The third and last debate in the UK election campaign has now taken place. In the week before this debate, 11 firms have polled UK electors, with the following results (numbers from 9 firms with polls wholly after the 2nd debate in parentheses; the other 2 firms had a small fraction of their polling before the 2nd debate):
CON - 34.0% (34.2%)
LD - 28.6% (28.3%)
LAB - 26.4% (26.4%)
Comparing these results to the average of the corresponding polls from a week earlier gives:
CON: +1.9% (+2.2%)
LD: -1.8% (-2.2%)
LAB: -0.3% (unchanged)
So basically the Lib Dems lost 2 points to the Tories in the past week, while Labour kept its ground. It will be interesting to see whether Bigotgate and the third debate generate some movement. An instant poll suggest that the 3rd debate is likely to have little effect on voting intentions.
A uniform swing with the current numbers would give something like CON 270, LAB 255, LD 95. As I've cautioned before, this likely overestimates Labour's count and underestimates the other two parties'. FiveThirtyEight takes into account and projects something around CON 300, LAB 200, LD 120, which looks reasonable.
Thursday, April 29, 2010
EKOS: Tories lead by 5.3
The weekly EKOS poll is out, and shows little change nationally for the top two parties. However, the NDP now gets 17.6%, up 1.3% from last week: not as great a result as the 20% in this week's Harris-Decima poll, but still a very encouraging one. This is a bad poll for the Liberals: they trail the Tories marginally in Atlantic Canada and Ontario, and have a measly 19.3% in Québec. For the Tories, this poll isn't great either: they trail the NDP marginally in BC (just like in the H-D poll), and only got 38.2% in MB/SK.
The aggregate projection is now:
CON - 130
LIB - 83
BQ - 54
NDP - 41
The Bloc's projection is a new high on this blog, and indeed matches their best-ever electoral results in 1993 and 2004. The Tories have been stuck in the 130-133 range for two months now, while this is the fifth consecutive projection where the Liberals get exactly 83.
The average Conservative national lead is 4.5%.
The aggregate projection is now:
CON - 130
LIB - 83
BQ - 54
NDP - 41
The Bloc's projection is a new high on this blog, and indeed matches their best-ever electoral results in 1993 and 2004. The Tories have been stuck in the 130-133 range for two months now, while this is the fifth consecutive projection where the Liberals get exactly 83.
The average Conservative national lead is 4.5%.
Tuesday, April 27, 2010
Harris-Decima: NDP Surges, Tories Under 30
The newest Harris-Decima poll puts the Conservatives at a startlingly low 29%, just 2 ahead of the Liberals, and 9 ahead of the NDP. In this poll, the NDP lead the Tories 31-30 in BC, challenge them (31-39) in MB/SK, and score a respectable 19% in Ontario. Are some Canadians giving a second look to our own leftish third party, as the British are doing?
Meanwhile, in Québec, this poll has the Bloc at an adscam-like 45%, while the Tories would be essentially wiped off the map with their 10%.
This NDP surge gives them 2 extra seats at the expense of the Conservatives in the aggregate projection, while the Bloc consolidates its 53 projected seats:
CON - 131
LIB - 83
BQ - 53
NDP - 41
The Bloc and the NDP are at their highest levels since the creation of this blog, while the Tories see their average national lead over the Liberals shrink to 4.4%.
Since this Harris-Decima poll is a bit wacky, I decided to also make a projection based on this poll alone, just for fun. I get: CON 110, LIB 90, BQ 61, NDP 46, GRN 1. If the NDP goes up even more at the expense of the Tories, we could get all parties below 100 seats!
Meanwhile, in Québec, this poll has the Bloc at an adscam-like 45%, while the Tories would be essentially wiped off the map with their 10%.
This NDP surge gives them 2 extra seats at the expense of the Conservatives in the aggregate projection, while the Bloc consolidates its 53 projected seats:
CON - 131
LIB - 83
BQ - 53
NDP - 41
The Bloc and the NDP are at their highest levels since the creation of this blog, while the Tories see their average national lead over the Liberals shrink to 4.4%.
Since this Harris-Decima poll is a bit wacky, I decided to also make a projection based on this poll alone, just for fun. I get: CON 110, LIB 90, BQ 61, NDP 46, GRN 1. If the NDP goes up even more at the expense of the Tories, we could get all parties below 100 seats!
Saturday, April 24, 2010
UK Election After Second Debate
First, here's the average of the last poll before the second debate from 10 polling firms. Of these 10 firms, 7 have already conducted a poll after the second debate; the pre-debate average of these 7 firms is in parentheses.
CON - 31.9% (31.9%)
LD - 30.5% (30.3%)
LAB - 26.4% (26.7%)
Here's the post-debate average, with change from the pre-debate 7-firm average in parentheses:
CON - 34.3% (+2.4)
LD - 29% (-1.3)
LAB - 26.6% (-0.1)
These post-debate numbers give, according to UK Polling Report (change with respect to pre-debate 7-firm average in parentheses):
CON - 270 (+31)
LAB - 254 (-14)
LD - 95 (-17)
Other - 13 (0)
The Tories have taken the lead, but do not have a majority. The usual caveat still applies: Labour is likely to get fewer seats than predicted by a uniform national swing, while the Lib Dems and the Tories are likely to get more. It looks very likely that if an election were held now, the Conservatives would be the largest party, followed by Labour and the Lib Dems.
Could a Labour-Lib Dem coalition arise? The above prediction gives them 349 seats, against 270 for the Tories and 31 for others (13 from GB + 18 from Northern Ireland). But in reality, Labour-Lib Dem should get fewer than 349 seats. If they fall below the majority threshold (theoretically 326, but in practice around 323 due to Sinn Fein not taking its seats and the Speaker not voting) without the Tories rising above it, Labour and Lib Dems would need the support of small regional parties. Alternatively, the Tories could try to woo the Lib Dems or maybe the small parties into government.
The UK could also get a minority government like in Canada. I doubt Labour would try, or be able, to stay in power without a formal coalition if they finish third in popular vote. However, as the largest party, the Tories could try to do that.
This election will be fascinating to follow!
CON - 31.9% (31.9%)
LD - 30.5% (30.3%)
LAB - 26.4% (26.7%)
Here's the post-debate average, with change from the pre-debate 7-firm average in parentheses:
CON - 34.3% (+2.4)
LD - 29% (-1.3)
LAB - 26.6% (-0.1)
These post-debate numbers give, according to UK Polling Report (change with respect to pre-debate 7-firm average in parentheses):
CON - 270 (+31)
LAB - 254 (-14)
LD - 95 (-17)
Other - 13 (0)
The Tories have taken the lead, but do not have a majority. The usual caveat still applies: Labour is likely to get fewer seats than predicted by a uniform national swing, while the Lib Dems and the Tories are likely to get more. It looks very likely that if an election were held now, the Conservatives would be the largest party, followed by Labour and the Lib Dems.
Could a Labour-Lib Dem coalition arise? The above prediction gives them 349 seats, against 270 for the Tories and 31 for others (13 from GB + 18 from Northern Ireland). But in reality, Labour-Lib Dem should get fewer than 349 seats. If they fall below the majority threshold (theoretically 326, but in practice around 323 due to Sinn Fein not taking its seats and the Speaker not voting) without the Tories rising above it, Labour and Lib Dems would need the support of small regional parties. Alternatively, the Tories could try to woo the Lib Dems or maybe the small parties into government.
The UK could also get a minority government like in Canada. I doubt Labour would try, or be able, to stay in power without a formal coalition if they finish third in popular vote. However, as the largest party, the Tories could try to do that.
This election will be fascinating to follow!
Ipsos: Conservatives Lead by 6
A new Ipsos poll is out (via ThreeHundredEight.com), and shows a 6-point national lead for the Tories. This poll is actually a pretty good one for the Liberals: an improvement in every region from the last Ipsos poll two weeks ago (in which the Tory lead was 10%), a 4-point lead in Ontario, and a weaker Bloc in Québec. Not much else to note, except that the Conservatives' suspiciously good BC result and the Liberals' and NDP's suspiciously poor one from two weeks ago did not repeat themselves.
All this puts the Liberals closer to some seats in Québec and Ontario, but not enough to change the projection:
CON - 133
LIB - 83
BQ - 53
NDP - 39
Despite accounting for about 2/3 of my projection weight, this week's EKOS and Ipsos polls did not cause any change. Canadians might be too busy following the NHL playoffs to be updating their voting intention :-)
The average Tory national lead is down to 5.1% due to the much reduced weight given to the old Ipsos poll as a result of the new one.
All this puts the Liberals closer to some seats in Québec and Ontario, but not enough to change the projection:
CON - 133
LIB - 83
BQ - 53
NDP - 39
Despite accounting for about 2/3 of my projection weight, this week's EKOS and Ipsos polls did not cause any change. Canadians might be too busy following the NHL playoffs to be updating their voting intention :-)
The average Tory national lead is down to 5.1% due to the much reduced weight given to the old Ipsos poll as a result of the new one.
Wednesday, April 21, 2010
UK Update
I have now seen polls from 9 firms since last week's UK debate. Averaging the most recent one from each gives (change from Monday's post in parentheses, +/- 0.1% due to rounding):
CON - 32.3% (+0.2%)
LD - 30.6% (+0.2%)
LAB - 26.6% (-0.3%)
3 of the 9 polls considered are the same as Monday, while 3 firms have released updated polls; of course, 3 new firms were added to the mix. The corresponding seat count according to UK Polling Report (which, again, is probably too optimistic for Labour and too pessimistic for the Lib Dems):
LAB - 259 (-6)
CON - 246 (+3)
LD - 114 (+3)
Other - 13 (0)
Let's see what movement the second debate will yield... (And you've got to give it to the British: after not having any televised leaders' debates until now, they're going with 3 in 15 days!)
CON - 32.3% (+0.2%)
LD - 30.6% (+0.2%)
LAB - 26.6% (-0.3%)
3 of the 9 polls considered are the same as Monday, while 3 firms have released updated polls; of course, 3 new firms were added to the mix. The corresponding seat count according to UK Polling Report (which, again, is probably too optimistic for Labour and too pessimistic for the Lib Dems):
LAB - 259 (-6)
CON - 246 (+3)
LD - 114 (+3)
Other - 13 (0)
Let's see what movement the second debate will yield... (And you've got to give it to the British: after not having any televised leaders' debates until now, they're going with 3 in 15 days!)
EKOS: Conservatives Lead by 4.6
This week's EKOS poll bears few surprises. The only notable result is the dismal 13.8% Tory number in Québec, which is to be expected given the "Québec bashing" by Conservative MPs from Québec, and amid talk of the Conservative legislation reducing Québec's weight in the Commons.
In fact, the poll suggests so little movement that although it accounts for almost 40% of my projection (due to the small number of other recent polls), it did not cause any change:
CON - 133
LIB - 83
BQ - 53
NDP - 39
Conservative average national lead = 5.4%.
In fact, the poll suggests so little movement that although it accounts for almost 40% of my projection (due to the small number of other recent polls), it did not cause any change:
CON - 133
LIB - 83
BQ - 53
NDP - 39
Conservative average national lead = 5.4%.
Monday, April 19, 2010
UK Election
6 different firms have completed polls following last Thursday's historic debate. Doing a straight arithmetic average of the polls gives:
CON - 32.2%
LD -31.2% 30.3%
LAB - 26.8%
However, according to UK Polling Report, assuming a uniform swing gives the following distribution of the 632 British seats (18 more are allocated to Northern Ireland - none of them are expected to go for any of the three major parties):
LAB -261 265
CON -242 243
LD -116 111
Other - 13
(The BBC's calculator gives the same result for each party, +/- 3 seats.)
If the popular vote pattern stays stable (which is of course a huge 'if', given the tremendous shift due to the debate), these seat projections are likely to be too pessimistic for the Lib Dems. The reason is that when a small party's support increases, assuming that the increase is uniform disadvantages it in a first-past-the-post system. Consider the following stylized example:
Before the surge
Seat A: Lab 40, Con 30, LD 20
Seat B: Lab 40, Con 30, LD 20
Seat C: Lab 35, Con 35, LD 20
Seat D: Con 40, Lab 30, LD 20
Seat E: Con 45, LD 25, Lab 20
Total: Con 36%, Lab 33%, LD 21% (Con 2.5 seats, Lab 2.5 seats, LD 0 seats)
After an uniform surge (+10 LD, -5 Con, -5 Lab)
Seat A: Lab 35, LD 30, Con 25
Seat B: Lab 35, LD 30, Con 25
Seat C: Lab 30, Con 30, LD 30
Seat D: Con 35, LD 30, Lab 25
Seat E: Con 40, LD 35, Lab 15
Total: Con 31%, LD 31%, Lab 28% (Con 2.33 seats, Lab 2.33 seats, LD 0.33 seats)
A uniform rise in Lib Dem fortunes allows them to finish second in ridings A, B, D and E, which were Labour or Tory strongholds. This of course doesn't give the Lib Dems any seats, so all they can hope for is that the 3-way race for seat C goes their way.
However, as we saw with the Conservatives in Québec during the 2006 general election, rising fortunes may well be distributed unevenly. Suppose the Lib Dems' support doesn't change in seats B and D, but their surge doubles in seats A and E. Then we get:
After a skewed surge
Seat A: LD 40, Lab 30, Con 20
Seat B: Lab 40, Con 30, LD 20
Seat C: Lab 30, Con 30, LD 30
Seat D: Con 40, Lab 30, LD 20
Seat E: LD 45, Con 35, Lab 10
Total: Con 31%, LD 31%, Lab 28% (LD 2.33 seats, Con 1.33 seats, Lab 1.33 seats)
In reality, the Lib Dem's increase in popularity is unlikely to be nearly as uneven as in the above example: if their popular vote stays where it is, they will likely end up with the most 2nd place finishes (and the fewest wins). However, it also probably wouldn't be as bad as the standard uniform swing model predicts.
With 2 debates left, this will be an interesting election to follow!
CON - 32.2%
LD -
LAB - 26.8%
However, according to UK Polling Report, assuming a uniform swing gives the following distribution of the 632 British seats (18 more are allocated to Northern Ireland - none of them are expected to go for any of the three major parties):
LAB -
CON -
LD -
Other - 13
(The BBC's calculator gives the same result for each party, +/- 3 seats.)
If the popular vote pattern stays stable (which is of course a huge 'if', given the tremendous shift due to the debate), these seat projections are likely to be too pessimistic for the Lib Dems. The reason is that when a small party's support increases, assuming that the increase is uniform disadvantages it in a first-past-the-post system. Consider the following stylized example:
Before the surge
Seat A: Lab 40, Con 30, LD 20
Seat B: Lab 40, Con 30, LD 20
Seat C: Lab 35, Con 35, LD 20
Seat D: Con 40, Lab 30, LD 20
Seat E: Con 45, LD 25, Lab 20
Total: Con 36%, Lab 33%, LD 21% (Con 2.5 seats, Lab 2.5 seats, LD 0 seats)
After an uniform surge (+10 LD, -5 Con, -5 Lab)
Seat A: Lab 35, LD 30, Con 25
Seat B: Lab 35, LD 30, Con 25
Seat C: Lab 30, Con 30, LD 30
Seat D: Con 35, LD 30, Lab 25
Seat E: Con 40, LD 35, Lab 15
Total: Con 31%, LD 31%, Lab 28% (Con 2.33 seats, Lab 2.33 seats, LD 0.33 seats)
A uniform rise in Lib Dem fortunes allows them to finish second in ridings A, B, D and E, which were Labour or Tory strongholds. This of course doesn't give the Lib Dems any seats, so all they can hope for is that the 3-way race for seat C goes their way.
However, as we saw with the Conservatives in Québec during the 2006 general election, rising fortunes may well be distributed unevenly. Suppose the Lib Dems' support doesn't change in seats B and D, but their surge doubles in seats A and E. Then we get:
After a skewed surge
Seat A: LD 40, Lab 30, Con 20
Seat B: Lab 40, Con 30, LD 20
Seat C: Lab 30, Con 30, LD 30
Seat D: Con 40, Lab 30, LD 20
Seat E: LD 45, Con 35, Lab 10
Total: Con 31%, LD 31%, Lab 28% (LD 2.33 seats, Con 1.33 seats, Lab 1.33 seats)
In reality, the Lib Dem's increase in popularity is unlikely to be nearly as uneven as in the above example: if their popular vote stays where it is, they will likely end up with the most 2nd place finishes (and the fewest wins). However, it also probably wouldn't be as bad as the standard uniform swing model predicts.
With 2 debates left, this will be an interesting election to follow!
Thursday, April 15, 2010
EKOS: Tories Lead by 2.4
As the mystery of the Guergis affair engulfs Ottawa, the Conservative lead has slipped according to this week's EKOS poll. The big news out of the poll is the 5.5% Liberal lead in Ontario (last week, the Conservatives had a 7.7% lead there). Also, the Bloc has dropped 4.7% from last week, and now sits at 34.7% in Québec - this change (surprising given the scandals in both Ottawa and Quebec City) is within the margin of error.
Not much change in the aggregate projection: Grits take back one Ontario seat.
CON - 133
LIB - 83
BQ - 53
NDP - 39
Tory average national lead is back down to 5.3%.
Not much change in the aggregate projection: Grits take back one Ontario seat.
CON - 133
LIB - 83
BQ - 53
NDP - 39
Tory average national lead is back down to 5.3%.
Monday, April 12, 2010
Ipsos: Tories Up 10; Léger: Bloc Up 17 in Québec
Two new polls out today. First, the Ipsos poll (hat tip, as usual, to ThreeHundredEight.com) has the Tories leading by 10% nationally. The big news from this poll is the Conservative 46% in BC, compared to just 20% for the NDP and 19% for the Liberals - a far cry from last week's EKOS poll which had the NDP leading the Tories there! The Tories also did well in Ontario, where they lead the Liberals by 3%, and in MB/SK, where they got 59%. The Bloc had a strong 42% in Québec.
The Léger poll confirms the Bloc's strength, putting them at 38% against a worrisome 21% for the Liberals. The Tories are still weak (but steady) at 17%, while the NDP managed to tie them a strong performance.
Including this poll in my aggregate projection gives:
CON - 134
LIB - 82
BQ - 53
NDP - 39
The Bloc's 53 seats ties a mid-December high for the Bloc. As expected, the NDP pulled back a little, but is still very strong. The Liberal slide that started in mid-February continues: they haven't been this low since the Jan. 7 update. Conversely, the Tories are at their highest since that same Jan. 7 update - but they've been stuck in the low 130s for over a month now.
Tories lead my national poll average by 5.8%.
The Léger poll confirms the Bloc's strength, putting them at 38% against a worrisome 21% for the Liberals. The Tories are still weak (but steady) at 17%, while the NDP managed to tie them a strong performance.
Including this poll in my aggregate projection gives:
CON - 134
LIB - 82
BQ - 53
NDP - 39
The Bloc's 53 seats ties a mid-December high for the Bloc. As expected, the NDP pulled back a little, but is still very strong. The Liberal slide that started in mid-February continues: they haven't been this low since the Jan. 7 update. Conversely, the Tories are at their highest since that same Jan. 7 update - but they've been stuck in the low 130s for over a month now.
Tories lead my national poll average by 5.8%.
Saturday, April 10, 2010
Adjusted Final NHL Regular Season Ranking, 2009-2010
As I explained in this post, to gauge the relative performances of NHL teams during the 2009-2010 season, a reasonable adjustment for the differing strengths of the two conferences is to add 3 points to Western teams, and subtract 3 points from Eastern teams.
Below are the final standings with this adjustment. I broke ties between teams in different divisions by considering the strength of the other teams in the division relative to the strength of the conference (since the difference in the latter has already been taken into account). If I had actually computed the appropriate divisional adjustments, all would be less than 1 point, and most would be less than 0.5 point. Ties between teams in the same division are resolved the usual way.
1. Washington 118
2. San Jose 116
3. Chicago 115
4. Phoenix 110
5. Vancouver 106
6. Detroit 105
7. Los Angeles 104
8. Nashville 103
9. New Jersey 100
10. Pittsburgh 98
11. Colorado 98
12. Buffalo 97
13. St. Louis 93
14. Calgary 93
15. Anaheim 92
16. Dallas 91
17. Ottawa 91
18. Boston 88
19. Minnesota 87
20. Philadelphia 85
21. Montreal 85
22. NY Rangers 84
23. Columbus 82
24. Atlanta 80
25. Carolina 77
26. Tampa Bay 77
27. NY Islanders 76
28. Florida 74
29. Toronto 71
30. Edmonton 65
7 of the top 8, and 12 of the top 16 teams are in the West - only 4 Eastern teams "deserve" to make the playoffs (WSH, NJ, PIT, BUF), while only 3 Western ones "deserve" to miss it (EDM, CBJ, MIN). The 8th seed in the West, Colorado, ranks ahead of Buffalo, the Northeast division champ!
Now, wouldn't it be awesome if the 21st regular-season team in the league wins the Cup?
Below are the final standings with this adjustment. I broke ties between teams in different divisions by considering the strength of the other teams in the division relative to the strength of the conference (since the difference in the latter has already been taken into account). If I had actually computed the appropriate divisional adjustments, all would be less than 1 point, and most would be less than 0.5 point. Ties between teams in the same division are resolved the usual way.
1. Washington 118
2. San Jose 116
3. Chicago 115
4. Phoenix 110
5. Vancouver 106
6. Detroit 105
7. Los Angeles 104
8. Nashville 103
9. New Jersey 100
10. Pittsburgh 98
11. Colorado 98
12. Buffalo 97
13. St. Louis 93
14. Calgary 93
15. Anaheim 92
16. Dallas 91
17. Ottawa 91
18. Boston 88
19. Minnesota 87
20. Philadelphia 85
21. Montreal 85
22. NY Rangers 84
23. Columbus 82
24. Atlanta 80
25. Carolina 77
26. Tampa Bay 77
27. NY Islanders 76
28. Florida 74
29. Toronto 71
30. Edmonton 65
7 of the top 8, and 12 of the top 16 teams are in the West - only 4 Eastern teams "deserve" to make the playoffs (WSH, NJ, PIT, BUF), while only 3 Western ones "deserve" to miss it (EDM, CBJ, MIN). The 8th seed in the West, Colorado, ranks ahead of Buffalo, the Northeast division champ!
Now, wouldn't it be awesome if the 21st regular-season team in the league wins the Cup?
Thursday, April 8, 2010
EKOS (Small Sample): Tories Up by 6.3
This week's EKOS poll only had 909 respondents, three times less than usual, due to the Easter holiday. The small sample size produced some wacky results: Tories lead by 7.7% in Ontario, but Liberals are up by 2.4% in MB/SK, while the NDP has a 0.7% lead over the Conservatives in BC. Needless to say that I find all this rather suspicious, and would wait before proclaiming that Ontario has gone blue, the (eastern) Prairies red, and that the federal NDP is leading in BC.
Nevertheless, adding this poll to the mix (with a lower weight) did hurt the Liberals and help the small parties:
CON - 130
LIB - 85
BQ - 52
NDP - 41
This projection is the worst for the Liberals since early January (down 20 from their February peak), but also the worst for the Conservatives since late February (though their best in this period was not much higher: 133). It is the best for the Bloc since mid-January, and a new high for the NDP since the creation of this blog.
Why so? Some of this is probably simply due to chance fluctuations: I wouldn't be surprised if the Bloc and NDP numbers slide back a little in the next update. Most of this strength, however, appears to be real, which is consistent with Canadians being utterly unimpressed with either Harper or Ignatieff. But regional factors also play an important role in the inordinate strength of the Bloc and the NDP: the Bloc numbers reflect the unpopularity of the provincial Liberal government and the Tory proposal to reduce Québec's weight in the House of Commons; the NDP numbers in BC probably have to do with the HST.
Tory national average lead: 5%.
Nevertheless, adding this poll to the mix (with a lower weight) did hurt the Liberals and help the small parties:
CON - 130
LIB - 85
BQ - 52
NDP - 41
This projection is the worst for the Liberals since early January (down 20 from their February peak), but also the worst for the Conservatives since late February (though their best in this period was not much higher: 133). It is the best for the Bloc since mid-January, and a new high for the NDP since the creation of this blog.
Why so? Some of this is probably simply due to chance fluctuations: I wouldn't be surprised if the Bloc and NDP numbers slide back a little in the next update. Most of this strength, however, appears to be real, which is consistent with Canadians being utterly unimpressed with either Harper or Ignatieff. But regional factors also play an important role in the inordinate strength of the Bloc and the NDP: the Bloc numbers reflect the unpopularity of the provincial Liberal government and the Tory proposal to reduce Québec's weight in the House of Commons; the NDP numbers in BC probably have to do with the HST.
Tory national average lead: 5%.
Wednesday, April 7, 2010
New Harris-Decima Poll: Conservatives by 3
Poll here. Short on time today, so here's the aggregate projection:
CON - 131
LIB - 88
BQ - 50
NDP - 39
National average: Tories up by 4.7%.
CON - 131
LIB - 88
BQ - 50
NDP - 39
National average: Tories up by 4.7%.
Sunday, April 4, 2010
Which 16 NHL Teams Would Make the Playoffs with a Balanced Schedule?
The last inter-conference game of the NHL regular season was played yesterday - the next time that teams from the East and the West meet will be in the Stanley Cup Finals. Teams from the Eastern Conference only won 115 of the 270 inter-conference games this season, a success rate of 42.6%.
The West being stronger than the East and yet getting the same number of playoff spots is a sore point for some Western hockey fans. For example, the Flames now only have a 5.5%-chance of making the playoffs according to Sports Club Stats even though their record would guarantee them a spot in the East. Some have therefore proposed to simply allocate playoff spots to the top 16 teams, regardless of conference.
However, the actual situation is worse: Western teams also suffer from facing stronger Western teams for 64 games, and weaker Eastern ones for only 18 games. How many points is that worth?
Well, to a first approximation, to make the schedule fair, each team should play 23 fewer games against teams in its own conference, and 23 more against ones in the other conference. In those 23 games, Eastern teams would have a success rate of 42.6% instead of 50%. On average, they would therefore win about 1.7 fewer games.
Each win is not quite worth 2 points, since the losing team gets 1 point for making it to OT. That has happened in 23.6% of games played this season. Thus, a win is worth 2-0.236 = 1.764 points.
Bottom line: One should subtract 3.005 points from each team's total in the East, and add 3.005 for Western teams.
Of course, ideally one would also adjust for the strength of divisional opponents (against whom each team plays 6 times instead of 4 against non-division rivals in the same conference), and the strength of the 3 teams from the other conference a team plays against twice instead of once. However, those adjustments are bound to be much smaller, so to a first approximation, adding 3 points to Western teams and subtracting 3 from Eastern teams is appropriate. Equivalently, we could simply add 6 to Western totals.
(The appropriate adjustment for now is not 6, but about 5.5 since the season isn't over yet.)
Looking at the current standings, counting each game left beyond 3 as 1.1 points (i.e. teams with 4 games left get +1.1, while the Predators get -1.1 for having just 2 remaining), and adding 5.5 to Western teams, my homemade NHL ranking is:
1. Washington 115.1
2. San Jose 112.5
3. Chicago 111.6
4. Phoenix 107.5
5. Vancouver 105.5
6. Nashville 102.4
7. Los Angeles 101.6
8. Detroit 101.5
9. New Jersey 98.1
10. Pittsburgh 98.1
11. Colorado 97.6
12. Buffalo 97.1
13. Calgary 94.5
14. Anaheim 91.6
15. St. Louis 91.6
16. Ottawa 91
17. Dallas 89.5
18. Minnesota 86.5
19. Montreal 86
20. Boston 85.1
21. Philadelphia 84
22. NY Rangers 83.1
23. Columbus 82.5
24. Atlanta 81
25. NY Islanders 77.1
26. Carolina 76
27. Florida 75.1
28. Tampa Bay 75.1
29. Toronto 72
30. Edmonton 62.6
Instead of being already guaranteed to win the President's Trophy, Washington would be in a tough fight with San Jose and Chicago for the top spot, especially if you additionally adjust for divisional strength. Fully 7 of the 8 best teams in the league are in the West, and the East has 10 of the 12 worst teams! The Devils, Penguins and Sabres suddenly look a lot less likely to hoist the Cup in June: they are only about as good as the Avalanche, ranked 8th in the West. Overall, 11 Western teams and 5 Eastern teams deserve to make the playoffs.
In terms of Canadian teams, instead of having Ottawa in, Montreal almost in and Calgary almost out, we'd have Calgary almost in, Ottawa on the bubble (but looking good) and Montreal almost out.
I will update the homemade standings at the end of the regular season, next week.
The West being stronger than the East and yet getting the same number of playoff spots is a sore point for some Western hockey fans. For example, the Flames now only have a 5.5%-chance of making the playoffs according to Sports Club Stats even though their record would guarantee them a spot in the East. Some have therefore proposed to simply allocate playoff spots to the top 16 teams, regardless of conference.
However, the actual situation is worse: Western teams also suffer from facing stronger Western teams for 64 games, and weaker Eastern ones for only 18 games. How many points is that worth?
Well, to a first approximation, to make the schedule fair, each team should play 23 fewer games against teams in its own conference, and 23 more against ones in the other conference. In those 23 games, Eastern teams would have a success rate of 42.6% instead of 50%. On average, they would therefore win about 1.7 fewer games.
Each win is not quite worth 2 points, since the losing team gets 1 point for making it to OT. That has happened in 23.6% of games played this season. Thus, a win is worth 2-0.236 = 1.764 points.
Bottom line: One should subtract 3.005 points from each team's total in the East, and add 3.005 for Western teams.
Of course, ideally one would also adjust for the strength of divisional opponents (against whom each team plays 6 times instead of 4 against non-division rivals in the same conference), and the strength of the 3 teams from the other conference a team plays against twice instead of once. However, those adjustments are bound to be much smaller, so to a first approximation, adding 3 points to Western teams and subtracting 3 from Eastern teams is appropriate. Equivalently, we could simply add 6 to Western totals.
(The appropriate adjustment for now is not 6, but about 5.5 since the season isn't over yet.)
Looking at the current standings, counting each game left beyond 3 as 1.1 points (i.e. teams with 4 games left get +1.1, while the Predators get -1.1 for having just 2 remaining), and adding 5.5 to Western teams, my homemade NHL ranking is:
1. Washington 115.1
2. San Jose 112.5
3. Chicago 111.6
4. Phoenix 107.5
5. Vancouver 105.5
6. Nashville 102.4
7. Los Angeles 101.6
8. Detroit 101.5
9. New Jersey 98.1
10. Pittsburgh 98.1
11. Colorado 97.6
12. Buffalo 97.1
13. Calgary 94.5
14. Anaheim 91.6
15. St. Louis 91.6
16. Ottawa 91
17. Dallas 89.5
18. Minnesota 86.5
19. Montreal 86
20. Boston 85.1
21. Philadelphia 84
22. NY Rangers 83.1
23. Columbus 82.5
24. Atlanta 81
25. NY Islanders 77.1
26. Carolina 76
27. Florida 75.1
28. Tampa Bay 75.1
29. Toronto 72
30. Edmonton 62.6
Instead of being already guaranteed to win the President's Trophy, Washington would be in a tough fight with San Jose and Chicago for the top spot, especially if you additionally adjust for divisional strength. Fully 7 of the 8 best teams in the league are in the West, and the East has 10 of the 12 worst teams! The Devils, Penguins and Sabres suddenly look a lot less likely to hoist the Cup in June: they are only about as good as the Avalanche, ranked 8th in the West. Overall, 11 Western teams and 5 Eastern teams deserve to make the playoffs.
In terms of Canadian teams, instead of having Ottawa in, Montreal almost in and Calgary almost out, we'd have Calgary almost in, Ottawa on the bubble (but looking good) and Montreal almost out.
I will update the homemade standings at the end of the regular season, next week.
Thursday, April 1, 2010
Harper's Slightly Modified Representation Plan
This is a long post concerning Bill C-12, which changes the formula for allocating seats for future elections.
Part I - General discussion
Part II - Analysis of the bill
Part III - Estimates of seats after the next redistribution (these are likely to be more accurate than the government's, for reasons explained in bold in Part I)
Part IV - Likely electoral impact
Part I
Here is the text of Bill C-12. There are two changes with respect to the previous Tory proposal, which I analyzed here:
1. The size of districts in ON, BC and AB is no longer be the population of Québec divided by 75. Rather, it is set at 108,000 for the coming redistribution, and increases according to the rate of increase of the 10-province population for subsequent redistributions.
2. The seat numbers for ON, BC and AB (for the coming redistribution, population divided by 108,000) are rounded up, rather than to the closest integer. So 108,000 is the maximum district size.
How many seats would this law give to ON, BC and AB? CBC and most media, citing a government press release, say 124 (+18), 43 (+7) and 33 (+5) respectively (and usually without any acknowledgment of the underlying uncertainty). These estimates are based on the most recent population projections by Statistics Canada. Unfortunately, those projections are outdated (based on July 2005 population estimates), and correct for census undercounting (which we do not want to adjust for in this case, since raw census numbers are used for redistribution). Hence, the numbers above are inaccurate. I believe that those estimates are too high for ON and BC, and too low for Alberta, and outline my calculations in Part III.
Part II
Before going into the nitty-gritty of estimates, let me analyze the broad impacts of this law, as well as compare it to the previous version, and to my proposal:
1. It gives ON, BC and AB fewer seats than the previous Tory proposal, which would have given those three provinces representation in proportion to Québec's. This is because Québec's population in Census 2011 will likely be around 7.8 million, which is lower than 75*108,000 = 8.1 million. Thus, Québec will remain overrepresented with respect to ON, BC and AB.
2. However, it still gives those provinces more seats than under my proposal, which would have given Québec representation in accordance to its demographic weight in Canada. In order for Québec to be proportionately represented with 75 seats, the House size would have to be around 323 seats. Under current law, it is likely to be about 315 after the next redistribution; under C-12, it will likely be around 335. Thus, C-12 makes Québec go from being overrepresented to underrepresented with respect to the rest of Canada.
3. Assuming that Québec grows more slowly than Canada as a whole (not true lately due to the Ontario-concentrated recession, but likely to be true in the long run), the rate of growth of the size of the House is lowest under current law, higher under my proposal, and highest under the old Tory proposal. Under C-12, the rate of growth will be similar to the one under current law. Thus, while C-12 will increase the size of the House significantly in the next redistribution, it will lead to small increases in subsequent ones. This means that C-12 is a one-time patch, and in the long run:
a) AB, BC and ON will resume losing relative representation as fast as they do now if their population growth keeps outstripping the national average;
b) Québec will eventually once again become overrepresented if its growth stays below the Canadian average.
Point 3 is the major factor distinguishing C-12 from old Tory proposals. Point 3 is also why I still favour my own idea: it should be reasonably durable by slowing the rise of representational inequality (unlike current law and C-12), while not resulting in an exploding size of the House (unlike the old Tory proposal).
Part III
Now, let me discuss why I think C-12 would increase the size of the House by about 27 instead of the widely reported figure of 30. First, let's look at the number of seats C-12 implies, based on the most recent population estimates, which are for Jan. 1, 2010:
ON - 122 (121.6)
BC - 42 (41.6)
AB - 35 (34.4) (remember the rounding up rule under C-12)
The intuitive thing to do next would be to add 1-2% to each of these numbers in order to account for the likely population growth between now and the next census, in May 2011. That would bring us to 123-124 for ON, 43 for BC, and 35-36 for AB. However, as explained above, population estimates adjust for census undercounting, while the redistribution of seats does not. Thus, figures above should also be reduced for expected undercounting. By how much? Well, across Canada, the May 2006 census count was 2.7% lower than the April 2006 population estimate. Doing that adjustment brings us to the following numbers:
Projected Seats Under Bill C-12
ON - 120-121 (+14 or +15, 3 or 4 fewer than the government projection)
BC - 42 (+6, 1 fewer)
AB - 34-35 (+6 or +7, 1 or 2 more)
Canada - 334-336 (+26 to +28, 2 to 4 fewer)
The above numbers have a fair degree of uncertainty: the 2.7% undercount in 2006 was actually 4.2% in BC, but just 1.3% in AB. However, regional variations are likely due to inaccurate April 2006 population estimates rather than the census actually missing 3 times more people in BC than in Alberta. That's why I applied a uniform 2.7% rate, since it's anyone's guess in which direction the January 2010 estimates are biased.
Also, I cannot help but notice that it's politically convenient for the government to cite lower gains for Alberta and higher ones for Ontario. While it can defend itself for using 2005 population projections because those are the latest official ones, it certainly knows that the actual calculations will give fewer seats to Ontario and more to Alberta than announced.
By the way, under current law, I get the following:
ON - 108-109 (+2 or +3)
BC - 37-38 (+1 or +2)
AB - 31 (+3)
(The government's estimates are 110, 38 and 29.)
The government projects that Ontario would get 14 more seats under C-12 than under current law, BC 5 and AB 4. My calculations suggest that the actual numbers are about 12, 4-5 and 3-4. So viewed this way, the government is maybe not deliberately using a suboptimal methodology after all.
Just for fun, my proposal would give:
ON - 113-114
BC - 39
AB - 32-33
Part IV
Finally, let's look at the likely electoral effects of Bill C-12, relative to current law.
In a tight election between the Liberals and the Conservatives, the 12 extra Ontario seats would probably go 7L, 4C and 1NDP; the 4-5 extra in BC might go 2C, 1L and 1NDP; the 3-4 in AB would likely all go Tory. On the whole, the 20 seats split roughly 10C, 8L and 2NDP, so the Tories gain a small edge over the Liberals.
However, in an election like 2008 where the Conservatives are strong, and the Liberals weak, the splits of the extra seats would be something like 6C, 4L and 2NDP in ON; 3C, 1-2NDP and 1L in BC. Then you'd get 12-13C, 5L and 3-4NDP in the aggregate. Thus, C-12 facilitates the formation of a Conservative majority government.
What about an election where the Liberals are strong enough to flirt with a majority? The splits might then be 9L, 2C and 1NDP in ON; 2C, 1-2L and 1NDP in BC; 3C and 0-1L in AB. That's 11L, 7C and 2NDP. Thus, C-12 marginally facilitates the formation of a Liberal majority government.
Obviously, the reason why a majority is made easier for both parties is the decreased weight of Québec.
It'll be interesting to see what emerges in the end.
Part I - General discussion
Part II - Analysis of the bill
Part III - Estimates of seats after the next redistribution (these are likely to be more accurate than the government's, for reasons explained in bold in Part I)
Part IV - Likely electoral impact
Part I
Here is the text of Bill C-12. There are two changes with respect to the previous Tory proposal, which I analyzed here:
1. The size of districts in ON, BC and AB is no longer be the population of Québec divided by 75. Rather, it is set at 108,000 for the coming redistribution, and increases according to the rate of increase of the 10-province population for subsequent redistributions.
2. The seat numbers for ON, BC and AB (for the coming redistribution, population divided by 108,000) are rounded up, rather than to the closest integer. So 108,000 is the maximum district size.
How many seats would this law give to ON, BC and AB? CBC and most media, citing a government press release, say 124 (+18), 43 (+7) and 33 (+5) respectively (and usually without any acknowledgment of the underlying uncertainty). These estimates are based on the most recent population projections by Statistics Canada. Unfortunately, those projections are outdated (based on July 2005 population estimates), and correct for census undercounting (which we do not want to adjust for in this case, since raw census numbers are used for redistribution). Hence, the numbers above are inaccurate. I believe that those estimates are too high for ON and BC, and too low for Alberta, and outline my calculations in Part III.
Part II
Before going into the nitty-gritty of estimates, let me analyze the broad impacts of this law, as well as compare it to the previous version, and to my proposal:
1. It gives ON, BC and AB fewer seats than the previous Tory proposal, which would have given those three provinces representation in proportion to Québec's. This is because Québec's population in Census 2011 will likely be around 7.8 million, which is lower than 75*108,000 = 8.1 million. Thus, Québec will remain overrepresented with respect to ON, BC and AB.
2. However, it still gives those provinces more seats than under my proposal, which would have given Québec representation in accordance to its demographic weight in Canada. In order for Québec to be proportionately represented with 75 seats, the House size would have to be around 323 seats. Under current law, it is likely to be about 315 after the next redistribution; under C-12, it will likely be around 335. Thus, C-12 makes Québec go from being overrepresented to underrepresented with respect to the rest of Canada.
3. Assuming that Québec grows more slowly than Canada as a whole (not true lately due to the Ontario-concentrated recession, but likely to be true in the long run), the rate of growth of the size of the House is lowest under current law, higher under my proposal, and highest under the old Tory proposal. Under C-12, the rate of growth will be similar to the one under current law. Thus, while C-12 will increase the size of the House significantly in the next redistribution, it will lead to small increases in subsequent ones. This means that C-12 is a one-time patch, and in the long run:
a) AB, BC and ON will resume losing relative representation as fast as they do now if their population growth keeps outstripping the national average;
b) Québec will eventually once again become overrepresented if its growth stays below the Canadian average.
Point 3 is the major factor distinguishing C-12 from old Tory proposals. Point 3 is also why I still favour my own idea: it should be reasonably durable by slowing the rise of representational inequality (unlike current law and C-12), while not resulting in an exploding size of the House (unlike the old Tory proposal).
Part III
Now, let me discuss why I think C-12 would increase the size of the House by about 27 instead of the widely reported figure of 30. First, let's look at the number of seats C-12 implies, based on the most recent population estimates, which are for Jan. 1, 2010:
ON - 122 (121.6)
BC - 42 (41.6)
AB - 35 (34.4) (remember the rounding up rule under C-12)
The intuitive thing to do next would be to add 1-2% to each of these numbers in order to account for the likely population growth between now and the next census, in May 2011. That would bring us to 123-124 for ON, 43 for BC, and 35-36 for AB. However, as explained above, population estimates adjust for census undercounting, while the redistribution of seats does not. Thus, figures above should also be reduced for expected undercounting. By how much? Well, across Canada, the May 2006 census count was 2.7% lower than the April 2006 population estimate. Doing that adjustment brings us to the following numbers:
Projected Seats Under Bill C-12
ON - 120-121 (+14 or +15, 3 or 4 fewer than the government projection)
BC - 42 (+6, 1 fewer)
AB - 34-35 (+6 or +7, 1 or 2 more)
Canada - 334-336 (+26 to +28, 2 to 4 fewer)
The above numbers have a fair degree of uncertainty: the 2.7% undercount in 2006 was actually 4.2% in BC, but just 1.3% in AB. However, regional variations are likely due to inaccurate April 2006 population estimates rather than the census actually missing 3 times more people in BC than in Alberta. That's why I applied a uniform 2.7% rate, since it's anyone's guess in which direction the January 2010 estimates are biased.
Also, I cannot help but notice that it's politically convenient for the government to cite lower gains for Alberta and higher ones for Ontario. While it can defend itself for using 2005 population projections because those are the latest official ones, it certainly knows that the actual calculations will give fewer seats to Ontario and more to Alberta than announced.
By the way, under current law, I get the following:
ON - 108-109 (+2 or +3)
BC - 37-38 (+1 or +2)
AB - 31 (+3)
(The government's estimates are 110, 38 and 29.)
The government projects that Ontario would get 14 more seats under C-12 than under current law, BC 5 and AB 4. My calculations suggest that the actual numbers are about 12, 4-5 and 3-4. So viewed this way, the government is maybe not deliberately using a suboptimal methodology after all.
Just for fun, my proposal would give:
ON - 113-114
BC - 39
AB - 32-33
Part IV
Finally, let's look at the likely electoral effects of Bill C-12, relative to current law.
In a tight election between the Liberals and the Conservatives, the 12 extra Ontario seats would probably go 7L, 4C and 1NDP; the 4-5 extra in BC might go 2C, 1L and 1NDP; the 3-4 in AB would likely all go Tory. On the whole, the 20 seats split roughly 10C, 8L and 2NDP, so the Tories gain a small edge over the Liberals.
However, in an election like 2008 where the Conservatives are strong, and the Liberals weak, the splits of the extra seats would be something like 6C, 4L and 2NDP in ON; 3C, 1-2NDP and 1L in BC. Then you'd get 12-13C, 5L and 3-4NDP in the aggregate. Thus, C-12 facilitates the formation of a Conservative majority government.
What about an election where the Liberals are strong enough to flirt with a majority? The splits might then be 9L, 2C and 1NDP in ON; 2C, 1-2L and 1NDP in BC; 3C and 0-1L in AB. That's 11L, 7C and 2NDP. Thus, C-12 marginally facilitates the formation of a Liberal majority government.
Obviously, the reason why a majority is made easier for both parties is the decreased weight of Québec.
It'll be interesting to see what emerges in the end.
Subscribe to:
Posts (Atom)