Quick post-mortem for Ontario 2018

The Progressive Conservatives proved the polls right (well, mostly, see below) and got their majority. I don't have all the results per riding but I have enough to write a quick post-mortem to an election which ultimately was a good one for me and people making projections.

1. The polls.

Polls weren't bad but they also weren't great, at least not in average. The IVR polls (Forum, Ekos and Mainstreet) did really well while online polls (and live callers) missed the PC-NDP gap by quite a lot.  And one firm, Abacus, missed completely. Mainstreet had the best overall accuracy but they didn't have the Green within their Margins of error (note: MoE vary with the level of support for a party. A party at 5% does NOT have the same MoE as a party at 50%. Polls usually only report the MoE for a party at 50%). Forum had all 4 parties within the MoE... so hard to say who won. Let's call it a tie between Forum and Mainstreet. Funny enough, these are the two firms often criticized online (I have some annoying people on Twitter for instance who always want me to ignore these firms lol).

See the table below for the detailed performances:



I had mentioned the IVR-online gap before but since the two methods had been equally good at Canada over the last decade, I couldn't pick who was right. With that said, even IVR polls underestimated the PC and overestimated the NDP. Should I have had the PC higher given the late trend? Maybe but the late polls were also mostly all IVR (except Research Co.) and I didn't want to trust one method more than the others. Hindsight is always 20/20 but I'd do the same again. I'm not in the business of trying to guess who is right, I follow the average and the numbers.

People will likely remember this election as one the pollsters got right (they were predicting a PC majority and got one) but going from a 2 points PC lead to a 7 points one isn't close at all. Abacus was the worst one by far and they'll argue that it's because they finished polling 6 days before the election. While it obviously doesn't help, it also shouldn't matter to you. First of all, Abacus could have polled later. They simply chose not to. And I strongly suspect they did so to have a way out in case they were wrong. Second of all, Abacus was showing a bigger NDP lead than most other pollsters anyway.

I think IVR pollsters got it right. Online ones were off.

A note about the Mainstreet riding polls. They ultimately were not bad but I feel vindicated in the point I had been making for a week: they were consistent with a province-wide result where the PC is largely ahead of the NDP. It's interesting that the riding polls were showing this situation way before the provincial polls.


2. The projections.

I'm quite happy with my forecast. The average deviation per party is 3 seats, which is one of the best forecast I saw out there.

I correctly called 110 ridings out of 124, that's a 89% success rate (it pisses me off to miss on the 90% mark!). And 1/3 of the mistakes are from Scarborough! I'm two three mistakes away from Earl Washburn from Ekos, the guy I consider as being the best in this country (and he does have access to riding polls from Ekos that we don't get to see). So I'm quite happy. With the actual vote percentages, I'd have actually made 16 mistakes! But I'd have had the PC at 82 seats and the NDP at 40, so closer than my normal forecast. I'd have missed the Liberals even more though.

Here are the ridings I missed. For each, I also discuss whether I could have done better.

Brampton Center: was projecting a close race with the PC winning by 0.6pt, the NDP won it by 0.26pt! Let's be clear here, getting such a close race right is just luck, nothing else. I'm quite pleased with the forecast.

Brampton-West: bigger miss there but I had the PC with around 20% chances of winning, so not a shocking surprise. The polls underestimating the PC lead didn't help though. By the way, the riding polls also had it going NDP...

Chatham-Kent-Leamington: Pretty big miss percentage wise, but the probabilities were showing the PC had a 36% chance of winning. Interestingly, I had noticed that my regional average in the SW was becoming too biased in favour of the NDP compared to the polling average over the last 2 weeks. I corrected it for it but not fully. That explains part of the miss.

Don Valley West: Wynne won by a small margin and proved me wrong. I don't have much to add here, it's a miss for sure. The riding poll from Mainstreet had it right.

Kenora-Rainy-River: the riding poll was right and I did include it, but ultimately my model still had it going NDP. A clear miss too.

Orléans: one of the seats the Liberals managed to save. Look, I was expecting the OLP to be stronger than expected in some seats, the question was to identify (guess?) which ones. The riding polls were even more off than me so I don't feel too bad.

Ottawa South: similar story here, including the riding poll being wrong. I'll say one thing though: the one trend over the last week for the Liberals was an increase in the East. So I was expecting the Liberals to possibly keep more seats than forecast in the Ottawa region but I didn't include this subjective reading into the model (beyond adjusting the regional numbers of course). So it didn't shock me when the Liberals kept it.

Sarnia-Lampton: similar to Chatham-Kent-Leamington. The riding poll was better.

Four mistakes in Scarborough: Not much to say, this is one region where I got it wrong (or got unlucky). With that said, the Liberals keeping Scarborough-Guildwood is the biggest surprise of the night and I haven't seen anyone projecting it to remain red. The chances of winning were only 9.24%. Still, when your biggest mistake is one where the chances were 9.24%, you know you had a good night! After all, if my probabilistic model is correctly calibrated, you need some people projected with only 10% chances to actually win!

Thunder Bay-Superior North: the North is always more unpredictable, although most were expecting the Liberals to keep the other Thunder Bay riding.

Toronto St. Paul's: I knew the NDP had its chances there but given the advance turnout that wasn't increased compared to 2014, I thought the NDP surge in the core of Toronto wouldn't take this riding. Oh well.

So, any mistake I regret? Maybe the two in the SW (Sarnia and Chatham) as there was some data showing they would go PC. Maybe Don Valley West too (although advance turnout showed a big increase, which isn't usually good for the incumbent). For the rest, I consider correctly predicting almost 90% of the ridings correctly to be a good night. I don't think it was really possible to anticipate 7 OLP wins. Not without doing some crazy subjective adjustments.


The probabilistic model worked again very well.

Projected chances of winning Actual win percentage
0% 0%
1-20% 2%
20-40% 29%
40-60% 52%
60-80% 64%
80-99% 99%
100% 100%

Adding this election to the others  (so we have more observations as some bins/cells are quite low for one election only), the model has now the following accuracy:

Projected chances of winning Actual win percentage
0% 0%
1-20% 9%
20-40% 30%
40-60% 50%
60-80% 64%
80-99% 88%
100% 100%

This is a well calibrated model. The only one not perfectly matching is for the 60-80% range. But it's still close enough.

That's it for now. I'll not write as much in the coming weeks as I want to take a little break (and there is nothing to write about anyway lol). I hope you enjoyed my coverage during the election and see you next time.

Final projections for Ontario 2018: A Conservative majority

ติดการพนัน ภาษาอังกฤษ
Note article published at 6:47pm PT. I might make small changes between now and tomorrow morning. But I need a break for now. The general outcome will remain the same anyway.

Here we are, finally! After around a month of campaigning, Ontarians have to decide who they want for their government. Based on the polls and past election results, the most likely outcome is by far a Conservative majority. There are, however, reasons to believe a surprise victory from the NDP of Andrea Horwath is possible. Only certainty is the reign of the Liberals will end tomorrow.

This is a long article. So I put some key points summary at the beginning (for you busy people) and then I dig deeper into the numbers and the analysis.

First of all, here are the final projections with the riding by riding breakdown. That's what a big majority of you want to see on this site.

Voting intentions; Seat projections with confidence intervals at 95%; Chances of winning the most seats

I used all the polls published in the last week. I allocated undecided evenly 1/3 between all three major parties (as to account for the general underestimation of the incumbent and the overestimation of the smaller parties).

And now the riding by riding projections. Remember that projections can never be perfect. I recommend that you focus more on the chances of winning than on the raw percentages.



Key points:

- Progressive Conservatives enjoyed a late campaign bump, pushing them back at 38%-39% percent.

- The NDP stalled after a continuous rise for the first 2-3 weeks.

- The popular vote lead combined with higher vote efficiency means the Tories are by far the favourites tomorrow. It really isn't debatable. Yes polls can be wrong and surprises can happen but it remains that a NDP win would indeed be a surprise.

- Elections in this country are often won in the suburbs. If the trend were to continue tomorrow, there is no question "Ford nation" will bring a lot of blue to the GTA and the 905 in particular.

- Depending on how you measure uncertainty, the NDP's chances vary between 10 and 15%.



1. The polls

This election has mostly seen two dynamics. The starting situation was the PC comfortably ahead at 40% and the NDP and Liberals fighting for second place. Then the NDP started rising, apparently taking votes both from the OLP and PCPO.


This rise turned into a pseudo "orange wave" when some pollsters started showing the NDP up by as much as 14 points! The polling average started putting the New Democrats ahead for quite some time, although never enough to be projected with the most seats (see vote efficiency below).

Finally, the final dynamic happened during the last week with a PC rebound. Not sure what happened exactly but the Tories definitely increased at the end. One possible explanation is some Liberal voters switched. The PC rise at the end is correlated with the decrease of the Liberals. Also, if you look at the second choices (from Mainstreet for instance), you see that PC voters are now more evenly split between NDP and OLP as 2nd choice. That wasn't the case earlier this case when Tory voters were much more likely to have the NDP as 2nd choice. Therefore this is consistent with the theory that OLP voters switched at the end. Maybe when they saw that the NDP could form the government? This is pure conjectures at this point.

So the Conservatives are back around 38% and ahead of the NDP. Most pollsters agree with a close race (with a slight edge for the Tories). Some pollsters are more bullish for the PC. Mainstreet and Ekos (along with Ipsos to some extend) are showing a very comfortable PC lead (over 5 points for Mainstreet, 4 for Ekos). They agree with the PC around 38% but they disagree with the NDP and OLP. On the other side of the spectrum is Abacus which put final numbers with the NDP ahead by 4 points (although their poll is slightly older than the others).

We had a IVR versus online poll discrepancy not too long ago (and actually all campaign long) and the final numbers tend to confirm this divide. Ekos and Mainstreet show the two biggest leads. Given that my research has shown no significant difference in the accuracy of IVR or online polls, I'm happy to have an almost equal mix of all types on my final average.

[Update] Both Forum and Research Co. released late polls today. The latter is online and has a very small sample size. It has the PC at 39% and the NDP at 37%. Forum is IVR and has over 2000 respondents. Forum is showing the PC ahead by 5, 39 to 34, thus confirming the difference between IVR and online polls. We'll see tomorrow who is right and who is wrong. I won't update the projections for now for that one poll. I ran the numbers and it doesn't change anything anyway. It does push the Tories past 38% in the polling average though. So since I published already, I'll just leave it as it is.

In short, from the most to least bullish for the PC, we have Mainstreet, Forum and Ekos, then Ipsos, next come the online (and live callers) polls (Pollara, Leger, etc). Finally Abacus is alone in showing the NDP not only ahead but comfortably so (4 points). Abacus will either be very right or very wrong tomorrow. This election might also help us decide which methodology is superior.

How much can we trust these polls? When I looked at past elections in this country, I found the following:


There were two big misses during that time: Alberta 2012 and BC 2013. The rest was mostly fine, although usually not perfect. What this means is that, at the very least, you should be prepared for the PC to be anywhere from 36 to 40% and the NDP between 34 and 38%. And that's going with the average accuracy only. Want to be more confident? Then you need to use the margins of error. The graph below represents the possible outcomes for the popular vote based on the polls and past polling accuracy. This is also the base for my simulations and what gives us the probabilities.

This means the range for each party is the following:


Yes this is pretty wide! Hopefully the polls will be accurate tomorrow. If not, you can't tell you weren't warned of the possible ranges.

Notice that the simulations I do randomize the popular vote (the graph above) but also at the riding level. There is a double randomization process in order to (try to) account for the vote distribution.


1.1 Riding polls versus province-wide ones?

I have been very vocal during this campaign regarding the riding polls from Mainstreet (available to subscribers only). Mainstreet did an amazing job with providing us with a ton of riding polls. The issue I have is the numbers in these polls weren't consistent with the province-wide ones. They had the PC systematically higher and the NDP systematically lower than what you'd expect. I maintain this opinion. With that said, the riding polls are at least mostly consistent with Mainstreet's provincial numbers. The discrepancy became smaller with the late campaign jump for the Tories though.

Let's be very clear here: if Mainstreet is right (both provincially and at the riding level), this will be a very comfortable Tory majority. Mainstreet is showing the PC almost sweeping the 905, including Brampton (4 polls, 2 PC and 2 close races). They are also showing the PC very competitive in ridings you'd never expect them to win such as University-Rosedale (difficult to imagine the Conservatives winning that riding tomorrow!).

I made the choice to follow the provincial average and I therefore ignored most of the riding polls. I used the numbers to make some adjustments such as in Guelph or in some Northern ridings. I also tried to estimate a "Ford nation" effect in the outer rim of Toronto proper, an effect that could indeed expand to Brampton (and also to account for the loss of Jagmeet Singh whose previous riding got dissolved into 4 others). But overall I decided to trust my model. The riding polls in BC last year cost me 2 correct calls (including the 3rd Green seat) and I really just trust province wide numbers more.


2. Seat projections

I want to begin by mentioning that models like mine are more suited to transpose votes into seats when we have small swings. The BC election last year was a good example. It gets way more complicated when one party collapses -like the Ontario Liberals- and end up in an historically low support level. Similarly, modeling a NDP rise by possibly over 10 points (there as well, record level, possibly going higher than under Bob Rae) is more difficult.

We are possibly about to witness a drastic change to the electoral map. The GTA that had been so good for the Liberals and so hostile to the PC for over 10 years appears on the verge of doing the exact opposite. My model does account for that but I'm way less confident about my accuracy than usual.

So, using the polling average, we get the projections above. This is good enough for a Conservative majority and Doug Ford as Premier for the next four years. As you can tell from the probabilities, there isn't much uncertainty. At least according to the model. I'm already including more uncertainty than the CBC tracker (which definitely doesn't have enough) but I suspect that I'm underestimating the NDP's chances. Using the previous election's vote distribution might not be as valid when parties are up and down so much. A model with some socio-demographic variables might be better here but unfortunately I don't have one.

With that said, while the NDP can indeed win, it would be absolutely wrong to pretend that it's a 50-50 race or that "anything can happen". This is simply not the case based on the information currently available. And even if Andrea Horwath is Premier on Friday, that wouldn't mean she was favourite and I was wrong. It'd mean a surprise happened.

The only thing certain at this point is that the Liberals cannot win. I'll go as far as saying that if they somehow do, I'll stop making projections. Not because I dislike this party but because that would mean the polls are useless and therefore so are projections. The Liberals also have a 70% chance of winning 7 seats and fewer, meaning they wouldn't be recognized as an official party! Oh how the mighty have fallen.

If you want the full range of possible seat outcomes, here it is:



Notice that the Liberals can in theory save more than 3 seats. Possibly being the key term.

So why are the Conservatives so heavily favourite? Most of it has to do with vote efficiency.


2.1 Regions and vote efficiency

Our electoral system being what it is, not all votes are equal. For the longest time during this campaign the projections were showing the NDP winning the popular vote and the PC winning a majority of seats. It seems that we might at least avoid this situation tomorrow. It remains that the PC is much better at turning votes into seats.

The main reason is the 905, the Toronto suburbs. The same way the GTA pretty much gave the Liberals a mandate 4 years ago, the suburbs are about to give Ford enough seats to win a majority. The NDP simply doesn't seem capable of winning many ridings there. My model might actually be underestimating the Tories there with a possible strong "Ford nation" effect as mentioned before. Unfortunately for the New Democrats, even if they beat their polling numbers by a few points, it will likely not be enough to win a large number of 905 seats.

The Ontario map will likely look like this tomorrow:




In the North, we can expect the NDP to possibly sweep it. Ridings such as Kenora-Rrainy River, Thunder Bay-Atikokan or Sault Ste. Marie will likely be the only competitive races. Polls have been a little bit all over the map in this region. My estimations also show that this is by far the most volatile region every election. Still, the NDP won it easily 4 years and they are up big times province wide, so I'm projecting a NDP sweep minus Sault Ste. Marie.

In Central Ontario, this is a PC sweep with high certainty. Not much to debate here.

In the East, the PC will crush it in rural ridings while the NDP will try to win some urban ones (should take Kingston and the Islands, will be in contention for some Ottawa ridings). Ottawa Vanier is often seen as THE riding that would remain red no matter what and we'll possibly get confirmation of this tomorrow. Projections and riding polls seem to agree Kathleen Wynne should salvage this seat but it's not an absolute certainty.

In the Southwest, the NDP should win the urban ridings (Windows, London) while the PC will hope to sweep the rural areas including Huron-Bruce.

We then have the Kitchener-Guelph area of midwestern Ontario. The Green could finally win a seat in Guelph in what could be a close 3-way race with the Tories and New Democrats. This would be pretty big for Mike Schreiner and one feels like if he doesn't succeed this time, there'll never be a better opportunity. The rest of the region is looking good for the Tories.

As mentioned before, the 905 is going to be almost all blue, from Ajax to Mississauga. 26 PC seats versus only 4 NDP. Barely any gain for the NDP compared to 2014. For the NDP, this year might be more a transitional one where it sets itself up to fight for the win next time.

Finally, maybe one of the most interesting battlegrounds is surprisingly Toronto. I'm saying surprisingly because urban centers aren't usually that competitive, in particular for right wing parties. Doug Ford appears to be boosting his party there while the New Democrats will hope that former Liberal voters will join them. The core of Toronto (Spadina, etc) should be all orange while the outside (Etobicoke, Sacroborough North) should follow the Ford nation. Then we have the middle with the Don Valley ridings where it's essentially a 3-way race. Those are also the ridings the NDP absolute needs to win if they want to have any hope of forming the government. The projections currently have 2 OLP in Toronto, 10 PC and 13 NDP.

So Central cancels the North in our PC vs NDP matchup. The 905 and the East provide Ford with the roughly 20-25 seats lead giving him the job of Premier. And because most of the 905 PC seats are safe, we are confident about our calls.

Here below I have estimated the probabilities of winning for the Tories as a function of the gap in the popular vote. A positive gap of +2 means the PC wins the popular vote 38 to 36 for instance. See this graph as representing the uncertainty due to the electoral system and vote distribution. It does NOT represent the uncertainty due to polling errors.



As you can see, the chances become 50-50 when the NDP wins by around 3.5%. That seems high and I agree but this is what the model gives me. Anyway, this inefficiency might not come into play too much tomorrow if the Tories do indeed win the popular vote. That was more an issue a week ago.

If the popular vote is indeed tied tomorrow, chances are overwhelmingly in favour of the PC (the 0% on the x-axis). This sucks if you are a NDP supporters but this is what happens with our electoral system.


3. Outcomes and how surprises can happen



So, as described above, the most likely outcome is a Conservative majority. But this isn't a sure thing. Ford could end up with possibly only a minority. Notice that due to the Liberal collapse, a minority of any kind isn't very likely. That's normal. For a minority to happen, you need a strong third party winning enough seats.

The big question is really: what if the PC only wins a minority? Just ask Christy Clark! While I don't  foresee a deal/coalition between the NDP and OLP as likely (unless maybe the seat count is really close), I also can't imagine a Ford minority surviving very long. It'd last for as long as the OLP would need to find a new leader. Ford would also be under pressure in his own party as he could be seen as having "dropped the ball". Let's not forget his lead entering this campaign. At 22%, the chances of something else than a PC majority are not insignificant.

Beyond the majority/minority thing, the other uncertainty we need to talk about is regarding the NDP and its path to victory. I'll use the 2016 US election as an analogy. Before the election every expert and pundit was saying Trump didn't have a path to 270, that the electoral map just didn't work for him. And yet he won while losing the popular vote badly. How did this happen? Because a bunch of States we thought were safe flipped. As a matter of fact, 4 hours into the electoral night and CNN (and others) were saying Trump had more path to 270 than Hillary.

My point is here is that things can happen. Ridings and regions can change surprisingly. When a party like the NDP rises from 24% to possibly 36% or more, things will happen. I'm sorry if I can't formally model how the electoral map could change but I'm at least aware that my model might not be as robust as I want it to be. With that said, my model does account for the fact that when the NDP rises so much above its natural level, it starts rising more in regions traditionally not favourable to it. So it's not like I'm completely oblivious to this fact.

So, how can the NDP win exactly? Three ways.

1) The scenario just mentioned where the electoral map is changed. Polls show that Ford might be winning a slightly different demographic than traditional right wing parties. Along with the NDP rise, this is a possibility. I can't, however, quantity it.

2) Beat the polls. This I can do. The question really is: given the polling average what are the odds that the NDP would win the popular vote by 3, 4 or 5 points? Using the historical accuracy of Canadian polls, we get the following chances, respectively: 14%, 9% and 6%. Let's assume my model has a slight bias and a 3 points lead would suffice, it means the chances of the NDP are closer to 15%. Still not great but slightly better.

As discussed under vote efficiency, winning the popular vote isn't enough for the NDP to win the seat count but it makes the path a lot easier.

3) Polls are right but the NDP increases its efficiency. In this scenario, the NDP surges in urban ridings in Toronto and to a lesser extend Ottawa. Many of the 416 seats projected to go PC have a low margin of victory. A good GOTV along with a last minute rally of some undecided (and Liberal voters wanting to prevent a Tory majority) and we could see most of the 416 orange. I went into more details a few days ago and I stand by my analysis, even though the numbers are a little bit outdated at this point.

Note that a combination of these three scenarios is totally possible. A little bit of 2) for instance would make 3) a lot easier.


Conclusion

After a campaign full of swings and changes, we might actually be back to the original planned outcome: a PC majority! While the NDP can win, that would be a fairly major surprise. It'd require a combination of the polls being quite wrong as well as the NDP overcoming its vote inefficiency. The last one appears difficult mostly because of the 905.

Still, the chances of a Tory majority are "only" around 75%. We've seen surprises like this in the past such as Trump or Scheer over Bernier.

Advance voting appears to indicate that competitive urban ridings in the 416 and 905 will attract more voters. We might observe the clash of two movements: Ford nation versus the rise of the NDP.

Enjoy election night and I hope you appreciated my coverage during this election. If you did, don't forget to follow me on Twitter @2closetocall.



[Update] Since I'm not always right in my predictions and since an octopus used to be so good at predicting soccer games, I thought I'd try another method:



So looks like my dog and I disagree!

Last day of the campaign!

Last day of the campaign!
A very short article this morning as I'll be working during the day to publish the final projections. We should get a few more polls this morning. So far they all seem to show the same trend -OLP decreasing, tight race between the PC and NDP around 37-39%- except Mainstreet that keeps seeing a fairly decent PC lead.

Just for fun, I thought I'd look at which ridings we could use as "barometer".

1. Barometer for the province-wide voting percentages

I looked at the results from 2003 and calculated the absolute deviation for each party. I then summed it by riding and look at which one was, in average, was the closest to the overall popular vote.

The best three ridings are:

1. Brantford-Brant
2. St. Catherines
3. Kitchener Center

In average these three ridings have the top 3 (or 4) parties within 2 points of the province wide results. So on election night, if you want to know whether the polls were right or not, keep an eye on these three.

2. Barometer for who will win the most seats

Maybe a better barometer is a riding that tend to switch along with the winning party. I'm talking here of a riding voting NDP if the NDP wins the election and PC if Ford wins.

Looking at my current simulations, I have found these three ridings:

- Lampton-Kent-Middlesex: chances of winning are roughly similar to the 80-20 of the overall projections. It's a riding in the Southwest, one of the regions where some surprises could happen (the trend there in the polls wasn't the best recently for the NDP). This riding goes to the party winning the most seats overall more than 80% of the time.

- York-Simcoe

- Nipissing or Hasting-Lennox and Addington.

It might be surprising not to see a riding of a close race in the 905 but it actually makes sense. We are looking for riding the PC is expected to win, same as the election. So that's three (or four) ridings in different regions.

Of course, one way to determine what is happening on Thursday will simply be to look at the close races and see if they turn one or another. But I thought some of you might like to have other "barometers".

Alright, enough for now, I have some writing to do today.

Advance turnout shows busy GTA

2 days to go! Only 2 days to go in this Ontario election. Yesterday we got a bunch of new polls from Forum, Mainstreet and Pollara. Mainstreet is the only one still showing a decent lead for the PC (more than 3 points) while the others have a super close race. We also got the stats about advance voting and turnout. More on this after the projections.

It's nice to have a lot of recent polls with decent sample sizes. Also, the new update from Forum, showing a much tighter race, kind of broke with the trend that IVR polls were showing a decent PC lead while online (as well as live callers) polls were showing a tied race.

Anyway, here are the most up to date projections.

Voting intentions; Seat projections with confidence intervals; Chances of winning the most seats

As for the possible outcomes:



As usual, riding by riding projections are available at the bottom.

I won't go into much detail here mostly because the projections haven't changed much. We are still in a situation where the Tories are favourite but it isn't hard to imagine a path to 63 seats for the NDP. We have something more interesting to talk about this morning: advance turnout!


Advance turnout

According to Election Ontario, about 768,895 citizens have already voted. This is more than the 647,261 of 2014. That year was marked by a fairly weak turnout overall at 51% (still better than the 48.2% of 2011). So, does it mean we can expect a higher turnout this time around? Not necessarily. Last year in BC, advance turnout was on fire but the overall turnout barely increased. There is simply a trend where people take more and more advantage of the possibility to vote early. So hard to tell if overall voter participation will be up. If it is however, I wouldn't expect a giant jump like what we saw for the 2015 federal election. Which is good news for polls watchers as they tend to be more accurate when turnout is similar.

Maybe more interesting for someone like me trying to predict the results at the riding level is to look at the advance stats per electoral district.

Here below you have the top 10 ridings with the more advance voters. 

1.Kingston and the Islands
2. Simcoe-Grey
3. Renfrew-Nipissing-Pembroke
4. Orléan
5. Guelph
6. Simcoe North
7. Northumberland-Peterborough South
8. Ottawa Centre
9. Toronto-Danforth
10. Burlington

Out of these ten, 6 are projected to go PC, 2 NDP, 1 OLP and 1 Green,

But total number of voters isn't that interesting. Not only are some ridings bigger, turnout can also tend to simply be higher in some of them years after years.

Instead let's look at the top 10 ridings where the advance turnout increased the most. When this happens, this is usually a sign of changes. In BC last year for instance, we could clearly see many ridings in the Lowe Mainland -in particular Surrey- where the NDP ultimately made many gains.

Top 10 ridings with biggest increase in advance turnout

1. Kitchener South-Hespeler
2. Brampton West
3. Markham-Unionville
4. Brampton South
5. Aurora-Oak Ridges-Richmond Hill
6. Markham Stouffville
7. University-Rosedale
8. Davenport
9. Etobicoke North
10. Toronto-Danforth

The list continues with other 416 and 905 ridings mostly. Note that this is the top 10 if measure in percentage change. If I had used the increase in the number of voters, the list would include most of the same ridings but Kingston and the Islands would be there.

We see a clear pattern here: increased turnout in Toronto or in the Toronto suburbs. Out of the 10, 5 are projected blue and 5 orange. It shows that close races will most likely attract higher turnout.

So, can we infer anything else? Both in 2015 and 2017 (BC), the increased turnout was clearly helping the one party increasing (LPC in 2015, BC NDP 2017). This time around is more difficult since we have two parties clearly rising compared to 2014.

What this list shows me for sure is that the Liberals won't win these seats. Increased turnout isn't a good news for a collapsing governing party. If they were hoping to salvage some seats in downtown Toronto, they are most likely mistaken (well, unless the concession of Wynne somehow works). After, we can possibly interpret this list as the PC and NDP mobilizing their voters. Riding polls have shown the PC upsetting the NDP in the Brampton ridings (2 wins and 2 super close losses) while my projections are showing the NDP well ahead except in Brampton-South. At the end of the day, it might come down to which party gets its vote out the most.

I can also see this list as showing that the scenario where the NDP would sweep the very urban riding (core Toronto) while the Tories would win in the suburbs can happen. If that's the case, the electoral night could be long. The increases turnout in Davenport, Toronto-Danforth or University-Rosedale shows me the NDP is surging there while the Markham and Etobicoke ridings are pointing to the PC finally breaking into the GTA.

What says you? Do you have a better reading of the situation?



With 3 days to go in the Ontario election, polls disagree with each other

The 2018 Ontario provincial election is coming to an end this Thursday. We are starting to get the final polls from many firms and it seems we are observing another case of methods disagreeing with each other. Namely, phone polls IVR, automatic phone calls) do not show the same thing as online ones.

Before looking into this though, here are, as usual, the latest projections. I have added the polls published yesterday into the mix. As you can see, for the first time in a long time, the Conservatives are ahead in the popular vote.



Trend just isn't good for the NDP. On top of that, the trend by region is also negative (NDP decreasing in the Southwest for instance). Riding by riding projections are available at the end of this article.

Online versus IVR polls

The latest Abacus, research Co. and H+K Strategies polls, all done online, show the NDP either slightly ahead or tied with the PC. On the other hand, the recent Ekos, Mainstreet (tracker) and Forum, all IVR, have the Tories between 3 and 5 points. Pollara, using a mix of online and live callers, is more similar to the online polls. Let's leave it aside for now but we'll get back to it.

The graph below shows you the polls during the campaign. For the Mainstreet tracker, I only included the data once every 3 days (since the tracker is a 3-days rolling average). I separated the polls by method. It's a messy graph but I don't believe I can do better. Also, I had to put one of the NDP in green because orange and yellow are too similar. Click on it for a bigger size.



Couple of observations:

1) Online and IVR polls have had a systematic difference during the entire campaign. IVR having the PC higher and the NDP and OLP lower than online polls. The gap has remained constant for the PC and NDP.

2) Both methods give us the same trend and dynamics. At least!

3) Because of 1), we see that online poll had the NDP and PC tied around May 19th while IVR poll only reached the same level almost a week later (and at that point, online polls naturally had the NDP ahead). The Mainstreet tracker was especially slow to pick up.

4) Right now, we can see that the online polls have a tied race while IVR polls have the PC ahead. But again, same late campaign dynamic, just different levels.

5) At least there is convergence for the OLP

6) The IVR graph is heavily influenced by Mainstreet which published way more polls. So keep that in mind.


So, who is right and who is wrong? In order to answer the question, I went back to my data set of polls I used to estimate the average accuracy. I compared IVR versus online polls. The answer? The average absolute deviation of both methods (with respect to the actual outcome) is almost identical. So IVR and online polls have performed similarly over the last few years. IVR polls did better in 2015 but this is mostly because they were the last ones to poll and there was a late campaign shift in favour of the Liberals.

So I see no reason to believe one more than the other. If we limit ourselves to the last Ontario election (using the numbers among all voters and not the crappy "likely voters" pollsters were trying at the time and that proved very wrong), the best two polls were online (Angus-Reid and Abacus).

Maybe we have other polls to help us determine which one is right? The latest Pollara was a mix of online+live callers (most of the sample being live callers) and the results are in line with the online polls. Live calls polls have the best accuracy in recent years, although a lot of it is due to Nanos being right on the spot during the federal election. Still, between 2014 (online better in Ontario) and Pollara, that's two arguments in favour of online polls.

We also got the latest Ipsos poll that was 2/3 online and 1/3 by phone (live caller). This poll was more in line with the IVR ones! Thus not providing any tie breaker (actually making things more complicated).

Finally, Innovative actually did two polls, one online and one by phone (live callers) with very similar results. This would confirm the online+live call group.

So at the end of the day, I'll simply do an average. But I'll be careful in case the latest polls (like published on Wednesday) are all of one type. Aggregators that tend to give a much higher weight to the very last poll may introduce a bias whereas, instead of measuring a late campaign swing, they are merely capturing the effect of the method.

One thing to keep in mind is that it might not matter as much this time around. Because not matter if you use online or IVR polls, you get a Tory majority as most likely outcome. Of course online polls would give more chances to the NDP but the PC is favourite in both cases.


Extra: riding by riding projections.

What could a NDP victory look like?

Polls are currently showing a Conservative majority with probabilities around 70%. This means a NDP victory next Thursday is possible but would require some surprises. In this blog post, I will go through one scenario that ends with Andrea Horwath as Premier. It isn't a purely fantasy scenario where the NDP gets 50% of the vote or anything. No, I'll use the current projections and make some minor changes.

But first, here are the most up to date projections. I just added the Mainstreet tracker of yesterday (showing the PC back on top) as well as accounting for the riding poll in Guelph showing the Green leader first. At the bottom of this article you can find the riding by riding projections that will serve as our baseline case.

Voting intentions; Seat projections with confidence intervals; Chances of winning the most seats

Alright, the NDP needs 13 more seats compared to the projections. Let's see the ridings that could be changed. The theme for this scenario will be the NDP surging in very urban areas. We'll leave most of the 905 to the Tories. We'll also not go with the scenario where the NDP is severly underestimated in the polls (although this is a possibility). We'll simply flip some ridings here and there.

Ok, here we go.

1. Guelph. Mainstreet published a second riding poll yesterday and it confirmed that Mike Schreiner was ahead, so I updated the model (even though I continue my decision not to use most of the riding polls as I believe they don't match up with the province wide ones. In this case however, we got two polls that showed the same results. Also, I can absolutely imagine the Green benefiting from the collapse of the OLP there). But the lead of the Green leader (3.9%) is small and smaller than the margins of error of the poll. Also, the riding polls of Mainstreet have been underestimating the NDP consistently. So let's assume that the NDP manages to grab this riding. It could be helped by some undecided ultimately going for the best option to defeat Rob Doug Ford.

2. Two of the three Don Valley (East, West and North). Current projections give the NDP some low chances there. Don Valley East is by far the easiest pick but those are urban ridings. The NDP mostly needs half of the Liberal voters to rally around them. Maybe Kathleen Wynne admitting she knows she won't win would be enough to move these voters.

3. Eglington-Lawrence. Projections aren't super favourable but that's one of the riding polls where the NDP was higher than in my model. Keeping with the surge in urban environment, this could make sense.

4. Scarborough-Guildwood. Currently one of the two Scarborough riding not going NDP, the NDP just needs a little bit of a push to win there.

5. Toronto Centre and St. Paul's. The former is almost a perfect toss up in the projections but riding polls were putting the NDP easily ahead. Similar story for St. Paul's. These are two ridings that could go NDP with the collapse of the Liberals.

Ok we are halfway there!

6. Willowdale and York-Centre. We are getting a little bit far from the city core though. But that's still the 416 and it still fits our scenario.

Not gains but for this scenario to work, I'm assuming all of Brampton goes orange.

7. One extra riding in Mississauga or Etobicoke-Lakeshore. There as well riding polls aren't very good but let's assume one surprise victory. The model seems to indicate that it should be possible in Mississauga.

8. Three Ottawa ridings; Vanier, West-Nepean and South. Both the model and riding polls are showing close races. Let's factor the usual collapse of the Liberal vote and a rally to the NDP and these can flip.

Here we are, the NDP is now at 63 seats and forming government. What did it take? A relatively basic scenario where the NDP is surging on Toronto proper and Ottawa, as well as getting lucky in one or two ridings in the suburbs. In most of the ridings here, the NDP's chances are already projected around 30%, so it's not like winning would be a monumental surprise. I did include some ridings with fairly low chances though (around 5%). So let's see if we can have a little bit of a buffer.

Let's add Sault St. Marie to the list. We can also put Kitchener South-Hespeler. They can replace two of the ridings above.

Look, of course I'm looking at a very optimistic scenario for the NDP. But my point is to show that it's not that hard to picture a path to 63 seats for this party. It doesn't require to flip seats where the NDP is 30 points behind.

Let's put it this way: remember last US election when so many pundits where saying there was no path to 270 for Trump? And then, 3 hours into the night, we started realizing that Trump actually had more path than Hillary simply because he flipped a couple of states in the midwest? I'm not comparing the NDP to Trump, I'm simply saying that we often think there is no way and then be surprised. My scenario here above would allow the PC to win most of the suburb and win almost everything in the rural part of the province (except the North). That's not a stretch. The fact the current chances of winning are almost exactly the same as for Hilary Clinton and Trump is just an amusing coincidence.

Look, whenever a party increases significantly between two elections, there are always ridings that flip that we never thought would. If the NDP really wins the popular vote by jumping to 37, 38 or 39%, I'm convinced such surprises will happen. They always do.

My model isn't even smart enough to actually include correlations across ridings that are similar (outside of being in the same region; But no socio-demographic variables unfortunately) and yet it gives the NDP around 30% chances of winning! So when I see the CBC poll tracker giving the NDP only 9.1% chances, I find it ridiculous.

Finally, uncertainty works both ways. I could easily do the same exercise for the Conservatives and have a scenario where Ford wins a super majority of 80 seats. Do not interpret this post as me saying that this is what will happen. I simply wanted to illustrate how possible a NDP majority was. In other words: do not assume that this election is over.

[Poll] What do you think of Kathleen Wynne's strategy?

[Poll] What do you think of Kathleen Wynne's strategy?
Interesting strategy from the Ontarian Premier Kathleen Wynne to admit she won't win next week. Not very common. What do you think of it?

What do you think of Kathleen Wynne's strategy to admit she won't win?
I love it, this is genius
I like it but it won't work
Meh
It's stupid, her party will now drop in the polls
Created with SiteListener

Projection updates for June 2nd: The PC rebound confirmed?

There was a time during this campaign, about the end of last week, were people genuinely wondered if we were witnessing an orange wave in Ontario. This reached a peak when Pollara published a poll putting the party of Andrea Horwath at 43%, 11 points ahead of the Conservatives of Doug Ford. At that time, my projections were literally showing a 50-50 race.

Since then, however, things have changed slightly against the NDP. The NDP didn't really drop in the average, but it stopped increasing and seems to have got stuck around 36-37%. The PC on the other hand enjoyed a small rebound. A rebound somewhat confirmed by looking at the recent trend in the Mainstreet tracker or the most recent polls from Ekos and Forum. At the very least you can say the PC managed to stop its fall.

If the election were tomorrow, that would be the best projections I'd have:

Voting intentions; Seat projections with 95% confidence intervals; Chances of winning the most seats

[Small update] Mainstreet released the new riding poll in Guelph and it confirmed Mike Schreiner ahead. So my next projections will have the Green leader as favourite there, although it remains a close race. Here below the updated projections. It is still a mix of projections and the numbers from the poll.




A couple of remarks about my model and my poll aggregation. I don't waste my time trying to weigh polls based on whether they were done 24 hours earlier or if they have an extra 200 respondents. I mostly use the polls over the last week and give everyone the same weight. You can think I'm lazy but my polling average has been better than the CBC one for multiple elections. I have published research showing that these factors don't matter significantly regarding polling accuracy.

What I do differently is regarding undecided. Pollsters usually allocate them proportionally. So if one party is at 40% among decided voters, they assume this party would get 40% of the undecided. This means they are making a very implicit assumption: undecided will ultimately break out the same way as decided voters. My point is: then why were they undecided in the first place? Alternatively, the proportional method is equivalent to dropping all the undecided and assuming none of them will vote.

I dislike both assumptions and find them unrealistic. Instead I allocate the undecided non-proportionally among the top parties (Liberals, PC and NDP). The small parties don't get undecided. This allows me to take care of the usual overestimation of small parties. Additionally, I allocate proportionally more undecided to the incumbent. Why? Well because a careful look at polls over the last years showed that the incumbent was systematically underestimated. Also, I can justify this by saying that undecided voters will be more likely to go with the safe option, the one they already know.

In the case of the Ontario election, I started the election by giving 40% of the undecided to the Liberals (instead of the 25-30% they would get with the proportional method), 40% to the PC and 20% to the NDP. But since the NDP has now taken the lead, I changed to 1/3 for everyone. It boosts the OLP a little bit and decrease the Green. It mostly leaves the NDP and PC where they'd be. Yes my 1/3 rule is arbitrary but you have to make an assumption at some point. I find mine no more stupid than going with a proportional allocation.

On top of this, as I was explaining the other day, the riding polls from Mainstreet simply do not line up with the province wide ones. And the difference is still present in the last few days. So after spending way too much time looking into them, I dropped the adjustments made based on the riding polls. I believe the riding polls are biased towards the PC and against the NDP (compared to the province wide polls that I trust more) and using their information is therefore introducing this bias into my projections. It's possible these polls are right but if they are, then the provincial average will be way, way off.

I kept 4 riding polls. Three in the North (Sudbury, Sault Ste. Marie and Kenora-Rainy River), a region traditionally very volatile (and in the case of Sault Ste. Marie, the poll confirmed that the by-election results should most likely be used as baseline). I also kept the one in Guelph because it was the only way to measure the Mike Schreiner effect. Mainstreet will re-poll there and publish tomorrow, so we'll see if the Green really have a shot. All the other ridings polls have been discarded.


The state of this race



Let's focus on the chances of winning instead than on the seat total. We basically have a 70-30 race. Anecdotally, this is similar to the state of the race at the end of the 2016 US election. Don't interpret this as me saying you should expect a surprise NDP victory! With that said, there are similarities in the sense that it looks like there are very few paths for the NDP to win. The NDP vote being less efficient (in being converted in seats) than the PC vote, mostly because of the 905, it means the NDP either needs to hope for a severe underestimation in the polls (could be caused by a good turnout of the NDP voters) or a massive change to the electoral map. One path that would allow Horwath to become Premier would be a NDP surge in the core urban ridings of Toronto and Ottawa. This could be accomplish by a massive transfer of OLP voters to the NDP (strategic voting anyone?).

If nothing changes in the last few days, we'll have an election night where Doug Ford will be favourite but a NDP win shouldn't be considered as super unlikely or shocking. Well, that's assuming you believe my probabilities and not the CBC's ones since the CBC is currently only giving the NDP 9.1% chances of winning. I already explained why I think this is plain wrong. This is either seriously underestimating how wrong polls have been in the past or overestimating how good the model is.

Many think that for Ford, it's majority or burst. They might have a point. First of all, winning only a minority for the PC would be barely above blowing a 4-1 lead. After all, let's remember that the PC started this campaign with a very comfortable lead. I'm sure many would criticize Ford for dropping the majority. Also, if Ford fails to get a majority, we can't exclude the NDP and Liberals making some sort of deal. Anyway, if you are in the camp that thinks it's majority or burst, then you should know that the chances for such a majority are only 52.9%. So a toss-up basically.

Speaking of majority, the collapse of the Liberals means a minority is not the most likely outcome. It's simply arithmetic really. For a minority to happen, you need a third party winning enough seats. That's not currently the case for the Liberals, although the confidence intervals shows that if the Liberal vote was to hold better than expected, then this party could win significantly more than 5 seats.

What are the best and absolute worst case scenarios for each party?

For the PC: 22 seats is the worst, 93 is the absolute best (think of this scenario as the one where the Tories were underestimated by the polls and its vote was incredibly efficient).

OLP: 0-36 (note: around 2% chance of 0 seat)

NDP: 27-99.

Green: 0-1 (Note on Mike Schreiner: it's always hard to measure a local effect such as what might be happening in Guelph. The conditions look good for the Green leader -no incumbent, a split vote meaning the threshold to win the seat is low, etc- but I can't just go ahead and boost Schreiner. If the Mainstreet poll today confirms he's ahead (like in the first poll), then I'll make adjustments. For now, my take is that this is a 4-way race and he has some good chances.

Maybe the best way to see this uncertainty is with the distributions for each party:


As you can see, the only certainty seems to be that the Liberals will finish 3rd.

Notice that the projections above don't fall exactly in the middle of each distribution. The PC is currently projected slightly on the higher end of its interval while the OLP is in the low tail. This mostly comes down to close races and the PC is currently winning a majority of those. Still, if you instead prefer the average scenario as projections, then it's OLP 7, PC 63 and NDP 52. That should convince you that a majority for Ford is far from guaranteed at this point.

In conclusion, this will probably remain a close race until the end. I'm starting to be of the opinion that the real uncertainty will be whether Ford gets a majority though (because of how difficult the 905 is for the NDP; Elections in this country have often been won in the suburbs in recent years and I feel the suburbs have decided to change Wynne for Ford). I know that many things can happen during the last few days, but I feel that if there was supposed to be an orange wave, it would have happened by now. There is something preventing the NDP from reaching the 40% mark, or at the very least from taking a commanding lead over the PC. I'm not sure what it is exactly (fear of the  Rae days? etc) but I don't expect a late surge for this party. That's just my reading of the race and I'm usually wrong, so take it with a big grain of salt.

Finally, the riding by riding projections. Remember that you can make your own using the simulator.