October 21st 2013: Forum poll shows Liberals are up for some gains in by-elections

Four federal by-elections will take place on November 25th. The four ridings are Toronto-Centre, Bourassa, Provencher and Brandon Souris (in Manitoba). During the last federal election, these ridings elected a Liberal in Toronto and Bourassa, and a Conservative in the the Manitoban riding (all four ridings were won comfortably).

Forum just released four polls in these ridings. As usual with riding level polls, the sample size is quite small, varying from 287 to 855 observations. If you ask me, it's almost ridiculous to publish a poll with 287 observations (in Provencher), but I guess it's better than nothing. To put that in perspective, it means the 95% confidence intervall for the Liberals in Provencher is 23.7%-34.3%.

Here are the findings of the poll. In bracket, I put the results of the last election as well as the latest projections based on an Ipsos poll.


CPC: 18% (22.6%; 18%)
LPC: 45% (41%; 49%)
NDP: 30% (30.2%; 29%)
Green: 7% (5%; 4%)

As we can see, the Liberals are up a little bit but overall, all the parties are pretty much constant. For the Liberals, it's actually not very good news as they should be higher. Or, if you prefer, it might means the Liberals aren't up that much provincially in Ontario. More about this below.


CPC: 7% (8%; 7%)
LPC: 47% (40.9%; 51%)
NDP: 18% (32.3%; 20%)
Green: 12% (1.6%; 0%)
Bloc: 15% (16.1%; 21%)

So here, the Liberals are up, the NDP is down. The Green are also up a lot, probably thanks to their star candidate and former NHL player, George Laracque. Compared to the projections, the Bloc is down and again, the Green are up a lot (In the Ipsos poll, the Green party was at 1% in Quebec. Projected in this riding, I had the Green party at 0%. So George Laracque seems to benefit from quite a sizable bonus here). Overall, and it we neglect the Green for a moment, it shows what I've been saying many times: Trudeau did improves the Liberals in Quebec, but mostly by taking votes from the NDP (in Montreal especially).


CPC: 56% (70.6%; 57%)
LPC: 29% (6.7%; 18%)
NDP: 9% (17.9%; 18%)
Green: 6% (3%; 7%)

Here we clearly see the Liberals again taking some votes from the NDP. I understand that it looks like good news for the LPC, but if I were them, I'd be a little bit worried about the inability to take some votes away from the Conservatives. After all, if the Liberals want to win the next election, they can't simply take back their votes lost to the NDP in 2012.


CPC: 35% (63.7%; 49%)
LPC: 39% (5,4%; 16%)
NDP: 12% (25.2%; 25%)
Green: 12% (5.7%; 10%)

This riding is more interesting because of the controversy about the Conservative candidate and some possible interferences from the party. It does indeed seem that the CPC might be losing a riding that should be a safe bet. And again, the NDP is down more than projected.

While these numbers are interesting on their own, we need to remember that by-elections are more difficult to predict. The turnout is usually a lot lower and who actually get out to vote are different from the general voters.

For me, the real interest in these numbers is to compare them to the projections that are based on provinvial polling. In Manitoba, we see the Liberals up much more than what the provincial polls would lead us to believe. This suggests that in this province, the Liberals might actually be much higher than in the Ipsos poll. Specifically, in order to replicate the results (you can try in the simulator by yourself), we need to input the Liberals at around 35-40% in the Prairies (in the model, Manitoba and Saskatchewan are combined, as in most polls). On other hand, these riding numbers would indicate the NDP is actually closer to 17-20% in this region. If this was true, the Liberals' chances of winning the next election would be greatly improved as they'd grab quite a lot of seats in this region (and that's before accounting for the new map and seats).

So good news for Trudeau and his party in Manitoba for sure (again, keeping in mind the small sample size as well as the potential other effects in Brandon-Souris). But in Quebec and Ontario, the numbers don't look that good. In Quebec, it probably means the Liberals are around 25%, behind both the Bloc and the NDP. In Ontario, the Grits are likely around 35%. A big improvement over 2012, but not enough to be first.

Of course, we need to account for the possible effect of losing long time star candidates for the Liberals. Specifically, in Bourassa, they had Denis Coderre (who will likely become the next mayor of Montreal). In Toronto-Center, it was Bob Rae. The model currently doesn't account for that. But let's say that losing these two candidates had an adverse effect on the Liberal vote in these ridings, then the Forum numbers could be seen as really good news. The problem here is of course to estimate the "Coderre and Rae" effects. Moreover, it's not like the Liberals aren't running strong candidates this time around.

Make no mistake, I might sound hard on the Liberals, but these numbers are good news overall. It's just that for the two big provinces, they are in line with what we could expect based on most recent polls. And these polls not only showed the Liberals down from the highs reached after the arrival of Justin trudeau, they also had the Conservatives as being favorites to win the next election.

Overall, the Forum polls show that the recent federal polls were probably right in Quebec and Ontario, but not necessarily in Manitoba. At this point though, it's impossible to conclude anything based on numbers with so much uncertainty.