The BC elections was technically on Saturday October 24th. But with over half a millions of mail-in ballots (representing almost a third of the total votes) not counted for at least another week (don't ask why, BC law is weird), I thought I could try to update the 'projections'.


The final projections did very well. Let's face it, it wasn't a hard election to predict. The biggest surprise was the Green winning a third seat in West Vancouver. Otherwise everything was pretty much as expected with maybe the exception that the BC NDP won the popular vote by 'only' 10 points. This is significantly below the projected 49% in my final numbers. It's possible the polls were overestimating the NDP (and underestimating the Green) but I believe this is mostly due to the mail-in ballots not being counted. I actually believe that once they are, the polls will prove to have been accurate for the NDP but to have overestimated the BC Liberals (most likely due to many Liberal voters staying home).


Update: Election BC has released the number of mail-in ballots per riding and it seems there were, in total, over 644k ballots! Way more than we thought. Looking at my estimates below, you can likely add 1-3 ridings flipping NDP. Langara is now flipping based on my method 3. Fraser-Nicola and Kamloops-North Thomson are very close, so is Surrey South. So best case scenario is now 62 for the NDP but average scenario is still 'only' 59.


Here a quick summary:


- All evidence point to mail-in ballots being very favourable to the NDP.

- The exact magnitude of the impact depends on many assumptions, but it's safe to bet the NDP will ultimately flip a few more seats and finish with more than 55 seats.

- The best estimate is for the NDP to finish with 59 seats, the Liberals with only 25 and the Green to keep the 3 from election night.

- Abbotsford-Mission, Surrey-White Rock and Vernon-Monashee are likely to flip while Vancouver Langara, Fraser-Nicola and Kamloops-North Thomson are more uncertain (the last two mostly because of the relatively low number of mail-in ballots).




1. Estimating the mail-in ballots


We know that Elections BC has received 525,000 mail-in ballots (out of 724,000 requested). Given that 546.877 voted on election day and 670,324 voted in advance (plus 85,000 absentee ballots), the mail-in ballots represent almost 30% of the total. This is a significant number and more than enough to change the popular vote province-wide as well as flipping some seats.


You probably heard that those ballots will help the BC NDP. How do we know that? Mostly because some polls gave us this indication. Specifically, Ipsos and Insights West both had numbers broken for mail-in, advance and election day. Mainstreet and Angus-Reid only had the early (mail-in + advance) vs same day but it's still providing us with some info.


First, let's compare the early votes to the election day one. Early being the mail-in and advance votes.


As you can see, they mostly all agree (well Ipsos a little bit less). The early vote is likely more favourable to the NDP than the votes on e-day. The issue for us is naturally that the advance votes were counted but not the mail-in, so 'early' isn't super useful to us.


Let's look at the mail-in ballots:


So the NDP is projected between 54 and 59% in those mail-in ballots, an almost 30-points lead over the Liberals!


Maybe a better approach is to compare the ratio of (the voting intentions of) mail-in to election day (advance + same day). Doing a ratio can potentially eliminate the error of some pollsters (a poll might have been over or underestimating one party but as long as the bias is the same, doing a ratio will work).


Using a few assumptions, I also estimated this ratio for Mainstreet and Angus-Reid. What is remarkable is how similar those ratios are across firms. So, all in all, I am quite confident ins saying that the NDP will increase its share of votes once we count the mail-in ballots. By how much? My best estimates is that the NDP will finish around 48%, the Liberals around 32-33% and the Green at 15-16%.


2. How many seats will flip?


Method 1: use the newly estimated shares of votes, plug them in the model and see what happens. This is the least sophisticated method but let's start here.


Doing so gives us the following:

BC NDP: 61 seats

BC Liberals: 24

BC Green: 2


Let's improve those with some of the results we already have. For instance the model was clearly underestimating the Green in West Vancouver. The model was also overestimating the NDP in Cariboo-North (or the Interior in general). Skeena also went better for the Liberals than expected. So if we fix those errors (as well as Columbia River-Rivelstoke), a better projection would be:


BC NDP: 59

BC Liberals: 25

BC Green: 3


Compared to the current results, Vancouver Langara is now NDP (well it was already NDP in my original projections), so is Fraser Nicola. Vernon also flipped so did Kamloops-North Thompson. Abbotsford-Mission didn't flip, neither did Surrey-White Rock.


Method 2: apply the province-wide swing to each riding


The swing caused by mail-in ballots is about +3% for the NDP, -2-3% for the Liberals. If we do this in every riding (using the current results as base), we get 6 ridings to flip Liberals to NDP (Vancouver Langara, Kamloops-North Thompson, Fraser-Nicola, Abbotsford Mission, Surrey-White Rock and Vernon), so:

NDP: 61

Liberals: 23

Green: 3


Similar results as the first method. The caveat is that the swing caused by mail-in ballots won't be uniform, especially since some ridings have a much higher share of mail-in ballots.


Method 3: Use the estimated ratio and apply it to every riding, taking into account how many mail-in ballots there are.


From the ratios above, we know the mail-in ballots should be, in average, more favourable to the NDP. Specifically, if the NDP was at 10%, it should be at 10%*1.24=12.4% in the mail in ballots. Assuming the percentage of returned mail-in ballots is the same in every riding as it is at the provincial level (so 525k/724k=72%), we can thus estimate the swing in every riding.


Let's use Abbotsford Mission as our example. The current results are 38.3% for the NDP and 39.3% for the Liberals. That's out of 17,983 votes in total. We also know there were 8,119 mail-in ballot requests in this riding. If 72% returned their ballots, it means an extra 5,846 votes to count.


Since the NDP was at 38.3% with the advance+same day results, this party should be at 38.3%*1.24=47.5% in the mail-in ballots. Given there are 5,846 of them, that's an extra 2,776 votes for the NDP. Doing similar calculations for the Liberals, we get an extra 1,677votes. Given that the Liberal candidate had a lead of 188 votes, the mail-in ballots easily allow the NDP candidates to win with an overall score of 40.6% of the vote compared to 36.7%. So we have NDP climbing by 2 points and the Liberals dropping by 3 for a total swing of 5, very similar to the other method (it was expected).


Using this method, we get the following seats to flip: Abbotsford Mission, Vernon-Monashee, Surrey-White Rock and Vancouver Langara.


Fraser-Nicola is not flipping and the Liberals kept the seat with 40.6% versus 40.1%! Mostly due to a low number of mail-in ballots. Kamloops-North Thomson would also remain Liberals by the smallest margin (less than 1%). We get a few 'odd' results with this method as we would expect Fraser-Nicola, where the BC Lib have a lead of 3%, to flip before Vancouver Langara (lead of 5%). But the mail-in ballots represent 38% in Langara but only 14% in Fraser-Nicola. So the seat doesn't flip.


Edit: Surrey South might also flip given the really high number of mail-in ballots. Unlikely but not impossible.


Overall, the results would be:


NDP: 59 seats

Liberals: 25

Green: 3


I believe this is the most realistic results. While the NDP might flip 1 or 2 more, we also need to remember that I have so far only looked at ridings that could flip from Liberals to NDP. The polls really don't seem to indicate this as a possibility but we can't ignore a few odd ridings out there. While it's reasonable to assume the overall mail-in ballots will help the NDP, we can't exclude that some ridings will see mail-in ballots more favourable to the Liberals. Richmond-South Centre for instance could well flip NDP to Liberals. Similarly, Michael Lee in Vancouver Langara has so far done better than projected. So it's not unreasonable to expect him to also do better among the mail-in ballots and ultimately saving his seats. So we could have Langara remaining red but Fraser-Nicola flipping, thus cancelling each other out. That's why 59-25-3 is the most likely scenario.


Conclusion


It's almost guaranteed that the mail-in ballots will advantage the NDP over the Liberals. We can expect this party to increase its popular votes by anywhere from 1 to 4% while the Liberals will likely end up below 33%. The Green should move much.


Regarding seats, there are really 6 ridings to watch: Vancouver Langara, Surrey-White Rock, Abbotsford Mission, Fraser-Nicola, Vernon-Monashee and Kamloops-North Thomson. The NDP should be able to flip anywhere from 3 to 6 of those.

Unless we observe one of the worst polling failures in Canadian history, John Horgan will have won his gamble of triggering a snap election in the middle of a pandemic in the hope of getting a majority. After 3 years of governing thanks to the Green, he'll now have full control for the next 4 years.


Yes I'm that confident. Some people will remind us of BC 2013 when polls were wrong but there are many reasons not to expect a repeat (more on that below). The real uncertainty is whether the NDP will get a majority or a very large majority. There are naturally a few close races here and there but the overall situation is very clear.


For the busy readers, here are the projections.

Adjusted polling average; Seat projections with 95% confidence intervals; Chances of winning the most seats.


The riding by riding projections are here:

BC Final Projections 2020 by bryanbreguet

Just a reminder that I do NOT simply aggregate the polls. In particular, I allocate undecided differently. Instead of allocating them proportionally (which is, in my opinion, a bad assumption since it comes down to either saying all undecided voters will vote like decided ones or, alternatively, that none of them will vote), I allocated 50% to each of the NDP (incumbent) and Liberals (other big party). None for the Green and Conservatives. This method has worked very well, in particular to correct the typical overestimation of smaller parties by the polls. I have also adjusted the Conservatives down due to only running 19 candidates out of 87 ridings. Their vote is redistributed mostly to the Liberals but also to the other parties.


Chances of winning come from simulations where I randomize voting intentions (to account for the uncertainty of the polls) as well as at the riding level (to account for the distribution of the votes). Such simulations have proven to be very accurate and the model is calibrated.


Reminder that you can make your own projections by using my model here. It asks you to enter province-wide numbers but is using regional adjustments based on the polls.


1. The polls and the campaign


It hasn't been a particularly exciting campaign. Polls, as shown below, have been very stable. It makes sense, if people don't care much and are happy with the government, they won't really look for an alternative. I think it would always have been difficult for the BC Liberals to make a convincing case to change government. The fact that Andrew Wilkinson was absolutely terrible during the campaign (multiple scandals he couldn't deal with promptly, lack of charisma, etc; Even the Liberals supporters don't seem to like him with only 32% of them declaring that Wilkinson has ran the best campaign according to the latest poll from Leger.) obviously didn't make things better.


If people were expecting a tightening towards the end, they must be very disappointed. Most polls published today (Research Co, Mainstreet, Ipsos) had the NDP above 50%! The exceptions being Leger where the NDP was at 'only' 47% and a weird, last minute poll from Forum where the NDP was at only 44% and the Green at 18% (as well as projected to win 4 seats)! With possibly more than half the votes already cast, one might wonder if the projections should be based on a weighted average across the campaign. Thankfully the polls have been stable enough that it doesn't matter much. I do believe the overall accuracy of the polls will depend on the turnout tomorrow, Saturday 24th.


I think the best indication of the low interest is Google Trends. Usually, we'd see a trend for all parties and leaders as we get closer to election day. There is one but it's weak, with the only real spike being around the leaders debate. On top of that, many people chose to vote by mail (over 700k ballots requested), so they likely voted early and tuned out.



It was even more flat if we use the leaders. We saw a somewhat similar situation during the NB election. I really believe people aren't in the mood for an election and, as long as the government is okay, they'll just reelect it. It'd be interesting to see a federal election soon as Trudeau, while ahead, is nowhere as popular as Horgan was. So that might generate a more interesting campaign. Maybe.


To the Liberals credit, they did have a platform that 'made sense'. They tried to focus on jobs and the economy and had interesting proposals (PST, ICBC, childcare). But this wasn't going to be an issues-focused campaign. It was more of a referendum on Horgan and his government and people have been happy with it for a while. A much better leader for the Liberals would have limited the NDP to a smaller majority though. The PST promise was also maybe just 'too much'. I understand the Liberals tried to hit big and they succeeded for one day, but it was so costly that the opposition was just able to accuse them of giving tax cuts to rich people. Weirdly enough, it might have been more effective had they promised to only cut it in half. But that's just my intuition. Polls have shown people more responsive to the promise of opening ICBC to competition. Also, in Surrey specifically, it is possible that the promise to allow a referendum on getting rid of the RCMP might move some votes. This is super local and hard to measure though.


As for the Green, losing Andrew Weaver was always going to be a big loss. The fact he literally supported Horgan during this election didn't help. I know many people got a positive image of Sonia Furstenau but as a leader of a party, she spent way too much complaining about the fact we shouldn't have an election. The Green platform is also weirdly light on specifics. But the main issue is really that a pandemic election isn't good for third parties. People don't want instability and uncertainty. If they can get their centrist, competent government with Horgan, no reason to go with the Green. The Green just really need to keep at least one seat, going back to zero would be devastating. And they'd be become the first Green MP or MLA in Canada not to be reelected. Last minute trend was better and mail-in ballots (as well as advance polls) looked good for the Green, so they might actually keep their two seats. I'm curious how Forum arrived at 4 seats projected though.


2. Mail-in ballots and uncertainty


This election will have one particularity: a good share of the votes won't be counted until at least two weeks after the election. By law, Elections BC can't start counting the mail-in ballots for 14 days. In normal times this isn't an issue as those ballots only represent a small fraction (and even there, we had to wait for them in 2017 in a few close ridings). But with over 700k requested (and 66% have been returned already), the situation is obviously very different this year. Given that there were 1.9m votes three years ago and that we expect a lower turnout in general, those mail-in ballots could represent as much as 40% of the total.


A quick look at the number of requests for such ballots, along with the info from some polls, show that mail-in ballots were especially popular in urban environments. We also see a strong correlation (even after controlling for other factors) with the share of Green votes in 2017. Here below you have a simple regression (using riding level data from Elections BC and the census) of the determinants of mail-in ballots requests.


Bottom line: urban environments where there are more renters and high income requested more mail-in ballots. Green ridings and, finally, older voters too (this last one is likely just a general turnout effect, although we might expect older voters to particularly want to avoid crowded places). Outside of the age factor, most variables are correlated with higher NDP and Green votes instead of Liberal ones. Polls have also shown that NDP and Green voters were more likely to have chosen to vote by mail. What this means is that the results tomorrow night (where they'll count the advance votes as well as the votes cast on Saturday) might be generally more favourable to the BC Liberals. This is especially true at the aggregate level as the Interior of BC is likely to represent a bigger share than it should. It might not cause too much of a bias at the riding level, although I'd personally think that any seat with the Liberals narrowly ahead tomorrow night will likely flip to the NDP 2 weeks later.


Notice the lack of correlation with the advance turnout. So mail-in ballots didn't displace advance votes as I'd have thought. Also, while those variables here can explain 86% of the variation in mail-in ballots requests, it failed spectacularly at explaining the variations in advance turnout. Not sure what to do of this info but I thought I'd share it. Speaking of which, a record 681k people voted in advance, representing about 20% of registered voters. All in all, I suspect that over 50% of the votes were cast in advance (in-person or by mail).


Every single poll that has measured the voting intentions among those who had already voted have shown the same thing: the BC NDP is well ahead in this group. The Mainstreet poll was particularly bad for the Liberals. We are talking of a 50-30 lead, if not more. The Liberals really need to get their vote out tomorrow if they want to avoid a collapse similar to the Quebec Liberals in 2018.


In terms of uncertainty, this is one of the most certain and confident forecast I wrote. The BC NDP came within distance of winning the election in 2017. All they need is a few wins here and there and they get a majority. Given that the popular vote swing is showing a NDP at +10 overall, I think it's an absolute no-brainer that the NDP will win a majority. Any other results would require a massive polling failure.


Speaking of which, BC 2013 remains the second worst polling failure in Canada (behind Alberta 2012). The BC NDP had been ahead in the polls consistently and everybody thought they were cruising towards a majority. And yet the Liberals won. So why am I discarding such a possibility? I'm not, my simulations absolutely allow for such polling mistake (and even worse). But the situation is very different. First of all, the Liberals had the trend going for them during the 2013 election while it hasn't been the case this year. By the end of the campaign, the NDP's lead was about 6-7 points and that's without adjusting the BC Conservatives down for not running a full slate of candidates. Therefore, the NDP lead this year is bigger and more consistent. If anything, the last minute trend was very favourable to the NDP with some polls putting this party above 50%. Maybe more importantly, we don't see age as an important factor for voting intentions. BC voters skew very old (50% of the voters were over 55 in 2017 and over 58% in 2018 for the referendum on electoral reform). In 2013 and 2018, polls showed the 55+ being much different from the other age groups (supporting the Liberals in 2013 and against PR in 2018). While the Liberals still perform better with older voters today, the gap is far less pronounced. Using the 2017 age-turnout and reweighting the polls, the 13 points lead of the NDP over the Liberals shrink to 12. Using the 2018 turnout shrinks it even more to 11.4. Therefore, even if the age turnout is as skewed as in 2018 (when polls were dead wrong), we still get a NDP majority, just a smaller one. So while there is uncertainty with turnout and mail-in ballots, I don't believe there is enough uncertainty to think Wilkinson can pull a win.


3. Seats to watch


I have 21ridings where one candidate is projected with 100% chances of winning and my model has never ever made a mistake in those situations. All of those are for the NDP. Then let's add the 8 ridings where the chances are not 0.0% but like 0.1%. So the NDP pretty much starts at 29 ridings. The BC Liberals have 2 guaranteed seats (the two Peace River) and another one almost guaranteed.


If we add the ridings where the NDP has at least 98% chances of winning (9), we have 38 ridings. Therefore the NDP only needs to win 6 'competitive' ridings to win a majority.


I have 14 ridings where the outcome is decided by less than 5%.


As for the Green, province-wide (or even regional) polls show a sharp drop from the Andrew Weaver-led party of 2017, although the last minute trend was better. Polls show that 30-40% of 2017 Green voters have now moved to the NDP. Hopefully for the Green, the bleeding out is less pronounced in the few ridings they hold. It's really hard to make projections when local factors are in play (we'd need riding polls but somehow Mainstreet didn't publish a single one). Other metrics (fundraising, early voting) point to the Green doing okay and saving what they have but we can't ignore the clear trend. On the Island, the Green are down by 5 points and the NDP is up 9 points. Unless the Green can concentrate their support in their 2 ridings, they'll be wiped out. As for the 3rd seat, Oak Bay - Gordon Head, I believe the Green have no chance after losing their star candidate and not replacing him with another big name. But hey, we'll see, the mail-in turnout was really high at 37%, something that looks good for the Green. I think the drop among older voters might really hurt them, in particular since the ridings they compete in have some of the oldest median voters in the province.


A quick word about the BC Conservatives who are running 19 candidates. Keep an eye on some of those ridings as they could split the vote in a few (like Langley East) to allow the NDP to win. The party's leader, Trevor Bolin, is running the BC's most right wing riding in Peace River North. Compared to 2017 where they only ran 7 candidates, this party seems more organized and should get 1-3% of the vote overall, more in the Interior.


In Langley, my model is flipping what I feel will happen. I wouldn't be surprised to have Langley East going NDP (the Lib MLA was in trouble for voting against a rainbow crosswalk and the Conservatives are running a good campaign there) while Langley might just remain Liberals. The fact John Horgan went there on October 21st (while Wilkinson hasn't gone at all!) probably shows the NDP knows they can flip at least one riding there.


Can we expect any Liberals gain? I doubt it but I'd look at Surrey for such surprise. Local issues (RCMP) plus a softer NDP vote (2017 saw a big surge that might not repeat) could generate a surprise gain.


Ultimately, the seats to watch aren't there to decide whether the NDP will win a majority but on whether their majority will be huge, including massive gains in the interior. Personally, if the turnout tomorrow is low (for Saturday votes only), I'll start looking at seats like Abbotsford-Mission where we could see surprises.


Thanks for reading and see you at the next election (my guess is in the spring for the federal one). I'll write here and there in-between obviously.

Angus Reid just published a second poll in only a few days and, maybe for the first time since this campaign started, the numbers are slightly better for the Liberals. Don't get me wrong, they are still pointing towards a NDP majority but it's relatively better for a few reasons.


The BC Liberals are 'only' trailing by 10 points (45% to 35%) among decided voters (including many who have already voted in advance or by mail). Moreover, some of the crosstabs are more favourable for the Liberals. For instance we see this party well ahead in the Interior and higher among the 55+. This is more typical of BC politics but quite a contrast to what polls were showing previously (with Horgan making significant gains in the Interior and among older voters).


The one really good news for the NDP in this poll is the other confirmation that, among people who have already voted, this party is largely ahead (51% to 33%). We had seen this in other polls previously. Whether the NDP gets a small or super majority might well depend on how many people actually go out and vote on Saturday, as the in-person voters are likely to favour the Liberals.


We have seen such trends in previous polls as you can see in the next few graphs.







The good trend on the Island for the Green might be misleading as it is mostly due to the last few pollsters (Angus-Reid, Insights West) having this party higher than Leger or Mainstreet, a difference observed throughout the campaign.


We are still far from the BC Liberals being actually competitive but if such trends should materialize on Saturday, Wilkinson could at least save the furniture and not fall below 30 seats.


Here are the projections after adding this poll to the adjusted average. Notice that the BC Liberals no longer have an official candidate in Chilliwack Kent and the incumbent MLA is listed as independent after he was forced to resign. He'll still appear on the ballot under the Liberal name. Projecting such a situation is difficult but I looked at the few past precedents and concluded that Laurie Throness will likely perform at 50-60% of what he'd have otherwise. Most likely not enough to win but enough to be a strong second.







The BC election is officially on Saturday October 24th. However, many people have already voted by mail. Elections BC has statistics about the number of mail-in ballots requests, how many got returned already and the advanced polls turnout. They obviously won't tell us who those votes were for but I believe we are seeing some strong trends that will help us make sense of the results on Saturday.


First, let's look at who requested a mail-in ballot. I plotted the percentage of registered voters by riding against the projected voting intentions (based on my projections; I could have used the 2017 results for almost identical correlations). See the results:


In average, about 19% of RV have asked to vote by mail. Whether they'll all receive their ballots on time and will be able to return them by mail is a different question (note: if you have your mail-in ballot but haven't sent it back already, it's likely too late and you should drop your ballot yourself to one of the authorized locations by Elections BC). As we can see, ridings where the Liberals are projected to be high have had a much lower rate of requests for such mail-in ballots.


Elections BC said that while over 700,000 mail-in ballots have been requested, only 177,000 have been returned so far. Given that the 2017 election had just shy of 2 mio voters and I think it's reasonable to expect a lower turnout this time around, I think it's safe to assume that about half of the votes will be cast by mail. Which means we'll only have the results of half the votes on Saturday since Elections BC cannot start counting the mail-in ballots until 2 weeks after the election (that's the law). This is a very important element that people will need to keep in mind on Saturday when we start getting results.


We can also see the pattern of mail-in ballot requests by region where Vancouver Island is high (27%) compared to Metro Vancouver (21%) and the Interior (13%). We obviously always run the risk of ecological fallacy when using aggregate data to infer individual behaviors. We actually saw that during the referendum in 2018 where ridings with more young voters were also the ones who were voting more, yet the final results showed younger votes actually voted less. So I'll have to look into it more.


Does it match with the polls? Insights West was showing that Liberals voters were more likely to declare their intentions to vote in-person than NDP and Green voters. It also showed that older voters seemed more likely to vote in-person. Research Co. was showing that while half the voters planned to do so by mail, it was actually the 55+ who were declaring their intentions to do so. So we have contradicting evidence here. However, both polls were showing that among the people who had already voted, the NDP was ahead.


What about advanced in-person voting? According to Elections BC, 305,789 votes have been cast during the first 3 days (and it climbs to 380,000 if we include the fourth day but Elections BC hasn't updated their pdf yet). Over 600k voters had chosen this option in 2017. I don't believe we'll beat this number since many preferred the mail-in ballots route this year.


Do we see a trend in advance voting? Yes and this time, it's more positive for the Liberals. See below:



So it seems that this is consistent with Liberal voters preferring to vote in person while NDP voters would rather do it by mail. The BC Green party should be happy with both measures and they might be getting out their vote more than what the polls have indicated.


By the way, the correlation between mail-in turnout (measured by requests) and advance turnout is -0.04, so there isn't much of a correlation there. It's really when we look at each party that we see the negative correlation. I was expecting to see  a negative correlation as people choosing to vote by mail might have been the ones who would have voted in advance otherwise. That might not be true then or other things are happening at the same time.


What does all of this mean? On Saturday, the votes of the day (and advance polls) will likely favour the BC Liberals (depending on the ratio of mail-in ballots received and advance turnout) and we'll need to keep that in mind. On the other hand, if the Liberals are in trouble based on Saturday results only, then we could just go ahead and call the election for the NDP.


Just a final word: I expect Saturday's results to be, overall, more favourable to the BC Liberals than the full results. So if we look at the province-wide percentages, the Liberals will be higher than expected. However, at the riding level, the bias might not be that clear cut. Indeed, if we do a regression where we control for both the projected voting intentions and the regions, the voting intentions aren't significant anymore (although it's not far to be for the Green). What it means is that the distinction between mail-in and in-person might have more to do with the region than the parties. So, on Saturday, we'll get more votes from the Interior than the Island but within a region, if we look at an individual riding, we might get approximatively accurate results. With the caveat that the Green might likely get a lower same-day votes than mail-in and advance ones (which is also consistent with the idea that small parties simply get less votes from the undecided and late switchers).

Tuesday was the leaders debate of this 2020 BC election. a debate that was actually quite good (in my opinion) and saw all three leaders perform relatively well. It appears, however, that Green leader Sonia Furstenau likely performed a little bit better, especially if we measure as we should:  adjusted for voting intentions.


The new poll from Insights West tells us that John Horgan got the most 'likes'. We need to remember, however, that people tend to be biased when answering such questions and are more likely to find the leader of the party they support to have won.



If we control for the fact that the NDP is currently polling around 50% while the Green are barely at 12%, it puts Furstenau as the true winner. See it another way: while 81% of NDP voters found Horgan to be good or very good, 54% said the same for Furstenau. On the other hand, only 38% of Liberals voters and 30% of Green voters thought Horgan did good or very good. Sonia Furstenau therefore received some high praise from the voters of other parties.


I have done such analysis for years now, what I usually call the leader index. This has worked well in every single election to determine what post-debate trajectory we should expect in the polls. In the present case, it'd be logical to expect the Green party to increase. Of course, that is assuming enough people watched the debate and nothing else happens.


I had done my own Twitter polls on Tuesday and had found qualitatively similar results. It's weird how Twitter polls, as biased as they are, can actually work relatively well for such leader index. I guess the bias cancels out once we take the ratio.



As a reminder, Andrew Weaver had won the 2017 debate. It didn't really translate into an increase in the polls but maybe his win prevented a drop towards the end? The 2017 campaign was a lot closer and everybody was worried about vote splitting. So the Green party managing to actually get 16-17% and 3 seats was very impressive.


Will this debate help the Green this year? Hard to say. One caveat for Furstenau is the fact so many have already voted (by mail) even before the debate. Also, her best moments (including the answer on white privileges) came relatively late during the debate. Ultimately, I doubt it'll create much movement in the polls overall but it might be enough for her to win her seat. Google Trends shows the peak for Furstenau right after her answer. Unfortunately for her, and despite the best efforts of CBC, she didn't seem to generate much interest post debate.



Speaking of which, my model was, until now, giving her a 5% boost as new leader. Those things are always hard to measure and are mostly subjective. After the debate, and looking a little bit more into it (like previous Green leaders), I changed the bonus to 10%. It is enough to flip the seat back from the NDP in my projections but my confidence is low. Fortunately, it seems Mainstreet might do a few riding polls and I really, really hope Cowichan Valley is one of them.


If we go back to the Insights West poll, done entirely post-debate, the Green are actually down! From 16% to 14% (and from 26% to 24% on Vancouver Island) if we compare the two polls from this firm. There are other things that should worry this party. First of all, the Green are clearly doing better with the 18-34 group than the 55+. It was already the case in 2017 but not to the same extent. Given how voters in BC tend to skew older (50% above 55 in 2017 and 58% in 2018 during the referendum!), this could mean the Green will underperform their polling numbers (note: I already don't allocate undecided to them as they are a small party, exactly in order to correct the typical overestimation of small parties in the polls). Moreover, polls have consistently shown that Green voters are less certain to vote and are less certain of their choice. Finally, the Green voters are more likely to declare their intention to vote by mail. This is good but how many have actually requested their ballots and will return it on time? How many will just forget?


Anyway, I updated the projections. Not much change really except the Greens eat (but this is due to the change in the leader bonus, not the polling average). Also, no, I haven't incorporated yet the situation in Chilliwack-Kent, not sure how I'll as this is very tricky.



Adjusted polling averages; Seat projections with 95% confidence intervals; Chances of winning the most seats





 With two new polls published today right before the leaders debate (one from Ipsos and one from Leger), the situation in BC remains incredibly difficult for the BC Liberals. As a matter of fact, the more polls we get, the closer to 50% the BC NDP is.


I don't have the full details of the Ipsos poll but I used the top line numbers. I added the Leger poll.


You have the projections below. The BC Liberals are really, really behind. And as I previously showed, the age-turnout is unlikely to provide any much needed help as the NDP dominates every age group. The lead of the NDP is just too big and I honestly don't see anything will change this between and now and election day, especially as half the voters might cast a ballot by mail and might have done it already.


Adjusted polling average; Seat projections with 95% confidence intervals; Chances of winning the most seats



With Research Co publishing their second polls yesterday and confirming the trend, there is very little doubt right now that the BC NDP of John Horgan is ahead. So ahead actually that it's difficult to imagine them missing on a majority. As I was showing yesterday, the NDP is dominant in every region. I'm not updating the projections with the Research Co poll because adding it makes no difference overall (I believe it marginally flips one seat from liberals to NDP).


Still, as shown in my Twitter poll, you have a lot of people who aren't confident in calling for a NDP majority yet. And it's fair, there are still over two weeks of campaign ahead of us, including the leaders debate on the 13th.


There are also memories of 2013 where the polls were pointing towards a NDP majority and the Liberals ended up winning. To this day, the BC 2013 remains the second worse polling failure I have seen in Canada (the worst being Alberta 2012). Polls were also dead wrong for the referendum on electoral reform in 2018.


So, should we be concerned about the polls this year? The answer is mostly no and I'll explain why. With that being said, let's remember that this is a fairly unique election with what will most likely be a low turnout and most people voting by mail. So there are certainly sources of uncertainty.


Polls can be wrong for many reasons. Pollsters love to pretend people just changed their mind at the last minute but this is rarely true. No, usually the reason has more to do with a turnout that favoured one party. In BC, the turnout by age is incredibly skewed towards older voters. Polls weigh based on census data. So if the 18-34 represent (say) 30%, they will want to give their sample a weight of 30% to this demographic (polls also weigh based on gender and other characteristics). This is problematic if young people vote less and ultimately represent a smaller fraction of the electorate.


Why am not not that concerned this year then? Because there is remarkably little impact of the age on those. While it's typical to see the 18-34 voting very differently from the 55+, we don't see a strong pattern this year. See the table below.



This is very different from 2013 and 2018. In 2013, the NDP was dominating the Liberals by 20-30 points in the 18-34 age range but were trailing (or tied) in the 55+. Given the age turnout, it's actually not a big surprise the polls were wrong.


The 2018 referendum was even worse. Over 58% of the voters were above 55 years old, a proportion much higher than in the population and higher than the 50% of the 2017 election. Given how polls were clearly showing that younger voters overwhelmingly favored PR while older voters preferred FPTP, there as well we can see why the polls were so off.


To illustrate my point, I calculated the current voting intentions in BC if the age-turnout of 2017 or 2018 occurred. You have the results below:




Similarly to how the NDP dominates every region in BC, this party is ahead in every age group. So even if the turnout is low and skews older, we should still get a NDP majority.


Don't take it the wrong way, I'm not claiming the polls will necessarily be right, I'm just saying we at least don't seem to have to worry about the age turnout causing trouble.


Final note, the BC Green are currently doing better among the 18-34 than among older voters, in particular the 55+. This wasn't necessarily the case in 2017 and shows, again, what I've been saying: this party is losing voters to the NDP. I think going more left and losing Andrew Weaver might have a huge impact on older voters on the Island and this is why I'm not confident they'll keep their seats.

 No new poll yesterday, so my projections are still valid.


Today, I thought it'd be interesting to look at the evolution of voting intentions by region over the last 3 elections. Let's start with Metro Vancouver:


7 years ago, during an election that saw one of Canada's biggest polling failures, the BC Liberals of Christy Clark were still the first party in the Vancouver region. Over the last 7 years, the Liberals have decreased from 46% to 35% right now in my adjusted polling average. The NDP went from 41% to 53%. In 2017, Metro Vancouver is actually the only region where the NDP increased! John Horgan pretty much campaigned the entire time there and bet big on the suburbs -- and it paid off with many gains in Surrey and elsewhere. This year, we see the NDP just increasing its lead to unprecedented levels. The dominance is such that the Liberals could well be left with only a few seats.


Let's look at Vancouver Island (where Victoria is in case you aren't from BC and are rightfully confused).


Three years ago, that was the battleground with essentially a three way race. Ridings like Courtney-Comox ended up deciding the fate of the province based on fewer than 200 votes. The Green made a remarkable progression which allowed them to go from 1 seat (that was almost entirely due to their leader Andrew Weaver) to 3. As you can see, this year is very different and this party is back to its 2013 levels. Worse news for them, the NDP -- their main opponent in their 3 ridings -- are now comfortably ahead and above 50%. This is why my projections have the Green at zero seat in average. Could they save their 2 or 3 seats? Possible but it'll require strong personal effects from Sonia Furstenau and Adam Olsen. In Oak Bay - Gordon head, the riding of Andrew Weaver, I believe the NDP is clearly favorite. Weaver had one of the strongest personal effects I have seen (around 30 points) and his departure will hurt the Green. Although his riding is partially shared with the one of Elizabeth May at the federal level and is one of the Greenest ones in the country. Still, there is no way that losing Weaver won't hurt the Green, especially since the new candidate is far from a star candidate (sorry...)


Finally, the Interior (or the rest of BC).


Traditionally a Liberal stronghold, recent polling has revealed a big surprise: the NDP is essentially tied with the Liberals! This is pretty crazy and shows the current dominance of the NDP everywhere. If they are competitive in the Interior, they can not only win a majority but a very large one. Notice the BC Conservatives that were at 7% in 2013 back when they were organized and were running a lot of candidates. They are doing better this year than 3 years ago but it's still a far cry from their peak. Many have discussed a very strong urban-rural divide in BC but I tend to disagree. The graph above is clearly showing the NDP increasing the most in the Interior. Sure it might still be concentrated in the more 'urban' areas (Chilliwack, etc), but we are very far from the rural-urban divide we observe during federal elections. The fact the NDP, that lost by 20 points in the Interior 3 years ago, is now tied with the Liberals is remarkable and not something I was expecting to see.


So, what do we see? Well it's one of the clearest elections I have covered. The BC NDP is just dominant everywhere, including in the Interior. The Green are suffering a pretty serious setback compared to 2017. It's likely due to losing Andrew Weaver (who has literally endorsed Horgan...) as well as the covid-19 situation pushing progressives towards the NDP. The polls clearly showed the Green dropping around March-April this year. And the recent Angus-Reid poll was showing that as many as 40% of Green voters from 2017 were now voting NDP.


What would make this election more interesting? The Green need to defend at least 2 seats on the Island and the Liberals need to climb back in Metro Vancouver. If they can do that, they might just have a small chance. The NDP could be up provincially but some of the increase would be wasted in the Interior while they would lose some seats in Metro Vancouver. But that will require an overall shift of at least 5 points and, so far, polls have not shown this to be the trend.

Angus Reid published their first poll yesterday and it shows the BC NDP more ahead than ever. The party of John Horgan is at 49%, a full 18 points ahead of the BC Liberals. The Green are far behind at 14%.


The regional numbers are even more devastating for the Liberals of Andrew Wilkinson. Someone on Twitter said they used my model with the regional numbers and it was a complete wipeout in Metro Vancouver. According to this poll, the NDP is at 59% versus only 24% for the Liberals in this region, rich in seats. Obviously the sample size is small and the polling average is less favourable to the NDP (I have the NDP at 53% and Liberals at 35%) but the trend for the NDP is strong.


With less than 3 weeks to go, I believe we can start to genuinely wonder if there is any path to victory for the BC Liberals. Sure this Angus-Reid poll is particularly bad but every single poll points to the same thing: a NDP majority. Angus-Reid is also a polling firm that is often accused of having a right-wing bias, so you know things are bad for the Liberals when this firm is showing such a devastating poll!


The NDP is more ahead than ever in Metro Vancouver, has taken a lot of the Green votes on the Island and is even competitive with the Liberals in the Interior, the one region they actually decreased in 2017. At this point, the uncertainty isn't to know whether the NDP would get a majority, it's whether they would win over 50 or 60 seats. The battleground ridings are like Chilliwack, Cariboo North or even, to a lesser extent, Abbotsford!


So anyway, the Liberals are running out of time and even if they are clearly focusing on the Lower Mainland (Massey tunnel, referendum on the police in Surrey, etc), I'm starting to really doubt they can't prevent a NDP majority.


If you are thinking 'oh well, polls have been wrong before', you aren't wrong, but we are at the point where those polls would need to be spectacularly wrong. I'm talking worst failure in history type of wrong.


Here are the projections.

Polling average; Seat projections with 95% confidence intervals; Chances of winning the most seats












Now that Elections BC has published the final list of candidates, I have updated the model. It's a lot of work for nothing, in particular with the BC Conservatives party. They ran 10 candidates last time and have 19 this time around. However, they aren't running in the same riding. So I need to make some adjustments.


My data analysis shows that there is usually a stronger relationship between the BC Conservatives votes and the BC Liberals (logical). In 2017 for instance, the Liberals most likely benefitted partially from the drop from 56 BC Cons. candidates to only 10, especially in the Interior. On the other hand, if this party runs a candidate in a riding where it didn't in 2017, I need to 'input' an hypothetical score. I use the regional average for this.


You might think this is useless and you might be right. But while the BC Conservatives ran few candidates, they usually get around 5% in every riding they contest. So they could have an impact in Chilliwack or Boundary-Similkameen for instance.


Anyway, here are the updated projections. You also have the updated basic simulator here. I'll update the regional one soon. Not a lot has changed, it's still a NDP majority unless we get a monumental polling failure. As for the Green, they could keep their 2-3 seats in theory but would require strong incumbency effects and/or a polling failure as well.


Adjusted polling average; Seat projections with 95% confidence intervals; Chances of winning the most seats






Quick projections update since Ipsos published their first BC poll of the campaign yesterday. As opposed to the first two polls, Ipsos has the NDP significantly higher at 51% among decided voters, a full 18 points ahead of the BC Liberals! Here are a few thoughts about this poll and other stuff:


- Ipsos has a really high number of undecided (25%). Since I allocate undecided not proportionally (because I find this assumption stupid, it comes down to assuming that all undecided will either not vote or will vote exactly like the decided voters) and allocate 50% of undecided to both the NDP and Liberals (so no undecided for small parties), I have the 'adjusted' NDP lead in this poll at 'only' 13 points. You might hate my assumption but it has worked well in the past. The BC Liberals have a tendency to be underestimated and incumbent governments do too, so 50% each is logical to me.


- The last three polls have shown something maybe surprising: the BC NDP is especially up (compared to 2017) in the Interior/Rest of BC. They are up a full 10 points, more than the 6.3% increase overall. This could be important for their vote efficiency and could really hurt the Liberals. This is why if you use the advance simulator, you usually get higher seat total for the NDP than using the basic simulator (although the difference should be attenuated now that the basic simulator/model also has regional adjustments). Make no mistake, this is great news for the NDP and I'm curious if Horgan will campaign in the Interior more than in 2017. Ridings such as Chilliwack or Cariboo could be in play.


- The BC Green is down 8 points on the Island, this is really bad news for them. They need to hope their support is now concentrated in the three ridings they own. They also need to announce a good candidate in Oak Bay-Gordon Head if they want to prevent the NDP from taking this riding. I hear an announcement is scheduled for today, Wednesday.


Alright, that is all. We are supposed to get a Leger and Mainstreet polls soon, the later being the first phone (IVR) poll of this campaign. In the meantime, here are all the numbers, including a new presentation of min, max and the 5th/95th percentiles (called 'low' and 'high'). Advice: click on the picture to enlarge it.



Voting intentions, seat projections (with 95% confidence intervals), chances of winning the most seats








The BC election is underway, although I wouldn't blame you for not knowing. I'm pretty sure you heard more about the new Supreme Court nominee in the US than the various promises from Horgan, Wilkinson or Fursteneau.


We only had two polls so far and while they showed the BC Liberals more competitive than we might have thought (given pre-election polls this summer), it remains that the BC NDP is clearly ahead and significantly favourite to win. I currently have the NDP with a 89% chance of winning a majority and another 4% chance of winning a plurality. The Liberals only have a 5% chance of winning more seats than the NDP (and 1% of a tie). For the Liberals to win the most seats, it would require a fairly large polling error. Still, BC has had a history of polling failures (2013, 2018). Make no mistake here, I'm not saying the Liberals have a 5% chances of winning even if they really only get 35% of the vote while the NDP is at 43.5%. No, what those probabilities are showing is that if the BC Liberals are polling 8 points behind, given the traditional polling accuracy, they have a small chance if the polls are underestimating them and their vote is efficient.


So, how far are we from a 50-50 race? One that would be... too close to call?


I ran my usual simulations (you can too by the way now that I put my code on GitHub) and, using the results, I then estimated a logit regression showing the chances for the BC Liberals to win the most seats as a function of their lead in the popular vote (negative numbers mean the Liberals are behind obviously). Here's the results:




BC doesn't have much of a difference between the popular vote and the seats. At least not given the distribution of the votes in 2017. Essentially, it's a 50-50 race if the Liberals are tied with the NDP in the popular vote (the 0 on the x-axis). Wow, such great work Bryan, thank you! In all seriousness, this is relatively rare in Canadian politics. We usually see a difference between the two with one party being able to win with fewer votes (PQ in Quebec, LPC at the federal level). So it might seem like a whole lot of work for nothing -- it isn't!


The BC Liberals have their chances as long as they are within 1 or 2 points of the NDP. It'd naturally require better vote efficiency (in other words: having their votes in the right ridings) but it's far from impossible. In this scenario, it'd likely come down to fewer than 10 ridings (most of which in the Metro Vancouver area).


So, to answer the question of this post: I'd say that if polls start showing the BC Liberals within 5 points of the NDP, we will start seeing the projections with the Liberals with good chances. The historical polling accuracy in Canada, over the various provincial and federal elections of the last 10 years, is an empirical margin of error just shy of 5%. Therefore, if the Liberals are polling 5 points behind, the odds that the popular vote is a tie are becoming quite good. Don't misinterpret what I'm saying. It won't be a 50-50 race if polls start showing the Liberals within 5 -- it simply won't be one where a Liberal victory would require a massive polling error.


In other words: we are a 3-points shift in the polls away from having what I'd consider a contested election. If polls continue to have the NDP 8+ points ahead, Horgan can sleep nicely knowing his bet likely paid off. Although a major polling error can't be excluded given the possible low turnout of an election held during a pandemic (and, as opposed to NB, during a phase where cases are increasing). That's a topic for another day though.