Tuesday, September 28, 2010

Flaws of proportional representation redux

"This electoral system is so distorted!" many NDPers, Greens and other proponents of proportional representation likely screamed as the results solidified the evening of September 27. Fair Vote Canada was one of these groups, and they demanded New Brunswick forthwith implement the New Brunswick Mixed-Member Proportional (NB MMP) system advocated by the Commission on Legislative Democracy.

But if we had NB MMP in place, we might well have heard others screaming the same thing.

In 1947, Winston Churchill said "democracy is the worst form of government, except for all those other forms that have been tried from time to time." You might easily substitute in our specific type of democracy, first-past-the-post.

I re-ran the 2006 election results through the NB MMP model a while back, and explained why I didn't think PR proponents would care much for the result.

Running the 2010 results through NB MMP and comparing it to 2006, raises even more problems.

Here is what the results would have looked like if you apply the 2010 election results through the model using the same assumptions I used in my 2006 analysis.

2010 actual resultsProjected NB MMP seats
Seats% voteRidingRegionalTotal

Here you do actually largely achieve what PR proponents want - a seat share for each party relatively close to their vote share. But it also shows for all the distortions this particular brand of PR eliminates, it creates many of its own.

This is the second election we've run through NB MMP where a party that gets less than 50% of the vote still wins a majority government, something that PR proponents view as the most cardinal sin of first-past-the-post.

Also troubling for PR types is that under NB MMP, the NDP didn't win any seats with 5.1% of the vote in 2006, but the Greens won 3 seats with 4.5% in 2010. Talk about a distortion! And another distortion is that while the NDP polled 130% more votes than the Greens, they would only have gotten 67% more seats.

No electoral system is without flaw. And NB MMP certainly has its fair share.

I think that Chris Baker may have had the most succinct analysis of PR here.


My earlier analysis ignored the fact that under the NB MMP model, parties that get less than 5% of the province-wide vote are not eligible for seats. Therefore, the Green Party would not have won any seats under NB MMP if they got the same share of the vote as they did last night.

Here is a revised projection of the results under NB MMP:

2010 actual resultsProjected NB MMP seats
Seats% voteRidingRegionalTotal

Under this model, the Greens lose their Central seat to the Liberals and their other 2 seats to the NDP. This to me presents even greater problems. The Greens, while breaking 5% in two regions of the province and nearly hitting 5% province-wide are shut out.

A question has been asked, even though NB MMP is imperfect, is it better than the status quo? I would argue it is not. To move to NB MMP you give up a large degree of local representation and create two classes of MLAs. In exchange, you are supposed to get more parties represented and prevent parties who don't receive a majority of the vote from winning a majority government. This delivers partially on one point and not at all on the other.

Monday, September 27, 2010


Well that was quite something. I think that all signs pointed to a Conservative win, and while I thought they would win a modest majority, I did see the potential for 38 seats.

My sincere congratulations to Mr. Alward and the Conservative campaign who exceeded that with 42 seats (pending at least one recount).

The Tories won all but 3 of the 15 ridings to watch, but also won 7 ridings that I had thought to be Liberal victories. Here is a map with the Tory wins that I didn't forsee highlighted. Of these, considering the big sweep, none are a major surprise. I'll single a few out for comment.

  • Grand Falls-Drummond-Saint-André (+6)

  • Campbellton-Restigouche Centre (+20) - I really missed the ball here as this race wasn't even close to close

  • Miramichi Bay-Neguac (+5)

  • Quispamsis (+16) - this is a bell-weather and I started to get buyers remorse this afternoon when I realized that, but a 16 point margin is quite something, especially when an opinion poll had given Schryer a 9 point victory just a few weeks ago

  • Saint John East (+5) - this result was almost the exact same as 1999 - PC winning with less than 40% due to high NDP vote, I didn't expect the NDP to be able to score anywhere near the 24% they got here and the 28% they got in Saint John Harbour

  • Saint John Harbour (9 votes) - this is the first ever PC victory in this riding and with a final result of 30.7-30.5-27.6, this was quite a nail-biter all night
This map is also eerily similar to 1999, with the popular vote victory by Alward also similar (+14.5 vs +15.7 in 1999). Here are the maps side-by-side.

Saturday, September 25, 2010

Final prediction

Here is a list of all of the ridings, organized by my level of confidence. I offer a profile for the 15 ridings to watch. I had promised to limit myself to 10 ridings to watch, but I couldn't get myself below 12. I rounded it up to 15 for fun.

I am absolutely, 100% confident of the outcome

Albert (PC)
Caraquet (Lib)
Carleton (PC)
Hampton-Kings (PC)
Kings East (PC)
New Maryland-Sunbury West (PC)
Nigadoo-Chaleur (Lib)
Oromocto (PC)
Shediac-Cap-Pelé (Lib)
Woodstock (PC)

TOTAL: PC 7, Lib 3.

I am 99% sure of the outcome

Centre-Péninsule-Saint-Saveur (Lib)
Fundy-River Valley (PC)
Grand Falls-Drummond-St. André (Lib)
Kent (Lib)
Lamèque-Shippagan-Miscou (PC)
Memramcook-Lakeville-Dieppe (Lib)
Saint John Harbour (Lib)

TOTAL: Lib 5, PC 2;

I am confident

Dieppe Centre-Lewisville (Lib)
Edmundston-St. Basile (PC)
Fredericton-Nashwaaksis (PC)
Moncton North (Lib)
Restigouche-la-Vallée (PC)
Riverview (PC)
Rogersville-Kouchibouguac (Lib)
Rothesay (PC)
Saint John East (Lib)
Saint John Portland (PC)
Saint John-Fundy (PC)
Southwest Miramichi (PC)
York (PC)

TOTAL: PC 9, Lib 4;
RUNNING TOTAL: PC 18, Lib 12... these numbers I would say are the minimums for both parties and all of the numbers above I think you can take to the bank... an electorate can be unpredictable so I am less certain below.

In my mind I feel confident, but I am somehow uneasy

Campbellton-Restigouche Centre (Lib)
Fredericton-Fort Nashwaak (PC)
Grand Lake-Gagetown (PC)
Madawaska-les-Lacs (PC)
Miramichi Bay-Neguac (Lib)
Miramichi-Bay du Vin (Lib)
Moncton East (Lib)
Petitcodiac (PC)
Quispamsis (Lib)
York North (PC)

TOTAL: PC 5, Lib 5;

Toss ups - "nbpolitico's fifteen ridings to watch"

These 15 ridings will be the ones to decide the election and are the ones to watch on election night.

This riding has never gone Conservative since its creation in 1967. Prior to that it was part of the multi-member Gloucester riding which only went Conservative once. Notwithstanding that long Liberal history, it has been one of the closest seats in the province for the past two elections. This election will be the third. In 2003, 2006 and again in 2010 it will be a race between Liberal Brian Kenny and Conservative Nancy McKay. Will Kenny's position as a minister in the government help him? Will McKay's two consecutive losses hurt her? What will be the effect of the NDP which will probably poll between 10 and 20 per cent? These are not easy questions to answer, so that makes this a riding to watch. Leans Liberal

An opinion poll in the riding of Grand Lake-Gagetown showed People's Alliance leader Kris Austin capturing 23% of the vote. It is not hard to imagine PANB candidate John Craig achieving a similar result. The question is, will that vote come from the Conservatives or the Liberals or from both? Or is his support high enough for him to win himself? These are not easy questions to answer, so that makes this a riding to watch. Leans PC

Charlotte-the Isles
This is traditionally a Liberal stronghold. The old riding of Charlotte Centre (which makes up most of this riding) was Sheldon Lee's fiefdom from 1978 until it merged with mainland parts of Charlotte-Fundy to make the new riding of Charlotte in 1995. Lee held Charlotte until he retired in 2003. The riding stayed in Liberal hands in 2003, and again under its current name in 2006 when it absorbed the island of Grand Manan into its boundaries. The Liberals won the seat fairly comfortably in 2006. The Tories, however, enjoy a 19 point lead in southern and central New Brunswick in the most recent poll. Other polls straight back to the beginning of the campaign show them with a double digit lead in those regions. And that lead is likely amplified in rural ridings, especially those with a large number of NB Power employees (Point Lepreau is in this riding). Will the long tradition of voting Liberal save this riding from a Tory surge? Will the NB Power issue hurt the Liberals here as it appears to be in neighbouring Fundy-River Valley? These are not easy questions to answer, so that makes this a riding to watch. Leans Liberal

Dalhousie-Restigouche East
Generally a reliable Liberal seat, it went Conservative in 1999 and part of what is now this riding also went Conservative in 1974. In 2006, Liberal Donald Arseneault won by an astounding 3,500 votes. Since the last election, two of Dalhousie's three major employers have closed and the last will close within a few months. Since this spring, Donald Arseneault has been deputy premier and in that role has served as the Liberal's principal attack dog against the Conservatives. Will Donald Arseneault's senior role in the cabinet help him? Will his role as an attack dog hurt him? Will voters blame the economic collapse of the riding on Arseneault? And are any negatives enough to wipe out a 3,500 vote margin? These are not easy questions to answer, so that makes this a riding to watch. Leans Liberal

This was a new riding in 2006, taking large parts of the relatively new riding of Fredericton-Fort Nashwaak (created in 1995) and New Maryland. It is hard to judge its history. Greg Byrne returned to politics after a 7 year absence by running here in 2006 and winning convincingly. In that election, the NDP leader Allison Brewer also ran here taking over 1,000 votes. Over the past year Byrne has served as both finance minister and government house leader making him the face of the government on a large swath of issues. Will the NDP hold its vote or grow it? If the NDP vote slips will it go to Byrne or Conservative Craig Leonard? Will Byrne's role as face of the government on many difficult files hurt him? These are not easy questions to answer, so that makes this a riding to watch. Leans Liberal

Of the Fredericton ridings, this is probably the safest for the Liberals. It has the weakest Conservative candidate and it is the most urban. The riding has changed a lot over the years and the continued growth of Fredericton has caused it to split several times. It lost almost half of itself to Fredericton-Fort Nashwaak (now Fredericton-Lincoln) in 1995 and another smaller chunk in 2006 to Fredericton-Lincoln. It was the only Fredericton proper seat to vote Conservative in 2003, but thanks to a high NDP vote, the Conservatives won with only 41% of the vote that year. Will the NDP return to its previous tradition of getting in the high teens or low twenties or even higher? If so, does that NDP vote come at the expense of the Liberals or the Conservatives? Will a relatively weak Conservative candidate thwart the Tory tide that is surrounding the riding? These are not easy questions to answer, so that makes this a riding to watch. Leans Liberal

Kent South
This riding has been among the best for the Tories in terms of francophone seats not in Madawaska county. It went blue in 1974 (its first election) and 1982. It was held by outgoing premier Camille Thériault (just barely) in 1999, and then returned to Tory hands in a 2001 by-election where it has remained ever since. But with polls showing the Liberals doing very well in a) southeastern New Brunswick, and b) among francophones, this could be the best chance the Liberals have had in quite sometime to snatch it back. Is it possible for Claude Williams to be beaten even with the above factors on the Liberals' side? Is the Liberal lead in the southeast really as big as the polls suggest? These are not easy questions to answer, so that makes this a riding to watch. Leans PC

Miramichi Centre
This riding has gone Conservative twice before - in 1999 when the Tories swept the Miramichi region and in 1982. In 1991, the riding was a three-way split between the Liberals (who won), the Conservatives and the populist-right wing Confederation of Regions party. The Telegraph-Journal polled this riding and showed a result within the margin of error of PC 46 to Liberal 43. Will the downturn of the Miramichi economy hurt Liberal John Foran? Will his history working for the popular MP Tilly Gordon help Conservative Robert Trevors? These are not easy questions to answer, so that makes this a riding to watch. Leans PC

Moncton Crescent
This riding was created out of the Tory stronghold of Petitcodiac in 1995 and has been has been strong every cycle for Conservative John Betts since he ran in 1999. The strong showing for the Liberals in Greater Moncton shows this seat going handily for the Liberals based on the uniform swing however, and the Liberals are running a stronger candidate than last time around. Does the big Tory drop in Greater Moncton impact the anglophone ridings to Moncton's West or just he francophone ridings to its east? Will the Liberal investments in Atlantic Baptist University translate into votes for the Grits? These are not easy questions to answer, so that makes this a riding to watch. Leans PC

Moncton West
From its creation in 1974 to the Liberal sweep of 1987, this riding was in Conservative hands and in 1978 and 1982 the race wasn't even close. It was recaptured by the Tories in their sweep of 1999. They held the seat in 2003 and 2006, but by margins of less than 500 votes. Since the last election, Conservative Joan MacAlpine-Stiles crossed the floor to the Liberals, but she isn't re-offering. The two principal candidates are both big local names: Liberal Anne-Marie Picone Ford is co-owner of Ford's Apothecary and anyone that has ever listened to the radio in Moncton has heard her voice its ads. Conservative Sue Stultz is better known as "the turkey lady" having run for many years a major turkey drive for needy families. Will MacAlpine-Stiles' party switching hurt or help the Liberals? Will two well known women nullify the effects of the provincial campaign in this riding? And if so, is that good for the Liberals - down in the province-wide polls, or good for the Tories - behind the Liberals in the Moncton region? These are not easy questions to answer, so that makes this a riding to watch. Leans Liberal

This has traditionally been a Liberal riding, having only gone for the Conservatives in the sweep of 1999. However, the Liberals won it by only 400 votes in 2006. The NDP are strong here, having received one of their best ever results in 1999 when Gilles Halley got 28% of the vote. The battle in this riding is likely principally between the NDP and the Liberals but the Conservatives could easily sneak up the middle in such a race. Will the NDP's Pierre Cyr be able to meet or exceed Halley's 28%? Will a strong NDP vote come at the expense of the Liberals or the Conservatives? Does Liberal Cheryl Lavoie's last minute addition to cabinet help her or hurt her? These are not easy questions to answer, so that makes this a riding to watch. Leans Liberal

Tantramar has behaved almost like an American congressional district in its history. It re-elects incumbents but when an incumbent retires it is a close race. The NDP won the riding in 1982 by just 450 votes upon the retirement of Conservative Lloyd Falkins. Marilyn Trenholme won big majorities in 1987, 1991 and 1995 but after she retired Conservative Peter Mesheau won by only 39 votes in a 1997 by-election. Mesheau was re-elected by huge margins in 1999 and 2003. Conservative Mike Olscamp broke the trend somewhat when he won his first term by over 1,000 votes. However, I've heard from a number of sources that both the Liberal and NDP campaigns are much stronger this time around. Will history be on Mike Olscamp's side and the incumbency give him a huge boost? Will the riding return to its NDP tradition when it got over 30% of the vote in 1978, 1982, 1987 and 1997, and finished second in 1999? Does the large Liberal lead in southeastern New Brunswick extend into this riding? These are not easy questions to answer, so that makes this a riding to watch. Leans PC

The NDP have put on a big push in this riding held by the Conservatives since 1994. Conservative Claude Landry seeks a second term, while the Liberals hope to regain a riding they held from its creation in 1974 until Landry's former boss won it in a 1994 by-election. NDP leader Roger Duguay's decision to run here raised a lot of eyebrows, he had previously run in Caraquet, Centre-Péninsule and Miramichi Bay-Neguac - all of which would be considered better prospects for the NDP. Notwithstanding long odds, Duguay has a number of political professionals working full-time to help win the riding, he has spent almost all of his time there and polls show the NDP riding high in northern and francophone New Brunswick. Will the NDP be able to make a breakthrough here? Will Claude Landry be able to continue to maintain the strong support former Conservative MLA Elvy Robichaud enjoyed here? Will the Liberals be able to sneak up the middle in what is largely a Conservative-NDP battle? These are not easy questions to answer, so that makes this a riding to watch. Leans NDP

Saint John Lancaster
This riding was created out of parts of Saint John West (a Conservative stronghold) and Saint John Harbour (a Liberal stronghold). It went Liberal in 1995, 2003 and 2006, while it went Conservative in 1999. Abel LeBlanc is very popular in the poorer, harbour parts of the riding, while the Conservatives enjoy more support in the middle class residential neighbourhoods of the west side. Will the Conservative surge in southern New Brunswick reach into the city of Saint John? Has Abel LeBlanc's salt-of-the-earth approach outstayed its welcome? These are not easy questions to answer, so that makes this a riding to watch. Leans Liberal

This riding was traditionally a Conservative stronghold. Then in 1987, the Liberals won every seat in the province. The Tories expected to regain the seat in 1991 but were foiled by a vote-split with the Confederation of Regions Party who took 1,580 votes to the Conservatives' 1,664. In 1995, they tried again and lost by only 7 votes while COR took 600. After that, Liberal Larry Kennedy had become well entrenched and against poor Conservative candidates Kennedy won by more comfortable margins in 1999 and 2003. In 2004, the Conservative government's health plan promised to close both hospitals in the riding, which led Kennedy to win the highest share of the vote of any Liberal in the province. The Conservatives have been doing very well in rural, anglophone New Brunswick according to polling and anecdotal evidence. And David Alward is from a nearby riding with shared media coverage. Will voters have forgotten the hospital issue? Will the riding return to its traditional conservative orientation? And even if it does, will Kennedy's long tenure allow him to hold on anyway? These are not easy questions to answer, so that makes this a riding to watch. Leans Liberal


After massive blowouts in elections from 1982 through 1999, this looks like it will be the third consecutive close election, returning New Brunswick to its traditional electoral orientation. That said, the Conservatives certainly have the edge with 50% more safe seats than the Liberals and with 28 seats in the final prediction. With no polling in the past week, it is hard to say whether the trend of the election getting closer has continued, or if the Tory lead has expanded back to 10+ points.

Were the Tories to sweep the "ridings to watch" the result would be PC 38, Liberal 17.

If the Liberals have retaken the lead and were to sweep the "ridings to watch" the result would be Liberal 32, PC 23. That is almost exactly the same as the current standings but it would be a very different legislature with the Liberals losing 7 of the seats they currently hold, including three cabinet ministers.

The NDP is in play in somewhere between 2 and 4 ridings (Tracadie-Sheila, Nepisiguit, Bathurst and Tantramar) according to my analysis. They may also be in play in Saint John Harbour, Saint John East and Fredericton-Silverwood, though my analysis says no.

The People's Alliance is in play in probably two ridings - Grand Lake-Gagetown and Charlotte-Campobello. They should also finish respectably (over 10 per cent) in York North and Southwest Miramichi. Under my analysis they have only one (long) shot at a seat and that is in Charlotte-Campobello.

I do not see the Greens in play anywhere, but I expect to see them get respectable (high single-digit to low double-digit) showings in Fredericton-Nashwaaksis, Fredericton-Fort Nashwaak, Petitcodiac, Tantramar and possibly Charlotte-Campobello.

A divided New Brunswick: What a Liberal win would look like

While the odds favour a Conservative victory, a Liberal win is not out of the realm of possibility. But the result would make for a very unusual map. The Conservatives will win the majority of the geography of New Brunswick concentrated in the central, western and southern parts of the province.

The Liberal path to victory is one where they use their lead in southeastern New Brunswick to do a near sweep of that region and hold all or most of their seats in northern New Brunswick to counterbalance the sure gains for the Conservatives in the larger anglophone regions in and around Saint John, Fredericton and Miramichi.

The map to your right is not the most likely scenario for a Liberal win, but it is not impossible to imagine this map which shows the most extreme version of what a divided province yielding a Liberal win would look like. In this scenario, the Liberals are held solely to seats on the Northumberland and Chaleur coasts with a few seats in Greater Saint John, Fredericton and Victoria County.

And before I leave, here is some fun with numbers. I've taken a number of somewhat absurd "surge" scenarios for various parties of varying degrees of unlikelihood to get these possible results:

Liberal vote completely collapses:
PC 44, Liberal 11

Liberal vote collapses to minor parties:
PC 37, NDP 7, Lib 6, PANB 4, Ind 1

Tory vote completely collapses:
Liberal 44, PC 11

Tory vote collapses to minor parties:
Lib 39, PC 8, NDP 5, PANB 3

Tory vote collapses to minor parties, but in English New Brunswick it goes only to Greens and PANB:
Lib 37, PC 8, NDP 4, Green 3, PANB 3

NDP surge at expense of both major parties:
PC 23, Lib 22, NDP 10

NDP surge at expense of PCs:
Lib 39, PC 12, NDP 4

NDP surge at expense of Liberals:
PC 38, NDP 11, Lib 6

Green surge at expense of both parties:
PC 27, Lib 23, Green 4, NDP 1

PANB surge at expense of both parties:
PC 25, Lib 24, PANB 4, NDP 1, Ind 1

Assuming that this analysis is correct, then the NDP have made a fairly major strategic error. Had they focused their campaign on Liberal voters (i.e. this is the change you can trust, a rational, progressive alternative to these two conservative parties) as opposed to a focus on disconnected and disillusioned voters from both parties they would be in a position to do much better.

Friday, September 24, 2010

A "new" poll

Eric at 308dotcom has managed to get CRA to release to him the combined results of their last 3 days in the field (from their surveys for both the TJ and AN/RC/CBC). This includes a regional breakdown.

Here are the numbers as they compare to 2006 and the past two polls with regional breakdowns.

2006CRA/TJ poll
(Aug. 26 - Sep. 1)
(Sep. 15-18)
CRA combined sample
(Sep. 16-19)

2006CRA/TJ poll
(Aug. 26 - Sep. 1)
(Sep. 15-18)
CRA combined sample
(Sep. 16-19)

2006CRA/TJ poll
(Aug. 26 - Sep. 1)
(Sep. 15-18)
CRA combined sample
(Sep. 16-19)

This seems to show the Liberals taking the lead in the north and solidifying their lead in Moncton, with them losing a marginal amount of ground in a stabilizing race in south and central New Brunswick. If you were to assume a uniform swing in the Moncton region, the only PC seats would be Petitcodiac, Riverview and Albert.

In fact, just for fun, here is the result of a uniform swing in each region based on this poll and what the final results would be:

The uniform swing gives us PC 28, Lib 25, NDP 2. Of these, the swing suggests that 17 PC seats (shown in dark blue) are very safe (would be won by more than 1000 votes), 10 Liberal seats (shown in dark red) are very safe and 1 NDP seat is very safe (Miramichi Bay-Neguac). Now the uniform swing has a lot of drawbacks, most notably it is completely blind to local factors. It predicts, for instance, the NDP will get 0 votes in Tracadie-Sheila because they didn't run a candidate there last time. However, it is an interesting experiment nonetheless.

Please note that this is not my prediction, that will follow. However, looking at data in ways like this contributes to my prediction making process.

Wednesday, September 22, 2010

Even more new data

We have the final two CRA/TJ riding polls in the Liberal-held ridings of Miramichi Centre and Fundy-River Valley. At present, both of these ridings are down in the leans PC column, let's see if the data suggests they should be elsewhere.

Miramichi Centre: PC 46, Lib 43, NDP 6, PANB 3, Green 2 (Sept. 17-18)
Reader feedback I had been receiving suggested that this riding was going to be a Tory blow out. This new data shows that the race is well within the margin of error of +/- 4.6 percentage points. I am moving it to toss up (from leans PC).
Fundy-River Valley: PC 53, Lib 36, NDP 5, Green 3, PANB 3 (Sept. 17-18)
With a Tory lead of more than three times the margin of error, this riding belongs in the safe PC column (moving from leans PC).
We also have had a new provincial-wide poll by Abacus Data released yesterday. It shows the race as PC 35, Lib 32, NDP 9, Green 5, PANB 2 with undecided sitting at 17% and a margin of error of +/- 3.7 percentage points. Among just decided voters, it is PC 42, Lib 38, NDP 11, Green 6, PANB 2 with a margin of error of +/- 4.0 percentage points. Unfortunately, this poll does not have regional data so it is impossible to tell whether the Tory lead in Fredericton/Saint John and surrounding areas is falling, the Liberal lead in Greater Moncton is widening or if the Liberals are pulling ahead in the north. I will have to rely on your feedback from the ground to help gauge that.

There are a few other changes that I will make, however:

Moncton East: from toss up to leans Liberal
Based on the consistent wide margin for the Liberals in Greater Moncton, this seat should have been moved to leans Liberal some time ago.
Moncton West: from leans PC to toss up
Based on the consistent wide margin for the Liberals in Greater Moncton, this seat should move.
Nepisiguit: from leans NDP to toss up
Based on overwhelming consensus from reader feedback, I am moving this from to the toss up column.
Please keep your comments coming. A final projection on Saturday or Sunday will take the same format as in 2006, I'll eliminate toss ups and hold myself to 10 "leaning" seats which will be the ridings to watch.

Current projection: the Tories have the edge in a tight race, but the Tories have a base nearly twice as solid as the Liberals

Conservatives: 26 (17 safe + 9 lean)
Liberals: 21 (9 safe + 12 lean)
NDP: 1 (0 safe + 1 lean)
Toss up: 7

Monday, September 20, 2010

A wealth of new data

There is a whole lot of new data to chew on today. The latest CRA/TJ daily tracking poll, a new CRA poll from Acadie-Nouvelle/Radio-Canada/CBC, two new ridng polls by CRA/TJ, and several other CRA/TJ riding polls since our last update.

We've now seen 8 of the promised 10 riding polls, with Miramichi Centre to come tomorrow and another mystery riding some time after that.

Before I get into the polling results, I will briefly explain why I prefer to use "decided+leaning" voters whenever possible, as opposed to just decided voters. Most pollsters ask two vote intentions questions when they are polling elections. First, they ask who people intend to support. Then they ask those who claim to be undecided if they are leaning toward a particular party. This is done because for a variety of reasons, people may have their mind made up but they don't want to say so. The result is a clearer picture of the true results, and a truer picture of the number of undecideds. For some reason, CRA has always ignored this practice and instead asks only the first question resulting in undecided numbers always in the the 35+ range. This election season they have often relesed the answers to the first question, and the answers to the combined first and second question. When available, I will always use the latter.

An article that explains this phenomenon in detail (as it relates to polling party identification in the United States) can be found in here. Other benefits of asking the second (leaners) question include eliminating phenomena like the UK's "Shy Tory Factor" (when voters have decided to vote for a party that is not fashionable but don't want to admit it) and "Flora Syndrome" (when voters aren't going to vote for a party/candidate that is fashionable but want to say they are anyway).

With that behind us, let's recap the riding polls and see what impact they have on our projections.

Grand Lake-Gagetown: Lib 34, PC 34, PANB 23, Green 2*, NDP 1* (Sept. 8-9)
Current projection: Leans PC. No need to revise this one. While the poll shows a tie, it also showed that when asked for their second choice, the Liberals placed third. Another gut analysis to this one is that it will be hard for PANB to get out their vote, so their number will likely slip on election day (a common phenomenon for minor parties) and my expectation is their vote would split in the PCs' favour.
Fredericton-Fort Nashwaak: PC 41, Lib 37, Green 11, NDP 10 (Sept. 9-10)
Current projection: Toss up. No need to revise this one. While the polls shows a PC edge, with the Green and NDP vote high, there is room for it to collapse likely in the Liberals' favour. That would likely even the race up making it too close to call.
Caraquet: Lib 57, PC 31, NDP 8, Green 3 (Sept. 10-11)
Current projection: Safe Liberal. No need to revise this one (for obvious reasons).
Restigouche-la-Vallée: PC 53, Lib 40, NDP 5, Green 2 (Sept. 11-13)
Current projection: Toss up. While the Restigouche West portion of this riding has a history of strongly favouring incumbents (this was the Tories' best seat in the Liberal sweep of 1987, and one of the Liberals' best in the PC landslide of 1999), this is a significant lead for the PCs and it seems quite obvious that this one should go into at least the leans PC column.
Quispamsis: Lib 47, PC 38, NDP 5*, Green 3* (Sept. 13-14)
Current projection: Toss up. This is a fairly signifcant lead for the Liberals, but it is in single digits with a poll with a relatively large margin of error. It moves to leans Liberal but could easily slip back to toss up in the future.
Saint John Harbour: Lib 39, PC 24, NDP 24, Green 5*, Ind 2* (Sept. 12-14)
Current projection: Safe Liberal. While a result of 39% should be very disappointing for the Liberals, on a three-way split it isn't bad. And with them 15 points ahead of both of their competitors this remains in the safe column.
Moncton North: Lib 45, PC 34, NDP 12, Green 8, PANB 2 (Sept. 15-16)
Current projection: Leans PC. This poll result a surprise to me, having just moved Moncton North into the leans PC column. I will move it to leans Liberal though it is on the cusp of safe Liberal.
Dieppe Centre-Lewisville: Lib 45, PC 36, NDP 11, Green 8 (Sept. 15-16)
Current projection: Safe Liberal. As this is in the single digits of a poll with a relatively large margin of error, I will shift it to leans Liberal.
* results with an asterisk are for minor parties which include only decided voters as the decided+leaning numbers for smaller parties those ridings was not included in the Telegraph-Journal's reporting.

With respect to the provice-wide polls at our disposal, I have chosen to ignore the CRA/TJ tracking poll. It has a sample of 700 people from Sept. 12-18. The CRA/AN/RC/CBC poll is by the sample pollster (therefore using the same or similar methodology), has a larger sample of 1055, was done over a shorter period and was just as recent. This poll also (unlike the tracker) has regional breakdowns. It breaks into the same three regions (North, South and Moncton and area), as the first day of the CRA/TJ tracking poll when they had a bigger sample and provided regional breakdowns.

The result of today's poll is PC 34%, Lib 29%, NDP 9%, Green 5%, PANB 0% with 17% truly undecided and 6% refusing to say or not intending to vote. With the latter categories factored out the result is PC 44%, Lib 38%, NDP 12%, Green 6%, PANB 0%.

For analysis let's first take a look, by region, at that the 2006 results, the results of the CRA/TJ poll and the results of the CRA/AN/RC/CBC poll. To compare apples to apples, I will be using the decided and leaning voters (per above) with the undecideds and non-voters factored out. To your right is the table from today's Acadie-Nouvelle showing regional results with the undecideds and non-voters included.

2006CRA/TJ poll
(Aug. 26 - Sep. 1)
(Sep. 15-18)

2006CRA/TJ poll
(Aug. 26 - Sep. 1)
(Sep. 15-18)

2006CRA/TJ poll
(Aug. 26 - Sep. 1)
(Sep. 15-18)

The most interesting result of these numbers is in the Greater Moncton region where the Liberal support has held since the last poll, while the PC support has dropped significantly in favour of the NDP. My instincts suggest that this means that among francophones in this region, the anti-Liberal vote is shifting to the NDP from the PCs. However, as the NDP is not focussing on the region and has little political infrastucture here they won't be able to turn out that vote so some of it will shift back to the PCs and some of those voters will just not go to the polls. In this region, the PCs are down 8 points from their standing just a few weeks ago and 18 points from the last election.

A more significant development (though it is more to be expected so not quite as interesting) is the Liberal collapse in "southern New Brunswick". We have polls from three urban/suburban ridings and one rural riding near Fredericton suggesting that the Liberals are holding their own, so this may be more symptomatic of Liberal support collapsing in ridings that the PCs already hold, but it is impossible for the Liberals to gain seats on a 49-31 spread, and very unlikely that they won't lose some seats as well. In this region, the Liberals are down 8 points from their standing just a few weeks ago and 16 points from the last election.

One might assume that these two regions would largely cancel each other out with gains and losses in each being offset by opposite shifts in the other region. To some extent that is true, but the bad news for the Liberals is that that south region is much larger than the Greater Moncton region.

That brings us to northern New Brunswick. Here, the Liberals are down 9 points from 2006, while the PCs are down 3 from 2006. Both are down 3 from the last poll in late August. This continues to bode relatively well for the NDP who need to be taking votes from both parties in order to win seats up there. However, the NDP may now regret not putting a greater focus on the Moncton region where the francophone vote held by the Conservatives under Lord seems to have shifted to the NDP even without their working that demographic.

I would therefore make the following adjustments.

The North (Victoria, Madawaska, Restigouche, Gloucester and Northumberland counties)

Victoria-Tobique - leans Liberal to toss up
At this stage in the game, with a PC lead of 49-31 in our closest proxy for anglophone New Brunswick, it is hard to put a rural anglophone seat in the Liberal column. If this were any other riding it would be down as leans (or even safe) PC, but because it survived the 1999 PC landslide, I think it should remain a toss up for now.

Restigouche-la-Vallée - toss up to leans PC
The reader feedback was pushing me in this direction anyway but, per the poll above, this obviously belongs on the PC side of the ledger.

Dalhousie-Restigouche East - leans Liberal to toss up
I'm receiving persistent reader feedback that this is more of a race than I suspect in my gut. I will tentatively move it into the "toss up" category until I can learn more.

Bathurst - toss up to leans PC
With the Liberal vote well down from the 2006 election, it is likely that this seat witnessing the third battle in a row between Liberal Brian Kenny and PC Nancy McKay is in play. The last two elections in this traditional Liberal stronghold were nail-biters, we should expect this election to be no different.

(This brings the standings in the "north" - as seemingly defined by CRA - to PC 7, Lib 7, NDP 2, toss up 2.)

The South (Kings, Queens, Saint John, Charlotte, York, Sunbury and Carleton counties)

Quispamsis - toss up to leans Liberal
The poll done by CRA for the TJ in this riding shows Mary Schryer with a relatively comfortable lead of 9 points.

Rothesay - leans PC to safe PC
With an 18 point led in the south, it is impossible to imagine this seat falling out of Tory hands in this election cycle.

Saint John Portland - leans PC to safe PC
With an 18 point led in the south, it is impossible to imagine this seat falling out of Tory hands in this election cycle.

Charlotte-the-Isles - safe Liberal to leans Liberal
The promise by the Conservatives to eliminate tolls on the Grand Manan ferry would seem to suggest that they believe they have a chance in this riding, even though for the most part it has been held for the Liberals since 1978. However, with an 18 point lead for the PCs in the south, it is hard to imagine that there are any truly "safe" Liberal seats here.

York - leans PC to safe PC
With an 18 point led in the south, it is impossible to imagine this seat falling out of Tory hands in this election cycle.

York North - leans PC to safe PC
With an 18 point led in the south, it is impossible to imagine this seat falling out of Tory hands in this election cycle.

(This brings the standings in the "south" - as seemingly defined by CRA - to PC 15, Lib 7, toss up 1.)

Greater Moncton (Kent, Westmorland and Albert counties)

Kent South - safe PC to leans PC
While this is traditionally a relatively reliable seat for the Conservatives and Claude Williams has proved his mettle, unless the Liberals are making major inroads into ridings like Riverview, Albert, Petitcodiac and Moncton Crescent this seat needs to be downgraded to explain the 18 point drop for the Tories versus their 2006 result.

Tantramar - leans PC to toss up
While this is has been a reliable seat for the Conservatives since 1999, unless the Liberals are making major inroads into ridings like Riverview, Albert, Petitcodiac and Moncton Crescent this seat needs to be downgraded to explain the 18 point drop for the Tories versus their 2006 result. It has traditionally (1978-1987 and in the 1998 by-election) been a bit of swing district with with all three parties being in contention.

Dieppe Centre-Lewisville safe Liberal to leans Liberal
The CRA/TJ poll in this riding showed the Liberals leading by 9 points. That keeps it in their column, but a single digit lead in a poll with a relatively high margin of error is not "safe."

Moncton North - leans PC to leans Liberal
The CRA/TJ poll in this riding shows the Liberals ahead by double digits. It belongs in the Liberal column.

(This brings the standings in the "Moncton area" - as seemingly defined by CRA - to PC 6, Lib 6, toss up 2.)

Projection summary - with lots of changes below the surface, our top level numbers are unchanged - PC 28 to Lib 20. However, 3 more seats are in the PCs' safe column and 1 less are in the Liberals'.:

Conservatives: 28 (16 safe + 12 lean)
Liberals: 20 (9 safe + 11 lean)
NDP: 2 (0 safe + 2 lean)
Toss up: 5

Tuesday, September 14, 2010

Two new polls at a critical juncture

Two very interesting polls out today on a very critical day. I have been frustrated that the Telegraph-Journal hasn't yet put out any more regional data since they began their tracking poll, as regional data is much more helpful to me than province-wide data. However, with their poll today showing an 11-point (49-38) PC lead, it is time for an update. That number is eerily close to the 1999 result (51-37) which led to Bernard Lord sweeping 44 seats. This sets the stage as the leaders head into two days with four critical debates.

At the same time, Acadie-Nouvelle has released a poll from an Ottawa pollster called Abacus Data. It shows a nail-biter of a race with the PC's holding a lead of 34-33. The poll is just of "Francophone New Brunswick," though it excludes Dieppe and includes Miramichi. They used the boundaries of four federal ridings (Madawaska-Restigouche, Acadie-Bathurst, Miramichi and Beauséjour) as a proxy for that subset of the province. That gives us about half of the provincial ridings, with the following 22 ridings being entirely (or almost entirely) within those boundaries:
Madawaska-les-Lacs, Edmundston-St. Basile, Restigouche-la-Vallée, Campbellton-Restigouche Centre, Dalhousie-Restigouche East, Nigadoo-Chaleur, Bathurst, Nepisiguit, Caraquet, Shippagan-Lameque-Miscou, Centre-Peninsule-St-Saveur, Tracadie-Sheila, Miramichi Bay-Neguac, Miramichi Centre, Southwest Miramici, Miramichi-Bay du Vin, Rogersville-Kouchibouguac, Kent, Kent South, Shediac-Cap-Pelé, Tantramar, Memramcook-Lakeville-Dieppe.
What this tells as is that the Conservatives must have a massive lead in the rest of the province if they led by only one point in this portion of New Brunswick.

First, let's compare the 2006 vote in "Francophone New Brunswick" (as defined by the poll) to the poll result. The poll reads as PC 43, Lib 41, NDP 13, Other 4 with a margin of error of +/- 4.8 when the undecideds are removed.

 2006AN poll

That suggests that similar to the CRA/TJ poll, the Liberals are suffering to the NDP's benefit. However, unlike the CRA/TJ poll, this poll suggests that the NDP are riding a bit higher and while taking most of their support from the Liberals, they are also taking votes from the PCs.

At a micro level, the TJ has also polled the ridings of Grand Lake-Gagetown, Fredericton-Fort Nashwaak and 8 other "battlegrounds." The first of those polls is out today, with the second to come tomorrow. Today's poll shows the race in Grand Lake-Gagetown as a tie of 34% each for the Liberal and the Conservative, with People's Alliance leader Kris Austin coming in relatively close at 23%. It also shows the Liberals coming in third among people's second choice consideration. In my view, this supports the current projection of "leans PC" for the riding, as Austin will have the weaker organization of the three and will likely underperform his polling position on election day, with his voters not making it to the polls and breaking off in higher numbers to the Conservative candidate.

Now to dig into the CRA/TJ poll. First, to help me do that, I will add another layer to the Abacus/AN poll. That is, how did the vote turn out in that same region in 1999?

 2006AN poll1999

This suggests that the Conservatives are doing even better in "Anglophone New Brunswick" than they did in 1999. They are showing as 7 points off of their showing in the north and eastern part of the province, but overall they are only 2 points off of 1999. Therefore things must be looking very good for them in the rest of the province.

Nepisiguit: from toss up to leans NDP
Because the AN poll, I now expect that the NDP will win at least a seat, if not more than one seat in the north. This is for a number of factors: 1) the poll shows the NDP doing significantly better in the northern and eastern regions of the province than they did in 1999; 2) the NDP nearly won several northern seats in 1999; and 3) the NDP is now taking votes away from both the Tories and the Liberals on a net basis which is critical to their path to victory. This seat was the closest for the NDP in 1999 in these parts, and they are running a strong candidate here to whom they've been giving profile. It was also one of the closest seats in the province in both 1999 and 2006. All of this leads me to believe that this is a prime pick up opportunity for them.

Tracadie-Sheila: from leans PC to leans NDP
Again, if the NDP are able to get 13% of the vote across northern and eastern New Brunswick, they are likely polling in the 30s in the Acadie-Bathrust region. They are now taking support from the Tories who show as 3 points off of their 2006 result and 7 points off of the 1999 result. These are the kind of numbers Duguay needs to win.

Miramichi Southwest: from toss up to leans PC
I've received a lot of reader feedback suggesting that this riding should be in the leans PC column. It is a bit of a historical enigma. On paper, it should be a conservative-leaning seat, but it has been generally safe for the Liberals. But with what appears to be a massive surge of PC support in Anglophone New Brunswick it is difficult to see this seat not being within the PC's grasp.

Moncton East: from leans Liberal to toss up
With the PCs seemingly enjoying a 15 or better point lead in southern and western New Brunswick, urban Moncton seats seem likely to be at least in play, if not leaning Conservative.

Moncton North: from toss up to leans PC
With the PCs seemingly enjoying a 15 or better point lead in southern and western New Brunswick, urban Moncton seats seem likely to be at least in play, if not leaning Conservative.

Moncton West: from toss up to leans PC
With the PCs seemingly enjoying a 15 or better point lead in southern and western New Brunswick, urban Moncton seats seem likely to be at least in play, if not leaning Conservative.

Moncton Crescent: from leans PC to safe PC
If the Liberals were to have picked up 5+ seats, this riding might have been in reach. With the current electoral situation, it is not.

Quispamsis: from leans Liberal to toss up
In the close elections of 2003 and 2006, this seat was a nail-biter with the Liberals barely losing in 2003 and the Liberals barely winning in 2006. It seems to be getting a reputation of something of a bell weather and based on these polls it should probably be in the PC column. Thanks to Mary Schryer's high profile, I'll leave it in the toss up pile for now, but if the current trends continue, it will be the next to go blue.

Saint John East: from safe Liberal to leans Liberal
If there is an 11-point PC win on Sept. 27, there will be very few "safe" Liberal seats in English New Brunswick.

Saint John Lancaster: from safe Liberal to leans Liberal
If there is an 11-point PC win on Sept. 27, there will be very few "safe" Liberal seats in English New Brunswick.

Fundy-River Valley: from toss up to leans PC
This seat is traditionally among the Conservative's safest. It went PC when they won only 6 seats in 1995, and survived the Liberal wave in Saint John in 2003 and almost did again in 2006. If there is a Tory wave, this seat can't stay Liberal.

Charlotte-Campobello: from toss up to leans PC
The Liberals' only hope in this riding is that the strong People's Alliance candidate splits the anti-Liberal vote. But if this poll holds, even on a big split it would be difficult for the Liberals to sneak up the middle.

Fredricton-Fort Nashwaak: from leans Liberal to toss up
If the Telegraph-Journal has identified this as one of its "battlegrounds" (likely based on the cumulative sample of its rolling poll which would be up over 2000 now), it is probably close. This one will be subject to revision based on the poll results released tomorrow.

Fredericton-Silverwood: from safe Liberal to leans Liberal
If there is an 11-point PC win on Sept. 27, there will be very few "safe" Liberal seats in English New Brunswick.

Grand Falls-Drummond-St. Andre: from safe Liberal to leans Liberal
I've heard from some quarters that David Alward has a strong presence here from his time as agriculture minister. This riding went big for the Liberals in 2003, but was closer in 2006. Despite being a majority Francophone riding, it wasn't included in Acadie Nouvelle's poll. I don't have enough evidence to suggest that this riding is in play, but I do have enough to suggest that it is worth at least a closer examination.

This election is now very clearly David Alward's to lose. The next two days of debates will be critical for the Liberals. If they can't do some damage to Alward while at the same time boosting themselves, it will be very difficult for them to draw a map to victory.

Also, if in the coming days, if the CRA/TJ poll continues to show numbers like these (it was 5 points yesterday by 7 points on Sunday), there will likely be more slippage of Liberal seats in Anglophone New Brunswick and in urban New Brunswick.

Projection summary - PCs are more likely than not to win a majority government:

Conservatives: 28 (13 safe + 15 lean)
Liberals: 20 (11 safe + 9 lean)
NDP: 2 (0 safe + 2 lean)
Toss up: 5

Monday, September 06, 2010

New CRA/TJ poll looks good for PCs

A new poll on Friday means some big changes for my predictions. No fewer than 9 seats are affected.

First, the poll analysis...

My biggest pet peeve is when pollsters don't state the margin of error for the sub-samples that they break out.

Here is the meat on the CRA poll, with those margins.

Decided vote (sample of 646)
PC 42%
Lib 41%
NDP 10%
Green 4%
MOE +/- 3.9%

Decided and leaning vote (sample of 1065)
PC 26%
Lib 25%
NDP 6%
Green 2%
Won't vote/refused 14%
Undecided 25%
MOE +/- 3.0%

Decided and leaning by region*

"Northern NB" (sample of 367)
PC 26%
Lib 26%
NDP 6%
Green 1%
Won't vote/refused 13%
Undecided 27%
MOE +/- 5.1%

"Southern NB" (sample of 462)
PC 26%
Lib 24%
NDP 7%
Green 2%
Won't vote/refused 13%
Undecided 25%
MOE +/- 4.6%

"Moncton and area" (sample of 236)
PC 26%
Lib 25%
NDP 5%
Green 4%
Won't vote/refused 16%
Undecided 24%
MOE +/- 6.4%

* these regions are not defined by the pollster or the newspaper. For the purposes of this analysis I will assume that the north is approximately Victoria, Madawaska, Restigouche, Gloucester and Northumberland counties, the south is Kings, Queens, Saint John, Charlotte, York, Sunbury and Carleton counties and Greater Moncton is Kent, Westmorland and Albert counties. The number of ridings for each group is equivalent to the relevant proportion of the total sample that CRA has assigned to the regions.

What is probably the most interesting thing about this poll is that the voter preferences are relatively even across the province. I think that many people would have assumed the Liberals and particularly the NDP would be stronger in the north, but that is not the case.

The best way to give these results some context is to compare them by region to the results in 2006, here goes... These tables show the 2006 results and then takes the won't votes and undecideds out of the sample:

 2006CRA/TJ poll

 2006CRA/TJ poll

 2006CRA/TJ poll

Things look very good for the PCs to make gains in the north, with the NDP splitting the vote with the Liberals, and the Tories holding their 2006 vote there. Now, the Liberals won most of their ridings up there by healthy margins, so the split won't allow a big PC wave (unless it worsens), however it should make for an interesting result on the close seats.

In the south (Greater Fredericton and Greater Saint John), the Tories are down from 2006 but within the margin, while the Liberals are down by a statistically significant amount.

In Greater Moncton, the Liberals are holding their 2006 vote, while the Tories are well off of their vote - this was to be expected with Bernard Lord off of the ballot and won't really affect my predictions there.

The result is 9 prediction changes all in the PCs favour.

Madawaska-les-Lacs: from leans PC -> safe PC
This seat would always have been a long shot for the Liberals. Without a huge sweep of the north, this seat is almost certainly out of reach for them so I am moving it to safe PC.

Restigouche-la-Vallée: from leans Liberal -> tossup
The PC candidate is very strong, and is the younger sister of still-popular former MP Bernard Valcourt. All things being equal, I would give Liberal Burt Paulin the slight edge, but with the Liberals down 6 points from their 2006 result, it belongs in the toss up category.

Bathurst: from leans Liberal -> tossup
I have received a lot of emails from readers suggesting that this seat is definitely in play and that Nancy McKay may be able to make the third time the charm. That advice combined with the poll result pushes it into the toss up column.

Tracadie-Sheila: from toss up -> leans PC
Several readers have written to indicate what I also thought to be true - that Roger Duguay has made a big mistake by choosing this riding to run in. If he had run in Miramichi Bay-Neguac and taken votes from the Liberal he would have won. But here even if he takes all of the Liberal votes, he would still be well short of victory. The poll showing that the PC vote is holding to its 2006 levels in the north makes a Duguay victory and a possible Liberal victory (by sneaking up the middle) less and less likely.

Miramichi Centre: from toss up -> leans PC
This was the only seat on the Miramichi to go PC in 1982 (to that time, the best PC result ever). It still remains the most PC-leaning seat in the area.

Saint John-Fundy: from toss up -> leans PC
This should be, on paper, a safe PC seat any time. But the Liberals have won it in 5 of the last 6 elections on vote splits and the strength of Stuart Jamieson's candidacy. Jamieson has retired, but the Liberal to replace was his choice and is a strong candidate. Notwithstanding all of that, if the PCs are to pick up any seats in Greater Saint John, this is almost certainly the first one to fall.

Fredericton-Nashwaaksis: from toss up -> leans PC
In a traditionally Tory city, this is the traditionally strongest seat. Having gone COR in 1991, it clearly is a ("small c") conservative-leaning riding. It was also a bit of a nail biter in 2006 as compared to the other Fredericton-area ridings.

Grand Lake-Gagetown: from toss up -> leans PC
The Liberals barely hung on in this new riding in 2006, after a massive blowout in the old Grand Lake riding in 2003. There best hope here is that PANB leader Kris Austin splits the vote with the Tories, but based on this poll showing the PCs doing relatively better in the south/central regions of the province than last time, ridings like this that lack an incumbent and were close anyway have to be moved to leans PC.

York: from toss up -> leans PC
This is a traditional swing district and it bears watching, but with the Tories up in this region of the province and Urquhart now having the benefit of incumbency, I'll put it in the leans Tory column.

Projection summary - PCs have a slight edge:

Liberals: 23 (15 safe + 8 lean)
Conservatives: 24 (12 safe + 12 lean)
Toss up: 8