1. The Green party can attract Liberals voters
Let's look at 2013. The Green technically got a smaller share of votes than in 2009 (8.13% vs 8.21%), but this is highly misleading because they were running fewer candidates - only 61 out of 85, while they had a full slate in 2009. This is important because it means the Green swing was actually positive in many ridings.
In the 24 ridings without a Green candidate in 2009, the Green got 31,358 votes. That's about 23% of all the votes received by the Green party that year. You should see the importance of taking that into account while estimating coefficients or looking at the results. If the Green managed to stay around 8% province-wide while losing 23% of their votes in some ridings, it must mean this party went up in the other electoral districts, significantly in many places
Without the votes in these 24 ridings, the Green would have been at only 6.3% province-wide in 2009. So when they got 8.13% in 2013, it actually means an average swing of almost 2 points!
The big question here is obviously: what happened to these votes in 2013? Did the people who voted Green in 2009 in these 24 ridings simply didn't vote? Did they vote for another party? And this is where I become useful and where statistics and regressions can be used. Finding the answer to this question can give us some indications as to who these Green voters are and which other party they could support.
I tried to look at it with different approaches First of all, I looked at the results of 2013 in each riding and tried to explain the share of votes of the Liberals and NDP with the share of votes of each party in 2009. Doing so revealed that the share of a Liberal candidate in 2013 was highly correlated with the share in this riding 4 years ago, that's logical. It's the same for the NDP. But for the share of votes of the Green, I found significant effects on the shares for the Liberals but not the NDP. Specifically, my estimations indicate that about half the votes for the Green in these 24 ridings went to the Liberals in 2013 while almost none went to the NDP. There are a lot of potential issues with my method here, I'm fully aware of it -I've done enough econometrics in my life to be aware of that- but it's still interesting.
Then I tried to use variations instead (so the swing in each riding between 2009 to 2013). There as well the results showed me that the BC Liberals got about half the votes while the NDP didn't get anything. This is very significant because this explains, at least partially, while the BC Liberals increased in some ridings/regions while dropping overall province-wide. One way to make sense of this is to realize that the BC Liberals maybe received some votes from the Green voters in ridings where the Green didn't have a candidate anymore -in the Okanagan valley for instance, while the Liberals dropped more in other ridings/regions. In ridings where there was a Green candidate both in 2009 and 2013, the Liberals dropped much more than in the ridings without a Green in 2013 (where they actually increased in average!).
Again, the econometrics is very limited but the fact that I got the same results with both methods is encouraging.
Again, the econometrics is very limited but the fact that I got the same results with both methods is encouraging.
In average, it appears that the missing Green candidates gave a bonus to the Liberals of 3.7 points (again, in the 24 ridings).
This can seem very counter-intuitive. Most people usually assume that Green voters have the NDP as second choice. But polls have shown Green voters to be more divided than that. The final Mustel poll in 2009 was showing exactly that: Green voters were split regarding their second choices.
Obviously the current polls are showing a very different situation. Ipsos Reid shows that 42% of Green voters have the NDP as their second choice while only 14% have the Liberals. The latest Mainstreet polls even has the NDP as the second choice of 74%(!) of Green voters -a number much higher than in previous polls from the same firm though. On the other hand, many Mainstreet polls have shown the Liberals voters with the Green as their main second choice.
What this should at least shows is that Green voters don't simply come from or go to the NDP. The communicating vases are more complex than that. The simple fact that current polls show the NDP at its 2013 level while the Liberals are down should be another evidence of that.
Another example of this is Oak Bay-Gordon Head, the riding of the Green leader, Andrew Weaver. He won his seat in 2013 thanks to an incredible personal effect. My estimations show that his result was 27 points above what a "generic" Green candidate would have been expected to receive. It was obviously thanks to him and the incredible campaign the party ran for him there. Still, as far as I can tell, this is the biggest personal effect I've seen, bigger than the one for Elizabeth May.
Yet, the same estimations show this effect was taken equally from the Liberals and the NDP, both losing between 13-15 points. And remember this effect is estimated while also including regional effects. In Victoria Beacon-Hill where former Green leader Jane Sterk was running, I also find that her personal effect (around 12 points) was taken much more from the Liberals than the NDP.
Of course, there can be many explanations. For instance, it's possible that BC Liberals voters from the Island are more likely to vote Green than say Green voters in the Interior. Or, in the case of Victoria-Beacon Hill, some Liberals might have tried to make the NDP lose this safe seat by voting Green. But again, it at least shows that the Green Party seems capable of attracting votes from both parties.
2. Implications for the projections this year
There are mostly two elements to this part. The first one is the inverse of the first part of this post: what will happen in the ridings where the Green are now running a candidate and they weren't 4 years ago? Surely the swing there will be different from the swing elsewhere. Will we observe a "catching up" effect where the results in 2017 would be more along the line of "what should have happened in 2013+swing between 2013-2017"? If that was the case -and looking at what happened to the BC Conservatives in 2009 in ridings with new candidates, we have reasons to believe we'll indeed observe that- then it means the provincial swing for the Green party will be misleading again.
The Green Party is running 83 candidates this year (out of 87 ridings). That's a significant increase in the number of candidates. So when we see the Green being polled 10-12 points above their score of 2013, some of this increase will likely only come from the ridings with a new candidate. This could be fairly significant for the chances of the Green to get MLAs this year. If one is making projections only looking at the provincial swing, this will likely overestimate the swing in ridings where there was already a candidate -and those ridings are the ones the Green can win. I admit I haven't fully accounted for this yet and I'll do so by the end of the week. Make no mistake, this is a lot of work and it requires some assumptions.
The second part: in the ridings with a new Green candidate, where will the vote come from? Could we see the opposite of what we saw between 2009-2013? Meaning the Liberals could be the party most affected by these new candidates. There as well, I'll try to adjust the model by the end of the week.
Conclusions
There are really two things to take away from this blog post. First of all, the Green Party is likely attracting voters from both parties. It is indeed taking seats from the BC NDP on the Islands, but the voters aren't all coming from the NDP.
Second of all, the fact the Green party is running more candidates than last time likely means projections could overestimate the Green swing in ridings where they already had a candidate. Given that every single seat the Green are aiming for is in this situation, you can see the potential for making mistakes.
