I'm a little bit late for this analysis but you know how it is with Christmas and all.

The results for the referendum on electoral reform in BC were pretty shocking. While literally every single poll published before, during or right after the campaign had the two options pretty much at 50-50, the current system won with over 60% of the votes!

I guess this is fitting that, in a year that has been terrible for the polls (massive underestimations of the winner in Ontario and Quebec), we would also get polls be very off for a more uncertain exercise. Still, it doesn't make the results any less shocking. Beyond the polls, I really can't explain how a mail-in referendum with a simpler option (MMP instead of STV) with the active support of two of the three main parties wouldn't at least do better than in 2009.

So, why were polls so wrong? If you are expecting a clear cut answer in this article, you should stop reading. The short answer is still: I don't know. But I can try to provide some conjectures. If you don't have time to read, here is a point summary:

- Polls were wrong everywhere but especially wrong in the Lower Mainland

- It could be because polls didn't reach non-English voters enough as PR got really destroyed (and was overestimated) in the suburbs

- People aged 18-34 did vote more than usual but it wasn't enough for PR to win

Also, I'll ignore the magical polls published after the results (one from Research Co. and one mentioned by Ipsos) that somehow had FPTP clearly winning. I can give the benefit of the doubt to Ipsos since they were actually providing numbers to the NO side, even though I couldn't explain why Ipsos would have had a better methodology than other firms as Ipsos hasn't historically performed better. But for Research Co, they literally had another exit poll before and it was still mostly 50-50, so give me a break. Also, I should mention the exit poll from Angus-Reid, published right after the official results were made public. This exit poll had the two options pretty much at 50-50 but Angus-Reid made sure to use a headline hiding this fact. I found it amusing.

Anyway, I think the biggest discrepancy between results and polls can be observed regionally. Polls mostly all showed the same pattern: PR was ahead on Vancouver Island and the Lower Mainland while FPTP was clearly ahead in the Interior. The table below shows the polling average (same weight to every poll; Playing with weights doesn't make much difference) and the actual results.

Polls were overestimating PR everywhere but the bias is especially strong in the Lower Mainland. Is it because polls didn't correctly reach non-English speakers? Is that why Ipsos did better? Maybe.

If you look at the results by ridings, you see that PR won where we thought it'd (in some very urban Vancouver ridings) but it got destroyed way more than expected in most of the suburbs.

As for the general overestimation of PR, one easy explanation is that undecided ultimately chose the status quo. This would make sense but doesn't match with the fact that polls kept showing a 50-50 race as the campaign went on (and the number of undecided decreased). Exit polls (from Angus-Reid or Research Co, or even my own) were also good for PR and had no undecided. Ipsos mentioned on Twitter that FPTP was around 60% the whole time. So this theory really doesn't match with the numbers.

If we look at the second exit poll from Research Co (the one where they magically got the right results all of a sudden), we see that while BC Liberals voters did vote, as expected, to keep FPTP (with a margin even higher than expected), NDP and Green voters didn't remotely support PR as much as what the polls were showing. Polls had these voters over 70% in favor of PR but the exit poll only show PR at around 60%. So while I'm skeptical of a poll that all of a sudden got the right results, this might be the explanation: the BC Liberals got their vote out more and managed to convince more people while the NDP clearly didn't motivate its suburban votes to vote YES.

Polls had shown that this referendum was heavily split with regard to age with the 18-34 clearly in favor of PR while the 55+ were incredibly opposed to it. Regression-based analysis (with the final numbers) still shows what I had been saying: the turnout was higher in ridings with more young voters (relatively speaking, especially compared to 2017). Elections BC doesn't provide us with the results by age but I suspect that the NO won among the 55+ even more than what the polls were showing while the 35-54 most likely ended up voting NO as well despite the polls showing otherwise.

By the way, my measure of when a riding received the ballots (see previous blog posts about it) is still there and significant with the final numbers. If PR had lost by a tiny margins, I feel this should become a big issue. But given the margin of victory for the NO side, I guess we can forget about it even though this is troubling.

During the campaign I tried various methods to estimate the results (age, region or party based). I liked the method based on past votes the most as I thought this was the best "one variable" method available, one that would capture other effects as well (vote is correlated with region and age too). So I re-did these estimations and compared my results to the actual ones, riding by riding. Obviously my method was biased overall  since I had PR slightly ahead, but I'm more interested in finding patterns of where the method was more wrong. By the way, the overall correlation between estimated % for PR and actual ones is 0.83, so not too bad.

I found that the method was overestimating PR in ridings in the Lower Mainland and on the Island (and/or underestimating FPTP in the Interior). I was underestimating PR in ridings with more voters aged 18-34 and with a bigger share of English speakers. Maybe here is the most interesting piece of information: the method was underestimating PR in ridings that had a higher turnout this year but it was overestimating PR in ridings that had a higher turnout last year.

All in all, I maintain what I was saying: this referendum motivated more voters aged 18-34 to vote. I'm not saying the turnout in this age group was actually the highest of all age groups, but it was definitely higher than in previous elections. And this group likely voted for PR by a healthy margin. But that just wasn't enough to allow PR to actually win. The YES campaign (as illustrated by Horgan's performance at the debate) was weird. I gave them the benefit of the doubt during the referendum as I thought that they maybe had data showing it was working, but in hindsight, it was a mistake. Focusing on the young, progressive vote (it's "lit") was a bad idea. These voters were going to support PR anyway. What the YES side had to do was convincing older voters that PR was a good idea. Offering three systems without fully defining one just opened a giant door for the NO side who could spend 30 days saying it was "too complicated" or too risky. It also seems the NO side was successful in making this referendum feels "rigged" even though I don't really see valid reasons behind this argument.

Proportional representation is now a dead in BC for many years (as it should be after such a loss). Hopefully Quebec leads the way next year and adopts PR. Legault has mentioned he doesn't intend on doing a referendum and while I believe one could be successful in Quebec, I also never thought PR would get less than 40% in BC. So no referendum might be the safest way to get PR, although I feel a referendum is most likely needed nowadays to change the electoral system... But I'm seriously tired of losing referendums! So go Quebec and go PEI next year!

The referendum voting period ended on Friday at 4:30pm. Over 41% of British Columbians ultimately voted, a respectable turnout for a referendum on such a question. It is also slightly higher than the average turnout of the last municipal elections in the province.

Elections BC is now busy processing the last ballots received and will start counting them soon. Results are expected by December 20th, although there has been no official announcement.

While we wait, here are the results of the exit poll that some of you funded. More generally, see this article as the reasons why the YES side likely won. I'm not putting a specific percentage on the chances here and it remains close with both sides able to win. I do believe however that we have more evidence in favour of the YES.

1. My exit poll (+the last campaign poll)

Last week I reached out to you, readers, to fund a small Google Survey of 1300 respondents to see what option they had chosen. The idea being that it'd be better to ask people after the vote since we wouldn't have a large number of undecided.

Here are the result:

Margins of error aren't provided since it's an online poll but the equivalent ones with a probabilistic sample of that size would be 3.3% (or 4.6% among voters only).

If you donated to the GoFundMe campaign, email me if you want access to the raw data.

Notice that I only got 905 valid observations out of 1300. This is because Google Survey sometimes doesn't know the age or gender of the respondent and therefore assign a weight of zero. I personally feel like this is a little bit ridiculous since they charged me for these respondents, but whatever.

This exit poll gives a relatively healthy lead to the YES side. Better than all the 50-50 polls during the campaign.

I know what some of you will say: this isn't the best poll out there and there are margins of error. You are 100% right for the latter. As for the former, this is debatable. My previous experiments with Google Survey have proven successful. With that said, I'd totally admit that I don't necessarily trust such a $250 poll as much as a full fledged one from Mainstreet or other reputable firms. And yes this poll is overestimating the turnout. But this is normal for polls, they always do. People who vote are just more likely to answer polls I guess. I'm actually quite happy with how close to the actual turnout it ended up.

Still, this exit poll should at least give us some information (assuming it's not completely wrong). First of all, the fear that undecided would ultimately choose the NO side (you know, the status quo) might not have materialized. Polls for previous mail-in referendums in BC have been overall quite accurate and I suspect this is mostly because undecided simply end up not voting.

Second of all, this confirms what we have been saying for a while: this referendum will likely be won based on who turns out to vote (versus actually convincing people tho change their mind on the topic).

I'd also like to point out that the very last poll of this campaign (done so late it could actually be considered an exit poll as well) from Insight West had the YES side a 52%. Among people who had already voted, it was a perfect 50-50 race but it was clearly shown that people who hadn't voted yet were overwhelmingly in favour of PR. Given that the poll was done from November 29th to December 3rd, that was leaving a couple of days for these people to vote and ultimately tilt the scale on the YES side. So my exit poll here is quite consistent with the 52-48 situation of the Insight West poll.

Polling wise, that's two arguments for the YES so far.

2. Estimates of the vote based on the turnout data

Throughout this referendum, I have often provided estimates of the number of votes for the YES and NO sides. I mostly used two methods: the first one was using regional as well as turnout by age (in 2017), as well as general polling averages, to estimate how many votes were for each side in each riding. The second method used instead the past vote (in 2017) and the polling averages as well (i.e: BC Liberals voters were in favour of keeping FPTP at around 80%, so a riding that voted 50% BC Lib last year would contribute 80%* 50%* votes in this riding this year] to the NO side.

The second method is most likely the best as this referendum turned out to be quite a partisan affair. Yes age is a strong determinant of the vote but voting behavior will capture this as well (older people vote BC Liberals more).

Here is a table of the various estimates for the YES side based on various methods (and variations of these methods) using the turnout data published on Monday (which aren't complete with only 38% of ballots processed out of more than 41% received; Elections BC indicated yesterday that they'd likely not provide further update until the final results).

It's a mixed bag but there are reasons to believe method 2 is the superior one. Also, method 1 is most likely underestimating the YES side. Here is why. Regression-based analysis of the turnout (see below) have shown some strong patterns. One of them has been that the 18-34 voted more than usual, relatively speaking. The 35-54 were not interested at all while the 55+ also voted less than in 2017.

My estimates here use the turnout by age of the 2017 election, by riding. In average, the 18-34 represented only 18% of the voters. The 35-54 were at 32% and the 55+ at 50% (much higher than their actual share of the population. This is thanks to a turnout of 69%, much higher than the others).

The regressions aren't actually absolutely proving that the 18-34 were voting more, it's only capturing that ridings with a higher share of 18-34 had a higher turnout this year. So it's technically possible that some ecological fallacy is at play here and that it's actually the 55+ in ridings with more young people that are voting more. It seems unlikely but it's possible.

If we assume that the 18-34 voted more (they were very pro PR), the 35-54 voted a lot less (they were very 50-50 on RP) and the 55+ voted slightly less (while still representing the biggest share; Very against PR), then my estimates above are likely underestimating the YES. It's capturing some of this increased turnout simply because more votes are literally coming from ridings with more young voters, but it doesn't actually adjust at the riding level.

Let me try to explain it another way: Imagine that the 18-34 represented 30% of the voters this time and the increase of 12 points was uniform across the province. When I run my regression, I'll capture some increase but not the precise, representative number (because the computer will think that the increased votes coming from ridings with many 55+ is because this group voted, not because the 18-34 in that riding got out to vote).

It's really difficult to try to adjust the turnout but I tried nonetheless. Using the regressions from below, I estimated the following shares of voters this year:

18-34: 24%

35-54: 25%

55+: 50%

Yes the 55+ would still represent 50% despite a drop in their turnout. The reason being that the 35-54 dropped so much, the remaining 55+ represent a bigger share.

So it's well possible that the 18-34 actually represented a bigger share of the voters than the 35-54 this time around. This is remarkable. Remember how Trudeau won so many surprise ridings thanks to a large increase in youth turnout in 2015? You got to wonder how much of this was due to the promise of electoral reform. But anyway, let's go back on topic.

If I'm right here and the polling average by age is also correct, that would give us a very close finish right around 50%. Of course that's using an average of polls, some of them done during the campaign and an overall level of support lower than the latest Insight West or my exit poll. So there might be a slight bias in favour of the NO there, possibly.

3. Analysis of the turnout

The first two points use polls (and other tricks or assumptions). What about a straight up analysis of the actual published turnout? Let's use a regression for this. I'm using the turnout of processed ballots as published Monday morning by Elections BC (so still missing about 3.2% of the ballots (that's over a 100k...).

Here are the results for the two models I used throughout this campaign.

If you read this blog from the beginning of this referendum, you should remember that the regression coefficients didn't always look so favourable to PR. In the first 2 weeks we would see the % of votes for the Liberals as significant. We would also see the 55+ voting more. But slowly and surely we started seeing positive trends. Now, with almost final turnout data, the entire table is pretty positive for PR.

We see that the only age group that voted more is the 18-34, the group heavily behind PR. We also see that the BC Green party got out the vote in ridings where it got a high percentage last year. Vancouver Island is where the turnout is the highest (and that's compared to 2017 where it was already the highest region; Plus this is on top of the effect of the % for the Green).

It's not 100% good news however. We do observe that the % for the BC Liberals is still positive. Not statistically significant but positive nonetheless. While it's negative for the %NDP. This shows that the NDP most likely didn't get its vote out everywhere. Surrey comes to mind (might be a good thing for PR as Surrey quite against PR in 2009).

The number of days of voting (measured by how many days the riding has been over the 1% threshold) is still significant and still around 0.3% per day. This is unfortunate and raise the question of fairness. With that said, my attempts at correcting for this have shown no impact on the overall results.

Conclusion

Let's be clear, this remains a close race. But we are seeing more signs pointing to a victory for the YES than the NO. When polls started showing a 50-50 race at the BEGINNING of this referendum, I honestly thought it was lost. You'd think undecided would ultimately pick the status quo. I feel pretty confident in saying that it doesn't look it happened. Worst case scenario is it remained a 50-50 race and turnout will choose the winner. Best case scenario? The turnout again plus a small edge for the YES as indicated in the two exit polls.

My various methods give different results but the one that is the most likely to capture the most effects is the one based on past votes. And this one shows a win for the YES. This is despite the latest Insight West poll showing a somewhat low support for PR among Green supporters (lower than among NDP, which is unexpected).

Would I bet money at this point? I hope Elections BC will publish the turnout by riding fully, but even if I had to make the bet now... Yes I'd. I wouldn't bet $1000, but I would put down a $20 on the YES.

Note: my exit poll is collecting repsonses right now. If you want regular updates, follow me on Twitter at @2closetocall

This morning Insight West decided to publish a new poll for the referendum on electoral reform. And it confirmed what I've been saying all along: this is super close. With that said, I actually think we are starting to see more and more positive signs for proportional representation.

Ok, the poll first. Conducted online between November 29th and December 3rd, it has 965 respondents. It shows PR is favoured by 47% of voters while keeping the current system (FPTP) is at 44% and 10% aren't sure. With that said, among the people who declared having voted already (the vast majority of the sample), the two options are essentially tied at 50-50 (Keeping the current system has 346 respondents versus 340 for PR).

This poll is useful to us in many ways. First of all, it gives us actual vote decisions rather than pure voting intentions. Similarly to my own exit poll currently underway, this can show us how undecided voters ultimately voted. It seems they went slightly more in favour of PR (that is huge if it's true). So that allows me to update my polling averages that I use when estimating the votes (see below). It's especially useful to me because Insight West provided the breakdown by the 2017 votes (i.e: people who voted BC Liberals in 2017 are 80% against PR, etc).

Secondly, it confirms what my estimates have been showing all along: early votes where overwhelmingly for FPTP but the trend has been in favour of PR. As a reminder, the deadline is literally this afternoon (December 7th) at 4:30pm. If you are reading this and want to vote, you'll need to drop your ballots at one of the Referendum Services Offices (go on Elections BC website).

The poll has two weird numbers though. It has Van Island not that much in favour of PR (less than 53%), much lower than other polls. It also has the Green voters supporting PR at "only" 68%, there as well lower than the previous averages. The two results are more likely correlated. With that said, it also has the BC Liberals slightly less against PR and the BC NDP voters more supportive of PR, so the net effect is positive for PR in my estimates.

By the way the turnout is now at 41.1% (and 37.1% of processed ballots). It has increased very slowly this week and it seems it'll remain around that mark. Maybe a late surge today will push it above 42 or 43%. In any case, this is a very decent turnout for such a referendum in my opinion. It's already much higher than the turnout for this referendum in PEI last year and PEI usually votes more than BC.

Ok, let's look at the updated estimates.

I continue to do averages my age and region (and 2017 votes). I gave the Insight West poll a bigger weight since it's more recent. Specifically, I went with 40% for the age and regional numbers (and thus 20% weight for the Angus-Reid, Mainstreet and Research Co. polls). As for the averages by 2017 votes, I gave 40% to the Insight West and Research Co. and 20% to the older Angus-Reid.

Method 1 - Based on age and regional numbers

The YES is behind by around 12k votes. If I also assume that the newly registered voters (8487) are voting more for PR, then the deficit is only 7300 votes (and remember that's with still over 130k ballots yet to be processed).

Remember that these estimates are calculated using the new polling averages, so that's why they are slightly different from my previous ones.

Method 2 - Based on the 2017 votes by party

This method is now giving the YES side ahead and that's without adjusting for the newly registered voters. If I did, the lead for the YES would grow to over 18k (and be at 50.7%). Very good sign for PR.

Reasons to be optimistic for PR

I feel there are many reasons to be optimistic at this point. First of all, regression-based analysis is really confirming that voters aged 18-34 are voting a lot more than in 2017 while the other age groups aren't. Young people are definitely in favour of PR.

Regionally, we also see Van Island so much higher than the others and it's growing. Ridings like the two Victoria ones or the two Saanich ones have high turnout and you can be pretty sure most of these votes will be pro PR. The Lower Mainland is also catching up.

Also, for the first time in my regressions, it seems the the NDP vote of 2017 is getting out. Specifically, the variable "% of NDP votes in 2017" is no longer significant (it's still negative though). At this point, my regressions show the Green party is getting its votes out.

The Insight West poll also showed PR actually more ahead than before overall. And the fact it was a 50-50 race a few days ago among people who had voted is a very good sign. Add to this the late turnout we have seen and that might just be enough.

I'm being bullish on the YES side here but I fully realize the NO can still win. If the main determinant of the vote is the age, then the NO will likely win. The 55+ don't support FPTP the same way the 18-34 support PR but there are a ton more voters in the 55+ group. With that said, it's hard to fully predict at this point. Polls will weigh based on the census and therefore will underestimate the NO (cause the turnout among the 55+ is higher). But young people are voting more this time, so it might cancel out. I need to do some calculations on this. But that's for later.

Finally, without going into complex calculations, polls show 2 out of the 3 regions are pro-PR and these two (Lower Mainland and the Island) represent roughly 75% of the voters. You also have two age groups in favour (18-34 and 35-54 vs 55+) and you have two partisan groups against one (NDP and Green against Liberals). All of this should in theory be enough to push the YES side ahead.

First of all, a giant thanks to those of you who decided to contribute to the Go Fund Me campaign I started yesterday to get a post-referendum poll. This is amazing. I'll start the poll tomorrow and will keep you updated.

Ok, let's use the most up-to-date turnout data from Elections BC. Turnout is now above the 40% mark (so even BC Liberals leader Andrew Wilkinson would have to recognize this referendum as valid...) while the processed ballots are at 35.4%. It hasn't increased much this week (for the overall turnout based on received ballots). Google Trends was showing a decrease in interest in proportional representation for the last few weeks. I feel that people who wanted to vote did so at this point. With that said, given the potentially small margin (see below), trying to get extra votes tomorrow might be important.

Here is a graph from David P. Ball showing where we stand:

PENULTIMATE #PROREP turnout chart! (…until results out ~Dec 20)

40.5% ballots returned (>.6% or ~20,000 since Wed). 30 HRS for @ElectionsBC to RECEIVE ballots (Friday 4:30pm)

Voting PLACES—https://t.co/tcANrgL1LU@StarVancouver's explainers/explainers—https://t.co/01saEkS44c pic.twitter.com/FYOClNqbOm

— David P. Ball (@davidpball) December 6, 2018

Ok so let's try to estimate who is ahead.

Please remember that all the numbers below are estimates. They are based on demographic, turnout and polling data. They are the best I can do at this point but they are still estimates. I saw people spreading these numbers on social media as if they were hard counts. In particular, if the polls are wrong (for instance if the undecided broke down for the YES or the NO side), then these estimates will be wrong as well.

Method 1: using a mix of age and regional data.

Specifically, using the polling averages by regions (Lower Mainland, Van Island and the Interior) as well as by age (18-34, 35-54, 55+; combined with the turnout by age for 2017).

It has been super stable for a while now. The YES is climbing super slowly but remains behind. In this scenario the YES side would lose be fewer than 25k votes (out of over a million!).

All hope isn't lost for the YES side. Regression-based analysis of the turnout show that the 18-34 are voting far more than last year while the 35-54 are voting a lot less. The 55+ (very much against PR) are also voting less. So while the estimates should already take that into account partially (since more votes are coming from ridings with more voters aged 18-34), it might still be underestimating the YES votes. This could be the case if the 18-34 are voting more everywhere, including in ridings with more 55+. I will try to adjust for this in a future post.

Method 2: using the votes in 2017 (for each party)

In this estimate, I use the polling data by party (percentage of BC Liberals who support PR, etc; I'm using an average of the Research Co. and Angus-Reid polls with a higher weight for the former as it's more recent. If I were to use a straight up average, the YES would be ahead in the graph below). This might actually be the best method as vote patterns might capture a lot of effects, much more than accounting separately for the age and the region. Plus this referendum campaign has ultimately turned very partisan.

This is just insanely close. The difference is 2300 votes!

If I also adjust for the newly registered voters (around 8400) which can reasonably be assumed to have registered more in order to vote for PR (so assuming they vote PR at 75%), then I have the YES side AHEAD by 1700 votes! Imagine winning this referendum with a 50.07% margin! Take that 1995 Quebec referendum!

My conclusion remains the same: it's incredibly close. Maybe our post-referendum poll will help us there (hopefully, maybe we'll see that young people chose the YES even more or that the undecided that were over 55 also rallied behind the YES).

Is the turnout still affected by when a riding received the ballots?

A regression-based analysis is required here. As a reminder, not every riding got their ballots at the same time. Elections BC is only providing the date at which it was supposed to all be received. However, I came up with my own measure because even among ridings scheduled on the same day, you can clearly see differences (possibly due to the rotating postal strikes). I use the numbers of days the ridings has been above the 1% turnout threshold.

Results are below:

So the answer is yes, the variable "number of days of voting" (measured as the number of days above 1%) is still significant. The coefficient hasn't changed much recently either. This could potentially be hugely problematic. Elections BC still has a lot of ballots to process and it's possible it'll even out at the end, but we are one day before the deadline and it hasn't yet. Coefficient of 0.03 means a turnout higher by 0.3% for every additional day of voting. And that's after controlling for the region and other factors.

I even ran robustness checks (with the help of Jameson Quinn, Ph.d student in statistics at Harvard) to make sure the effect was there. Specifically, I identified the ridings that were less "enthusiastic" or motivated to vote and used "number of days above 0.5%" for them. The idea being that my measure could have been biased since ridings that are less motivated would also reach the 1% mark later. Doing so changed nothing to my regression (in one version it decreased the coefficient from 0.03 to 0.027).

So in average, and after controlling for multiple factors, I still find that a riding that got its ballots 10 days sooner has a turnout about 3 points higher. This is potentially important because there are a large group of ridings (19 to be exact) that were scheduled for November 2nd and that only crossed the 1% mark 19 days ago while 17 ridings scheduled for Oct. 24-26 (mostly in the interior) crossed the 1% mark 32 days ago!

With that said, I also ran estimations of the impact of this turnout on the votes and found nothing. Specifically, I removed the effect of the number of days on the turnout (so I simulated the turnout of every riding if they all had the same number of days) and re-ran my estimates above. There was no change in method 2 and a tiny increase (of 0.04%) for the YES with method 1.

So while I'm fairly convinced the turnout was indeed affected by when a riding got their ballots, it seems that the overall impact on the YES and NO side is zero.

The voting period for the referendum on electoral reform is coming to an end this Friday at 4:30pm.

We won't know the results for weeks. In 2011 for the HST referendum, Elections BC took almost 3 weeks to count all the ballots and publish the results.

Are you like me and you'd want to at least get an estimate sooner?

So here is what I'm attempting: let's crowdsource a poll! Google Survey allows us to run actual surveys/polls with actual sampling. I'm not talking here of those non-scientific polls done on websites. No, this is a proper poll with proper sampling and weighting, all done by Google.

It's also the cheapest option. At 20 cents per respondent, we can get a decent sample size for $200-$250 in total.

Such Google Surveys have been used in the US with great success. I myself ordered 3 this year. One as a test during the Ontario election (with results quite good to the final ones, although my poll was done in the middle of the campaign), one about Maxime Bernier's infamous tweet (can't compare to other polls) and finally one earlier during this referendum. This last one should convince people of the validity of the method as it gave results almost identical to the "regular" polls from other firms.

Yes I could order this poll myself but, as I said, I already ordered 3 this year. My regular job pays me well but I also don't want to keep spending hundred of dollars just to run this blog. Also, I believe this could be an interesting experiment: can we crowdsource the funding for a poll in Canada?

And just to be clear: if you think I'm making a ton of money off Google ads or whatever, you are wrong. This blog does generates some revenue during an election (like in October in Quebec) but it's nowhere near making a living and it has been super low during this referendum. So I'm not asking you to pay for a poll that will ultimately generates a ton of cash for me.

So I'm asking for your help, you fellow political nerds. You can pledge very small amounts of $5-$10. We don't need that many people to contribute to reach $250.

Once the funding is reached, I'll order the Google Survey. Ordering 1000-1200 respondents will take a few days. By experience it's slow at first, then picks up and is really, really slow for the last 50 observations. But by then we'd already have over a 1000 observations which is better than most other polls. I'll obviously only target respondents in BC aged 18 and over.

I intend to ask the following question:

"There was a referendum on electoral reform in British Columbia that ended this Friday. Which option did you vote for?"

3 choices:

"I voted to switch to proportional representation"

"I voted to keep the current system"

"I didn't vote"

Now, let's be clear, we'll get a lot of "I didn't vote". This is inevitable. This is why we want a large sample size to get enough voters. But we won't have undecided and this could be important. My daily estimates of the votes show that this could be a very, very close race.

If you want to help, here is the page. Don't hesitate to share it. Ideally I'd like the funding secured by Friday so that we can start right away. If it doesn't work, oh well, I tried.

Alright, let's use the newly published turnout numbers from this morning to estimate the YES and NO side again. with different methods. All methods use some form of polling averages (by age, region, etc). Remember, those are estimates. You could call them educated guesses and I wouldn't be too mad.

Quick note: Elections BC's numbers are the same for December 1st, 2nd and 3rd. So the graph stops at December 1st technically.

1. Using age and region

It has been stable for a long time now. Yes the Lower Mainland and the Island have been catching up after receiving their ballots later (Van Island is ahead actually now) but it doesn't seem to influence the overall shares much. And that is despite the fact that my regressions show the 18-34 are now the only ones voting more than in 2017. The 55+ are now voting less and this is statistically significant. Good news for PR but maybe not by enough.

Remember, these estimates here are the most conservative ones and the most "against PR" I have, mostly because it's a function of age and in BC, the 55+ just represent a large share of voters (and they hate PR).

2. Using age and region but assuming the two are independent from each other

See my previous blog post for details. Similar to the one above but I try to account for age and region at the same time using assumptions.

3. Using the votes in 2017 by party (and using the Research Co. and Angus-Reid polls only with a weight of 25% for the latter as it's older)

The nail biter one! The YES would actually be ahead if I used a 50-50 average with the older Angus-Reid poll. This shows how close it is. At this point, it could come down to the (around) 8000 newly registered voters. Or it could come down to how the BC Conservatives and independent voters side.

If I assume 75% of the newly registered voters will vote YES, then my estimates would be ahead by fewer than 500 votes!!

Conclusion

Still a tough (impossible?) one to call. Some things look good (turnout on Van Island, the fact the ridings with more 18-34 voters are voting more) but it remains that the voting intentions by age makes it hard for PR to win. Yes the regression shows the 18-34 voting more but that might not be enough to beat the NO vote of all the 55+ (who represent a bigger share of the electorate overall).

Technical note: I still find that when the riding received its ballots is a strong determinant of the turnout. This is crazy! Although it might really be capturing more the effect of the random strikes of Canada Post than anything else at this point.

In my estimates that are the most against PR, the YES side trails by 24,000 votes. Elections BC has received 39% of the ballots but processed only 32.6%. This means there still are 209,579 ballots not processed. And more will come until December 7th. All it takes for the YES to win is that that the new ballots are around 55% in favour of PR.


Update:

Some have asked me to post my regression tables updated with today's numbers. Here there are.

Notice that I'm now using the % for each party in 2017 instead of simply accounting for whether the riding was won by the BC Lib, BC NDP or BC Green. The percentages provide more information but it also make the regions less important (so for instance it's not so much that people on van Island are voting more, it's that ridings with a high percentage of Green voters do; The two are obviously correlated).

Alright, it's almost over! One week exactly and Elections BC will stop accepting ballots and will start counting them. Turnout is now at 37% (received ballots; it's 31.1% of processed ballots). So the good news is that the final turnout will be decent. More importantly, by being above the 40% mark, it ensures the Liberals leader Andrew Wilkinson won't be able to call the result illegitimate. Now the questions is really: who will win?

The turnout is actually still ahead of the one for the HST referendum:

New #prorep #s: 37.2% of ballots received (not incl those still at CanadaPost after 3 days of local strikes early in week). Slight uptick after lull; 7 days left for all to be IN to ElectionsBC

Baffled by #bcpoli ballot? https://t.co/Ao1WZAcvMX

SUBSCRIBE https://t.co/dwLfce4FBI pic.twitter.com/7KmeyQU9wX

— David P. Ball (@davidpball) November 30, 2018

Although we are closer to the deadline than we were for the HST vote. If we instead look at the number of days remaining, we are slightly behind:

@2closetocall pic.twitter.com/KDY5BPqenW

— PR Meme Stash (@PrStash) November 30, 2018

It seems a finish around 45% is likely and this is much higher than I'd have thought.

Here below is the current turnout by riding. Remember that some regions received their ballots later and/or were more affected by Canada Post's strike.





So, who is likely winning?

Throughout this campaign, I tried to estimate the number of votes cast for the YES and NO side. Ever since I switched to using the turnout by age provided by Elections BC, I've had the NO side ahead but with a decreasing lead. It has been quite stable the last few days.

Remember that these are estimates. They are subjects to margins of error, in particular from the polls they are based on (technically I'm using turnout, demographic and polling data to provide these estimates).

So first, let's see how sensitives these estimates are to a change of specifications.

1. What if we apply some margins of error to the polling averages

I'm using the four published polls during this campaign (from Angus-Reid, Mainstreet, Insights West and Research & Co.; I'm not using my private Google Survey even though the results were very similar, mostly because I don't have as much details regarding the location of the respondents).

But those are polls, they have margins of error even when we aggregate them. The issue here is that I'm not sure how accurate these polls are. For federal or provincial elections, I have done the research and know the empirical margins of error. For referendums however, we don't have many data points. In general polls have done quite well for the 2009, 2011 or 2015 transit plebiscite, but it doesn't mean it'll always be the case. The number of undecided this year is much higher as well.

These polls have also been remarkably similar. So let's use the standard error of these four polls and create an interval by adding or subtracting 1.96 times this standard error. That gives us margins of error of plus or minus 1.5% (and more by age or regional breakdown but let's ignore this for now; Errors are likely to be correlated anyway. For instance if the polls underestimated the YES in the Lower Mainland, they'd have done the same on the Island).

Ok, so let's re-run the estimates above but using either the upper or lower bounds for the support for PR.

% for PR with upper bound: 50.4%

% for PR with lower bound: 47.4%

So the YES can win but the interval shows it's not the most likely outcome.

2. How important is the age turnout?

These estimates work this way: using the turnout by age of 2017 (thus telling me the percentage of voters aged 18-34 in each riding for instance) and the region, I look at the number of votes in one riding and "assign" them to the YES or NO based on the demographic data (the age thingy) as well as the levels of support for this age group in the polls.

The issue here is that around 50% of voters in 2017 were actually above 55 (this is a crazy stat as the 55+ do not represent remotely close to 50% of the population). Given that polls show the 55+ want to keep FPTP in majority (63% exactly), this gives the NO side a huge boost.

But here's the thing: polls always weigh their samples based on census data, not turnout one. At first it seems like a perfect recipe to be wrong (since you'll underestimate the impact of the 55+ and over estimate the 18-34 for instance) but in practice it has been relatively fine. Look at the polls for the BC election last year. They did a fairly good job. If you re-weight the polls with the age-turnout data, you usually get a higher number for the Liberals than what the polls showed. This is typical in pretty much every single election where one party does better among the 55+. But again, in the real life, we don't observe such parties to be systematically underestimated.

What I'm saying here is that my method above might actually be quite harsh against the YES side.

There is also the issue that regressions (see below or previous blog posts) have shown a clear trend: the 18-34 are voting more than in 2017 while the 35-54 are voting a lot less (the 55+ were voting more at first but that's not the case anymore). So there as well I might be underestimating the YES side (note: my method should technically account at least partially for this since more votes are coming from ridings with more voters aged 18-34, but I'm not making any additional adjustments such as increasing the share of the 18-35 at the riding level).

Finally, I would technically need the polling data broken down by age AND region (like what is the level of support for PR for the 18-34 in the Lower Mainland?). Polls don't provide me with this. They provide the average by age OR the region. My method above basically runs the estimation using each and does and does an average.

I contacted the pollsters to try to get the data I want (because they have it), we'll see if they nicely oblige. In the meantime, we can try one thing: let's assume the two (age and region) are independent from each other. So if we compare the levels of support for the 35-54 to the 18-34 in the Lower Mainland, we should get a proportionally similar result as comparing the 35-54 to the 18-34 on the Island for instance. This assumption will amplify the differences. For instance the 18-34 (more in favour in general) living on the Island (generally more pro-PR) will therefore be really, really pro-PR. On the other hand the 55+ in the Interior will be incredibly against it. That might actually be a good thing as I suspect that the two campaigns are motivating the more "extreme" voters (the ones really in favour of PR and those really against it).

If I do this instead, I get the following graph:

Same trends but the YES is now ahead. Interesting. The assumption of independence is really strong (and likely invalid) but at the same time, is it worse than doing both separately and averaging? Not sure. Please somebody smarter than me could let me know.

3. What about new voters?

8487 people registered to vote by the deadline of Elections BC. That's not a large number (it's 0.26% of the registered voters). However, it's reasonable to assume that a majority of these voters did so to vote for PR. Why would people who weren't registered decide to go and vote to keep a system they weren't even using? I'm sure there are some people like this (maybe to protect the province from all the Nazi parties that PR will inevitably brings if you listen to some people lol) but the majority of these 8487 should go into the YES side.

If we assume that 75% of these new voters will pick the YES side, then my estimated percentage for the YES increases by a little bit less than 0.2%. Minor? Yes, but that could well be the difference at the end.

4. What about using the votes for each party in 2017?

Maybe instead of trying to mix age and regional data, the best and easiest method is to take the 2017 results by ridings and use those as benchmark. So a riding with a majority of votes for the BC Liberals will vote against PR. This should capture both the age and regional effects.

Unfortunately for us, most polls did not provide the levels of support for each side by the political choice in 2017. Only Research Co. and Angus-Reid did so and we can see that the BC Liberals voters really, really hate PR (81% in favour of keeping the current system) while NDP and Green are mostly in favour (around 70-75%).

So now my calculations are as follow: imagine a riding voted at 40% for the Liberals. Therefore, 40% of the votes for the current referendum will be assumed to come from the BC Liberals voters and 81% of them will vote NO. The regressions below show that the 2017 turnout is a strong predictor of the current turnout, so this method might actually be quite valid.

If I use this, I get the following graph (yes I know this post has the same graph many times but that's the point of this article):

One potential problem with this method is that it doesn't account for how the other voters will vote this time (in particular the BC Conservatives). Still, we see similar patterns but here again the YES side seems narrowly ahead. The trend for PR is quite steep but the scale is misleading as it shows results really, really close to 50%. Interestingly this is the method that gives us the closest race so far and the smallest absolute variations (it stays between 50.8 and 49.2%).

5. Regressions

What the polls have shown is that people in BC are mostly split 50-50. So using the turnout by riding so far, we can try to find patterns (and stop trying to actually estimate and count votes). Here are the results with the latest numbers:

If you've been reading my blog during this referendum, you already know that the same regressions were initially showing a positive coefficients for the share of 55+. Over time it has switched from a positive and significant effect to a negative but not statistically significant one. In the less simple model, we are actually getting close to getting statistical significance.

As for the regions, the Lower Mainland and the Island are above once we adjust for the number of days they had to vote (which is, again, still a determinant of the turnout even this late into the referendum, thus showing once again that Elections BC was right to extend the deadline).

The major impact of the % of people with English as mother tongue is just confirming to us that the Vancouver suburbs of Richmond And Surrey aren't very interested in this process so far.

If I had used the percentage for each party (instead of looking at whether the riding elected a Liberal or NDP MLA last year), then the share of 55+ is negative and significant. And in this case the higher the % for the Green, the higher the turnout (significant at 10%) while it's the opposite for the % for the NDP (thus again showing that this party isn't getting its vote out as much, although this could be partially explained by Surrey).

Conclusion

It still a very close race but the trend remains favourable to the PR side. Margins of error make predicting this referendum almost impossible but there are reasons to be optimistic if you want to change the electoral system.

The regressions show that the 18-34 are voting more than they have in the past while it isn't the case for the 55+, at least using aggregate data (so subject to a possible ecological fallacy - Google it if you want to know more). Regionally, the effect of when the ballots were received (and/or of the postal strikes) are decreasing and we gradually see the Lower Mainland catching up the Interior while Vancouver Island is now firmly ahead. Both trends should be positive for PR.

On the other hand, ridings with a BC Liberals MLAs continue to vote more (we see it with Vancouver Quilchena for instance) and this is good news for the current system. I heard that the BC NDP was now making phone calls to motivate their voters to get out and mail their ballots. Hopefully this isn't too little too late.

Personally I'm becoming cautiously optimistic for the YES side but I'll wait to see the final turnout to really make a call.

Simple, interactive map of the turnout (processed ballots) as of November 30th 2018.

Four categories, two below the average and two above.

That is all. Hope you enjoy.

Interesting new turnout data published by Elections BC yesterday. Let's take a look.

Summary for busy people

Using all the information available so far (polls, turnout by riding, data about the turnout by age in 2017), here is my best estimate of the current situation in the referendum on electoral reform in BC.

The NO side (keeping FPTP) is still ahead but the trend is favourable to the YES side (switching to proportional representation). The gap was 4.8% at the beginning and is now only 2.3%. Over the last few days alone, the YES side has progressed by an estimated 1 point.

How do I estimate these votes? Knowing the number of votes from each riding and knowing the age profile and regional location of these ridings, I estimate how any of these votes are likely a YES or a NO. Without going into the details, a riding in the interior with a high share of voters over 55 is likely to vote in majority to keep the current system. The exact calculations involved the polling averages and the turnout by age of 2017.

If you are among the dozen of people who actually read my blog on this topic, you should remember that my previous post was saying the YES might have been ahead already. So what happened? Well I was using the census data to get an age distribution by riding. But Elections BC has something better: the turnout by age by riding in 2017 (and before). This is a more accurate measure of who is potentially voting, but this is also a measure that hurts the PR side because older voters represent a disproportionate share of the voters (see below). So it isn't that I changed my mind or trying to generate clicks on my site, it really is just an updated version of what I was doing. The good news is, I believe, that my model will now remain the same. If I were to use the census age profile of my last post, then PR would be ahead 50.6% to 49.4%.

1. Current turnout

Based on the data published by Elections BC on Friday, the turnout is already 30%, although it falls to 18.9% if we only count the ballots that have already been processed. The roughly 12-points gap seems to be fairly consistent (i.e: both measures increase at a similar rate). Extrapolating the trend for processed ballots until the 30th and assuming the gap remains the same, it seems the turnout could end up close (or even above) the 40% mark (which, incidentally, is the arbitrary threshold set by the BC Liberal leader Andrew Wilkinson to recognize this referendum as "legitimate"). This is higher than I was predicting at first. One reason is because my early predictions were based on the turnout among processed ballots (Elections BC only mentioned they had way more ballots than the ones processed relatively recently). It's actually not impossible that we end up with a higher turnout than for the HST referendum!

Using Google Search as a measure of the interest for this referendum, we can see a declining search activity since the peaked around the debate (November 7th). So I'm not sure we should expect a late surge in turnout, but we never know. This whole analysis is also made more complicated because of the possible impact of the rotating strikes of Canada Post.

Overall though, the turnout will go from decent (low 30s) to quite good (40% and more). For a referendum coming in November, after a municipal election and regarding a topic that isn't the most interesting for many people, I consider this a small victory. Fears of ridiculously low turnout that would make this referendum invalid can be set aside it seems.

The current turnout is still heavily influenced by when the riding was scheduled to received the ballots, although the November 2nd group is clearly catching up to the 23 ridings that got their ballots early.

As explained in a previous post, the scheduled date from Elections BC isn't perfect and we know for a fact that ridings on the same schedule didn't received their ballots at the same time. This is why I prefer using my measure of "days where the turnout has been over 1%" as my measure of how many true days each riding has had to vote.

There is still a strong relationship but it's converging for sure. There shouldn't be any relationship by the end of this process (if there is, then this could be problematic as it'd mean the ridings that got lucky and received their ballots earlier voted more).

2. Analysis of the preliminary turnout

First, let's look which region is voting more (the Lower Mainland, Vancouver Island or the Interior). As of Friday, here were the turnouts (screened ballots):

Lower Mainland: 17.2%

Island: 21.1%

Interior: 19.9%

As comparisons, during the 2017 BC elections, these rates were, respectively, 59.7%, 66.4% and 60.2%. Also, and maybe more importantly for this referendum, the last few days have seen a dramatic increase in the Lower Mainland. It seems the many ridings that got their ballots later started voting (or Elections BC started counting them). Just 3 days ago the turnouts were, again respectively, 8.8%, 14% and 15%. The regions that are known to be more pro PR (Lower Mainland and the Island -where the BC Green are the strongest) have caught up with the interior (and actually surpassed it for the Island).

This trend is the big reason for the trend in favour of the YES side in the graph above. If it were to continue until Elections BC is done receiving and counting the ballots, I think the chances of a victory for the YES would become quite good.

The region isn't the only important factor. Polls have all shown that age is a crucial determinant of the level of support for proportional representation.

Looking at the 2017 election, we can see how much the 55+ voted more than the 18-34. Here is the turnout by age group for that election:

18-34: 49.3%

35-54: 57.7%

55+: 69.9%

In average, the 18-34 represented only 18% of the voters while the 55+ accounted for 49.8% of the ballots cast. So there is no question that the fact FPTP is the most popular option among the 55+ crowd is making switching to proportional representation difficult. Turnout in 2013 were also fairly similar.

3. Regression

We obviously don't know who is voting in each riding, we just know the total number of votes so far. Still, we can try to see some patterns. But for this, we need a regression. If you don't know anything about statistics, see a regression as correlations measured while holding other factors constant. So since ridings in the interior also have more older voters, it's hard to disentangle which variable is affecting what. A regression solves this problem. The significance level is simply a statistical measure of how confident we are of the results. If there is no * next to the coefficients, it means it isn't significant and we consider that there isn't a relationship between the turnout and this variable. The more stars, the more confident we are (* for 10% uncertainty, ** for 5% and *** for less than 1%).

What I want to know now is if, after controlling for the number of days each riding had to vote and the region, the current turnout is still higher in ridings with more voters aged 55 and over.

The table below shows you the results:

As mentioned above, I'm now using the actual turnout by age to determine the age composition of each riding. Thus instead of using the proportions of people aged 18 to 34, I'm using the share of voters that were 18-34 in 2017. I'm including the turnout in 2017 as a control. I really want to measure differences between this referendum and the previous election. Why? Because if we were to observe a significant different (for instance 55+ voting much more), then my vote estimates of above would likely be incorrect.

What we see is that ridings with a bigger share of 55+ are NOT voting more anymore. At least if compared with the 18-34 group (this is really how you should interpret the coefficients). It seems this referendum is motivating two groups to vote: young people (who want PR) and older folks (who want to keep the current system). The 35-54 (who are the most 50-50 on PR based on polls) seem less interested.

Just a technical note here: I'm using aggregate data (at the riding level) and therefore my regressions are subject to the typical ecological fallacy. What is this? Let's use an example. In the US, traditionally, richer people vote Republican (that might have changed with Trump but let's ignore him). At the same time rich states (NY, California) vote Democrat. So the correlation at the aggregate level doesn't match with the one at the individual one. In BC, we observe a similar effect where ridings that vote NDP have a higher proportion of Asian people but polls have shown that Asian people are more likely to vote for the BC Liberals. What does it mean for the current regression? Well hard to tell, but you should at least keep in mind that I'm not actually observing the age of the current voters, only that ridings with more older and younger voters are voting more. For instance, imagine a scenario where this referendum is motivating young people in ridings with a higher share of older voters, then the higher turnout from those ridings would actually be favourable to PR as opposed to my assumptions.

The results here are quite different from the ones a few days ago. If I used the turnout of just 3 days ago for instance, I'd find that the 55+ were voting more and the Lower Mainland was lagging behind!

If I try to include more variables (percentage of people who have English as mother tongue or whether the seat was won by one or the other), I find the numbers of the right columns. It doesn't increase the R squared coefficient much (i.e: how much of the variation I can explain with these variables) and I find the same effects for the age or the regions. I do find that Liberals-held ridings are doing better than NDP-held ones. This isn't new and you have to wonder how much the NDP is getting its vote out.

Overall, the new regressions increase my confidence in my estimates above. The fact the previous turnout (in 2017) is a strong predictor means my age-turnout based on 2017 is most likely valid as base. After, we do observe the 18-34 and 55+ voting more than the 35-54. The net effect is unclear as the 18-34 are heavily behind PR whole the 55+ are in majority against it. There are more people in the 55+ group though. All in all however, the two might cancel out.

Remember that my vote estimates above are valid to the extend that the polls are right. In particular, the polls have to be right regarding the regional numbers as well as the levels of support by age. Additionally, these polls have found a large number of undecided and I'm assuming here that these undecided will either not vote or will break similarly to the decided voters. That'a big assumption and an error here (let's imagine undecided ultimately choosing the status quo) would make my estimates all wrong.

Conclusion

If the vote count was ending today, I think I'd bet on a small victory of the NO side. But with a full two weeks to go (note: Elections BC actually announced an extension to December 7th) because of the strike of Canada Post) and looking at the trend, I think proportional representation has a chance. It remains an incredibly close race and trying to predict the result is just a recipe to be ultimately wrong, but I'm okay with this! This is just occupational hazard for me at this point.

Can the 18-34 (in the Lower Mainland and on the Island) continue to vote so much and make the difference? Regionally, the large increase in the Lower Mainland is nice but I wish (as a PR supporter) I'd see ridings such as Point Grey or the two Victoria ones (on the island) getting higher turnouts than Quilchena for instance (Wilkinson's riding and the one that voted the most NO in the 2009 STV referendum).