Saturday, April 30, 2011

COMPAS: Tories by 20

COMPAS has the Tories at 46%, the Dippers at 26%, and the Grits at 17%. Needless to say, this poll is grossly out of line with everything we've seen. Because the previous COMPAS poll was also pretty crazy, I will simply assume that COMPAS' methodology is flawed and ignore this survey.

Note that this decision will counteract a significant fraction of the Tory ballot box bounce included in the model. Of course, that bounce is relative to "normal" polls, which this one is not.

Frank Graves is a Tease

Apparently, EKOS numbers, although fairly stable, show some very interesting things that might explain last night's revelation about Layton (I'm not going to dignify that story by linking to it).

Has the NDP surged into the lead in BC? Or is it now taking more from the Tories than the Liberals in Ontario? Neither appears particularly likely, but either would seriously damage Harper's chances at a majority.

What we do know is that EKOS numbers don't point to a Conservative majority. For the rest, gratification is denied until tomorrow, when, as appears likely, a massive orgy of polls will be upon us, climaxing in a final projection and ropes of accompanying posts in the wee hours of Monday morning. Frank, you tease!

(So yeah, I will post regularly tomorrow, but wait until after midnight to make a final projection. It is possible that some pollsters release data from tomorrow's polling at night - that has been done in the past.)

Pontiac Poll and Today's Projections

CROP has the Tories tied with the NDP in Pontiac. So the projection there remains an NDP win.

Here are the updated projections around the web. As usual, I encourage you to visit these websites, which are all linked from the sidebar. I will update this post as more updates are made.

Most Current
151 C, 93.6 N, 45.2 L, 17.6 B, 0.6 I (Calgary Grit - update)
148 C, 97 N, 47 L, 15 B, 1 I (Riding by Riding - update)
153 C, 92 N, 49 L, 14 B (Canadian Election Watch)

Including some polls from today

140 C, 96 N, 57 L, 15 B (LISPOP)
144 C, 59 N, 65 L, 40 B (ThreeHundredEight.com)
Average of 5 projections: 147 C, 88 N, 52 L, 20 B
(Without ThreeHundredEight.com: 148 C, 95 N, 50 L, 15 B)

Including polls from yesterday
149 C, 79 N, 55 L, 24 B, 1 I (democraticSPACE)

Including polls from two days ago
142 C, 88 N, 64 L, 14 B (Too Close to Call)

Including polls from three days ago
143 C, 73 N, 69 L, 22 B, 1 I (The Mace)
Average of 8 projections: 146 C, 85 N, 56 L, 20 B
(Without ThreeHundredEight.com: 147 C, 88 N, 55 L, 17 B, 1 I)

Ranges excluding highest and lowest:
CON: 142-151
NDP: 73-96
LIB: 47-65
BQ: 14-24

Léger: Tories by 5

We have another piece of evidence, by Léger Marketing, that the Conservative lead is around 5%. My raw polling average has it at 5.5%, but after adjusting for a likely Tory ballot box bump, it is 8.0%.

This poll has the Tories leading the Liberals by 11% in Ontario, which is consistent with the polling average. That's good news since Léger is usually pretty middle-of-the-road. The NDP is 8% behind the Conservatives there.

In Québec, this survey has the Bloc a little higher than others. Still, that's just 27%, 13% behind the NDP. In the Atlantic, it's a tight three-way race, though the Liberals trail the others slightly. Out West, there's nothing surprising.

We keep hearing about an NDP surge in BC. Léger still has them 10% behind the Conservatives, and the pre-adjustment polling average, 9%. While that's less than the 18% gap recorded in 2008, it's unlikely to cost the Tories more than 2-3 seats.

In the projection, the Bloc gets an NDP seat in Québec, the NDP gets a Tory seat in Ontario, and the Tories get a Grit seat in PEI, which gives:

CON - 153
NDP - 92
LIB - 49
BQ - 14

Update:
Interestingly, according to this poll, still just 8% of Canadians think the next government will be led by Layton, while 66% think Harper will remain PM. This doesn't mean Canadians prefer Harper - he trails Layton by 4%, 30 to 34, on the best PM question.

Angus Reid: Tories by 4; Nanos: Tories by 8.4

Today's Angus Reid and Nanos polls show stability in the national Tory-NDP gap (though Angus Reid has the Grits dropping to 19%), but have very different implications due to different regional splits.

Ontario: Angus Reid has the Tory-Liberal gap widening from 7% to 15%, while Nanos has it shrinking from 6.5% to 3.6%. The former gives Harper a good shot at a majority, while the latter puts him nowhere close.

Atlantic Canada: Both polls agree that the Liberals and Conservatives are roughly tied. However, Nanos puts the NDP about 12% behind them, while Angus Reid has the Dippers roughly 20% ahead...

Québec: Nanos suggests that the NDP rise has stopped, and shows a small Liberal rebound. Angus Reid has the Grits flat and the Dippers still going up, now at 45%.

British Columbia: Angus Reid has the NDP up to a statistical tie with the Tories, while Nanos shows no such bump.

Overall, the Liberals plunge in my Ontario polling average, giving 3 seats to the Tories. The NDP picks up 4 seats in Québec, 3 from the Bloc and one from André Arthur. The Dippers also gain a seat from each of the other two parties in NS.

CON - 153
NDP - 92
LIB - 50
BQ - 13

The average Conservative national lead is 8.1%.

Friday, April 29, 2011

Conservative Majority Out of Reach? NOT SO FAST!

The Toronto Star reported this morning that the Conservatives think it is "almost impossible" for them to win a majority. Specifically, they say they need to win 74 Ontario seats in order to get to 155. Are these sources telling the truth, that they're at just 81 outside Ontario?

Here is the number of ex-Ontario seats projected for the Conservatives by each of the 8 websites that use a polling average:

Calgary Grit: 98 (from Monday)*
ThreeHundredEight.com: 92
LISPOP: 91
Canadian Election Watch: 91
Too Close to Call: 87
The Mace: 87
democraticSPACE: 85
Riding by Riding: 85 (from earlier today)*

*The latest projection's breakdown is not available.

Is the Conservatives' internal polling having them a lot lower than public polls? Do they use a seat model that's significantly less generous to them than all the ones above?

Unlikely. They are probably playing the media, perhaps hoping that fewer left-wingers in the GTA vote if they don't think a Tory majority is possible. Update: Or maybe they're trying to lower expectations, so that they can have more legitimacy if they "unexpectedly" come very close to 155.

To be fair, just 81 seats outside Ontario isn't insanely low, but it's just not what you'd expect, on average, based on current Conservative numbers.

Updated Trend Graph

Here are the trends updated with today's points.

Two things to note: first, the Liberals have stabilized around the mid-50s. The key number for them is their Ontario vote share, and most recent polls put them at around 28%, except for Ipsos' 21%.

Second, accounting for methodological change (that's the discontinuity on the graph), the Tories are at their lowest point of the campaign. They are still very close to a majority, however, so unless things change drastically in the next two days, there will be a lot of suspense on election night. Don't believe the media saying there's virtually no chance of a Tory majority.

Today's Projection Averages

As usual, I encourage you to visit these websites, which are all linked from the sidebar.

Most Current
149 C, 79 N, 55 L, 24 B, 1 I (democraticSPACE - update)
143 C, 94 N, 52 L, 18 B, 1 I (Riding by Riding)
151 C, 86 N, 54 L, 16 B, 1 I (Canadian Election Watch)
Average: 148 C, 86 N, 54 L, 19 B, 1 I

Including some polls from today
149 C, 78 N, 57 L, 23 B, 1 I (Calgary Grit)
144 C, 53 N, 70 L, 41 B (ThreeHundredEight.com)
Average of 5 projections: 147 C, 78 N, 58 L, 24 B, 1 I

Including yesterday's polls
142 C, 88 N, 64 L, 14 B (Too Close to Call)
Average of 6 projections: 146 C, 80 N, 59 L, 23 B, 1 I

Including polls from two days ago
147 C, 69 N, 60 L, 32 B (LISPOP)
143 C, 73 N, 69 L, 22 B, 1 I (The Mace)
Average of 8 projections: 146 C, 77 N, 60 L, 24 B, 1 I
(Without ThreeHundredEight.com: 146 C, 81 N, 59 L, 21 B, 1 I)

Ranges excluding highest and lowest:
CON: 143-149
NDP: 69-88
LIB: 54-69
BQ: 16-32

Ipsos: Tories by 5, Grits at 18%; Eight Riding Polls

Ipsos Reid confirms (updated link) the 5% Tory lead over the NDP that many others have observed. The Liberals are at a historic low, but Ipsos has consistently shown the Grits lower and the Tories higher than other pollsters.

In Ontario, the NDP has vaulted into second place at 34%, just 6% behind the Conservatives. The Liberals languish at 21%. Although the NDP is close to the Conservatives in the popular vote, it would likely only win around 26 seats, leaving 69 for the Conservatives, and just 11 for the Liberals. The NDP needs to win the Ontario popular vote handsomely before making significant inroads.

The NDP leads in Atlantic Canada, 10% in front of the Tories and 19% ahead of the Grits, who would fall out of contention in the last region where they have a shot at the lead. The Tories polled at 55% in SK/MN, 23% ahead of the NDP, and at a whopping 74% in Alberta.

The Québec numbers are pretty run-of-the-mill: NDP at 42%, 16 in front of the Bloc, with the Grits and Tories well back around 15%. BC numbers have the Tories and Grits slightly stronger than the polling average, so the Dippers are just 3% ahead of the Liberals and 13% from the Conservatives.

For all these funny numbers, this poll only caused a one-seat change in the projection: the NDP takes an Ontarian seat from the Liberals. Based on this poll alone, however, I have the Tories winning a bare majority: 157 C, 108 N, 31 L, 12 B.

Next, ProjectDemocracy has released eight riding polls conducted by Oracle. Two of these are in the same ridings as CROP polled, and unfortunately, the results conflict. In Portneuf--Jacques-Cartier, while CROP showed a tight three-way race, this one has André Arthur leading the NDP candidate by 6%, both well ahead of the Bloc. In Charlesbourgh--Haute-Saint-Charles, where CROP showed an NDP-Tory race, Oracle says it's an NDP-Bloc race.

Two more of these polls are in Québec: Lévis--Bellechasse, where the Conservatives have a whopping lead (this conflicts in a major way with an earlier poll which had the NDP much closer in an early stage of its surge), and Pontiac, where the NDP leads by 6%.

The Tories lead in the four other ridings polled: Saskatoon--Humboldt, Desnethé--Missinippi--Churchill River, Esquimalt--Juan de Fuca and Nunavut. Harper's efforts in the North have really helped Leona Aglukkaq, who polled at over 70%. The only competitive riding is Esquimalt--Juan de Fuca, where the Tories lead the NDP by 5.8%, almost exactly as the model expects.

All these riding polls result in one change: André Arthur retakes his seat by a fraction of a point. This one could bounce back and forth over the weekend... The new projection is thus:

CON - 151
NDP - 86
LIB - 54
BQ - 16
IND - 1

The average Conservative national lead is 8.8%.

Why Andrew Coyne is wrong, and the NDP will win at least 30 seats in Québec

Plenty of commentators, like Andrew Coyne (whose most recent article is an absolute must), are still skeptical about the NDP's ability to convert its polling numbers into a seat landslide in Québec. (See here and here.) Here are two common explanations for their skepticism, and why I think they're wrong.

Update, 6:45pm: Andrew has changed his mind from last night, and now agrees that the NDP will win over 30 seats in Québec. However, some still believe that they will only get 5 to 7 (see here and here).

1. The Bloc vote will turn out more than the NDP's.
Contrary to what many have asserted, the sovereignists are notoriously bad at turning out their vote. In almost every provincial and federal election, there is a prime à l'urne (ballot box bounce) for the federalist party/parties. The 2008 provincial election, when the PQ did unexpectedly well, was an exception and probably due to Quebecers being unenthusiastic about the incumbent government (Charest's Liberals). Such an anti-incumbency effect would clearly not hurt the NDP. While the NDP may not get a bounce because of its inferior organization, it is unlikely to be much worse than the Bloc at turnout.

2. The Bloc vote is efficiently distributed.
This is patently false. The Bloc vote was efficiently distributed in 2008 because it is quite uniform across French-speaking regions of Québec. This allowed it to win a lot of ridings by moderate margins, instead of wasting tons of votes on ultra-safe strongholds. This efficiency is predicated on the Bloc winning the Francophone vote. If the Bloc were tied with the NDP province-wide, it would still win Francophones, and Duceppe would trounce Layton seat-wise.

However, with the NDP 13% ahead of the Bloc, the Dippers are clearly also in the lead among Francophones. Both parties have quite uniformly spread votes. This means that the Bloc will lose dozens of races by 10% or less. Because the Bloc is no longer first, the same feature that made its vote very efficient in the past now makes it inefficient.

Now, obviously, there may be pockets of the province where the Bloc retreats less than elsewhere, which would make its vote less uniform and help it survive in some areas. The projection actually takes this into account through regional adjustments based on riding polls: if I applied a straight uniform swing without regional adjustments, the Bloc would be down to 9 seats.

Bottom line: If the election were yesterday, I'm confident the NDP would have won, at a bare minimum, a majority of Québec seats. Taking into account a potential recoil over the weekend, maybe that lower bound is around 30 seats for Election Day. However, while I think the NDP has probably stopped going up in Québec, I don't think that a voter recoil is much more likely than a further Bloc collapse. The best bet is that the Dippers win a majority of Québec seats, and that the Bloc loses at least half its caucus. I wouldn't even be surprised if the Bloc falls below 12 and loses official party status, though the odds of that are currently under half.

Riding Poll: Bloc Holds on by 7 in Berthier--Maskinongé

Cible Recherche has conducted a riding poll in Berthier--Maskinongé. This staunchly pro-Bloc riding was projected to be a tough fight, with a 3% NDP edge. This survey suggests that the Bloc vote may be holding up better in this region than elsewhere.

The Lanaudière region and nearby areas are the most sovereignist places in Southern Québec. Most of the Bloc's close races are there, as it is literally fighting for its survival. Along with far Eastern Québec, this is where the Bloc will make a last stand. Thus, this riding survey is very significant, and suggests that the Bloc might have a better chance at staying an official party than provincial polls imply.

As a result of this riding poll, I will make an adjustment not just for Berthier--Maskinongé, but also for other ridings on the North Shore of Montréal. As a result, the Bloc also retakes Laurentides--Labelle and La Rivière-du-Nord, both of which were projected NDP by under 2%.

CON - 151
NDP - 86
LIB - 55
BQ - 16

Update:
Bernard von Schulmann has alerted me to a Yukon riding poll by DataPath. The Liberal incumbent is comfortably in the lead, as expected.

Update 2: Skoblin points out that Berthier--Maskinongé is the riding where the NDP candidate can barely speak French. I had missed that! Obviously, that means that this poll may not be representative for what's happening in the region. However, the adjustment I made for nearby ridings is a lot smaller than the one applied to
Berthier--Maskinongé. I will keep it since it also serves as a correction for the fact that the NDP was winning lots of close races, but losing very few when I had the Bloc at 13.

Nine Québec Riding Polls and a Halifax Area Poll

By now you've all come to expect it: Skoblin points out more riding polls.

CROP conducted a collection of six riding polls in the Quebec City area and three in the Saguenay--Lac-St-Jean region. All 6 Quebec City ridings are tight races involving the NDP, just as the model expected:
- NDP vs. Tories: Louis-St-Laurent (tie) and Charlesbourg--Haute-Saint-Charles (NDP by 4);
- NDP vs. Bloc: Québec (tie) and Louis-Hébert (NDP by 2);
- Three-way: Beauport--Limoilou (NDP by 6 over Bloc, 9 over Tories) and Portneuf--Jacques-Cartier (NDP by 2 over André Arthur and 3 over Bloc).

The projection formerly included a downward NDP adjustment because previous riding polls had indicated a weaker NDP rise in the area. All of these polls are consistent with a more refined story: the absence of such an adjustment in ridings with a Conservative incumbent, but a stronger adjustment in ridings with a Bloc incumbent. After this change, the NDP takes Charlesbourg--Haute-Saint-Charles and Portneuf--Jacques-Cartier, both of which it was losing by less than 3%. The projected NDP leads in Québec and Louis-Hébert both shrink, as does the projected Tory lead in Louis-St-Laurent.

In the Saguenay--Lac-St-Jean, CROP has the Tories with a healthy lead in Roberval--Lac-St-Jean and the Bloc ahead in Chicoutimi--Le Fjord. The race in Jonquière--Alma is tight, with the Tories holding a slight edge. All of these are consistent with the model, which takes into account last week's Segma polls in the same ridings, though the Bloc lead in Chicoutimi--Le Fjord is larger than expected.

Finally, MarketQuest-Omnifacts has a Metro Halifax poll. It shows the NDP surprisingly low in the area - in between the leading Grits and trailing Tories, while last time, it held a 10% lead over the Liberals and 20% lead over the Conservatives. However, all four seats in the region are so safe that none of them switches hands, though as indicated in the article, the race in Halifax (the riding) may be closer than expected.

As a result of the changes in and around Quebec City, the projection now gives:

CON - 151
NDP - 89
LIB - 55
BQ - 13

Update:
Reader Dave Roberts mentions an Environics Alberta poll that shows no orange wave there. (Reader JeffS points out the poll was taken before the NDP surge spilled outside Québec.) Really, Environics, or whoever ordered the poll? Alberta?

EKOS: Tories by 4.8; Nanos: Tories by 5.2

Two more polls suggest that the NDP rise continues at a much slower pace, but tell different stories about the Liberals.

First, today's EKOS survey has the Grits down to 20% nationally, a drop of 2.3%, while the NDP rises by 2.2% to 29.7%. In Ontario, the NDP rises by 3.4% to 26.2%, pulling slight more from the Grits than the Tories, the gap between whom grows to 12.3%. It's now a virtual tie for second place in Ontario. In Québec, the NDP ascension continues, and they build a 13.4% lead over the Bloc. The Liberals tank to 13%. Regionals for other areas will become available later.

Update: EKOS regionals outside Québec and Ontario are now up. The NDP now trails the Tories by just 5.8% in BC.

In the Nanos poll, however, all parties are essentially flat nationally. In Ontario, the Grits gain several points from the Tories (the gap shrinks to 6.5%) while the NDP is flat, which is very different from what EKOS shows. EKOS' sample is much larger, however. In Québec, the Bloc drops to 23.6%, 17.8% behind the NDP. The NDP jumps by 8.3% in BC, and are now within 8 points of the Tories.

In the projection, the NDP takes 5 more seats from the Bloc, which now risks losing official party status. The Dippers also gain one from the Tories in BC. (You may notice that the Tory count went up. That's because I had made mistakes in the previous update, now corrected. Sorry.)

CON - 152
NDP - 87
LIB - 55
BQ - 13
IND - 1

The average Tory national lead shrinks to 9.1%.

I will post another update shortly incorporating a bunch of new riding polls, so check back soon!

Thursday, April 28, 2011

Around the Web: Projection Roundup

A reminder that links to all of these websites can be found on the left. Only projections based on a polling average are listed below.

Including today's polls
142 C, 88 N, 64 L, 14 B (Too Close to Call)
146 C, 74 N, 63 L, 24 B, 1 I (democraticSPACE, updated)
146 C, 92 N, 51 L, 18 B, 1 I (Riding by Riding, updated)
153 C, 81 N, 55 L, 18 B, 1 I (Canadian Election Watch, corrected)
Average: 147 C, 84 N, 58 L, 18 B, 1 I

Including yesterday's polls
150 C, 69 N, 59 L, 29 B, 1 I (Calgary Grit)
143 C, 47 N, 74 L, 43 B, 1 I (ThreeHundredEight.com - also includes today's Nanos)
147 C, 69 N, 60 L, 32 B (LISPOP)
143 C, 73 N, 69 L, 22 B, 1 I (The Mace)
Average of all 8 projections: 146 C, 74 N, 62 L, 25 B, 1 I

Harris-Decima: Tories by 5; Trois-Rivières Riding Poll

Update: Two Tory seats were mistakenly allocated to other parties. See corrections below.

The latest Harris-Decima poll confirms what other pollsters are seeing on the national level. All five pollsters that have polled this week show a 3-7.5% Tory lead over the NDP in their most recent survey.

If you're a Conservative, however, this is a terrible poll: the Liberals have a 1% lead in all-important Ontario. Under this scenario, it would be impossible for Harper to aspire to a majority.

If you're a Bloc voter, this is also a terrible poll: it has the NDP 20% ahead of the Bloc. Under uniform swing, such an outcome would leave the Bloc with a grand total of 1 (yes, that's one) seat. Obviously, in reality, swings vary by riding, so the Bloc would probably keep a few more. However, it would probably lose official party status.

The changes in the projection this time around are:
- Liberals take one Conservative seat in Ontario.
- Tories make it up by reclaiming the NS seat it dropped to the NDP earlier today.
- The NDP gains a seat from each of the three other parties in Québec the Bloc and the Grits. The Bloc seat is Gilles Duceppe's own. The Tory seat is Beauport--Limoilou, admitted to be vulnerable to a Conservative operative. (Correction: Beauport--Limoilou had already been allocation to the NDP in the last update!)

CON - 151 153
NDP - 82 81
LIB - 56 55
BQ - 18
IND - 1

The average Conservative lead decreases to 10.1%.

Once again, our friend Skoblin points us to a riding poll, this time from Trois-Rivières. It shows a 14-point NDP lead over the Bloc, in a riding where the Dippers got 9% of the vote in 2008. This is one of the seats the Bloc lost to the NDP in the last projection update, so it's good to see confirmation.

EKOS: Tories by 7.3

Today's EKOS survey has the Tories marginally up, and the NDP and Liberals marginally down. The regional numbers didn't change by much: the Tories recover from a bad day in Atlantic Canada and make it a three-way race, which is consistent with what the poll average says. They are 2% up in Ontario, which pushes their lead to 11.5% - a good sign for Harper.

In Québec, the NDP is stable at a high 37.4%. As a result, the Bloc and Conservatives both lose ground, two and one seat(s) respectively in the projection, all to the NDP. The NDP is now projected to win a majority of Québec seats.

CON - 152
NDP - 80
LIB - 56
BQ - 19
IND - 1

The average Conservative lead is up a whisker, at 10.4%.

My Favourite Tweet of the Campaign

says:

@ endorses Canucks: . @ endorse the fucking Predators

Updated Trends and Poll Weights

Here is the updated trend graph that I've been promising. The discontinuity in the lines reflects methodological changes. Data in the days before the break has been revised: this is what they would have been without any GTA adjustment (which I have decided to keep at 4%) or change in the vote share formula. This makes it easier to see the parties' true progression.

The other change is that while I used to plot the projection at the end of each day, I have started yesterday to plot the projection at noon as well. So the graph will be a little more choppy from now on.

The story in the past day and a half has of course been the NDP's rise: it gained a whopping 25 seats! However, I don't think it's the main thing to watch going forward - it's fairly likely now that the NDP will finish second, and extremely unlikely that it'll finish first. Indeed, even if the NDP wins the popular vote, it would probably get fewer seats than the Tories, whose vote is very efficient.

Instead, I think that the thing to watch now is the Liberal number in Ontario. The Tories have been flat through the NDP progression because it has made up for losses to the NDP by gains from the Grits in Ontario. However, in the past few days, the Liberal support in Ontario seems to have hit a floor at 28%, which explains the dip in the blue line in the past day. If the Liberals hold on to that 28% of core supporters in Ontario and the NDP numbers from the past two days' polls hold, Harper will be denied a majority. But if the Liberal vote disintegrates in Ontario, the Tories have a good chance of achieving their goal.

Another thing to watch is the Bloc. I'd say they currently have a one-in-three (OK, 1/3 is probably a tad too high) one-in-five chance of losing official party status on Election Day.

Here are the weights on the national polls for the most recent projection. Hopefully this gives you a better sense of how the model works, which can help with understanding the trends:

Nanos 4/24-27: 12%
Nanos 4/20-23: 4%
Nanos 4/17-19: < 1%
EKOS 4/24-26: 19%
EKOS 4/18-20: 3%
Forum 4/26: 20%
Forum 4/20: 5%
Angus Reid 4/25-26: 16%
Innovative 4/21-25: 12%
Environics 4/18-21: 5%
Ipsos 4/18-20: 4%

In bold are the four polls conducted entirely this week. They account for 2/3 of the projection. Polls from last week carry 1/3 of the weight.

Currently, polls with a midpoint date within 2 days of the most recent one (Forum 4/26) are given full weight, which is the square root of the sample size. Polls up to 10 days old are given some weight. As I've indicated a few days ago, I am gradually lowering these thresholds; they were 3 and 12 for most of the campaign, and will be even lower than 2 and 10 for the final projection, which will be almost entirely based on this week's surveys. There is also additional depreciation for polls that are not a firm's most recent one.

Nanos: Tories by 6.2

Here's today's Nanos release. Nationally, the NDP keeps progressing at the expense of both traditional parties. In Québec, Nanos concurs with the latest Forum poll, and puts the NDP in the lead by 17%. Duceppe and the Bloc could really be wiped out! The Tories lose several points in Atlantic Canada and Ontario, which, as we'll see in the projection, hurts their chances of a majority. They do gain in BC, however.

This poll again causes many changes in the projection. The NDP takes 5 more seats from the Bloc, and one from the Tories in NS. Furthermore, the Liberals regain two in Ontario. Thus, the Conservatives fall out of majority territory, and Layton catches up to Dion's 2008 performance:

CON - 153
NDP - 77
LIB - 56
BQ - 21
IND - 1

The average Conservative national lead is 10.3%.

Take a look at the Cyberpresse article (in French), where a Conservative operative concedes that the Tories could lose Pontiac and Beauport--Limoilou. The projection has the NDP ahead in Pontiac, and literally 0.1% behind in Beauport--Limoilou. The Tories think they could make up for those losses in Abitibi--Baie-James--Nunavik--Eeyou, Richmond--Arthabaska and Mount Royal. The projection indeed has them ahead of the incumbent Bloc in A-BJ-N-E, but the current projected winner is the NDP. Richmond--Arthabaska had a riding poll putting the Tories a whopping 26% behind the Bloc last week - so that poll needs to be very wrong. Finally, the Conservatives lost Mount Royal by 28% last time, with less than half the Liberal vote alone, so this would seem like a tall order.

Wednesday, April 27, 2011

Too Tired

And too bummed out by Habs loss. Will post the updated trend graph some time tomorrow. Sorry.

For now, take a look at this Léger poll, which is not on voting intentions. Canadians and even NDPers overwhelming think the NDP is gaining on (and as we've seen, leaving behind) the Liberals due to Layton outperforming Ignatieff, and not due to its platform (and thus ideas). Also, a majority of Liberals support considering a merger with the NDP, but on the NDP side, there are more opponents than supporters.

Five Québec Riding Polls, Fun with Forum Numbers

Skoblin once again alerts me to riding polls, five of them this time!

Two of these, from Segma, were conducted in Eastern Québec (Gaspésie--Îles-de-la-Madeleine and Haute-Gaspésie--LM--M--M). They are very old (April 6-13, before the orange wave struck) and both give huge leads to the Bloc. Still, the NDP was already up a few points provincially, and the Bloc down a few. These polls instead had the Bloc gaining significantly, the NDP gaining less than elsewhere in Québec, and the Liberals down more than elsewhere. These surveys will help these ridings stay in the Bloc fold, and move HG-LM-M-M back from being narrowly projected Liberal.

Two more polls, by Connexion, were conducted April 12-16 in two Conservative ridings, Lévis--Bellechasse and Lotbinière--Chutes-de-la-Chaudière. Both had the Tories first, a few points in front of the Bloc, which was in turn a few points ahead of the NDP. Compared to contemporaneous provincial polls, they show the Tories dropping by more, the Bloc holding up instead of dropping, and the NDP rising slightly more than elsewhere. Nevertheless, because these seats were so safe, they remain Conservative for now in the projection.

In Hull--Aylmer, the NDP led by 13% over the Liberal incumbent according to a Segma poll conducted 4/21-23. This is virtually identical to what the model predicted!

As a result of the change in HG-LM-M-M, the projection is now:

CON - 156
NDP - 71
LIB - 54
BQ - 26
IND - 1

I've also played around with the Forum Research poll, and made a projection based on it alone. I turned off the popular vote adjustment, since I want to see what the projection would look like if the poll were entirely accurate. I did keep other features of the model, like the GTA adjustment and tweaks due to riding polls. The result: 143 C, 103 N, 54 L, 8 B. Forum's projection based on that poll is 137-108-60-3. The NDP vote outside Québec would be very inefficient, causing them to be well behind in the seat count despite being just 3% short of the Tories. So nothing we have suggests that the NDP may win the seat count... yet.

EKOS: Tories by 5.9; Forum: Tories by 3

One might have thought when the NDP surge started that it might compete for second place with the Liberals. Well, it's instead competing for first with the Tories!

First, today's EKOS shows little meaningful change from yesterday's. The Liberals did lose 2.6% in Ontario, mostly to the benefit of the NDP. This is far from statistically significant, but every point the Liberals lose in Ontario increase Stephen Harper's chance of winning a majority.

Next, Forum Research also has a new poll, all conducted yesterday. It shows the NDP with the largest lead yet in Québec: 40% to 23% for the Bloc. With such numbers, the Bloc risks losing official party status. The NDP also surges to first place in the Atlantic, and is close to the Liberals in Ontario. In this poll, it's the Conservatives and not the Liberals that are down in Ontario, which is bad news for Harper. Further bad news for the PM is that Forum shows the NDP tied with the Tories in MB/SK, and above 20% in AB where it would be competitive for a second seat. In BC, the NDP drops to the benefit of the Greens.

The NDP increases once again in the projection, this time by 4: it takes one seat from the Tories in SK, one from the Grits in ON, another from the Grits in QC, and one from the Bloc. The Tories also take a Liberal seat in ON, but the Grits take one back in PEI. All this gives:

CON - 156
NDP - 71
LIB - 55
BQ - 25
IND - 1

The average Conservative national lead is 11.4%.

Note that this projection does not yet incorporate some new riding polls that some very kind readers have brought to my attention. I will add them after the game, and will also update the trend graph.

Go Habs Go!

Delay in Posting

I might not be able to post until late tonight. Sorry. On the plus side, it's now (unofficially - still some paperwork to be completed) Dr. Election Watcher :)

Nanos: Tories by 10 over NDP, by 21.2 (over Grits) in Ontario

After a one-day hiatus, the daily Nanos poll is back, and how! It shows a 4.2% one-day national jump for the NDP, meaning that it got 12.6% more on April 26 then April 21. Not surprisingly, all other parties (save the Greens) are down.

Nanos has the NDP up 12% in Québec, but if you're a junkie like me, that's old news. The big story is that the NDP has eaten into the Liberal vote in Ontario, allowing the Conservatives to build a 21.2% lead even while losing 1%. This would give them around 70 seats in Ontario, and Harper would get a majority with room to spare.

Because the NDP is in the "paying zone" in Québec, this poll has a dramatic effect on the projection. The Dippers pick up 8 (typo) 7 seats in Québec alone, 5 (typo) 4 from the Bloc and 3 from the Liberals, both of whom are starting to lose Montréal fortresses like Hochelaga and Notre-Dame-de-Grâce--Lachine. Even Gilles Duceppe's own riding is in peril, and would already have gone NDP were it not for a Leader's effect. The NDP now has more seats than the Bloc in Québec, 29 to 26.

The Conservatives pick up a seat from the NDP in BC due to methodological change. They also gain two from the Liberals in Ontario, one of which is also due to methodological change. They would have lost a seat to the Grits in PEI if it weren't for the vote share adjustment modification. (Sorry, mistake when using old model.) So although the projection has the Tories picking up 3 seats, they are in fact flat only up 1.

CON - 157
NDP - 67
LIB - 57
BQ - 26
IND - 1

The average Conservative national lead over the NDP is 13%. That's a drop of 0.8%, since the previous update would have shown 13.8% using the new formula.

The Tories have now gained 7 seats in 5 days, putting them in majority territory. However, 6 5 of these seats were due to methodological change. Therefore, taking into account the apparently increased Conservative swing in the GTA and the increased greater motivation of Tory voters, the Conservatives have been virtually flat since the end of week 1, and were almost always in (or on the edge of) majority territory. I will have a trend graph tonight (or tomorrow, if I'm too drunk after watching Game 7) that makes it easier to separate out the effect of methodological changes in the projection.

I will be pretty busy today, so I might not be able to post again until around 5pm or so. But do check back then for an update!

Change in Vote Share Adjustment

It is now apparent that there is a large voter commitment gap between the Conservatives and the Liberals (not surprising given the way the Liberal campaign is going). See the latest EKOS and Angus Reid polls for evidence. Starting with the next projection, I will make the following methodological change:

Current Vote Share Adjustment: I reduce the vote shares of the Greens and Bloc, and distribute the deductions among the three main parties. Then I reduce the vote share of the NDP, and distribute the deduction among the Tories and Grits.

New Vote Share Adjustment: The first step with the Greens and Bloc remains the same. However, in the second step, I will give the NDP deduction entirely to the Conservatives.

This change effectively shifts half a point from the Liberals to the Conservatives. It doesn't affect the NDP's estimated share (which is still penalized about 1% relative to the raw polling average), but will (very modestly) help the NDP in races against the Liberals and hurt it in races against the Conservatives.

Tuesday, April 26, 2011

Canucks Win!

Angus Reid Regionals: NDP at 27% in Ontario

Angus Reid has obliged and promptly posted its regional breakdown. (Unlike Innovative Research Group...) It is very consistent with what we are seeing from EKOS: three-way race in Atlantic Canada (though Angus Reid doesn't put the NDP first), large NDP lead in Québec (9% here), Tories strong on the Prairies.

Angus Reid has the NDP still third, but at a high 27% in Ontario. That's actually an increase of just 3% over last week's survey: Angus Reid has usually had the NDP higher than everyone else in Ontario. But even if it were at 27%, it would not gain more than a handful of seats in Ontario.

One place where Angus Reid and EKOS disagree is BC, where Angus Reid shows a healthy Conservative lead while EKOS shows a tight race. Here Angus Reid is much closer to the polling average.

The NDP has hit the red zone in Québec. In the projection, it gains 7 seats there, six from the Bloc and one from the Grits. The Grits further lose St. John's South--Mount Pearl to the NDP. Finally, the Tories are down two seats in Ontario, one to the Grits and one to the NDP. (The Tories would have lost another seat, but I've further phased in the GTA effect, which caused no net change.)

The new projection has the Grits holding on to second place by the slimmest of margins:

CON - 154
LIB - 62
NDP - 61
BQ - 30
IND - 1

The average Tory national lead over the Liberals is flat, at 13.3%. However, the NDP is now in second place, 13.2% from the Conservatives. You can see that the icon following the "13.2%" on the top left has changed from the Liberal logo to the NDP logo.

So the first hour of the day has been interesting. We will get EKOS and Nanos updates later, so stay tuned!

Minor Correction: The further phasing in of the GTA effect in fact caused no net change. The model now has a GTA effect at 3%, and currently that translates to a net gain of 3 seats for the Tories.

Updated Trends and Riding Polls

Just five days left in the campaign! The updated trend graph of course shows a steep NDP climb, along with heavy Liberal and Bloc drops. The Conservatives' uptick has been modest, and half of it (3 seats) was caused by the phasing in of a change to the model.

I predicted the flatness of the Conservative seat count in the event of an NDP rise. However, if the Liberals collapse in Ontario, Harper would get a safe majority unless the NDP gets as high in Ontario as it is in Québec - the latter still appears very unlikely.

Two riding polls to report: one by Léger in Lac-Saint-Louis, where Conservative star candidate Larry Smith trails the Liberal incumbent by just 6%, down from 20% in an earlier CROP poll. This Léger poll appears to be dated (the linked writeup only says it was conducted after the debates).

The other poll, conducted 4/18-19, shows Elizabeth May 7% ahead of Gary Lunn in Saanich--Gulf Islands, with the NDP and Liberals both under 10%. Now, this is a poll commissioned by the Green Party and I'm not familiar with this pollster (Oracle). The Conservative support level, 38%, is right around what you'd expect, but one has to be skeptical that May has indeed pushed the NDP and Liberals under 10%. For all these reasons, I will not allow this poll to tip the projection unless there is independent confirmation.

Game 7!

Angus Reid: Tories by 5 over NDP, 13 over Grits

According to Robert Fife, Angus Reid will release a poll putting the NDP at 30%, within striking distance of the Conservatives. EKOS is not an outlier anymore.

EKOS: Tories by 6 over NDP, 9.9 over Grits, and more

The now daily EKOS update shows little change from yesterday on national numbers. In the Atlantic, the NDP is still in the lead, with an increased gap over the Grits and Tories. In MB/SK, the Liberals are down while both the Conservatives and NDP are up.

Numbers moved a little more in Québec, and for the first time in a while, the NDP is flat (actually half a point down). The Bloc is also down, to less than 25%, while the Tories are up a point. Ignatieff gets a boost - possibly because the last day of the sample came on the heels of his very successful visit to Tout le monde en parle - and gains 3 points.

We also have, through ThreeHundredEight.com, regional numbers outside Central Canada from the Environics poll. Given the small size of the poll, these don't mean much, but let's see what we have just for fun. In the Atlantic, the story is very different from EKOS: Environics has the Grits and Tories tied 18% ahead of the NDP. Across the Prairies, the Tories get over 60%, even reducing the Liberals to 8% in MB/SK. In BC, Environics calls for a replay of 2008.

ThreeHundredEight.com also reports on four Québec riding polls, three in the Saguenay region and one in the Coeur-du-Québec area. These surveys suggest very erratic swings: a 9-point increase in the Bloc lead over the Tories in Richmond--Arthabasca, but a 23-point increase in the Tory lead over the Bloc in Roberval--Lac-St-Jean, both to 26%. Good thing the gap is so big in these ridings, so the projection doesn't need to worry about them too much! The NDP has increased its support in both, but remains third.

The two other ridings are more interesting: the Bloc leads by 6% in Chicoutimi--Le Fjord over the Conservatives, just like in 2008. I had this riding going Conservative by 2%, so this is enough of a nudge to bring it back under the Bloc fold. However, the NDP is not too far behind, 12% from the Bloc, just 4-5% more than the model suggests. The projection model did reasonably well here: it hinted at a three-way race, and that's what we have.

In Jonquière--Alma, however, if the riding poll is right, the model completely flunks. The poll has this also as a three-way race, with the Tories leading, followed by the NDP and Bloc within 10. The model had it as a safe Conservative seat. At least it got the party right...

So there you have it: the Saguenay, a crazy region. Tories do much better than expected in Roberval--Lac-St-Jean, but worse in neighboring Jonquière--Alma and Chicoutimi--Le Fjord. The NDP surges in a major way in Jonquière--Alma, but almost not in Roberval--Lac-St-Jean. Only the Bloc performance has a semblance of consistency (well, the Liberal one as well - completely absent).

All this makes for the following changes in the projection:
- As stated, Chicoutimi--Le Fjord goes Tory to Bloc.
- But Bloc loses St-Bruno--St-Hubert to NDP.
- Liberals lose three seats to Tories: one in PEI and two in the GTA.

CON - 156
LIB - 63
NDP - 52
BQ - 36
IND - 1

The Conservative national lead is 13.3% over the Grits and 15% over the NDP.

One of the two changes in the GTA is due to the further phasing in of the GTA effect, which now affects 3 seats. That's currently the difference between a majority and a minority. For what it's worth, the effect is around 4% in the latest EKOS poll, and our friend Bryan over at Too Close to Call informs me that it's also around 4% in the latest Environics poll. That's less than the 10% earlier EKOS polls suggested. I currently apply a 2% effect in the model.

Around the Web: NDP Keeps Rising in Québec

This post will be updated throughout the day as projection sites refresh their numbers to account for yesterday's polls, showing the NDP potentially leading by double-digits in Québec. Once again, all links can be found on the left side of the page. As of 8am, four of the eight websites I'm tracking had already incorporated yesterday's polls:

144 C, 71 L, 55 N, 38 B (Too Close to Call)
144 C, 62 L, 64 N, 37 B, 1 I (Riding by Riding, now 155 C, 47 L, 70 N, 35 B, 1 I)
151 C, 68 L, 59 N, 29 B, 1 I (The Mace)
154 C, 66 L, 51 N, 36 B, 1 I (Canadian Election Watch, now 156 C, 63 L, 52 N, 36 B, 1 I)
146 C, 75 L, 42 N, 44 B, 1 I (ThreeHundredEight.com - update)
161 C, 57 L, 53 N, 36 B, 1 I (democraticSPACE - update)
155 C, 64 L, 50 N, 38 B, 1 I (CalgaryGrit - update)

Average: 151 C, 66 L, 53 N, 37 B, 1 I

CalgaryGrit and ThreeHundredEight.com will also update today. I'm hoping that LISPOP and democraticSPACE follow suit, so we get a full slate.

We also have two brave souls that ventured into predicting a final outcome for the election:

155 C, 69 N, 47 L, 36 B, 1 I (Riding by Riding)
152 C, 71 N, 52 L, 31 B, 2 I (BC Iconoclast - fixed, my bad)
153 C, 75 N, 54 L, 30 B, 1 I (BC Iconoclast - draft, adds up to 313 instead of 308)

Either result would mean Layton as the Leader of the Opposition, and very interesting by-elections... If I had to guess now, I'd predict a tighter fight for second place.

Ipsos: Tories by 13 in BC Lower Mainland

There is no Nanos poll today because of Easter Monday yesterday. However, Ipsos Reid has conducted a poll in the largest region of the country where all three main federal parties are competitive: BC's Lower Mainland. It has the NDP up 5% since 2008, while the Grits and Tories each lose 1%.

Based on this poll alone, we would see only one seat out of 20 change hands in the region: Surrey North, switching from the Conservatives to the NDP. This is exactly what my current projection shows, so this poll causes no change:

CON - 154
LIB - 66
NDP - 51
BQ - 36
IND - 1

The survey does suggest Vancouver South, currently Liberal, is more precarious than previously thought, while North Vancouver, currently Tory, is safer. However, throughout this campaign, Ipsos has produced much higher Conservative numbers than all other pollsters (except COMPAS). Taking that into account, this poll suggests that the Lower Mainland is moving almost exactly like the rest of BC.

Today should bring us new data from EKOS, who is planning daily releases from now on. We should also get the rest of the regional breakdowns from Innovative Research (MB/SK and AB) and Environics (all but QC and ON). Obviously, there might be other polls as well. So stay tuned!

Monday, April 25, 2011

Day's Recap and Big Picture

Four polls today made the NDP gain 8 seats in the projection, by far the largest one-day change for any party since the start of the campaign. The Conservatives gain 2 and are right on the fence between a minority and a majority, though it should be noted that their gain is due to a methodological change being phased in. The Liberals and Bloc suffer heavy losses: 5 seats each.

It's definitely time for the Liberal and Bloc campaigns to hit the panic button. The Liberals' goal is now to stay in second place. To do so, it must successfully defend the GTA and hope the NDP don't make any further inroads in Québec. A dozen of Toronto area seats could simultaneously determine who is Leader of the Opposition and whether the Tories win a majority.

As for the Bloc, the campaign is now a matter of survival. If it ties the NDP in the Québec popular vote, it keeps a respectable 35-40 seats - far more than any other party. But if it loses Québec by 9-10 points, as suggested by today's polls, its seat count may drop below 20, while the NDP could win a majority of Québec seats. Should the latter be the case, the Bloc would be at risk of withering away - who would want Duceppe's job?

The two key numbers to watch in the polls to come are:
- the Conservative lead in Ontario: 5% or below suggests a minority, while 10% or above suggests a majority;
- the Bloc vote share in Québec: around 25% implies disaster, while around 30% implies significant, but moderate losses.

It's clear where Harper and Ignatieff must focus: Metro Toronto. It's also clear where Duceppe should campaign: Southern Québec - those few Bloc-Conservative fights around Quebec City are not going to determine the Bloc's survival.

For Layton, things are more complicated. He obviously wants to win as many seats as possible, but he also needs to do his best to prevent Harper winning a majority. The reason is simple: if Harper gets his way, the Liberals will have 4 long years to rebuild, reestablish their credibility, and make sure that 2011 is a one-off. If Harper fails, the Liberals will have to choose between propping up a Conservative government - and alienating left-of-centre voters at every confidence vote for years - or give the NDP an important role in government.

Québec offers the most potential gains for the NDP, but it's in the West that Layton can impede Harper's quest to reach 155. One thing is clear, however: for once, Ontario shouldn't be on the NDP's priority list - the number of potential gains is very small, and a stronger NDP in Ontario helps Conservatives.

EKOS Regional Breakdown

Kudos to EKOS for its prompt posting of the regionals! (Go here for a discussion of the national numbers.) The big news is the NDP's 13.5% lead over the Bloc in Québec, 38.7 to 25.2, with the Liberals and Conservatives under 15%. Also, unlike Nanos, EKOS picked up NDP rises in the Atlantic (NDP first!), Ontario (NDP still a somewhat distant third) and BC (NDP a close second).

This poll also puts the Conservative lead over the Liberals at just 6.6% in Ontario. That appears bad for Harper's majority hopes, but the poll also has him 2% ahead in the GTA, a big change from 2008 that would allow him to gain several seats.

Adding this poll to the projection results in many changes:
- NDP gains one seat each from Grits, Tories and Bloc in Québec;
- NDP gains one seat from Tories in Ontario;
- NDP gains one seat from Tories in BC;
- NDP loses one seat to Tories in Saskatchewan.
Furthermore, I am starting to phase in the GTA effect. For now, this results in the Liberals losing two seats to the Tories.

The projection is thus:

CON - 154
LIB - 66
NDP - 51
BQ - 36
IND - 1

The NDP is now equidistant from the Bloc and Liberals seat-wise. It is still 1.9% behind the Grits in the polling average, mostly because I make an in-house adjustment reflecting the NDP usually performing worse at the ballot box than in polls. The Tory lead over the Liberals decreases slightly, to 13.3%.

As promised, I give you projection numbers based on this poll alone:

Canada: 138 C, 104 N, 60 L, 6 B. (Yes, that's 104 New Democrats and 6 Bloc.)
Québec: 57 N, 8 C, 6 B, 4 L. (Yes, that's 57 New Democrats.)

These numbers leave me speechless, so I'll let you do the commenting.

Update: EKOS' own seat projection is not too far from mine:

Canada: 131 C, 100 N, 62 L, 14 B, 1 I
Québec: 53 N, 14 B, 4 C, 3 L

EKOS: Tories by just 5.7 over NDP!

Prime Minister Layton? EKOS says maybe. Only the national numbers are out, so I will not update the projection yet, but yeah: C 33.7, N 28, L 23.7. EKOS suggests that its poll implies 131 Conservatives, close to 100 dippers and 69 Liberals - clearly enough for a coalition. When the regional breakdown comes out, I will say whether I agree...

The Bloc is at 6.2% nationally, which probably translates to around 26% in Québec - consistent with other polls. My model suggests that if the NDP gets around 30%, the Bloc would remain a significant political force, perhaps with 30-35 seats. But if the NDP gets around 40%, the Bloc might carry only a dozen seats and be well on its way to extinction.

Innovative Research: Tories by 13 over Grits, 14 over Dippers; Environics: Tories by 14 over Dippers, 17 over Grits

Andrew Coyne has been tweeting polls, one by Innovative Research (full results to be posted on Macleans), and one by Environics. Both these polls confirm that the Liberals have fallen and are now tied for second place with the NDP, and both show the NDP leading the Bloc by around 10 points in Québec.

However, Ontario, which will determine whether Harper wins a majority, is a point of contention. Environics puts the Conservatives ahead by 14% there, but Innovative pegs the lead at just 5%. Both agree with Nanos that the NDP remains uncompetitive there.

The partial regional breakdowns currently available (Québec and Ontario for Environics, and all but Prairies for Innovative) have been incorporated in the projection. (I'm assuming that the Innovative poll has the same sample size as last week's.) All changes come in Québec, where the NDP now has a 0.5% lead in the polling average. It gains two seats from the Bloc, and one from the Liberals, for a total of 11:

CON - 154
LIB - 69
NDP - 47
BQ - 37
IND - 1

The average Conservative national lead keeps increasing, and has reached 13.5%. The polling average still has the Liberals 2.9% ahead of the NDP.

Update: Just noting that the Bloc has fallen below the majority threshold in Québec. Also, EKOS is planning to release some numbers tonight, so stay tuned.

Nanos: Tories by 13.6, Liberals Collapse in Ontario

Today's Nanos poll points to a strong Conservative majority. It has the Tories above 44% everywhere outside Québec, where they drop to 14%. In particular, the Conservatives lead by a whopping 18.5% in Ontario, which would net them 72 seats in that province against just 20 for the Liberals. Interesting, the Liberals are collapsing on their own in Ontario: the NDP has made no progress at all there according to Nanos, and would also lose seats to the Tories.

The other big news is that Nanos now has the NDP ahead of the Bloc in Québec, by 2.8%. So not only is the orange wave there holding, it is actually strengthening. The Liberals actually manage 22% in La Belle Province, suggesting that the NDP is not drawing too much of their support.

In the projection, the Conservatives take two seats in Ontario, one Liberal and one New Democrat. The NDP more than makes this up by gaining two from the Bloc, which falls below 40:

CON - 154
LIB - 70
NDP - 44
BQ - 39
IND - 1

The Conservative average national lead hits 13.1%, right around where it was during the first week. This projection gives them an effective majority, as independent André Arthur almost always supports them.

Note that I have not yet started phasing in a GTA adjustment, which would further increase the Conservative seat count. If the election were today, I'd say the Tories would have a 60% chance at an outright majority.

Update: I know people might ask about this, so here it is: based on this poll alone, I get 169 C, 54 L, 47 N, 37 B, 1 I.

Sunday, April 24, 2011

Léger: Bloc by 13 in Chambly--Borduas

Skoblin alerts me to a Léger poll in Chambly--Borduas conducted Tuesday and Wednesday, showing the Bloc 13% ahead of the NDP, 37-24. This is a win for the model, and does not result in any adjustment: I had the Bloc lead at 14%.

The independent candidate in the riding, a humorist who often uses vulgar language, is at 15%, tied for third with the Liberal candidate and way ahead of the Conservative candidate at 7%. Francophone southern Québec shows no sign of giving Harper or Ignatieff another look.

GTA Adjustment To Be Phased In

Update: I mixed up "rest of Ontario" and "Ontario as a whole" originally. The post is now fixed, and the effect is even bigger!

I've looked back at EKOS polling, which breaks out large metropolitan areas, since the start of the campaign. There are now over 1,200 data points from the GTA, so we can say something meaningful about the swing there relative to the swing in the rest of Ontario by pooling those numbers.

The GTA is a region that the Liberals carried by 8-10% in 2008 (I don't know the exact number since riding boundaries don't match neatly with GTA boundaries). That's roughly 14-15% better than the 5.4% loss they suffered province-wide.

In the six EKOS polls to date, the GTA is only about 4-5% more favourable (actually, less unfavourable) to the Liberals than Ontario as a whole, a full 10 points less than in 2008. (Forum's latest poll actually finds the GTA 4% more favourable to the Conservatives. The COMPAS poll had the Tories leading by 17% in Ontario, by >15% in the 905, but the Liberals ahead in the 416, meaning the the GTA as a whole might be 10% more favourable to the Grits. These are the only two other GTA breakdowns I came across, and they pretty much cancel each other out.)

What does this mean? I now have the Conservative Ontario lead 3.3% higher than in 2008. If EKOS is right about the relation between the GTA and the province as a whole, we actually have, roughly, a 4.4% reduction in the Tory lead in the rest of Ontario, and a 13.3% reduction to the Liberal lead in the GTA (this makes it negative, so the Tories are now ahead).

If I fully introduce the above adjustment, the Liberals would regain four seats:
- Guelph
- Kitchener--Waterloo
- Kitchener Centre
- London West.
But the Tories would add a whopping 11 GTA seats:
- Mississauga South
- Eglinton--Lawrence
- York Centre
- Don Valley West
- Ajax--Pickering
- Bramalea--Gore--Malton
- Mississauga--Streetsville
- Richmond Hill
- Etobicoke Centre
- Etobicoke--Lakeshore (not including a leader's effect)
- Scarborough Southwest.
That's a net gain of 7 for the Conservatives, enough to deliver a majority.

Now, a sample of 1,214 since the start of the campaign pales in comparison to the sample of around 6,000 we have for Ontario as a whole just in polls released this past week. The margin of error for the Conservative-Liberal gap for a sample of 1,214 is around 5%, so there is significant uncertainty regarding the size of the GTA effect. One has to be skeptical that the GTA is indeed breaking 17.7% differently from the rest of the province - that's a huge deviation for such a large region. I am hoping for pollsters to conduct polls with bigger samples in the last week, in which case more of them might release GTA numbers.

For now, I will not yet incorporate a GTA adjustment in the headline projection. When tomorrow's EKOS update comes out, which will give us another couple of hundred observations, I will start phasing it in. Whether I end up with a full or a partial adjustment depends on whether other pollsters become more forthcoming with their GTA data.

The upshot is this: as I've said in a previous post, just before the NDP surge, the GTA carries the key to a Conservative majority. The NDP is not a factor in that region, except in a few Central Toronto ridings. The Conservatives' chance of getting a majority increases significantly if the orange wave reaches suburban Toronto. If I were Jack Layton, I would avoid Ontario, particularly Toronto, for the rest of the campaign, instead spending my time in Québec, Saskatchewan and BC.

Nanos: Tories by 12.7, NDP Flat

After a one-day hiatus, Nanos' tracking poll suggests that the NDP's rise may be done, or at least taking a break. Indeed, the New Democrats are flat everywhere in the country (and down in the Atlantic), meaning that their April 23 numbers weren't better than their April 19 ones. Still, the Bloc fell in Québec, so the NDP is now within 3.5% of the lead there.

This is a very good poll for the Conservatives, who take a 12% lead over the Liberals in Atlantic Canada and Ontario. These are exactly where Harper needs to make gains in order to win a majority. In both cases, because the NDP did not increase, these were likely direct gains at the expense of Liberals (it seems rather likely that it went LIB->NDP->CON). The Liberals did gain at the expense of the Tories in BC.

Many changes in the projection: the Grits lose three seats to the Tories, two in Ontario and one in New Brunswick. The NDP gains three seats, one from the Bloc, one from the Liberals in Québec, and one from the Conservatives in BC.

CON - 152
LIB - 71
NDP - 43
BQ - 41
IND - 1

Finally, the NDP surpasses the Bloc in the projection! The Liberals have relinquished almost all of their gains made in the first week of the campaign. The average Conservative national lead is now 12.7%, virtually the same as on Day 1.

Saturday, April 23, 2011

Proportional Representation

If the Tories win a minority, but the NDP gets more seats than the Liberals and more than the Bloc in Québec, will we get a majority in favour of electoral reform?

What If the NDP Replaces the Liberals?

Here's a question: what would happen if the NDP "replaces" the Liberals in the popular vote? I will look at two interpretations of this question:

1. The NDP gets to 26.3%, the Liberal vote share in 2008
The Dippers are at 21.2% right now in my polling average. I therefore increase their vote share by 5.1% throughout the country, drawing from the other parties as follows:
- Atlantic and Ontario: 3.5% from Liberals and 1.5% from Tories
- Québec: 3% from Bloc, 1.5% from Liberals and 0.5% from Tories
- West: 2.5% from Liberals and Tories
I'll assume the other 0.1% comes from the Greens.

The resulting popular vote would be roughly: 36.8 C, 26.3 N, 23.4 L, 7.2 B, 5.1 G.

North: 2 N, 1 L
BC: 20 C, 12 N, 4 L
AB: 27 C, 1 N
SK: 12 C, 1 N, 1 L
MB: 9 C, 4 N, 1 L
ON: 56 C, 31 L, 19 N
QC: 36 B, 14 N, 12 L, 12 C, 1 I
NB: 7 C, 2 L, 1 N
NS: 5 L, 4 C, 2 N
PE: 3 L, 1 C
NL: 4 L, 2 N, 1 C

Canada: 149 C, 64 L, 58 N, 36 B, 1 I

In this scenario, the Conservatives are virtually unaffected, while the NDP, unsurprisingly, draws from the Liberals and the Bloc. It would be a tossup as to who forms the Official Opposition, though the model gives the edge to the Grits. The Bloc still wins the Québec seat count by a large margin despite losing the popular vote by 2-3%. However, it would be the first time since its creation that it fails to win a majority of seats in La Belle Province.

If the Liberals hang on to the second most seats, this might not be a realignment election provided that their new leader is a bit more adept than Dion or Ignatieff. The Conservatives would almost certainly govern as a minority. Even if the NDP comes in second, it would likely not been legitimized to govern with around 60 seats.

2. The NDP leads the Liberals by 8.1%, the same as the Liberal-NDP gap in 2008
Here, I will consider an 8.5% NDP gain throughout the country, drawing from others as above, but scaled up by two thirds.

The resulting popular vote would be: 36 C, 29.7 N, 21.6 L, 6.7 B, 5 G. Note that the Conservative-NDP gap would be similar to the Conservative-Liberal gap in 2006.

North: 2 N, 1 L
BC: 18 C, 15 N, 3 L
AB: 27 C, 1 N
SK: 11 C, 2 N, 1 L
MB: 9 C, 4 N, 1 L
ON: 59 C, 27 L, 20 N
QC: 33 N, 23 B, 11 C, 7 L, 1 I
NB: 7 C, 2 L, 1 N
NS: 5 N, 3 L, 3 C
PE: 3 L, 1 C
NL: 4 L, 2 N, 1 C

Canada: 147 C, 85 N, 52 L, 23 B, 1 I

Once again, on the net, the Tories shrug. They get a much more comfortable minority than with a similar vote margin in 2006. Those 3.4 extra percentage points really pay off for the NDP, who get 27 more seats! Fully 19 of these are from Québec, where the NDP really lays into the Bloc by carrying the province by 8%. The Liberals are left with a grand total of 7 seats west of the GTA.

In this case, it's harder to tell what would happen. With 85 seats and nearly 30% of the vote, the NDP might feel comfortable trying to form government. This would present the Grits and Bloc with a very hard choice: prop up Harper and risk further alienating left-of-centre voters, or accept that Jack Layton becomes Prime Minister, which might relegate them to oblivion. Both the Liberals and Bloc would probably like to vote against Harper, and yet have a Conservative minority so that Layton does not become PM. This would get very very interesting!

Around the Web: How Much Has the NDP Surged?

Update, 4/24, 6:46am: Latest numbers from democraticSPACE included

Riding by Riding has clued me in on the existence of The Mace, another projection blog, which I have now added to my blog roll. I am now aware of 8 websites, including mine, regularly making projection based on polling averages. Links to the 7 others can be found on the left of the page. Here are the latest projections, from most to least recent:

Including Polls Published Friday and Earlier
157 C, 69 L, 39 N, 42 B, 1 I (democraticSPACE)
150 C, 75 L, 40 N, 42 B, 1 I (Canadian Election Watch)
150 C, 76 L, 36 N, 45 B, 1 I (ThreeHundredEight.com)

Including Polls Published Thursday and Earlier
145 C, 74 L, 47 N, 42 B (Too Close to Call)
147 C, 68 L, 49 N, 43 B, 1 I (Riding by Riding)
149 C, 68 L, 52 N, 39 B (LISPOP)
131 C, 73 L, 81 N, 22 B, 1 I (The Mace)

Average of 7 sites: 147 C, 72 L, 49 N, 39 B, 1 I
Average excluding The Mace: 150 C, 72 L, 44 N, 42 B, 1 I


I don't know how The Mace gets to 81 NDP seats, but for the rest, the number of NDP seats is directly correlated with how fast the site depreciates polls:
- LISPOP only included polls released on Thursday.
- Riding by Riding includes earlier polls, but its projection carries a forward looking component that extrapolates from past trends.
- Too Close to Call includes past polls; it has no forward looking component, but depreciates early polls quickly.
- I depreciate past polls more slowly than Too Close to Call.
- ThreeHundredEight.com uses the slowest depreciation.

Everybody agrees that the Liberals are slightly lower than their 2008 result. Apart from The Mace, there is also consensus that the Bloc will suffer moderate losses. Finally, everybody but The Mace and democraticSPACE thinks that the Conservatives are between their 2008 result and the majority threshold.

Calgary Grit has yet to include the very NDP-favourable polls published on Thursday: 150 C, 74 L, 35 N, 48 B, 1 I.

Elsewhere, Election Almanac offers projections for individual national polls. The average of the Friday projection based on Nanos and the three Thursday projections based on EKOS, Ipsos and Forum is: 150 C, 70 L, 52 N, 36 B. EKOS has for its own polls as well, and the latest is 134 C, 82 L, 60 N, 32 B.

Finally, BC Iconoclast's prediction of the final results is 152 C, 67 L, 48 N, 40 B, 1 I.

So how far has the NDP surged? If you believe the Thursday polls, it looks like the NDP is in the 50-60 range, from the mid-30s a week ago. My model is waiting for more confirmation (both of the magnitude of the move - Nanos still had the NDP trailing by 5.5% in Québec - and that the new NDP support holds) before allocating that extra dozen seats to the Dippers. I am particularly interested in seeing new Léger numbers, as it has the best track record in Québec.

P.S. A group called Fair Vote Canada said that based on the Ipsos alone, the Tories would get 201 seats, but the same projection has the Bloc at just 4, which is extremely implausible (it was at 27% in the poll, good for second in Québec). I do not view this as a credible projection.

No Nanos Today; Segma: Bloc by 6 in Brome--Missisquoi

Because yesterday was Good Friday, there is no Nanos update today as the pollster took a break.

I came across a Segma riding poll in Brome--Missisquoi. This poll is helpful, since I had the riding as a Liberal-Bloc tossup (Liberals marginally ahead by 0.3%), with the NDP 5-6% behind. Segma has the Bloc leading the Liberals and NDP 32-26-26, so I was not far off. Still, this poll does tip the riding to the Bloc in the projection:

CON - 150
LIB - 75
BQ - 42
NDP - 40
IND - 1

Segma notes that throughout the four days of polling, which was conducted Monday through Thursday, the NDP progressed at the expense of the Bloc, while the Liberals held steady. On the last day, the NDP was almost tied with the Bloc.

Looking at contemporaneous provincial polls, the NDP polled 20% better than in 2008, while its progress in Brome--Missisquoi was 17%. The Bloc lost 9% province-wide, but just 3% in this riding. The Liberals are down 7% both across Québec and in Brome-Missisquoi, while the Tory losses are 4% and 8% respectively. What we can see is that variations in Brome-Missisquoi are quite consistent with provincial variations - the differences can be mostly attributed to statistical noise. Still, the Bloc does seem to be retaining a bit more of its support in the riding. Two potential explanations come to mind:

- The Bloc often scores higher in polls than in elections. If this overvaluation of the Bloc in surveys is more pronounced in rural Québec, a Brome--Missisquoi poll will overstate its support more than a provincial poll.
- The Bloc candidate Christelle Bogosta was, in fact, the NDP candidate in 2008 and nominated by the NDP for this election. She defected shortly before the start of the campaign. Perhaps she is helping the Bloc bleed less to the NDP than it otherwise would.

(The Conservative campaign claims to have a poll showing the NDP ahead, with the Bloc in second, the Tories in third and the Liberals in fourth, 20% from the lead. This is, of course, completely unreliable, and could be an attempt to move moderate federalists from the Liberal to the Conservative camp.)

Trends 3/4 Through the Campaign

As I have done at the 1/4 and the 1/2 marks of the campaign, I now bring you the trends in my daily projections.

As you can see, the Conservatives have been essentially flat since the first week, just a few seats shy of a majority. The Liberals are down slightly: it's a loss of only two seats since the leaders' debates. However, the more precious loss by the Grits is momentum: they are now at risk of dropping precipitously if the NDP makes gains in Ontario.

The most salient features of the third quarter of the campaign are of course the Bloc's decline and the NDP's rise. The former was already apparent in the first half of the campaign, but accelerated in the past few days. The latter only came about this week. The NDP has yet to overtake the Bloc in third place, but unless their recent surge abruptly reverses, it will happen very soon.

The Conservative advantage over the Liberals has been essentially flat around 11% since April 5. The current uptick is almost exclusively due to the Ipsos poll showing a 22% difference.

Friday, April 22, 2011

Nanos: Tories by 11.7 Over Liberals, 14.1 Over NDP

Today's Nanos poll shows a continuing NDP increase, along with modest drops by the Liberals and Conservatives. The NDP rises everywhere except in BC, and are now tied with the Liberals for second in Atlantic Canada, 5% behind the Tories. Despite gains, the NDP is still shown 5.5% behind the Bloc in Québec and below their 2008 result in Ontario.

The Liberals are modestly down everywhere except on the Prairies. The good news for them is that the Tories are down even more both nationally and in Ontario. There is no sign (yet?) of a Liberal collapse in Ontario, which would deliver a Conservative majority.

The only change in the projection moves one Bloc seat to the NDP. Actually I still have the NDP ahead in only 3 ridings, but it is so close to winning several additional ones that in practice, it would in all likelihood get at least one of them.

CON - 150
LIB - 76
BQ - 41
NDP - 40
IND - 1

The average Conservative national lead is stable at 12.4% over the Liberals. I have the Grits at 4.9% over the NDP.

A note about poll depreciation. Given the avalanche of polls this week, I have restarted accelerating poll depreciation, as I had said I might do. Prior to today, polls up to 3 days old received full weight, while those up to 12 days old got some weight; I'll call this the 3/12 formula. Today's projection is based on a 3/11 formula. I don't yet know what cutoffs I'll end up setting for the final projection - this will depend on how many polls we get next week - though it'll be somewhere between 1/6 and 3/10. In any case, none of the polls published as of now will receive any weight in the final projection. I believe that this is appropriate as things are moving so fast that the situation on Election Day might not resemble what we have today, especially in Central Canada.

Thursday, April 21, 2011

Angus Reid: Tories by 10 Over NDP in BC

Angus Reid has released a BC poll, confirming the strength of the Conservatives and the NDP at the expense of the Liberals. At 42%, the Conservatives are almost as high as in 2008, while at 18%, the Grits are down even lower than under Dion. The NDP's 32% is strong, and if borne out on Election Day, would allow them to gain two seats.

In the aggregate projection, this poll moves the NDP closer to additional seats, but doesn't cause any immediate changes:

CON - 150
LIB - 76
BQ - 42
NDP - 39
IND - 1

Ipsos: Tories by 19 over NDP, 22 over Grits

As if the day didn't provide enough crazy scenarios, here's another, by Ipsos: the Liberals drop to 21% nationally while the Tories surge to 43%!

This poll is stellar across the board for the Conservatives: a 14% lead over the NDP (and 20% over the Liberals) in the Atlantic, a respectable 24% in Québec, 14% ahead of the Liberals in Ontario, over 2/3 support on the Prairies and 46% in BC.

It is dismal for the Liberals. Besides the numbers mentioned above, the Grits are at just 12% in BC. The silver lining is that in Québec, although they're in fourth place (20%), they're actually within striking distance of the leading NDP (28%) and Bloc (27%).

For the NDP, beside their slight lead in Québec, this poll also gives them strong results in Atlantic Canada (30%), Ontario (22%) and BC (32%).

Adding this poll to the aggregate projection moves North Vancouver from the Liberals to the Tories. In Québec, the Bloc loses one seat each to the Liberals and Tories, and gets closer to losing several to the NDP (though none yet).

CON - 150
LIB - 76
BQ - 42
NDP - 39
IND - 1

The average Conservative national lead increases to 12.4%, the highest since the first week of the campaign.

What would Parliament look like with the Liberals this low and the Conservatives this high? A one-poll projection gives me:

CON 175, LIB 49, NDP 49, BQ 34, IND 1

It's a tie for second place! The reason is that by winning Québec by just 1%, the NDP puts itself at the edge of the "Red Zone" without actually cashing in - it gets 9 seats in Québec, against still 34 for the Bloc. The only Liberal seats west of the GTA would be Vancouver Centre and Yukon!

Riding Polls: Good News for Grits in Newfoundland

ThreeHundredEight.com reports on the results of five riding polls: Sherbrooke, Gatineau, Avalon, St. John's East and St. John's South--Mount Pearl.

The two Québec polls yield little surprise. Sherbrooke is a safe seat for the Bloc, while the NDP now has a slight lead in Gatineau. Interestingly, the NDP rise in Gatineau is a bit less than its rise in contemporaneous province-wide polls, while the Bloc and Liberals did not fall in Gatineau, unlike elsewhere in Québec. This is consistent with an observation I made a few days ago, namely that because the Outaouais shares some characteristics with Ontario and because the NDP has not gained ground in Ontario, the NDP surge in the Outaouais is likely smaller than in the rest of Québec.

In Newfoundland, the three riding polls are extremely helpful: we knew from an Angus Reid poll in February that the effects of the 2008 ABC campaign had diminished, but because the Angus Reid poll had a small Newfoundland sample (300-400 respondents), we couldn't tell precisely by how much. These polls help us answer that question, and it now appears that the Tories' variation in Newfoundland are about 10% more favourable than in Atlantic Canada as a whole, while both the Grits' and Dippers' are 5% less favourable. The Conservative and Liberal numbers are smaller than suggested by the Angus Reid poll, which is what I had to go by until now.

As a result, extrapolating based on the provincial average now gives the Liberals a small lead in St. John's South--Mount Pearl instead of the NDP. This is consistent with the riding poll, so the seat reverts to the Grits in the projection.

For Avalon, extrapolating based on the provincial average gives the Tories an 8% lead, but the riding poll gives the Grits a 4% lead. Accordingly, this race is now still projected Conservative, but by a tiny margin.

The new aggregate projection is thus:

CON - 148
LIB - 76
BQ - 44
NDP - 39
IND - 1

Forum: Tories by 11 over NDP, 13 over Grits; NDP by 9 in Québec

And yet another poll confirms the NDP's meteoric rise, this one by Forum Research. This survey has the NDP vaulting past the Liberals nationally, 25% to 23%. Just like the Liberals, the Conservatives also lose 2 points, to 36%. Ominously for Ignatieff, his party falls to 28% in Ontario, 14% behind the Tories, while the NDP rises to 20%. And this poll shows the largest NDP lead yet in Québec: 34% to 25%.

In the Ontario projection, the Grits lose a seat to the Tories, who relinquish one to the NDP. But the Liberals make up for it by taking one from the Bloc:

CON - 148
LIB - 75
BQ - 44
NDP - 40
IND - 1

The average Conservative national lead is 11.6%.

The NDP does not gain any additional projected seats in Québec, but is now within 2% of tripling their seat count to 9. Currently, I have Québec at B 31%, N 25%, C 20%, L 19%.

Forum Research states that based on their poll alone, the projected seat count would be 149 C, 71 N, 64 L, 24 B. That PM Layton scenario I described earlier today just got more likely!

EKOS Update: Tories by 9.7, Liberals and NDP Tied, NDP by 4.2 in Québec

The latest EKOS poll, which is updated with one more day of data relative to what was available this morning, shows the Tories below 35%, with the Grits and Dippers exactly tied.

The good news for the Liberals is that they're just 4.1% behind the Conservatives in Ontario; the bad news is that they're doing poorly everywhere else, especially in Québec with 15.5%. For the Tories, this poll is pretty average in the East, but weak in the West: just mid 30s in MB/SK and BC, and "only" slightly over 50% in Alberta.

For the NDP, this survey is almost all good. The only area where they dropped relative to the EKOS poll from three days ago is BC, but the Tories lost an equal amount. They are statistically tied with the Conservatives in MB/SK, though that's probably due to the small sample. They are up in Ontario and Atlantic Canada, though only back to around their 2008 support levels. And, of course, the biggie is that they're leading in Québec.

Adding this poll to the projection mix makes the NDP gain a Conservative seat in Saskatchewan, while the Bloc takes one back from the Liberals:

CON - 148
LIB - 75
BQ - 45
NDP - 39
IND - 1

The Conservative national lead is back to 11.5%.

Now, since this is the first national poll with both a Liberal-NDP tie and an NDP lead in Québec, I know that many of you must be curious what parliament would look like if these results are borne out on Election Day. (Or even if you aren't, I was!) Here's what I get:

CON 135, LIB 76, NDP 64, BQ 32, IND 1

This scenario could well lead to a coalition, as LIB + NDP = 140 > 135 = CON. The Liberals remain slightly ahead of the NDP in the seat count due to their strength in Ontario, but there is real competition for the job of Leader of the Opposition/Coalition. And because EKOS is showing the NDP surge to be mainly outside of Ontario, it actually hurts the Tories more than the Grits. This would totally change if Ontarians move toward the NDP.

In Québec, I get: BQ 32, NDP 19, LIB 12, CON 11, IND 1. The NDP loses the seat count to the Bloc despite leading the popular vote by 4% because that's not quite enough to lead among Francophones, as I explained this morning. However, as I also mentioned, an extra 5% in the NDP-Bloc gap could swing an extra dozen seats to the Dippers. In that case, the Bloc would be on its way to irrelevance, while Layton could become Leader of the Opposition, or even Prime Minister - Prime Minister Layton! - in the event of a coalition. This is of course not likely by any means, but it's a possibility that cannot be dismissed anymore.

This is an exciting time in Canadian politics!