Since then, the Liberals have cobbled together a majority through floor-crossers and byelections, the Conservatives have seemingly lost their mojo and the New Democrats have replaced their leader. The odds of another election happening in the short or medium term has dropped significantly, even if the Liberals would easily secure an even bigger majority if they went to the polls today.

The Writ’s federal Vote and Seat Projection has the Liberals leading in 218 ridings and winning an average of 210 seats, a big gain from the 169 seats the party won a year ago. The Liberals have slipped back a little since last week, dropping 0.6 points to 45.7% and sliding in the seat projection, but that nevertheless keeps them in a very strong position.

The Conservatives have hardly budged at 34%, putting them 11.7 points behind the Liberals in the projection. The party is leading in 86 seats and wins an average of 103.2 seats in the projection, an increase of three and 3.6, respectively, since last week. That represents a big loss of seats compared to the 144 the Conservatives won a year ago.

The Bloc Québécois, at 6.3% nationally (27.8% in Quebec), is about where they were a year ago. They are leading in 22 seats and win an average of 19.6 seats in the projection, suggesting little change from April 2025.

For fully interactive charts, tables and maps, head to The Writ’s website:

While it pales in comparison to the number of seat gains the Liberals would make, the biggest proportional gain since a year ago goes to the New Democrats. The party sits at 9.4% (+0.6 since last week), is leading in 15 seats (+5) and projected to win an average of 9.1 seats (+1.3). This puts recognized party status (12 seats) in reach of the NDP, but the party’s expected over-estimation in the polls suggests it might not be as easy as it appears — the NDP is leading in too many toss-ups for all of them to go its way.

The last week of polling has been pretty good for the New Democrats, relatively speaking, but there is also good reason for the NDP not to get too carried away with some of the numbers.

Read more

Full regional breakdowns and tracking charts are available for subscribers below, as are complete riding-level projections. A full methodological explanation of how the vote and seat projection model works can be found here.

The Vote Projection is based primarily on a weighted average of all publicly-available national and regional polls. Complete listings of published polls can be found at 338Canada or on the Wikipedia page for polling in the next federal election. The weighted average is then used to project the outcome in all 343 ridings across the country, taking into account local candidates. Those individual riding-level projections are then weighted by the number of eligible voters in each riding and past turnout to determine the national vote projection for each party.

The Seat Projection shows the number of seats in which each party is projected to be ahead. Parties are assigned a range of likely vote share in each riding based on past polling and modelling errors and 10,000 simulations are run to assign to each party a number of ridings in which they are in contention. The High and Low ranges take into account 95% of all likely outcomes, while the Avg. Projection shows the average number of seats each party wins in these simulations, which takes into account the tendency for parties to under- or out-perform their polls.

Wondering how to interpret the Seat Projection and the Avg. Projection? While the Seat Projection is The Writ’s “official call”, the Avg. Projection is an important guide to understanding the projections, in conjunction with the high and low ranges. Seat projections are rarely right on target, so the Avg. Projection should set expectations. In other words, if the Seat Projection for a party is 25 seats but the Avg. Projection is 21.2, your expectation should be that the party will win around 25 seats, with a greater likelihood that the party under-performs that target rather than beats it.

The percentage of simulations in which each party wins a majority or a plurality of seats is shown below.

Complete riding and sub-regional projections are available for subscribers of The Writ:

A full breakdown of the methodology behind the vote and seat projection model can be found here:

Vote and seat projections broken down by region and tracking charts going back to the 2025 election are available for subscribers of The Writ below 👇

Read more

Detailed projections for all 343 ridings across the country, as well as vote projections for smaller regions within each province, are available for subscribers below 👇

Read more

There’s been reporting on the fears within the Conservative caucus that Pierre Poilievre’s poor polling numbers have put their seats in danger. It’s undoubtedly been one of the factors contributing to the loss of four MPs to the Liberals through floor-crossing, with perhaps more to come. But some MPs are in greater need of a bounce back by the party (and have more reason to be tempted by the life raft of the Liberal benches). Who are they?

As it stands, the Conservatives would likely see their holdings in the House drop significantly if an election were held today. This week’s update to The Writ’s Vote and Seat Projection puts the party ahead in just 83 seats, representing a decrease of 57 seats from the 140 the Conservatives currently hold. Even the party’s Avg. Projection, which takes into account the likelihood that the Conservatives would outperform their polls, would still have the party losing 40 seats.

There’s been little change in the projection since last week. The Liberal lead now stands at 12.4 points in the vote projection (46.3% to 33.9%), with the NDP stuck at 8.8%. The Liberals are projected to be leading in 226 seats, a gain of three since last week that has come equally from the Conservatives, Bloc Québécois (22) and NDP (10).

That things have not deteriorated further for the Conservatives might be a silver lining for the party — but for a lot of MPs, that just means that the party is currently bottoming out at levels that put their own seats on the chopping block.

For fully interactive charts, tables and maps, head to The Writ’s website:

Let’s get to those seats that are most at risk. The Riding Projections classify seats from Safe to Toss. There are a lot of Conservative seats projected to be Liberal Tosses, but those could go either way. Instead, let’s focus on the Conservatives who are in most need of a comeback for their party and are in the greatest danger of losing — even if they beat their polls.

Read more

That being the case, I figured I should update the Vote and Seat Projection Model a day early. Not only will this give us a chance to take a look at what the model says we should expect in the two Toronto ridings of Scarborough Southwest and University—Rosedale and the Quebec riding of Terrebonne, it’ll also give us a baseline against which to judge the results.

Before diving into the numbers, let me take this opportunity to invite you to tune in to our coverage of the byelection results, starting as soon as the polls close at 8:30 PM ET tonight on YouTube!

We’re expecting a bit of a late night. We should be able to get through the two Toronto ridings fairly quickly before turning our attentions to Terrebonne, where a special write-in ballot will delay the count in what is expected to be a closely-contested seat. We might have to stick around until every vote is counted!

By then, though, we should already know if the Liberals get their majority government if the two Toronto seats are called relatively early. But, if the Liberals can take it, Terrebonne will provide them more cushion. Conversely, if the Bloc can win it back, then they will deal Mark Carney his first real setback since the April 2025 election.

Before getting to the byelections, here’s where the latest projection puts things nationally:

The Liberals have slipped 0.4 points in the vote projection and two seats in the seat projection since last week, but otherwise are holding steady with 45.9% of the vote and 223 seats. The Conservatives are up a small 0.3 points to 34.1%, but are unchanged at 84 seats. The Bloc is up one seat to 23, while the NDP is also up one seat to 11 and sits at 8.8% of the vote. That share of the vote is virtually unchanged from last week — and this is after the inclusion of a few polls conducted after Avi Lewis become NDP leader. In short, we’re seeing no Lewis bump just yet.

For fully interactive charts, tables and maps, head to The Writ’s website:

That’s the big picture. What about the three ridings holding byelections today? Below are the results from the last general election, when the Liberals’ Bill Blair won in Scarborough Southwest and Chrystia Freeland prevailed in University—Rosedale. In Terrebonne, Tatiana Auguste defeated the Bloc’s Nathalie Sinclair-Desgagné by a single vote, a result that was overturned by the Supreme Court due to an error made by Elections Canada that could’ve impacted the result.

How do these three seats look today, and what benchmarks can we set to help us judge the results?

Read more

While there hasn’t been a significant amount of movement over the last week in the polls, what has occurred has pushed the Liberals above the 220-seat threshold in the projection. They are now projected to be leading in 225 seats in the model, up 10 seats from last week. Most of those gains have taken place in one province: Ontario.

But there’s a caveat.

The Conservatives have dropped down nine to just 84 seats, their lowest level in the projection to date since the last election. The Bloc is unchanged at 22 seats while the NDP is up one to 10.

In the vote projection, the Liberals have ticked up 0.5 points to 46.3%, followed by the Conservatives at 33.8% (-0.3) and the New Democrats at 8.9% (-0.3).

For fully interactive charts, tables and maps, head to The Writ’s website:

But the key number to look at in this week’s update is the Avg. Projection, which takes into account expected over- or under-performance of each party in an actual election. While the Liberals have jumped 10 seats in the number of ridings in which they are projected to be leading, their average projection has only increased by 2.9 seats to 212.1 seats, while the Conservatives have slipped only 2.8 in the average projection to 102.2 seats. What gives?

Read more

The Liberals sit at 45.8% in the vote projection, down only 0.2 points from last week. The party has been sitting around 46% since the beginning of March, a new cruising altitude for the Liberals since rising from around 40% in January. The party is projected to be leading in 215 seats (and averaging 209.2 seat wins when taking into account potential polling and modelling errors), again largely unchanged since the beginning of the month.

While the Liberals have levelled out, it’s not clear that the Conservatives have. They’re down just 0.3 points from last week to 34.1% in the projection, but unlike the Liberals they have been on a steady (if slow) downward trend. With the exception of a few marginal upticks, they’ve been dropping almost every week since January when they were at 38%. This has pushed them down to 93 in the number of seats they are leading in and an average of 105 seat wins.

For fully interactive charts, tables and maps, head to The Writ’s website:

We’ll have to keep an eye on those Conservative numbers. Some polls have suggested that the party might be rebounding, but for every (relatively) good poll we’ve seen for the Conservatives lately there have been some bad ones.

The New Democrats are sitting at 9.2% in the vote projection and 11 seats, but just 8.2 seats on average when taking into account their tendency to under-perform the polls. These numbers are drawn from polling conducted before Avi Lewis’s leadership victory on Sunday but, as we can see from the regional breakdown, the NDP has a lot of work to do.

Read more

The Conservatives stand at 34.4%, with both the Liberals and Conservatives virtually unchanged since last week’s update. The New Democrats, however, are up 1.2 points to 9%.

There’s been little change in the national seat projection, with the Liberals down one seat to 217, the Conservatives up three seats to 94, the Bloc down one to 21 and the NDP down one to nine. The Greens are holding at two.

In the Average Projection, which takes into account potential polling and modelling errors, the Liberals are still above the 200-seat threshold at 208.4. The Conservatives, who are expected to beat their polling, have an average projection of 107.8 — which certainly looks a lot better than being sub-100.

Visit the Vote and Seat Projection pages at The Writ to see all the interactive charts, tables and maps:

While there hasn’t been much movement at the national level, there has been some shifts regionally, the biggest having been in Alberta.

Read more

That’s what The Writ’s new model shows in its inaugural projection. Yes, the vote and seat projection model is finally live!

I’ve been working away at a new model since the last election and, after a flurry of activity over the last few weeks, it is now finally ready to present. A methodological explanation can be found here, if you want to dive into the details. The full national and regional projections can be found here, and full riding projections can be found here.

These will all be best viewed on the website, as the charts are interactive. But let’s take a brief look at some of the topline numbers as this projection is launched.

The Liberals are projected to be leading in 218 seats across the country, a significant increase of 49 seats from the last election. That puts them well over the threshold of 172 seats required for a majority government, hardly a surprise when they are projected to take 46.1% of the vote. The party’s odds of winning a majority if an election where held today are 92%. Their odds of a majority are 95%.

If things are going well for the Liberals, it’s because things are going badly for the Conservatives. They’re down to 34.6% of the vote and just 91 seats, a fall of 53 seats since the last election. The Bloc Québécois is holding at 22 seats, while the New Democrats might be able to take enough advantage of the Conservatives’ weakness to jump to 10 seats with 7.8% of the vote. The Greens, at 2.3%, have a shot at two seats.

Below you’ll see the Vote and Seat Projection table that you’ll find at the top of the Vote and Seat Projections page. Let me take you through how to read it.

In the first column is the Vote Projection. As it is reflects the sum total of all of the riding-level projections in the model, it isn’t a simple polling average (though that is where things start). It’s the best guess at what the result would be in a snap election based on where the polls are today and adjusted for any important dynamics in the 343 ridings across the country.

The next three columns show the Seat Projection with the high and low ranges. The Seat Projection is the number of seats in which each party is projected to be leading, based on the polls and any local riding dynamics (such as the presence or lack thereof of an incumbent or a floor-crosser, to name a few of the factors the model takes into account). The low and high ranges show 95% of all likely outcomes, taking into account potential polling and modelling error. You’ll notice that there is a much bigger gap between the Conservative projection and their high range vs. their low range, and vice versa for the New Democrats. This is not an accident, as the ranges take into account how parties perform in actual elections relative to their polls.

The final column on the right shows the Avg. Projection, which is the result of 10,000 simulations. Like the ranges, this takes into account potential polling and modelling error. The Conservatives often beat their polls; the New Democrats don’t. Accordingly, the Avg. Projection for the Conservatives is 106.6 seats rather than 91, because they are more likely than not to beat their polls and win more of the toss-ups than expected. The NDP Avg. Projection is 6.8 seats, as they are more likely than not to come up short in a few of the races in which they are projected to lead.

How should you read these numbers? The Seat Projection is The Writ’s “official call”, as it shows what the result of an election would be based on where the polls are today — it doesn’t assume the polls and the modelling will be wrong in one direction or another, as they aren’t always wrong in the same direction from one election to the next. But the Avg. Projection should also be taken into consideration, as it gives an indication of where to set expectations. If tomorrow was election day, the New Democrats should win 10 seats with where the polls are. But don’t be surprised if they would win closer to seven.

My plan is to update the model once per week, usually on Tuesdays, with short posts published to go over the changes in the update. But I might post updates more frequently if the need arises. The model would be updated daily during a campaign.

There’s so much more to dig into, as you’ll see if you head over to the website. Tracking data going back to the last election, riding-by-riding projections and full regional breakdowns. Subscribers to The Writ will have more to explore on their own, but let me take you through some of the highlights.

Already a subscriber to The Writ? You can skip ahead to some more charts and analysis below. But if you’re not a subscriber, here’s a little sneak peak of what is behind the curtain at The Writ’s new vote and seat projections: Charts!

Read more

While I’ve been building and running election models for over 15 years and ran the CBC’s Poll Tracker in the 2015, 2019, 2021 and 2025 elections (as well as for a handful of provincial elections), this model is newly-designed from the ground up. I wrote about some of my thought processes during the development of the model here:

The Writ’s vote and seat projection model was launched in March 2026. Prior data was loaded into the model to calculate projections stretching back to the April 2025 election for comparison purposes.

This projection model will always be in a state of evolution as changes will be made when required (and, if significantly changing the model, explained in full). While I’ve taken as many steps as possible to reduce the amount of my own subjective judgment that is applied to the projections, there might arise some situations in which there are few or no historical precedents to use as guidance. In those cases, some ad hoc adjustments to the projection model might be made by me.

Calculating the polling average

National and regional polls are weighted by three factors.

The first is the date of the poll, based on the mid-point of the poll’s field dates. Outside of a campaign, polls with a mid-point date within the last week are given a full weight, while during a campaign full weights are given to polls with a mid-point date within the last three days. Outside of a campaign, the date weight of a poll is reduced by 7% per day, while during a campaign it is reduced by 25% per day.

The second weight is the sample size, with larger sample sizes given more weight than smaller sample sizes, up to a limit of 1,500.1

The third weight is the track record of a polling firm, with polling firms with a better track record give more weight than polls with a worse track record.

The relative weights of these three factors is 70%, 25% and 5%, respectively, with the sample size and track record weight reduced at the same pace as the date weighting. After each poll is assigned a weight, it is then used to calculate an average in a given region. Polling averages are calculated for each region and sub-region, depending on what data is available.

Riding-level projections

For each region with a polling average, a baseline riding projection is calculated for each riding in that region. The riding projection is a 50/50 combination of a proportional and uniform swing model.

With proportional swing, a party’s result in a riding in the previous election is adjusted proportionally by how a party’s support has shifted in the region as a whole since that election.2

With uniform swing, a party’s result in a riding is adjusted uniformly by how a party’s support has shifted in the region as a whole.3

The average between the proportional and uniform swing is then calculated for each party in each riding, and then those numbers are adjusted to ensure support adds up to 100%.

The projection for each riding in each region with a polling average is then assigned a weight according to the weight of all the polls in the polling average for that region, and a weighted average riding projection is then calculated for every riding.

At this point, adjustments are made to the projection to take into account local factors. This include:

• Byelections: When a seat has changed hands in a byelection, the vote projection in the region where and when the byelection was held is used as the new baseline from which to apply the proportional and uniform swing. When a seat hasn’t changed hands, the previous general election result is still used as the baseline. Regardless of whether the riding changed hands or not, the error range for the riding’s projection is widened to increase the level of uncertainty.

• Incumbency: Based on historical precedent, the presence of an incumbent marginally boosts support for the Liberals or NDP, largely at the expense of the other party.4 The lack of an incumbent penalizes the party that lost the incumbent by between three to four points, with different parties getting a corresponding boost depending on which party has lost the incumbent.

• Floor-crossing: The party that has welcomed a floor-crossing MP receives a boost that comes at the expense of the party that lost the floor-crossing MP. The size of that boost can be adjusted according to the circumstances but is based on past precedent.

• Independents: The presence, or departure, of significant Independent MPs or candidates is taken into account.

• Party leaders: A party leader running in a riding for the first time gets a significant boost in the model, based on past precedents. A party whose leader has vacated a riding is penalized to a greater extent than when it is a normal incumbent MP that has retired.

• Candidates with previous electoral experience: New candidates who have previously sat in a provincial or territorial legislature or who were leaders of a provincial or territorial party are awarded a boost which comes at the expense of other parties, depending on the party of the candidate.

• Withdrawn or no candidate: When a major party withdraws a candidate but that candidate will still be present on the ballot as a candidate for that party, a smaller share of what the candidate would have normally been expected to get is awarded to the party. That reduction is awarded to other parties depending on which party withdrew their candidate. When a major party has no candidate on the ballot, the share that would have been projected for a candidate from that party is distributed to other parties in different amounts depending on which party does not have a candidate. When a party did not run a candidate in the last election, previous results are used to estimate that party’s support.

The size of boosts or penalties assessed is determined by looking at how past actual election results compared to expected results when applying the mixed uniform/proportional swing model to those past elections.

Once these adjustments are complete, a final projection for the riding can be calculated. These are the numbers that you can find on the Riding Projections page.

Vote projection

The national, provincial and regional vote projections are calculated by summing the riding projections. In other words, the weighted polling average and the vote projection are two separate things, and only the vote projection is presented.

The projection for each province is calculated by weighting each riding projection by the average of 1) the riding’s share of the province’s total eligible voting population in the last election and 2) the riding’s share of the province’s total actual voters in the last election. With each riding assigned a turnout weight, they can then be combined to calculate the vote projection for the province as a whole.

This means that the vote projection takes into account not only turnout dynamics but also the adjustments made to each individual riding. A party losing a large number of incumbents in a province, for example, would be projected to take less of the vote than the polling average would suggest.

To calculate the national (and, in the case of Atlantic Canada, regional) vote projection, the provincial vote projections are weighted by the share of each province’s national turnout in the last election. Different provinces have consistently and predictably had higher or lower turnout than the country as a whole.5

Since the national vote projection is dependent on the provincial vote projections, single-province or region polls can influence the national vote projection.

Sub-regional vote projections presented with the Riding Projections (such as for Toronto or Montreal) are a simple average of the individual riding projections in that region, with a few exceptions made for regions with ridings with significantly smaller populations or lower turnout.6

Seat projection

The Seat Projection is simply the number of ridings in which a party is projected to be leading.

However, each party in each riding also has a lower and higher range of likely support. These ranges are based on the past performance of this model when applied to elections between 2011 and 2025, meaning the ranges take into account both the errors in the model itself and in the polls. For that reason, the ranges aren’t a simple +/- calculation, with the likelihood of the result being higher or lower than the projection being equal. The precise projection for each riding does not take into account the past errors in the polls, but the projection ranges do take into account past polling error.

This means that parties that have historically beaten their polls (such as the Conservatives) have a higher high range and a lower low range than parties that have historically under-performed their polls (such as the New Democrats). These tendencies to under- or out-perform the polls is calculated for each party in each region. In some regions, for example, parties have largely matched their polls while they haven’t in others. Alberta and Saskatchewan for the Conservatives are good examples of where the polls have been off by greater amounts than in other regions.

These upper and lower ranges are then used to determine which seats are in contention for each party at each confidence level. More seats are in contention at the 95% confidence level, where the ranges are greater, than at the 75% confidence level, where the ranges are smaller.

The classification of a projection as Safe, Likely, Lean, Close, Edge and Toss reflects at which level of confidence does a seat become contested. If a seat is projected as “Safe”, it means that with the widest projected ranges for every party the leading party’s floor is higher than every other party’s ceiling at the 95% confidence level. “Likely” represents the 85% to 95% range, “Lean” the 75% to 85% range, “Close” the 65% to 75% range, “Edge” the 55% to 65% range, and “Toss” is 55% and less.

Since the upper and lower ranges differ for each party, in each region and at each confidence level, the classification of a projection is not uniformly based on the same margins. Two seats contested by the Conservatives and Liberals with the same margin between the two parties might be an “Edge” call in Ontario and a “Lean” call in Alberta because of how the ranges for the two parties differ from region to region. For example, a Liberal lead in Alberta is going to be more uncertain than a Conservative lead of the same size in the province, because the Conservative floor is higher than the Liberals’ floor.

The precise riding projections do not take into account these past polling performances, however, because polling errors are not consistent from one election to the next. Rather than risk adding to potential errors7, I have decided to let the riding projections reflect the seats each party should be able to win based on where the polls are, and let the projection classifications reflect the certainty (or uncertainty) of those projections.

Calculating probabilities and the Avg. Projection

Using the ranges at each confidence level, the number of seats being contested between two or more parties is calculated. Then, a proportion of these contested seats are randomly assigned to each party to give each party a total of seats won. The result is one election simulation that incorporates the uncertainty in the riding projections. This process is repeated 10,000 times to simulate 10,000 individual election results.

The presented “chances of winning” the election is the equivalent of the share of times one party has the most seats and/or crosses the 172-seat threshold for a majority government in those 10,000 simulations.

The high and low seat projection ranges reflect the 95th percentile of outcomes in those 10,000 simulations, while the Avg. Projection is the average outcome of those simulations. As the simulations rely on the projection ranges for each individual riding, which themselves are calculated by assessing the tendency for parties to out- or under-perform their polls and the model, the ranges and the Avg. Projection take into account the likelihood that parties will beat or under-perform their polls and the model.

• Any questions regarding the model? Leave them as a comment below or send me an email.

Leave a comment