What about SmartVoting.ca?

SmartVoting.ca and VoteWell.ca and all other strategic voting sites

Vote splitting has become such a scourge in our elections that new websites to help you vote strategically pop up every time. The latest of these is SmartVoting.ca although the older VoteWell.ca remains popular.

Let me be clear: We love SmartVoting.ca We love VoteWell.ca We're big fans of anyone who is trying to figure out how to get Doug Ford out of Queen's Park. And for the most part, these tools work pretty well.

"Look, these websites tell us how to get rid of Doug Ford."

VoteWell.ca essentially repackages data from the Canadian polling aggregator 338Canada, while SmartVoting.ca takes 338's data plus others, and then does a little extra math. Polling in Canada is generally very high quality, and 338 Canada examines the accuracy of the polls, and also publishes the accuracy of its own aggregation methods, which is around 90% or so. SmartVoting.ca claims they have even higher accuracy than 338 Canada based on their past results, but since they have only recently launched publicly, there's no way to confirm this. Still, I also have no reason to doubt them.

But no polling projection can be 100% accurate, and there are some inherent limitations, and whatever voting tool you use, you should be aware of that.

Let's go through it.

1) Polls don't predict the future

Philippe Fournier, the astrophysicist and polling analyst who created 338Canada, points this out often. Polls do not predict the future. The only tell us about the recent past. Changes happening at the last minute will missed by the polls, and therefore these tools.

More to the point, when polls show that Doug Ford may be elected to a majority, all that tells us is that as of a few days or longer ago, that's how people answered the polls. It doesn't mean that people won't change their minds tomorrow.

2) Polls aren't usually local

The seat projections made by 338Canada--how you or these other websites figure out your best best for a strategic vote--are usually not based on local riding polls. They did not ask everyone in your riding who they are voting for. In general, there's only a very limited number of local riding polls that happen during an election.

They are instead projections that takes broader polling across the city or region or even the province, and figures out what will likely happen based on history, demographics, trends, etc. As mentioned before, it's often pretty accurate.

Still, if there is something happening in your riding that bucks the typical trend, this won't be as accurate. If you have an independent candidate that is very strong or well-known, that might not be picked up by polls. Similarly, a candidate with extremely strong ground game, or (especially in rural Ontario) a candidate with very few connections to the community, might change things.

EXAMPLE: Scarborough-Southwest in 2022

We track 338's projections (among others) as part of our research, and just prior to the election, they had predicted this for Scarborough-Southwest

May 28, 2022: 338 Canada projection for SSW PC: 29%, Lib: 29%, NDP: 29%

We had endorsed Doly Begum, NDP, because our field research had uncovered her phenomenal ground game, but most of the other pollsters were calling it for the Liberals. Here are the actual results:

Election June 2, 2022 results for SSW

PC: 28%, Lib: 19%, NDP: 48%

3) Data issues

Polling in Canada is generally good quality, and certainly, 338 Canada looks at high quality polls and even then, adjusts their weighting accordingly.

That said, pollsters have to make their predictions based on certain assumptions. There's a lot of math and science involved in this, but when things don't go as expected, things can get a little wonky. The 2022 election had the lowest turnout in Ontario history, and even pollsters expecting and adjusting for low turnout would hardly have expected record-breaking lows. We don't know what turnout will be this year, and given that winter elections are rare, it's harder for the pollsters to know how to adjust for that.

Plus, in a highly-multilingual city like Toronto, there are many communities whose voices simply may not be fully represented in the polls. That can be adjusted for demographically, as SmartVoting.ca claims to do, but again, it's not always easy.

EXAMPLE: York-South Weston in 2022

We track 338's projections (among others) as part of our research, and just prior to the election, they had predicted this for York-South Weston

May 28, 2022: 338 Canada projection for YSW PC: 34%, Lib: 29%, NDP: 27%

We were extremely torn in 2022, as the incumbent NDP Faisal Hassan had great ground game and community connections, but also the polling trends had not favoured him. We ended up calling it a toss-up. In the end, it was heartbreakingly close, and we suspect that polling firms don't reach the Somali community in this area, who support Faisal specifically.

Election June 2, 2022 results for YSW

PC: 37%, Lib: 24%, NDP: 34%

4) 3-way races are tricky

A tight 3-way race makes things very difficult for anyone to predict, let alone point to the candidates most likely to defeat the PCs. That said, these are also the races where uniting around a single candidate can make the most difference.

The difficulty is that all polling has a margin of error, usually within a couple of percentage points, which is to say that the poll is still accurate even if the final numbers are off by one or two. When projecting a seat, that margin of error is generally a bit bigger. But when the difference between one strategic candidate and the next is only a percentage point or so, it's very difficult to know exactly who the right candidate is.

EXAMPLE: Humber River-Black Creek in 2022

We track 338's projections (among others) as part of our research, and just prior to the election, they had predicted this for Humber River-Black Creek

May 28, 2022: 338 Canada projection for HR-BC PC: 30%, Lib: 31%, NDP: 30%

Though it was always close, the Liberals had always polled highest through the election--even up to 6 points higher than the NDP--even though the NDP Tom Rakocevic was the incumbent. Despite our efforts, we also had less field research here than in other ridings, and so with the data we had, we decided to go with the Liberals. Post-election, we also wonder if this is a tricky riding to poll accurately; the Jane-Finch neighbourhood is within the riding and for various structural reasons, turnout tends to be lower here, which may affect polling assumptions.

Election June 2, 2022 results for HR-BC

PC: 30%, Lib: 31%, NDP: 35%

What we learned from this

In all of these cases, the riding went to the incumbent. It's well-known that incumbents have an advantage in elections. While we know 338 Canada accounts for this in their modelling, we suspect that may be stronger than we realized.

This is something we are accounting for in our recommendations this year, as well as noting that for these particular ridings, we might need to take a more critical look at the polling numbers.

What you need to know

We don't want to make it seem like the polling is never right. It's often right; the three examples I gave here were the only ones where polling was different from the election results amid all ridings in Toronto we were tracking.

Most of the time, looking up information on websites like SmartVoting.ca or VoteWell.ca will give you the correct candidate.

That said, we also think that there is a value to qualitative information and field research. That is something we are working on here in Toronto, but this takes time and resources, and so it's not something we're able to do across the province. (Our partners at Cooperate For Canada have done some of this work in some areas outside Toronto.)

This kind of information is not always easy to get. We ask candidates for information they usually don't want made public; some trust us more than others. Some candidates are more accessible than others as well.

So in the absence of this information, your best bet is to use one of these strategic voting websites.

But remember--you are not without knowledge of your riding and the candidates in it. The polling projections can't tell you if a candidate is a paper candidate (i.e., someone who isn't expected to actually win, they are simply there to have a candidate on paper), or if they are well-known in the area, or if their volunteer organization is well-organized or chaotic. But you can figure out much of this for yourself.

Still.... sometimes, these decisions are simply difficult to make; we can't guarantee that we will always make the right call, but we will explain why we make the decisions we have, and then leave it to you to make the decision accordingly. After all, it's your vote.

At the end of the day, Not One Seat, SmartVoting, VoteWell, etc, all have the same goals: a better government in Ontario, despite our absolutely bonkers voting system and Opposition leaders who will not work together.