The 2019 federal election wasn't particularly exciting with very little vision offered. That was particularly the case with the main two parties that had quite boring platforms. As a result, the campaign was mostly marked by scandals and tangentially related topics such as SNC-Lavalin or Doug Ford.
Polls don't usually allow us to measure the impact of one variable or event on the voting intentions. They can show us that, for instance, people didn't like Doug Ford, but they almost never provide any measure of how much it impacted the voting intentions.
Using the CES (Canadian Election Study) of 2019, I decided to try to answer some of those questions. This is the first post in a series. Today's topic: did Doug Ford cost Andrew Scheer the job of PM? And did SNC-Lavalin cost Trudeau his majority? I use the online sample that contained all the information I needed. I estimated the regressions below . My two variables of interest were, naturally, "premiergood" which is equal to 1 if the respondent expressed satisfaction with their provincial government (either "very satisfied" or "satisfied) and "sncgood" which measured the handling of the SNC-Lavalin story (again, either "Very well" or "well").
I naturally needed to control for many other variables susceptible to influence voting intentions. I included the usual (age, gender, having children, education, income), the satisfaction with the federal government as well as the ratings given to the LPC, CPC, NDP and Green and their respective leaders. There is also one variable measuring the left-right orientation of the respondent (self identified). I believe this is, overall, a pretty standard and robust specification. Would it pass peer-review for publications? Likely not. But as a simple test for a blog post? Sure, that'll do.
I estimated the regression for Ontario only, here is the table in Stata.
As you can see, for the Liberals, a few variables mattered. Income and university education are both significantly correlated with voting Liberals. Being satisfied with the federal government is by far the strongest impact (a whopping 18.5 percentage points!; Yes I'm aware I should be using a Logit or Probit here given my dependent variable, but I want to be able to have easily interpretable coefficients), so was giving a high rating to the Liberal party.
Being satisfied with the government of Doug Ford is associated with a 6 points drop in voting intentions for the Liberals. On the other hand, people who thought Trudeau handled the SNC-Lavalin story well were voting for him 6 points more.
Estimating the same regression for the Conservatives gave the following results:
Being satisfied with the Ford government increased voting intentions for the CPC by 12 points! the SNC-Lavalin affair had the same impact as for the Liberals (just in the opposite direction obviously).
The variable cps19_lr_scale_bef measures the left-right orientation. So right-wing people were more likely to vote CPC. Interestingly, this is not a significant variable for the Liberals or NDP (note: I tweeted yesterday that it was significant for the NDP and not the other two. That was indeed the case with a different regression specification, a simpler one)
Finally, here is for the NDP:
Interesting to see the negative and significant impact of being university educated. You'd think the NDP would do better among this demographic but it really seems that Trudeau is too strong among this group. As for being satisfied with the Ford government, the effect was -6 points and -5 points for SNC (the impact of this variable is very consistent across parties).
So, both stories mattered. What would have happened if the Ford government wasn't that unpopular? Looking across the provinces, the percentage of people satisfied was only 27%! Compared to an average of 54% in the other provinces. Had the Ford government been as popular as the other provincial governments, the coefficients above indicate that the results in Ontario could have been (actual results in bracket):
LPC: 40% (41.3)
CPC: 36.5% (33.1)
NDP: 15.2%
Essentially, the Conservatives would have been 3.4 pts higher, taken from the Liberals and NDP. What impact in terms of seats? Hard to say but a quick look with my model would indicate that it could have cost the Liberals between 5 and 10 seats. Not enough to win the GTA or for Scheer to become Premier, but not insignificant either. The impact could be higher if the people not satisfied with Ford were concentrated in the GTA. The CES dataset doesn't allow me to look into that. But it does partially explain why the Tories increased everywhere in Canada except in Quebec (unique situation) and Ontario where they literally did worse, in percentages, than in 2015.
On the other hand, only 15% thought that Trudeau handled SNC-Lavalin well. That likely cost him as much as 5 points! It is quite massive and likely cost him a majority.
Both those estimates are very, very rough and should be taken with a large grain of salt. But I believe they showed that both had a significant impact on the vote in Ontario. Does it mean that Trudeau can now get a majority if people have forgotten SNC-Lavalin? Or that O'Toole can expect gains in Ontario now that Ford is, surprisingly, popular? I wouldn't say so, this is one causal link too far for such a basic analysis.
