The NDP continues to dominate the federal voting intentions as well as my seat projections. It has now lasted for a while, so I think it's safe to say the NDP bump isn't simply a momentary thing. It doesn't mean it will last until the next October of course.

Using the four polls published this month (one Ekos, two Forum and one Abacus), the NDP is ahead of the Conservatives while the Liberals trail as third. If the election was tomorrow, the NDP would clearly be favorite. I've seen other sites pretending it's a 50-50 race at the moment but this is nonsense. The only way you get a 50-50 race is if you take 2 points away from the NDP and give them to the CPC in every single province. While I do think the CPC could be underestimated by the polls and I fully intend on accounting for this during the election, we are still in a situation where the NDP can actually afford the luxury of being overestimated in the polls and still win!

The projections using these four polls are as follow:

Voting intentions; Seat projections with 95% confidence intervals; Chances of winning most seats

The detailed, riding by riding projections can be found here.

Right now, the NDP clearly dominates in Quebec (with a lesser margin than in 2011 but still sufficient to win the most seats by far) as well as BC. Ontario is really the last major province still fighting Thomas Mulcair. It's a close three way race as far as the votes are concerned, but the seats simply aren't there for the NDP. It's an improvement over 2011 for sure, but the seat share is still below the vote share. Without a breakthrough in Canada's largest province, it's hard to imagine a path to a majority for the New Democrats. At the same time, and as has been the case for a while, no party is in a position to get 169 seats or more.

I realize the election is still a long time away and things can change. However, while my probabilities aren't meant to measure the uncertainty over time (it's there to account for the uncertainty of the polls as well as of the electoral system), we can kinda use them for such an interpretation. At the very least, it does give us an idea of the uncertainty that exists. Specifically, the simulations do put each party above and below their current average. For the NDP, it means that some simulations have this party as low as 27% for instance. Moreover, if we look at the riding level, we can see how many seats are currently safe for each party.

The graph below shows, for each party, how many ridings currently fall in each probability category. For instance, the 1-10% bin shows the number of ridings where one party has between 1 and 10% chances of winning.

The NDP has 28 ridings with absolute certainty of winning (again, if the election was tomorrow). It doesn't mean these districts won't elect a MP from a different party in October, but it'll take major changes for this to happen. As a reminder, the model has not made a mistake yet in the ridings with 100% (and 0%) probabilities. The CPC has 33 and the Liberals only 10. While the Tories have the highest number of safe seats, the NDP ultimately would win here thanks to a much higher number of very likely ridings (at 90% and plus).

Notice also that the Liberals are the party with the highest number of ridings with no chance at all. 130 ridings are in this situation for Trudeau. It's 125 for Harper and his party and only 69 for the NDP. Thus, the NDP is currently the party that is truly in the race everywhere. Sure, their MPs would mostly come from Quebec and BC, but it'd wrong to say the NDP has a very regionalised support. If anything, it's because its support is so widespread that the NDP isn't winning more seats. Again, the vote efficiency in Ontario for instance just isn't there.

At the end of the day, the current poll average clearly shows that the NDP is ahead. Quebec and BC would both provide a significant number of seats to Thomas Mulcair. The key now for him is to find a way to become the second or first party in Ontario. Make no mistake however, the Conservatives are still in this race and it wouldn't take much for Harper to win another time (with a minority though). As for Trudeau, the situation doesn't seem to improve but it's also not getting worse. Chances of winning the most seats are low but they aren't zero either. Moving from 3rd to 1st is never easy though as you necessarily need to leapfrog two parties.

Un billet sur l'effet du retour de Duceppe sur le Bloc Québécois. Pour le compte de l'Actualité. Donc voici le lien.

http://www.lactualite.com/actualites/politique/analyse-des-sondages-leffet-duceppe-sur-le-bloc-quebecois/

Electoral reform isn't a new topic in this country. Multiple provinces have tried to change their system (BC, Ontario, etc) but none has actually succeeded. At the federal level, the NDP and Green were the only one talking about this issue for the longest time. Things got a little bit more interested recently when Justin Trudeau announced that a Liberal government would change the system (among many other things).

I should mention that I'm one of the biggest fans of electoral reform (even though a proportional system would greatly diminish the usefulness of my model). I find the current system archaic and not adapted to a modern democracy and I'd favor switching to a proportional system (such as the German system or the STV that was proposed in BC). I find the idea of getting a majority with less than 40% of the vote crazy. So is winning more seats with less votes.

Trudeau said he's opened to multiple ideas but currently prefers Preferential Voting (also known as Alternative Vote) where people would rank their choices and if no candidate receives a majority (more than 50%) of the votes, we eliminate the last placed candidate and redistribute his or her votes. We keep doing that until one candidate has reached a majority.

Such a system tends to favor parties at the center, the type of parties who are good at getting second choices from other voters. This is the system that was proposed in the UK but was defeated in a referendum.

I decided to look at how this would change my current projections. Here's how I proceeded:

- I used the latest poll by Forum in which the main three parties are within 5 points of each other (something that is in line with the average of recent polls). I chose this poll because it's a recent one and it includes questions about the second choice.

- Now I completely realize these second choices could not represent how people would rank candidates on their ballot. A lot of details would be required here (do we force people to rank all candidates? etc). A new system could also change the voting intentions themselves (for instance, you could now support the Green candidate first, knowing your vote won't be "wasted" since it'll be redistributed if the Green doesn't win). Nevertheless, this is pretty much the best we can do right now.

- Forum provides these second choices separately by age, region and parties. The last one is the most important for me as this is how I'll redistribute the votes of the candidates finishing last. However, a couple of issues arise. First of all, while only 8% of NDP voters have the Bloc as second choice, we need to remember these 8% are all in Quebec. So we need to do some calculations and assumptions here. Doing so reveals that in Quebec alone, the Bloc is actually the second choice of 30% of NDP voters. Secondly, the second choices could vary by party and region at the same time (let's say Liberals in NB have the NDP as second choice but the Liberals in Alberta have the CPC - I'm making this up but you see what I mean). I unfortunately can't really anything about this here and I'll assume the second choices to be constant in every province (except Quebec of course).

-Specifically, here are the second choices for the rest of Canada and Quebec (row is the first choice, column is the second; Number are %) after adjusting for the fact that the Bloc only gets second choices in Quebec:

Outside Quebec:

	CPC	LPC	NDP	Green
CPC	0.0	14.3	8.7	8.6
LPC	22.7	0.0	53.3	26.9
NDP	22.7	61.2	0.0	26.9
Green	9.3	11.2	22.8	0.0
None	45.4	27.6	23.9	46.2

And in Quebec

	CPC	LPC	NDP	Green	Bloc
CPC	0.0	12.8	6.0	5.5	8.0
LPC	18.7	0.0	37.1	17.0	19.0
NDP	18.7	54.9	0.0	17.0	49.0
Green	7.6	10.1	15.9	0.0	11.0
Bloc	17.7	10.3	30.4	36.6	0.0
none	37.4	11.9	10.6	23.9	13.0

- The Forum poll didn't exactly offer the option "no second choice". We only have "other parties" and "undecided". For the purpose of this post, I treated both as people without a second choice and therefore they'd simply not vote if their first choice isn't elected.  This is an important part of this exercise as it's perfectly sensible for some voters to prefer not to vote if their first choice is eliminated (although you do have systems in some countries where we force people to rank all candidates). Looking at the numbers, we can see the Conservatives voters are the ones with the highest numbers of no second choice. The NDP is the top second choice of absolutely all parties, including the Conservatives (well, again, excluding the voters with no second choice). As usual, this is only one poll, but this is consistent with what I've seen from other companies. On the other hand, the Tories are doing terrible with pretty much every other party, including the Liberals (only 14% of Liberals have the CPC as second choice outside of Quebec).

- I then made projections using the Forum poll. You can see the results yourself by entering the numbers into the simulator. I then started redistributing the votes of the candidates who finished last in ridings where nobody got more than 50%. I went step by step, eliminating one candidate at the time. This is especially important in Quebec where I often had to eliminate 2 or even 3 candidates before finally having one candidate winning a majority (necessarily, since there were only two candidates left).

- How much of a difference would this system make? The results are below.

With Forum poll
	CPC	LPC	NDP	Green	Bloc
Current system	96	103	131	1	7
Alternative Vote	49	129	155	1	4
Number of winners changed			51

If you want the detailed results, you can find a (ugly) pdf here. The first set of number are the projections under our current system, the second set shows the results after redistribution (when needed).

-As you can see, both the NDP and the Liberals would benefit greatly while the Tories would drop almost by half. Given that the Conservatives are the only right wing party in this country, it's not very surprising. Again, the Alternative Vote system rewards parties that are good at getting second choices. 51 ridings have a different winner under AV compared to FPTP.

- Obviously, such a simulation depends on the voting intentions. I thus decided to re-do the exercise using a recent Ekos poll that also contained information about second choices. However, in this Ekos poll, the Tories were doing better (actually projected to win the most seats) and the second choices were also slightly different (you still have the same patterns though, with the NDP raking up second choices and CPC voters preferring not to vote). Here are the results using this Ekos poll:

With Ekos poll
	CPC	LPC	NDP	Green	Bloc
Current system	121	87	113	1	16
Alternative Vote	92	95	147	1	3
Number of winners changed			46

- This time around, the Conservatives suffer less. They still are the clear loser of such an electoral reform (along with the Bloc). What this shows is that there is considerable uncertainty to simulate the results of such a change. Nevertheless, it's pretty clear the Conservatives wouldn't benefit from a reform of the electoral system, whether it's preferential voting or proportional representation (not shown here but they wouldn't have won a majority in 2011).

- I want to say that I'm really not a big fan of this system and I really wish that if we finally implement electoral reform, we actually choose proportional representation. Preferential ballot is actually worse for small parties and as you can see here, can actually make single party majorities more likely (the NDP is getting really close to 169 here despite only having initially 32% of the votes!). The fact the Alternative Vote can actually give results with even bigger distortions than FPTP (something noted in the UK) should really lowers expectations as to the benefits of such a reform. But that's beyond the point of this post.

- I'll try to re-run these exercises for past election results. I'm fairly confident the Conservatives will always be the one suffering of such a reform. But it'll be interesting nonetheless.