership campaign launch on Oct. 2, as the most volatile since the last election. I calculated this by splitting up the last year-and-a-half into blocks of polls of similar size and calculating the standard deviation from the mean in the results for each party. The result of the calculation was that this current period of polling has shown the most volatility for the Liberals than any party has experienced:
You can also see from this chart that the standard deviation for the Liberals has been higher than for the Conservatives and New Democrats in almost every period. This is probably due to the lack of a permanent leader of the Liberal Party: respondents might be answering with a generic Liberal Party in mind, one led by Bob Rae, or one led by any of the candidates for the leadership. That introduces a degree of uncertainty that doesn't exist for the other parties. However, if that was entirely the case, we should expect the same sort of volatility for the NDP in the run-up to their leadership convention in March.

Another thing I looked at was the margin of error. Perhaps this volatility is much ado about nothing, and we're just looking at statistical wobbling. That could be, but too many of the polls' margin of error taken since the beginning of October fall outside of the average over that time.

This chart to the left shows how much the polls have differed, but also how many of them fall outside of each other's margin of error. The EKOS, Léger, Abacus, and Forum polls taken at the end of November and in December are virtually exclusive from one another in terms of their MOE bands (assuming a random sample for argument's sake, of course).

Another factor is methodology. Are live-callers returning different results than polls conducted online or by IVR? Now we might be on to something:
There has been a systemic difference in Liberal results between live-callers, IVR, and online polls. Telephone surveys with live-callers have the highest Liberal results, with online being the lowest. But the differences recorded in this latest period are well outside of the average variation between the methodologies.

As is usually the case, there appears to be no single factor that is causing such volatility in Liberal polling numbers. Methodology is definitely one factor, as well as the introduction of the leadership race as a wild-card and the events that have occurred over the last three months (by-election, Alberta gaffes, etc.).

It is interesting to me, and hopefully to others, to delve more deeply into the poll numbers from time to time - particularly when something unusual is happening. This degree of volatility makes it difficult to determine exactly where the Liberals are right now, but it can safely be assumed that these trends of volatility are likely to continue straight through to April's convention.

Friday, December 21, 2012

2011 vs. 2012 in polls

The year is almost over and the world has not ended just yet. So, what better time to compare the performances of the federal parties in the polls over 2012 to where they stood in 2011 after the election?

I wrote more generally about each party's report card for 2012 at The Huffington Post Canada here. Please check it out.

For this comparison of 2011 to 2012, I did a simple average of every poll that was conducted in each year. For 2011, I only looked at polls taken after the May 2 election. This gives us a very large sample of polls to draw from, and gives us a good picture of where the parties actually won or lost support in 2012. As the years roll on, it will be fascinating to take a look at how 2011 and 2012 compare to 2013, 2014, and beyond.
Nationwide, only the Conservatives lost support. They averaged 37.5% in 2011 but only 33.9% in 2012, a drop of 3.6 points.

But no single party capitalized on it. The New Democrats went from 30.2% in 2011 to 31.2% in 2012, while the Liberals also bumped their support up by a single point (from 21.2% to 22.2%).

Nevertheless, this does show that 2012 has been a bit of a rougher year for the Conservatives. At almost 38%, the party could hope for the kind of ballot-box boost that would give them another majority. At 34%, they would be virtually assured of a minority government.

Considering the surprise breakthrough of 2011, it is perhaps a very good sign for the NDP that they have managed to hold on to their support. While the Liberals might have hoped that 2011 was just a fluke, they can console themselves that they didn't collapse after the debacle of the last election.
The drop in Conservative support was not insignificant in Ontario, where the party went from 39.9% in 2011 to 37.3% in 2012. But the Tories were not alone in slipping, as the Liberals fell 1.5 points to 27.1%.

The New Democrats took full advantage, increasing their support by three points to 28.8% from 25.8%, putting them in second place in the province in 2012.
However, the New Democrats fell sharply in Quebec. The party had maintained most of its support in 2011, with an average of 39.8%, but that slipped by more than four points to 35.7% in 2012.

The Conservatives were also down, dropping to 15.9% from 19.5% in the province.

Both the Liberals and Bloc Québécois were the beneficiaries, with the Liberals up 3.4 points to 19.2% in Quebec in 2012 and the Bloc up 3.8 points to 24.2%, slightly above their May 2011 election result.
The Conservatives lost the lead in British Columbia, averaging 35.7% in the province in 2012 instead of the 39.3% of 2011.

The New Democrats moved ahead with a three-point gain, putting them at 36.3%. The Liberals and Greens held steady, at 18.4% and 8.5% apiece.
The Conservatives and New Democrats slipped a little in Alberta, the Tories falling 2.5 points to 59.9% and the NDP falling 1.5 points to 18.9%. The Liberals made the only gain, going from an average of 9.6% support in the province in 2011 to 13.3% in 2012.
The largest Conservative drop occurred in Atlantic Canada. The party averaged 35.3% in the region in 2011, giving them the lead over the NDP, but fell 5.6 points to only 29.7% in 2012.

The New Democrats upped their support to 34.9% from 31.4%, while the Liberals were up 2.1 points to 29.6% in the region.
The Tories were also down in the Prairies, falling 4.9 points to 44.1%. The NDP was up 2.6 points to 33.8% and the Liberals were up 1.5 points to 15.3% in 2012.

Altogether, 2012 has to be considered a bad year for the Conservative Party. In addition to shedding 3.6 points of support nationwide (or almost 1 out of 10 supporters from their 2011 average), the Conservatives dropped in every region. Some decreases were relatively modest or inconsequential, such as in Alberta and, to a lesser extent, Ontario. But other drops were quite large and would result in the loss of many seats - particularly in British Columbia, Atlantic Canada, and potentially the Prairies.

The four other parties had good years, relatively speaking. The Greens did make a gain of 0.5 points in 2012 and were up in four of the six regions. But they were stagnant in British Columbia and down in Alberta, and their gains in the rest of the country were minimal.

The Bloc Québécois rebounded from a dismal 2011, but at 24.2% support in Quebec in 2012 they are still very close to their disastrous result of the May 2011 election. They are treading water, but that is not quite good enough for a party holding four seats in the House of Commons.

The Liberals had a better year, making a small gain nationwide and upping their support in five of the six regions. However, only their increase in Quebec was substantial and their loss of support in Ontario is proportionately more important. The polls at the end of 2012 suggest that 2013 could be a much better year for the Liberals, though.

The year has to be handed to the New Democrats. Though they did drop in two regions, more than the Liberals, one of them was Alberta where they have few prospects. The slip in Quebec is significant but the party still held a lead in the province in 2012. The gains of three points or more in British Columbia, Ontario, and Atlantic Canada could more than make up for the losses in Quebec, and they are also well placed to win more seats in the Prairies. After such a whirlwind 2011, the New Democrats had the most to lose in 2012 - but kept it together.

With their leadership race coming to a close in April, the Liberals will have the most to gain in 2013. How will the next year play out?

Wednesday, December 19, 2012

PCs maintain advantage in tight race

An Abacus Data poll on the voting intentions of Ontarians shows that the Progressive Conservatives are still in front, with the New Democrats and Liberals tussling for runner-up status. The poll also gauged the Ontario Liberal leadership race, an exercise that drew a blank from most respondents.
This is Abacus Data's first poll of the Ontario provincial scene since the 2011 election (in which they performed quite well). The poll gave the PCs 35% to 31% for the New Democrats and 28% for the Liberals, with the Greens at 5% support.

Compared to the last election, the Tories were unchanged but the Liberals were down 10 points. Eight of those went to the New Democrats and two to the Greens. This aligns with how the poll breaks down past voting: the Liberals have lost about 20% of their 2011 support to the NDP, while only a very small portion (4%) have gone to the Tories.

It is worth noting how complete Abacus Data's report is for this poll. The report shows both unweighted and weighted sample sizes and decided and undecided results. This is exactly how every poll should be reported. It shows that, despite being an online poll, the number of corrections that needed to be made to the sample were relatively minimal. This isn't about double-checking a pollster's work, though - it is about being able to be confident in the results. When a polling firm is transparent, their work tends to be better and their polls are often more accurate.

Also, providing the detailed breakdown of undecided opinion gives us a better understanding of what is going on behind the numbers. For example, it shows that half of those who did not vote in 2011 and two-thirds of those who can't remember if they voted are currently undecided. That means that a large proportion of those who are undecided are highly unlikely to vote in the next election anyway. It also, quite intuitively, shows that Liberal voters are about twice as likely to be undecided as supporters of the PCs and NDP, perhaps because these voters are waiting to see how the leadership race plays out. This would suggest that the next leader of the party has a good chance of improving the party's numbers, but also that the choice the party makes could send more of their 2011 supporters to other parties.

The poll showed the Progressive Conservatives ahead by seven points among men and the NDP four points among women, with the New Democrats in front among voters 29 or younger and the Tories ahead among those 45 or older. Ontarians between the ages of 30 and 44 were split three-ways between the parties.

That split also occurred in and around Toronto, where the Liberals had 32% to 31% for the Tories and 28% for the NDP. Outside of the GTA, the Tories were ahead with 39% to 35% for the NDP. The Liberals were well behind with 22%.
With these numbers, the Progressive Conservatives would win 43 seats to 33 for the Liberals and 31 for the New Democrats.

The Liberals were able to win more seats with fewer votes than the NDP due to their incumbency advantage, which makes the difference in several close races (often PC-Liberal contests). But the margin is close enough that either the Liberals or the NDP could emerge as the second-place party with the results of this poll.

Abacus's assessment of the Liberal leadership race falls in line with what other surveys have shown. Gerard Kennedy is the most well-known and liked, while Kathleen Wynne and Sandra Pupatello also score better than the others.
But the real story of this poll is suggested by the yawning gap between the favourable and unfavourable impressions of each of the candidates.

Even though Kennedy has the highest name recognition, 31% of respondents still did not know who he was or were unsure of their opinion of him, while another 31% had a neutral impression (more suggestive of not knowing much about him than anything else). Wynne and Pupatello combined for 72% and 74%, respectively, on the unknown/neutral score, while all others were at or over 80%.

The winner of the race will have their work cut out for them in order to become better known but also in ensuring that those neutral opinions don't become unfavourable. These numbers do suggest, however, that Kennedy, Wynne, and Pupatello would come out of the starting blocks in the best position.
The canyon narrows a little among Ontarians who voted for the Liberals in 2011, but nevertheless the number of respondents marked as "unknown" in Abacus's polling is still at around 40% for all candidates except the top three. Kennedy has very good name recognition among Liberal supporters, with only 19% marked as "unknown". His 44% favourable impression rating is also well ahead of Wynne (25%) and Pupatello (22%). Of note is that Harinder Takhar scores the highest unfavourable impressions among Liberal voters and Ontarians as a whole.

While Kennedy might have the highest favour among voters, he does not seem to have the same sort of organization as Wynne or Pupatello. That could be a deciding factor in a delegated convention. If an election is called shortly after the leadership race comes to a close, however, neither Wynne nor Pupatello would have the sort of immediate impact that Kennedy would have, at least according to these numbers. That could complicate matters for the Liberals, unless Wynne or Pupatello (or one of the other candidates) turn out to be strong campaigners.

The Liberals need a bump if they are to move definitively ahead of the NDP and out of third place. The Liberals are almost certainly in third spot, as they have been ranked behind the New Democrats in 11 of the last 14 polls (the current aggregation puts them at 27.9% to the NDP's 29.2%). The Progressive Conservatives, meanwhile, have been rock solid at between 32% and 38% in the last 17 polls stretching back to March. The challenge the next Liberal leader faces is twofold: regaining those lost voters who have swung to the NDP and chipping away at the Tories' granite base. No easy task.

Tuesday, December 18, 2012

NDP drops in EKOS polling, but...

Yesterday, EKOS Research released its newest federal poll via iPolitics.ca, showing the Conservatives with a wide lead over the New Democrats and Liberals. The survey results are not without their eccentricities, however, and paint a very different picture in Quebec than another poll released over the weekend by CROP via La Presse/Le Soleil.
EKOS was last in the field June 27-Jul 5, around the tail-end of the NDP's high watermark in 2012. It is no surprise, then, that the New Democrats fell 6.5 points from that survey to 25.8%, putting them well behind the Conservatives. They were up 1.7 points to 31.9%. The Liberals were up 4.9 points to 24.4%, while the Greens were down 1.6 points to 8.4%. Another 6.7% of respondents said they would vote for the Bloc Québécois and 2.8% for other parties.

This 25.8% is the lowest the New Democrats have scored in any poll since Thomas Mulcair became leader of the party. Any extreme result will always raise a few alarm bells, but EKOS was in the field primarily in November when the NDP was not polling very well so these numbers may not be too out-of-step.

The 2.8% for the other parties is a little high, though, and this is something of a feature of EKOS polling of late. The average result for other parties, when reported at all, has been 1.2% in polls by all other firms since the May 2011 election. The average for EKOS has been 2.5%, or more than twice other firms, and in the last election other parties received only 0.9% support. EKOS's IVR method may be partly to blame, as respondents punch in "other" rather than "undecided" or some similar response. But Forum Research, which is the other major firm using IVR, usually does not have high "other" results. It could be that Forum is applying some weighting to their polls or making corrections to reduce their "other" percentages where EKOS is not, but there is little reason to suspect that almost 3% of Canadians would vote for a fringe or independent candidate.

The Conservatives held an even wider lead in this poll among men, 13.5 points up on the Liberals. The NDP was narrowly ahead among women, holding a 1.3-point edge over the Tories. The New Democrats led among voters younger than 25, while the Conservatives were ahead among Canadians 25 or older.

The Conservatives led in Ontario with 35.8% and were trailed by the Liberals at 28% (+5) and the NDP at 26.5%. The Conservatives also led in Alberta with 53.5% (though that is their lowest result in the province in the last 35 polls stretching back to March), and were followed by the Liberals at 15.3% and the NDP at 13.7%.

Because EKOS assembled a very large sample, they were able to break the results in the Prairies down. This showed that the Conservatives held a lead in Saskatchewan with 47.8%, followed by the NDP at 29.8% and the Liberals at 14.6%. Compared to the election, this represents an 8.5-point drop for the Tories, a 2.5-point reduction for the NDP, and a gain of six points for the Liberals.

The Conservatives were also ahead (but the lead was within the margin of error) in Manitoba with 35.7%, down almost 18 points since the election. The NDP was up four points to 29.8% and the Liberals were up 3.4 points to 20%.

The New Democrats were in front only in British Columbia, with 33.7% support to the Conservatives' 31.3% and the Liberals' 17.1%. The Greens were at 15.6% in the province.

The Bloc Québécois had the advantage in Quebec with 27.7%, a gain of 6.9 points since EKOS's last poll. The Liberals were up nine points to 25.4%, while the NDP was down 15 points to 24.1%. That is the worst showing for the NDP since February, stretching back 62 polls. The Conservatives were well behind with 14.7%.

And in Atlantic Canada, the Liberals were ahead with 35.9%, with the Tories at 30.9% and the NDP at 26.4%. This is the fourth poll out of the last five to show the New Democrats with less than 30% support in Atlantic Canada, and the fifth consecutive poll to give the Liberals the lead or putting them in a tie for it. Could this be a new trend after the NDP had led in Atlantic Canada following Mulcair's arrival on the scene?
With the numbers from EKOS's poll and using the proposed boundaries of the 338-seat map, the Conservatives would win 147 seats to 84 for the Liberals, 68 for the NDP, 38 for the Bloc, and one for the Greens.

The big anomaly in this poll is in Quebec. Because the NDP's support was very uniform in the province, placing third means they move behind in a lot of seats. They lose out to the Liberals in Montreal and to the Bloc outside of Montreal, often by narrow margins. The Conservatives are also able to move ahead in a few seats due to the NDP's drop. In fact, these sort of numbers result in a lot of odd projections with parties winning seats with a lot less than 30% of the vote.

But it is hard to argue that this is what is currently the situation in Quebec. A poll by CROP released over the weekend showed a very different situation, with the New Democrats still in full control.
CROP was last in the field in Quebec Nov. 14-19, putting their last two polls just before and after EKOS's. In those two polls, CROP had the New Democrats at 41% and 40%, making it difficult to reconcile with the 24% EKOS captured in between. Perhaps CROP is the odd-man-out here, though, as the four polls in Quebec prior to CROP's November missive had the NDP at between 30% and 34%. In fact, CROP is the only pollster in the last 18 polls (stretching back to August) that has put the NDP at 40% in the province.

CROP put the Bloc at 22%, up six points since their November poll, and the Liberals at 21%, up four points. They put the Tories down 10 points to 13%, but their previous result of 23% was certainly an outlier.

It muddies the water a little bit, and makes me want to lean more heavily on the aggregation. That currently situates the NDP at a little over 33% in Quebec, making EKOS a low result and CROP a high one.

In terms of seats, and using the 78-seat map, the New Democrats would win 59 of them and continue their dominance of the province. The Liberals would make big inroads, though, with 13 seats to three apiece for the Conservatives and Bloc Québécois. The difference between these two polls is starkly portrayed here: either 10 or 59 seats for the NDP and either 3 or 38 for the Bloc. That isn't exactly a clear picture.

Altogether, though, there are a few things that we can learn from the EKOS and CROP polls. It is clear that the Conservatives are leading nationwide, and that the New Democrats are trailing in second. Virtually every poll has shown that to be the case for months. Quebec is in flux but the NDP is still well positioned, as they are in British Columbia. The Conservatives still lead in Alberta and the Prairies, while their margin in Ontario has been reduced but not fatally so. Atlantic Canada could be flipping back to the Liberals again.

Focusing on the individual results of the EKOS poll - or any poll - is to miss the forest for the trees: the wider trends are not put into question by EKOS's national poll. Rather, they provide more evidence to back them up. For the most part, it puts us back to where we were when 2011 was coming to a close.

Monday, December 17, 2012

Liberals, CAQ tied as PQ maintains lead

Two polls from Léger Marketing (for Le Devoir) and CROP (for La Presse) were released on the weekend, showing the Parti Québécois with a wide lead over the Liberals and the Coalition Avenir Québec. In fact, both polls show the two main opposition parties in a tie. But while both also put the PQ in majority territory, it is only by a hair in the one by Léger.
Léger was last in the field Nov. 21-22 and since then the PQ remained unchanged at 33% support. The Liberals and CAQ were tied for second with 27% apiece, representing a five-point gain for the CAQ and a four-point loss for the Liberals.

Québec Solidaire was up one point to 9% while Option Nationale was down two points to 2%. The Greens and other parties were unchanged at 2% and 1% each.

The Parti Québécois led among francophones with 40% and were trailed by the CAQ at 27% and the Liberals at 17%. Among non-francophones, the Liberals led with 61% to the CAQ's 24% (+12) and the Greens' 6%.

The Liberals narrowly edged out the PQ in and around Montreal with 31% to 29%, while the CAQ was up to 22% support in the region. The party was ahead in and around Quebec City with 36%, trailed by the Liberals at 29% and the PQ at 20%. And in the rest of Quebec, the PQ was in front with 40% to 30% for the CAQ and 22% for the Liberals.
CROP showed some similarities to the Léger poll, but also some important differences.

The firm last conducted its routine 1,000-sample poll between Nov. 14-19, and since then the Parti Québécois picked up one point to hit 36% support. The Liberals were down two points to 25% while the CAQ was down four points, also to 25%.

Québec Solidaire was up four points to 10% while support for other parties (including ON and the Greens) was up one to 4%.

The main difference between these two polls is the result for the Parti Québécois. Though a three-point spread (33% to 36%) is not of much concern due to the different field dates and the inevitable sampling error, it does mean the difference between a six-point margin and an 11-point one. Aside from random variation in the polling, the cause could be the question asked: Léger listed the parties without their leaders, while CROP mentioned the names of the party leaders.

This brings two factors into play. Firstly, interim Liberal leader Jean-Marc Fournier is not as well known as either Pauline Marois or François Legault. The Léger poll showed that 52% of Quebecers did not know who Fournier was or had no opinion of him (more on that later). That score was 24% for Legault and only 13% for Marois. Secondly, it makes respondents consider their voting intentions as of right now, which means with Fournier as Liberal leader. Respondents to the Léger poll might not be taking into account Fournier's leadership, instead thinking of the party generically or considering someone like Philippe Couillard as the leader (which is the most likely outcome of the leadership race).

Nevertheless, the difference between the two polls is not enormous and, if they both had completely random samples, within the margin of error. What they have in common might be more important - both polls show the CAQ and Liberals in a tie, with the PQ ahead. Both polls show the PQ relatively stable and the Liberals slipping. Both polls show Québec Solidaire around 9% or 10% support and trending upwards. Both polls put the PQ at 40% to 42% among francophones with the CAQ in second at 27% or 28% and the Liberals with 17% among this linguistic group. Those are some significant similarities.

CROP gave the PQ the edge in and around Montreal, however, with 37% support to 30% for the Liberals and 22% for the CAQ. The CAQ was in front in and around Quebec City with 38% in this poll (-14 since CROP's poll of Oct. 17-22, the last one with regional data available) to 28% for the PQ (+13) and 27% for the Liberals. The PQ led in the rest of Quebec with 38% to 26% for the CAQ and 19% for the Liberals.
In terms of seats, the CROP poll would deliver a big majority to the PQ with 82 seats to 23 for the Liberals, 17 for the CAQ, two for QS, and one for Option Nationale.

The PQ would win its majority primarily in the suburbs around the island of Montreal, with 27 seats to only three for the Liberals. The PQ's dominance in the regions of Quebec is also felt here.
Léger's numbers would deliver the smallest of majorities to the PQ with only 63 seats. The Liberals would win 35, the CAQ 23, and Québec Solidaire four. Presumably, the PQ's majority could be made somewhat more comfortable with the support of QS.

Interestingly, this is a far different result from the projection for Léger's last poll, when the PQ held only a two-point edge over the Liberals. How can a wider lead result in a closer seat result? The answer is in the performance of the CAQ. Léger's last poll had the CAQ at only 22%. This made it much easier for the PQ to win a lot of seats north and south of Montreal. In a two-horse race with the Liberals, the PQ's advantage among francophones is crushing. But when the race becomes closer and the CAQ is over one-in-four support, the PQ's edge among francophones diminishes. This gives the Liberals a chance in closer races and puts the CAQ in a position to win seats from the PQ as well.

Pauline Marois's real foe in the National Assembly is thus François Legault. She has much more to gain by winning votes back from the CAQ rather than from the Liberals. According to the CROP poll, Legault is seen as the best person to be premier by 24% of Quebecers, compared to 23% who chose Marois. Françoise David of Québec Solidaire came up third with 10%, ahead of Jean-Marc Fournier's 9%.

Marois's numbers have held relatively stable on this question, but Legault's are on the rise. He was at 18% in CROP's poll in October and 22% in their November poll. Meanwhile, satisfaction with the PQ's government appears to be trending downwards: around 43% in November to 39% in CROP's poll. It has dropped from 37% to 32% in Léger's polling.

Support for sovereignty is relatively steady, however. CROP had it at 34% and 35% in their two November polls. This survey has it at 38%. That is an uptick, but it has been wavering back and forth around 38% for sometime, and this could just be a move back towards the mean.

Léger's poll included their bi-yearly "barometer" of Quebec's political leaders. The results of the top 25 figures, in terms of the best "good opinion" scores and colour-coded by party affiliation, are as follows:
Françoise David and Jacques Duchesneau each managed 48% good opinion, compared to 17% who had a bad opinion of them. François Legault was not far behind with 47% good opinion, but his bad opinion score was 29%. Pauline Marois and Marguerite Blais round out the top five, with 44% and 40% good opinion, respectively. Marois's bad opinion score was 43% to Blais's 15%, however.

(Blais was a former TV personality, which explains some of her popularity).

Also high on the list was Philippe Couillard (39% good to 29% bad) and Raymond Bachand (33% good to 23% bad), the two front runners for the Liberal leadership. Interim leader Jean-Marc Fournier was further down the list, with 27% expressing a good opinion of him and 21% a bad one.

Also of note on the list are Amir Khadir, who had the highest bad opinion score with 50%, and Nicolas Marceau, Minister of Finance. He had 23% good opinion to 14% bad, with 63% still unaware of who he is or unable to form an opinion about him.

In terms of the biggest winners, David's good opinion score increased by 23 points since Léger's last poll in June, while Marceau's was up 15 points and Marois's 10 points.

By party, Marois (89%) was the most popular among PQ supporters, followed by Léo Bureau-Blouin (71%) and David (66%). Only Couillard (54%) and Bachand (52%) had majority good-opinion among Liberal voters. Legault (85%), Duchesneau (70%), and the last leader of the ADQ, Gérard Deltell (66%) were the most favoured among CAQ supporters. Nothing unexpected here, but the ambiguous feelings towards the two men most likely to take over the PLQ leadership by their own supporters is certainly not a vote of confidence.

It is difficult to see much incentive for the CAQ and Liberals to pull the plug on the PQ in the spring. Though the CAQ's numbers are holding strong and François Legault's own ratings are moving in the right direction, it is not unusual for a third party to gain favour when then main opposition is distracted by a leadership race (for a recent example, see Bob Rae). The CAQ's numbers may drop when Couillard (or Bachand) becomes leader, but the Liberals will then have a long way to go among francophones if they have any hope of defeating the PQ in the next election - and the Charbonneau Commission is not going away anytime soon. Despite her relatively flaccid numbers, these factors may ensure Marois's tenure as premier continues through to at least the fall of 2013.

Thursday, December 13, 2012

Abacus puts gap between Conservatives and NDP at two points

The latest poll by Abacus Data for Sun Media shows that the New Democrats are keeping it close in the race with the Conservatives, as the Liberals remain mired in third place.

With a particular focus on the personal numbers of Stephen Harper and Thomas Mulcair, I analyzed the poll for The Huffington Post Canada in today's article. Here, I'll focus on the voting intentions numbers in the poll.
Abacus was last in the field Nov. 9-11, and since then the Conservatives dropped two points to 34% support, while the New Democrats were up three points to 32%. In addition to the gap between the two parties, neither of these shifts in support appear statistically significant.

The Liberals were unchanged at 22%, as were the Greens at 6%.

A few notes on methodology and reporting. Kudos to Abacus Data for including their unweighted regional sample sizes in their poll report, which puts them in rare company (Ipsos-Reid is the only other pollster who does so, though they go the extra mile by including the unweighted samples for other demographic breakdowns). Abacus uses an online panel for their polling, but it isn't mentioned in their report whose panel it is. We can safely assume it is Angus-Reid's, however, though Abacus designs its own survey and weighs the data according to their own parameters.

It also needs to be pointed out that Abacus over-sampled Ontario (821 in all, rather than the 320 or so that would have normally been polled) and gave the province an appropriate weighting in the national percentages. This gives the Ontario numbers a much smaller margin of error than they usually would have in a poll of 1,505 Canadians, but the other regions' results have the sort of margin of error that would normally apply in a national poll of 1,000 - assuming a random sample, of course.

The Conservatives led in this poll in Ontario with 38%, and were trailed by the NDP at 29% and the Liberals at 27%. The Conservatives also led in Alberta with 63% to the NDP's 19% and the Liberals' 12%. In the Prairies, the Tories were narrowly in front with 43% to 37% for the NDP and 15% for the Liberals.

The New Democrats led in Quebec with 39%, one of their higher recent results. They were up nine points since Abacus's last survey. The Bloc Québécois trailed with 25%, with the Conservatives at 17% and the Liberals at 16%. That is a low result for them, compared to other recent polls. Are the Liberals faltering in Quebec? This is the second consecutive poll to put them at below 20% support, after the eight previous polls put them over 20%. Something to keep an eye on.

The New Democrats were also ahead in British Columbia with 43%, their highest result in any poll in the province since June. The Conservatives trailed with 31% while the Liberals were at 14% and the Greens at 11%.

In Atlantic Canada, the Liberals were in front with 45%, a gain of 14 points since Abacus's last poll. This is actually an extraordinary, and likely outlier, result for the Liberals as we have to go back 137 polls to find a better number (in early April 2011). The Conservatives were second with 30% while the NDP had 24%, their lowest result in the last 45 surveys of the region. The results in Atlantic Canada, then, may not be worth very much, as is usually the case considering the small sample size.
With these numbers, and using the proposed boundaries of the 338-seat map, the Conservatives would win 141 seats. The New Democrats would fall just short with 128, while the Liberals would win 62, the Bloc six, and the Greens one.

This is one of the better results for the NDP we've seen, and demonstrates a much higher vote-efficiency than is usually the case. This is in large part because of two provinces: British Columbia and Quebec. Keeping close to 60 of Quebec's seats is essential for the NDP, while a strong showing in British Columbia and, to a lesser extent, the Prairies is also very important. If the NDP could do a little better in Ontario and with more plausible results in Atlantic Canada, a plurality of seats is clearly possible with little movement in the numbers.

That should be of great concern to the Conservatives, who normally have a bit of insulation in the numbers due to their dominance out west. Losing this many seats in B.C. and the Prairies, and taking less than 70 in Ontario, puts them in grave danger. A few shifts in the numbers due to polling error and a few variations in the seat projection due to its own inherent error, and you have the Tories winning fewer seats than the New Democrats (or a few more, of course).

But it has to be recognized that this poll, and the poll by Léger Marketing released over the weekend, is a little out of step with the consensus. Or, at least, the consensus from November when the Liberals were in a close fight for second. Have the Liberals really taken a hit over the last month or are methodological issues to blame for the discrepancy? We'll need more data to find out.

Tuesday, December 11, 2012

Léger poll on LPC leadership race

Over the weekend, Léger Marketing released a new poll via Le Devoir and the Montreal Gazette looking at the nationwide federal situation and the on-going Liberal leadership race. The poll shows that Justin Trudeau is still favoured and boosts Liberal support considerably, but that there may be a bit of a downturn in his appeal.

I wrote about what these numbers may mean for The Huffington Post Canada, but here I will go over the numbers in more detail.
First, the current numbers where Léger asked respondents who they would vote for without any reference to party leaders.

Léger was last in the field nationwide ages ago in early April. Since then, the Conservatives increased their support by three points to 35%, while the New Democrats were down three points to 30%. The Liberals were down one point to 18% and the Greens were up one point to 9%.

These shifts are not statistically significant but the lead the Tories held over the New Democrats was.
*****
Léger and Le Devoir, CROP and La Presse

A quick note about Léger Marketing and Le Devoir. Both organizations deserve a lot of credit for how they make their data available. Léger always puts their reports on their website very quickly, and Le Devoir goes the extra mile by posting the PDF of the report itself on their own website along with the article covering the poll. This is highly commendable and other media outlets should do the same thing. 

This is in sharp contrast to Léger's main competitor in Quebec. CROP often does not post their results on their website. This is because La Presse pays for the poll and so the results are "theirs", but La Presse nevertheless has a public responsibility to make the results available. If Le Devoir and Le Journal de Montréal, which also works with Léger, can do it, so can La Presse

In fact, in a federal election campaign it would be against the Elections Act to not make full reports available to the public. If Elections Canada deems this a necessity for the public good during an election campaign, it should be no different outside of one. 

La Presse should allow CROP to post the details of their polls to their website, if not immediately then at least 24 or 48 hours after the poll is published in the newspaper. At that point, the 'value' of the poll has been completely spent and La Presse loses nothing by allowing CROP to post the details.
*****
The Conservatives led in Ontario with 40%, followed by the NDP at 26% and the Liberals at 19%, down six points since April. The Greens were up four points to 13%, an unusually high result for them.

The Tories also led in Alberta with 63%, while the NDP was up to 21% and the Liberals were down to 10%.

The Conservatives were in front in the Prairies, where they dropped 14 points to 35%. The NDP trailed with 33% and the Liberals with 23%.

In British Columbia, the New Democrats and Conservatives were tied with 39% apiece, while the Liberals were down 10 points to 12% and the Greens were at 8%.

The New Democrats were ahead in Quebec with 35%, representing a 17-point drop from the 52% that Léger pegged the NDP at in their June Quebec-only poll. The Bloc Québécois was up 11 points to 29%, while the Liberals were at 17% and the Conservatives at 15%.

The Liberals had the edge in Atlantic Canada with 39% (+19), while the NDP was down 24 points to 25%. The Conservatives trailed in third with 17%, a very low result for them, while the Greens were at 16%, a very high result for them. But this is not exactly an anomaly - the Greens were at 15% in Léger's last poll. It would appear that their online panel in Atlantic Canada is somewhat non-representative, and a little unreliable. For example, when Marc Garneau and Martha Hall Findlay were mentioned as potential Liberal leaders support for other parties shot up to 9% and 10% in Atlantic Canada. It may be a good idea to put those numbers aside.
With these results and using the 338-seat proposed electoral map, the Conservatives would win 166 seats, just four short of a majority. The New Democrats would win 109 seats and the Liberals 45, with 17 seats going to the Bloc Québécois and one to the Greens.

This puts the Tories in a stronger minority position than most polls, as the NDP and Liberals would need the support of the Bloc to defeat the Conservatives.

The numbers in this projection are actually quite good for the NDP in the West, but the low result for the Liberals in Ontario delivers a huge amount of seats to the Tories in Ontario. The weakness of the NDP in Quebec also damages their chances of forming government.

But this poll is somewhat out of step with others - not so much in terms of where the NDP and Conservatives stand, but where the Liberals are. Out of the last 15 polls (stretching back to mid-July) before Léger's, the Liberals were pegged at 20% or more in 14 of them. Perhaps Léger is capturing a downward trend in Liberal support - we'll have to see what other polls say.

The elephant in the room is, of course, the Liberal leadership race. Léger asked respondents how they would vote with Trudeau, Garneau, or Hall Findlay as leader. It is difficult to read much into the results, as a lack of name recognition is a factor for Garneau and a major one for Hall Findlay. Half of the population did not know who Hall Findlay was and more than one in four did not recognize Garneau. Nevertheless, the results are interesting.
We'll start with Martha Hall Findlay's numbers, which are probably more indicative of what would happen if the Liberals completely collapsed rather than what would happen if Hall Findlay won the leadership (at that point, we can assume that she would be more well known).

In this scenario, the Liberals drop to 12%, losing their support about equally to the Conservatives and NDP. The collapse of the Liberals is particularly beneficial to the New Democrats in the Prairies, Quebec, and Atlantic Canada, but the Tories take advantage in Ontario. The Bloc also grabs some support and becomes a thorn in the NDP's side in Quebec.
Marc Garneau's numbers are not much different from those that Bob Rae manages on Léger's first question. The only notable differences are in Ontario, where Garneau takes from the Conservatives, and Atlantic Canada, where his leadership pushes supporters to the NDP.

With these results, Garneau looks like a bit of a status quo option. He won't hurt the Liberals but he may not do much better than Michael Ignatieff. However, his name recognition is rather low and he does have good personal approval ratings: 38% of respondents had a good opinion of Garneau against only 8% who had a bad opinion of him. This is understandable as Garneau appears to be a rather unoffensive, uncontroversial option for the Liberals. But that might not exactly galvanize support behind his party in an election campaign.
Justin Trudeau, however, does manage to shake things up dramatically. He increases his party's support to 31%, a gain of 13 points coming from the NDP (-6), Conservatives (-4), and Greens (-2). He dominates in Atlantic Canada, ties the NDP in Quebec, and is just a handful of points behind the Conservatives in the Prairies and Ontario.

But hold on - a tie? Forum Research has put the Liberals under Trudeau ahead of the Conservatives by nine points and has given them 39% support. A Harris-Decima poll showed that 42% of respondents were certain or likely to vote for a Trudeau-led Liberal Party (that same poll put Garneau's certain/likely support at 20%, almost identical to Léger's numbers). Have the recent controversies surrounding Trudeau whittled down his support? We will have to wait and see what Forum and Harris-Decima have to say in their next polls, but it appears plausible (and expected). However, the numbers that Trudeau was putting up before were always unrealistic.

In terms of seats, a collapse of the Liberal Party (as hypothesized by Martha Hall Findlay's results, though it has to be recalled that this assumes an election right now, where more than half of Canadians have never heard of her, rather than an election in 2015 when she would have had two years to introduce herself to Canadians) would benefit both the Tories and NDP. The Conservatives would get a very slim majority with 170 seats to 131 for the NDP. The Liberals would place fourth with 14 seats, behind the Bloc's 22.

Marc Garneau does about as well as the Liberals would do presently, though his weaker numbers in Quebec give the Bloc a few more seats.

Justin Trudeau would win 116 seats on the numbers attributed to him by Léger. The Conservatives would be reduced to 129 and the NDP to 63, while the Bloc Québécois gets their best result at 29 seats (winning the plurality of seats in Quebec). But if they were so inclined, the Liberals and NDP could combine for a majority in the House.

Those seats numbers spell things out quite clearly: Martha Hall Findlay is not well-known and Marc Garneau does not change the game for the Liberals in the way that Justin Trudeau would. It is no surprise, then, that 38% of Canadians chose Trudeau as the best option to lead the Liberal Party, ahead of Garneau (16%), Hall Findlay (3%), and Deborah Coyne (1%). All other candidates were rounded down to 0%. Among Liberals, Trudeau does better: 60% to Garneau's 20%. Hall Findlay manages 2% among Liberal supporters, while Coyne and David Merner get 1% apiece.

A Trudeau leadership was always going to be high-risk/high-reward, and he is still the favourite to win it (according to the data filed with Elections Canada, he is outstripping his rivals in fundraising by leaps and bounds). But this Léger poll could be the first indication that the shine is starting to wear-off. This will be something to keep an eye on over the next few weeks until the leadership race gets rolling with the first debate in January.

Monday, December 10, 2012

November 2012 federal polling averages

Four polls surveying a total of 5,930 Canadians were conducted during the month of November, suggesting a widening gap between the ruling Conservatives and the opposition New Democrats and Liberals. The news might be worse for the NDP, however, as they are now back to where they were in March when Thomas Mulcair won the party's leadership race.
The Conservatives averaged 34.1% during the month of November, up 2.1 points since October. The New Democrats were down 1.6 points to an average of 28.6%, their lowest result since March and their fifth consecutive month of stagnation or decrease.

The Liberals were down two points to 26.3% in November, but it was the second consecutive month in which the Liberals managed a better score than anything registered since the May 2011 general election.

The Bloc Québécois averaged 6.3% while the Greens were up 0.6 points to 4.1%. Support for other parties averaged 0.6%.

But it is problematic to compare November's results to October's. Only two polls were in the field nationwide that month. Instead, a better picture can be obtained by looking at how the polls have moved since the four firms that were active in November were last in the field at around the same time.
That brings us back to May 26 through to June 23, when Nanos, Forum, Ipsos-Reid, and Abacus all conducted polls.

The average of the polls from these four firms in May-June gave the NDP 35.9%, the Conservatives 33.4%, and the Liberals 21.2%. A similarly unweighted average of the polls in November gives the Conservatives 34.2%, an increase of 0.8 points, the NDP 28.6%, a drop of 7.3 points, and the Liberals 26.3%, an increase of 5.1 points. This time period is useful as May-June was the height of the Mulcair honeymoon. It is clear from how these polls have moved that the NDP has bled support to the Liberals since that time.

The Conservatives led in British Columbia with an average of 40% support, their best result since October 2011. The party is up 8.8 points since October 2012, but a quick look at the chart to the left shows how the month was likely an outlier. The NDP placed second with 34.3%, while the Liberals were at 19.5% and the Greens at 5.2%. That is the lowest result for the Greens on record, stretching back to January 2009.

In Alberta, the Conservatives averaged 60.1% to 18.2% for the Liberals, 16.7% for the NDP, and 4% for the Greens.

The Conservatives were ahead in the Prairies with 46.7%, their highest result since January 2012, while at 27.9% the NDP was at their lowest since September 2011. The Liberals were up for the third consecutive month, hitting 19%. The Greens were up to 5.6%.

The Conservatives had 37.7% support in Ontario and were trailed by the Liberals at 29.2%. The NDP managed 27.5%, but with the exception of a blip in July and August the party has been on a steady decline since May. The Greens averaged 5% in the province.

The New Democrats have been sliding in Quebec since June, and at 31.6% registered their lowest average support since March. The Liberals were up to 27.7% in Quebec, their best showing since September 2009. They have been steadily increasing since June, taking their support from the NDP. The Bloc Québécois averaged 24.7%, while the Conservatives managed 13.5%. The Tories have been relatively steady at this low level of support since March. The Greens averaged 2% in Quebec.

And in Atlantic Canada, the Liberals were at 34.2% and have been holding at their highest level of support since April 2011 over the last two months. The NDP averaged 32.8%, their worst since March 2011, while the Conservatives were in third at 27.4%. The Greens averaged 5.2% support.
With these numbers, the Conservatives would win 149 seats on the proposed boundaries of the 338-seat House (though not taking into account the final reports for some of the provinces released at the end of November). That is a gain of 21 seats over their October projection.

The NDP is down 24 seats to 91, while the Liberals are down seven seats to 85. The Bloc is up 11 to 13 seats, while the Greens would be shutout.

The major gains for the Tories occurred in British Columbia (+10 from October) and Ontario (+7), while the major losses for the NDP took place in Ontario (-16) and the Prairies (-7). The Liberals dropped eight seats in B.C. and seven in Ontario, but picked up six in Quebec in the projection.

Approval ratings
November was a bit of a troubling month for the New Democrats. Though they were still in second nationwide and were leading in Quebec, the trends are not heading in the right direction in Ontario, Quebec, and Atlantic Canada. The Liberals are eating into their support in Quebec (where, it should be repeated, the party put up its best numbers in three years) and are making it more difficult for the NDP to make inroads in Ontario.

The Liberal leadership race is likely one of the main contributing reasons for the party's uptick in support, particularly in Quebec. But the Léger poll released over the weekend should be cause for concern. Though Justin Trudeau still put up impressive numbers, tying the Conservatives at 31% if he led the Liberal Party, that is very different from the wide lead that has been given to him in other polls. This could be a methodological issue, but it might also be an indication that the shine is starting to wear off.

That does not leave the Conservatives out of the woods. They are still down six points from where they were on election night, and have suffered big losses in Atlantic Canada and are poised to lose a lot of seats in British Columbia and Ontario. Though the party appears to have hit a bit of a floor, their trend line is still rather negative since May 2011 and they are far from being in majority territory. Nevertheless, they are still in the most enviable position of the three major parties.

Friday, December 7, 2012

Liberals lead in Nova Scotia, but is it enough?

Earlier this week, the Corporate Research Associates released their quarterly poll on Atlantic Canadian politics, showing two very interesting races in Nova Scotia and New Brunswick. In both provinces, the governing parties are either tied are trailing the Official Opposition by a wide margin. This is especially problematic in Nova Scotia, where an election is expected some time in 2013.

As I've already done an analysis of these polls for The Huffington Post Canada here, I invite you to read that article. I'll quickly go over the numbers here before getting to the seat projections.
CRA was last in the field in August, and since then the Liberals have been unchanged at 41%. The New Democrats were down two points to 29% while the Progressive Conservatives were up five points to 27%. The Greens were down two to 3%. None of these shifts appear to be statistically significant, though the lead the Liberals enjoy over the NDP is.

While the gap widened between the Liberals and the NDP, it narrowed between Stephen McNeil and Premier Darrell Dexter. On who would make the best premier, McNeil dropped two points to 33% while Dexter picked up three points to hit 26%. Jamie Baillie of the Tories was unchanged at 17%.

But is the Liberal lead enough to give them a majority government? Maybe not. More on that in a moment.
New Brunswick is scheduled to go to the polls in 2014, so the numbers are less urgent. Since CRA's last poll, the governing PCs have been unchanged at 38%, but the Liberals were up six points to 38% after electing their new leader, Brian Gallant. The New Democrats were down five points to 19% while the Greens were down two to 4%.

Note that this is the smallest sample of CRA's four provincial polls, and when you take the undecideds out of the equation the margin of error becomes quite large. The significance of Gallant's gain may be nothing but a statistical wobble. CRA's next poll in three months will tell us whether that is the case or not.

Gallant's numbers improved dramatically over those of interim leader Victor Boudreau: they jumped 11 points to 29%, tying with Premier David Alward (down four points). Dominic Cardy of the NDP was down four points to 11%.
But the Progressive Conservatives seem to hold the advantage, as they would still win a slim majority government with a tie in the vote. It is difficult to say this with great confidence however, as there is no regional data available. The swing is being applied province-wide.

The Tories win 29 seats to 22 for the Liberals and four for the NDP. The PCs win seats in every part of the province, with the southwest (Saint John and surrounding areas) being their strongest. The Liberals win 15 of their 22 seats in the northeast and southeast (Moncton), while winning a slew of seats in Fredericton. The New Democrats win their seats in the northeast corner of the province and in Saint John.
The situation in Nova Scotia is also more favourable than it should be for the incumbent government. Despite trailing by 12 points, the New Democrats would still be capable of limiting the Liberals to a minority government of 24 seats to 20 for Dexter's NDP and eight for the Tories.

A few caveats, however. Again, this projection is being made without regional data and the boundaries are those that were in force in 2009. But the incumbency effect is important for the NDP as several ridings are very, very close - a few are even three-way races where third is separated from first by two or three points.

One of the biggest problems for the Liberals with this poll is the performance of the Progressive Conservatives. Almost half of the seats won by the Tories in this projection are Liberal-PC contests. In another five seats won by the New Democrats, the NDP and Liberals are almost neck-and-neck but the Tories have a large share of the vote. In effect, there is a vote-split among the opposition that chips away at the Liberals' chances of winning a majority outright. A lot of ridings are rather close, though, so if these were the numbers heading into election night the Liberals could still pull it off.

But while the Liberals' main rival is of course the NDP government, they may have more to gain by pushing support for the Progressive Conservatives back down. In fact, they had a better seat result in CRA's last poll when the NDP was at 31% and the Tories were at 22%. The Liberals need to hold at least a 10-point lead over the NDP in order to have good chances at winning a majority, but if the PCs start pushing 25% or more than even a double-digit advantage may not be enough.

Thursday, December 6, 2012

Ghiz and Dunderdale hold wide leads in CRA poll

Yesterday, Atlantic Canada was all a-flutter with the poll results of Corporate Research Associates' report, a  quarterly tradition on the East Coast. The two provinces furthest away from their next elections (Newfoundland & Labrador and Prince Edward Island) show that the incumbents hold wide leads over their nearest rivals. The poll results led, in part, to the resignation of Olive Crane as leader of the PEI PCs.
We will start there. CRA was last in the field throughout the month of August, and since then the Liberals picked up three points to hit 45%, well ahead of the Progressive Conservatives. They were down four points to 28%, while the New Democrats were up four points to 22%. Another 5% said they would vote for the Greens, down three points.

Fully 40% of respondents were classified as "undecided, no answer, do not plan to vote, refused to state", which is par for the course in CRA's polling. Thankfully, CRA increased their sample size in PEI from 300 to 600 in this poll, otherwise the sample of decided voters would have had a less useful margin of error of +/- 7.3%.

The Liberals have generally held stable over the last year or so, but they are down eight points from where they were in November 2011. The Tories have also been quite stable, with between 26% and 34% over the last 12 months - as good as a straight line considering the small samples. The New Democrats, however, have been on a steady increase, up from the 9% to 11% they were polling in November 2011 and February 2012. At 22%, the NDP are at their best polling level since at least November 2010, and likely for quite some time before that as well.

But the New Democrats are hard to classify as a major party in PEI. They only ran a half-slate in the 2011 election and were beaten by the Greens. If the NDP runs a full slate in 2015, they could do quite well.

On leadership, Premier Robert Ghiz was up three points to 36% on the Best Premier question, while Crane was down five points to 19%. That is her lowest result for some time, having polled between 22% and 26% over the last year. Likely this had something to do with her decision to step down as leader. She barely did better than new leader Mike Redmond of the NDP.

Satisfaction with Ghiz's government stood at 43% (+3) while dissatisfaction was down two points to 51%. It has to be pointed out that every shift in the polls on every question was within the margin of error, so it is difficult to definitively say whether anything has really changed in the last three months. The longer trends are clearer, however: the Liberals and PCs down a tick to the benefit of the NDP.
In Newfoundland and Labrador, the governing Progressive Conservatives were up one point to 46%, compared with CRA's last poll in August. The NDP was down two points to 31%, while the Liberals were up one to 23%. Status quo, then.

The PCs have stopped their recent decline, as they were as high as 60% twelve months ago. Things have stabilized, with the NDP polling around 30% since the last election. The Liberals have been slowly increasing their support, however, with 23% tying their best result since February 2011, when Telelink put them at that level.

Kathy Dunderdale was seen as the best option for premier by 36%, while Lorraine Michael was second with 29% and Dwight Ball, interim leader of the NL Liberals, was at 18%. Any shifts from August were within the margin of error, and generally the leadership numbers have followed the same trends as have the voting intentions.

Satisfaction with the government was rated at 58%, unchanged since August, while dissatisfaction was up three points to 37%.
In terms of seats, both Dunderdale and Ghiz would win another majority government on these numbers.

In PEI, the Liberals would win 24 seats to three for the PCs, with all of those Tory seats being won in the eastern part of the island.

In Newfoundland and Labrador, the Tories would win 29 seats to 11 for the NDP and eight for the Liberals, meaning that Michael would head-up the Official Opposition.

The Tories would win seats in every part of the province, while the NDP would be heavily concentrated in and around St. John's and the Liberals in the central and western parts of the island and in Labrador.

The thing to watch in Prince Edward Island is whether the New Democrats are really going to maintain this level of support and be able to run a full slate of candidates. One suspects that these polling numbers would help them find another 26 names to put on the ballot beside Redmond's. If that does occur, the fallout would be a little unpredictable, as the NDP has never been a real factor in PEI.

The NDP is also worth watching in Newfoundland and Labrador, as they have improved upon their 2011 election result and have maintained that support for almost a year. The Liberals aren't going away either, and could be revitalized after they choose their new leader. Dunderdale still has a wide advantage but that could slip away in the midst of a campaign. If the Liberals continue to pick-up support, a minority government in Newfoundland and Labrador in 2015 is not unthinkable.

Wednesday, December 5, 2012

B.C. Liberals close gap, but NDP still heavily favoured

A new poll by Ipsos-Reid was released last night, nudging the projection in the B.C. Liberals' favour by a small degree. The poll, which put the gap between the Liberals and the New Democrats at 13 points, results in the forecast now giving the NDP a 91.7% chance of winning the popular vote on May 14, down from the 95.7% chance forecasted as of Nov. 21. In other words, the odds of a victory by Christy Clark have been reduced from 20 to 1 to about 10 to 1.

But the projected support for the New Democrats has hardly moved at all, dropping 0.2 points to 48.5% and two seats to 60, still well above the 43 needed for a majority. In fact, the current seat projection range (50 to 71 seats) still puts the NDP entirely in majority territory. This is part of the reason why, if an election had been held Nov. 30, the New Democrats would still have had a 97.6% chance of winning it outright.

The Liberals picked up 3.4 points in the projection and now stand at 33.1% and 24 seats, but their high projected seat range of 34 still puts them only as the Official Opposition. The forecasted vote range, however, now overlaps with the New Democrats', at 46.1% forecasted high for the Liberals to a 44.5% forecasted low for the NDP.

The New Democrats and Liberals are the only two parties likely to be represented in the legislature, as the Conservatives are no longer considered in play in any ridings. Their forecasted high has dropped from 23 to 13 seats, even under a best case scenario. Their support has dropped 2.3 points to 10.5% in the projection. The Greens are also down by 0.9 points to 6.1%.

Regionally, the New Democrats were up in Metro Vancouver (47% support) and in the Interior and North (45.9%), but were down on Vancouver Island to 54.1% support. The Liberals were up in every region, but do not overlap with the NDP anywhere in the projected ranges. Their current projected results sit at 34.9% in Vancouver, 29.1% on Vancouver Island, and 33.6% in the Interior and North.
As to the Ipsos-Reid poll itself, the firm was last in the field Sept. 11-18 and since then the New Democrats were down one point to 48%, still giving them a wide lead over the Liberals. They were up three points to 35%, while the Conservatives were down three points to 9% (a statistically significant slip). The Greens were up one to 7%.

The recent Angus-Reid poll shows a similar trend. It was in the field about a week before Ipsos-Reid's, and they also had a poll from early September (Sept. 10-11). Their shifts over that period but the NDP up one, the Liberals up four, and the Conservatives down seven points. It is clear that the Liberals are making gains entirely on the backs of the Conservatives. This is good for them in the sense that it puts the Liberals closer to the NDP, but even if every Conservative voter swings to the Liberals the New Democrats would still hold a four-point edge in the polling.

But there are some odd demographic variations in the numbers. The Liberals actually hold a two-point advantage among men but trail among women by 29 points. That is enormous. The New Democrats also hold a wide lead (54% to 37%) among voters aged 55 and older - a cohort that turns out in big numbers.

Not all is bad news for Clark, however. Some 59% of the people still consider the economy to be in good or very good shape, and 46% think the province is headed strongly or somewhat in the right direction (42% think it is headed the other way). While those are good fundamentals for Clark and the Liberals, the party still trails the NDP by a wide margin and Clark herself polls very poorly at a personal level.

For example, Adrian Dix is chosen as the best person to be Premier by 39% (+4 since Ipsos's last poll) while Clark is chosen by 23%. If we take the undecideds and "none of the aboves" out of the equation, however, Dix's numbers improve to 54% to 32% for Clark. This suggests that Clark is dragging down the Liberals' vote while Dix has the potential to pull it up.

Their relative approval ratings also indicate that Dix is a net positive for his party. His approval rating stands at 53% (+2) with 34% disapproving of him. That is almost the mirror image of Clark's numbers: 34% approval to 59% disapproval. Of more concern for Clark should be the proportion who strongly disapprove of her: 35%, compared to only 17% for Dix. And Dix has 16% who strongly approve of him, compared to only 5% for Clark.

Note that I have added a tracking chart with a rolling average for leaders' scores on Best Premier and approval ratings to the B.C. projection page.

But Clark seems to be out of the woods in terms of being swamped by John Cummins of the Conservatives. Only 7% chose him as the best person to be Premier and his approval rating plummeted 10 points to only 13%, with his disapproval soaring 11 points to 51%. Fully 30%, or almost one-in-three British Columbians, strongly disapprove of him. Cummins might have been a good fall back option if Clark stumbled on the campaign trail, but he is becoming somewhat toxic for his party (though the party itself might share in the blame as well).

The odds of an NDP landslide are diminishing and the race could become a lot closer, particularly if the Conservative vote continues to drop. But there is not much time left for Clark to make this a truly competitive race.

Monday, December 3, 2012

BC New Democrats hold wide lead in first projection

The next election in British Columbia is less than six months away, and in ThreeHundredEight.com's first projection and forecast for the province the New Democrats hold a wide lead over the governing B.C. Liberals. They are also heavily favoured to win, even six months out.

An analysis of the projection results themselves can be read at The Globe and Mail here. Rather than go over the results a second time, I will explain some of the projection's new features.

A complete explanation of every aspect of how the projection and forecasting models work can be found here. For the most part, the basic methodology has not changed so the bulk of that explanation is covering what I have already described in detail before. In this post, I'll go over the features that have not been presented in past elections.

With this new model, I wanted to tackle a few problems but also take a somewhat different approach to how projections are presented. The most important problem is how to bridge the gap between what the polls show and what the election results end up being. The amount of data available in Canadian elections is relatively low, especially compared to the United States and particularly at the tail end of a campaign. It is common for the final polls of a campaign to be conducted days before the actual vote, leaving a big gap between the last information available and the time that people cast their ballot.

The projection now makes two calls. The first, which is what I call the projection itself, is the best estimate of a likely outcome if an election were held on the last day that the most recent poll was in the field. In the case of British Columbia, that is Angus-Reid's poll of November 21. The projection, then, is the best estimate of the outcome of an election held November 21.

In order to measure the degree of error that can be expected in the polling, high and low projection ranges are tallied. This is determined by applying the margin of error for the estimated "sample" of the projection itself. This sample is measured by comparing the weights and sample sizes of other polls to the most highly weighted poll in the projection. For example, let us assume that a projection has three polls in the database. All three have a sample of 1,000 people. One poll is rated at half the weight of the most highly rated poll, while the third poll is weighted at one-fourth of the most highly rated poll. The estimated "sample" of the projection would then be considered to be 1,750 people (1,000 people from the first poll, 500 from the half-weight poll, and 250 from the quarter-weight poll). That 1,750 sample is then used to determine the margin of error for each party, which changes according to the support a party holds. An easy demonstration of this is that a party with 2.5% support has a different margin of error than a party with 25%. If the margin of error in the poll is 3.1%, the range of results for the small party can't be as low as -0.6%.

By calculating the margin of error attributed to the projection's "sample", that is used to determine the likely ranges of the projection itself. Theoretically, this range is based on the assumption that the polls have accurately reflected the mood of the electorate, within their respective margins of error.

The current projection is heavily skewed towards Angus-Reid's last poll, since the amount of data is rather thin. This makes the "sample" of the projection relatively small, not much larger than the 800 people surveyed in that poll. For that reason, the high and low vote projections are rather wide. If more polls were available, that margin would be narrower. The high and low seat projection is based on these high and low vote projection ranges. Accordingly, that range narrows once more polls are available.

The second call made by the model is what I refer to as the forecast. Having a background in History, I am partial to the idea that the past can tell us a lot about the future. The forecasting model is based entirely on this premise.

In addition to the projection, the model also gives the plausible high and low results each party might be able to manage by election day, tentatively scheduled for May 14, 2013. Unless the polls begin fluctuating wildly, the margin will narrow as Election Day approaches.

The ranges are determined by measuring the degree of polling volatility in the past, with the period examined being equal to the amount of time before the next election. For example, the next election is scheduled for 174 days from the date of the projection, so the ranges are determined by the difference between the highest and lowest poll result for each party over the last 174 days. This is a measure of what kind of change in support is plausible based on how much that support has changed in the past. Of course, what is plausible does not equal what is possible - in theory, a party can get anywhere from 0% to 100% of the vote. That a party at 5% could win 75% of the vote six months from now is possible, and vice versa. It is not plausible, however, if that 5% has varied by only three points in the previous six months of polling.

This forecasting also cannot take into account completely exceptional events, but it is a best estimate of the plausibility of a party gaining or losing a certain amount of support.

The forecast will be continually updated as a measure of what should be expected, based on current information. Accordingly, it will change and the forecasts now may not overlap with the forecasts one month from the election. It is, instead, a best guess of what to expect based on what we know now.

In the same way that the seat projection model gives a range of likely outcomes based on the vote projection, the seat forecasting model gives a likely range of outcomes based on the vote forecast. These, of course, vary wide as the election date is far away. They give the range of plausible outcomes for the next election based on the information available to us right now.

Note that the seat forecast is not the same kind of assessment as the seat projection in terms of what seats are at play. Currently, the Conservatives are given a high forecast of 23 seats and the Greens of six seats. This does not mean that we should expect the Conservatives and Greens to win this many seats. In the case of the Conservatives, it means that if the party does end up at 25.8% support (their forecasted high, which would almost certainly mean the B.C. Liberals have dropped considerably), they could win as many as 23 seats. In the case of the Greens, it means that if the party ends up at 16% support (their forecasted high) they would be in play in as many as six seats. The one is dependent on the other. The Greens are not going to be at play in six seats at 7% support or even the high projected result of 8.7% support in the first projection. The model considers that they are at play in no seats at current polling levels.

One new feature added to the model is the probability that a call made by the seat projection model will be correct. This is based on an analysis of the seat projection model's performance in the eight elections that it has made projections for individual ridings. This probability is determined by the margin the projection model estimates the winner will win by. The following chart tracks how the projection has performed in the past, based on the projected winning margin in each riding.

Each red dot shows the percentage of calls that were correct, while the blue line shows the general trend. As the data is somewhat noisy (but the trend is still clearly visible), the trend line has been used to determine what probability of a correct call should be applied to every riding.

If the riding projection shows that a party leading in a riding by 12 points has a 73% chance of winning, that means that based on past performance the model will be right about 73% of the time when it chooses a winner by a margin of 12 points. It does not mean that there is a 73% chance that the projection for every party in the riding will be correct, or that the trailing party has a 27% chance of winning (a third place party could win as well). It is referring to the odds that the party projected to win will win.

Another new feature of the model is the ability to calculate the probability of a party winning the next election as of the date of the projection.

This is also based on the performance of the projection model in the past. In short, it determines the probability that the amount of error in the seat projection will be less than the margin between the leading party and other parties. A margin as large as the one of 40 seats between the Liberals and NDP in the current projection has been overcome in only 2.4% of cases. That means that if this was the final projection call before the election, the NDP would have a 97.6% chance of winning it. This takes into account the potential for an Alberta-sized error, but calling a winner by 40 seats would be right virtually all of the time.

However, this is not a forecast of the future. It is the confidence that can be placed on a call if the election were held the day of the projection. Forecasting the probability of a future event is an entirely different matter.

The greater challenge lies in being able to predict the probability that a party will win an election at a future date based on current information. The seat forecast shows what range of outcomes are plausible, but not which are most likely to occur.

After analyzing almost 6,000 pieces of data from polls conducted in over 20 federal and provincial elections since 2004, I have been able to determine the probability that the margin between any two parties can be overcome in a given period of time. As this analysis was based on the difference between polls and final outcomes, it takes into account both the past amount of error in the polls as well as the degree of real change that has occurred in voting intentions.

Using this model suggests how likely it is for the B.C. Liberals to overcome a 19-point deficit six months from the election, based on how often this sort of shift has taken place in the past. In this case, this sort of margin has been overcome in the six months prior to an election only 4.3% of the time, giving the NDP a 95.7% chance of winning the popular vote in May 2013 based on the polling of November 2012.

The calculations do take into account the role played by third parties and other parties, when they have a significant level of support. When the margin between the leading party and third or other parties is very large, the model assumes a (nearly) 0% chance that they could win. When three or more parties are a factor, the probability is calculated accordingly.

Calculating the probability of a party winning a future event should be very familiar to readers of Nate Silver's FiveThirtyEight blog. In his excellent book, The Signal and the Noise, Silver includes a chart showcasing the probability of a Senate candidate winning an election based on their polling on a given date. This gave me the opportunity to check my math, as it were. Here is a comparison of FiveThirtyEight's probability ratings to the ones that are employed by ThreeHundredEight:

As you can see, the probabilities are almost identical in most cases, and are often lower than FiveThirtyEight's. This may be a reflection of the complications caused by our three-or-more-party system. The numbers in the above chart assumes a two-party race, which of course is not always the case in Canada. But it is also possible to use these numbers to calculate the odds of a party winning in a multi-party race.

This calculation is based solely on the probability of a party winning the popular vote. In U.S. Senate races, that is also what determines who wins the election. That is not the case in Canada, where a party can win the most seats with fewer votes. The amount of variables at play to determine the winner of the most seats based on popular support in polls months before an election is, of course, enormous. But generally speaking, the party with the most votes will win the most seats, so while the probability of a party winning the popular vote may not necessarily determine their probability of winning the election, it is a close enough proxy. This is particularly the case in British Columbia, where neither the Liberals nor the NDP appear to have an intrinsic advantage in vote efficiency.

The projection will be updated continuously as the next vote approaches, with the forecast probabilities changing as new information becomes available. The B.C. model has been launched six months from the start date as it is the only upcoming election actually scheduled. Nova Scotia will be heading to the polls at some point soon, and that province's model will launch in the first half of next year if an election isn't already called. Ontario and Quebec could be heading back to the polls potentially even earlier than British Columbia, and this model will be tweaked accordingly and applied to those campaigns. But as B.C. is the only province we know is having an election in the next six months, the site will focus on that province's politics.

It should be an interesting campaign. The odds heavily favour the New Democrats, as a swing of 19 points even six months before an election is very rare. Much of the Liberals' hope lies with their ability to put the Conservatives back in their place. They aren't in much danger of being overcome by the Conservatives anymore, but at 12.8% the party makes it impossible for the Liberals to beat the New Democrats. Christy Clark will need to get those disaffected Liberals back into the fold if she is to have any hope of staying on as Premier.