How the polls did in the Ontario election

PC majority victory was foreseen, but Liberals under-performed

It was quite a result on Thursday night. Doug Ford’s Progressive Conservatives won a bigger majority in Ontario by going through some of the NDP’s traditional strongholds while the Liberals flopped. Turnout, too, was well below already low expectations.

I’m still processing the results and I hope to take a deeper dive into what happened next week.

But first I wanted to take a quick look at how the polls did in this campaign.

In short, they did really well — a particularly impressive feat considering that the lower turnout meant the voting population was even less like the general population than is usually the case.

Here’s the breakdown showing the final poll of each pollster in the field in the final weeks of the campaign. I’ve rounded the election results and the poll numbers to make it a easier to compare apples to oranges (and, let’s be honest, decimal points don’t make much sense in polls with a margin of error that can be three points or more).

The best performing polling firms were Forum Research, Ipsos and Léger, all with a total error of less than four points, or less than one point per party.

These three polling firms had the PCs around 40-41% and had a gap of only one or two points between the Liberals and New Democrats. That’s a great showing, considering the PCs (as of Friday night) stood at 40.9% of the vote, with the Liberals at 23.8% and the NDP at 23.7%.

With the exception of the Innovative Research Group, whose separate telephone and online polls were a fair bit off, all pollsters did quite well with an error of less than two points per party. The polls pointed toward a majority PC government and a tight race between the Liberals and NDP for second spot in the popular vote. Projections that were derived from these polls indicated the NDP had a much more efficient vote than the Liberals, suggesting that the pollsters were also effective in gauging the regional distribution of each party’s support.

Every pollster except Forum under-estimated the PCs, something which has become rather common in elections when it comes to the main right-of-centre party. But the error wasn’t as big as we’ve sometimes seen elsewhere. On average, the PCs were under-estimated by two points, but a lot of pollsters had the PCs within a single percentage point and several had them over 40%.

The NDP was accurately pegged by the polls with an average under-estimation of just 0.3 percentage points. The range was pretty tight, with nearly all pollsters putting the NDP between 23% and 25%, an almost perfect call.

The Liberals were over-estimated across the board and by an average of two points overall. This is a curious result, as turnout effects should not have hurt the Liberals this much — their support is generally not all that different among younger or older voters.

It suggests to me that a lot of voters might have either stayed home or went elsewhere when it finally came time to make up their minds. The Liberals just didn’t get their vote out — or they failed to keep it right up to the last moment.

The Greens were somewhat over-estimated (by an average of 0.8 points), but the Greens habitually under-perform their polls. This was actually a decent performance for them in terms of living up to expectations.

What about my CBC Ontario Poll Tracker? As should be expected with an aggregate, it ranked somewhere near the top but was beaten by a few individual polls. As with the polls overall, the aggregate under-estimated the PCs and over-estimated the Liberals.

In terms of the seat count, the Poll Tracker projected 80 for the PCs, 26 for the NDP, 17 for the Liberals and one for the Greens. The result of 83 PCs, 31 NDP, 8 Liberals, 1 Green and 1 Independent fell within the likely ranges for all parties (71-90 PC, 20-31 NDP) but the Liberals (10-25) and Independents (0).

The Liberal result did fall in the maximum band (4-37), but that is supposed to represent extreme outcomes.

And that is, perhaps, how best to define the Liberal result. It is remarkable that they took as much of the vote as the NDP but only managed to win a quarter as many seats — an extreme example of how first-past-the-post can distort results.

This failure for the Liberals appears largely driven by Toronto and northern Ontario, as the likely projected ranges in the Poll Tracker envisioned the Liberals as low as zero seats in the Greater Toronto Area, Hamilton-Niagara region and southwestern Ontario, which is what they got. But it also projected two seats in northern Ontario (they got none) and at least five seats in Toronto (they won four).

Coming up short in the GTA, where six seats were considered in play for the Liberals, was another big problem.

There will be plenty to sift through to figure out what went wrong for the Liberals, and why their vote was so inefficient compared to the NDP. For the PCs, though, this is a big victory — and one that was predicted by the polls from Day 1.