From national polls to 343 ridings: regions, Quebec and the swing model
Abstract. How the Canada forecast turns national shares into riding shares. Seven regions are each read two ways, by national swing from 2025 and from the pollsters' own regional figures measured against the same release's national figure, blended by how many pollsters report the region and scaled to add back up to the nation. Each riding's 2025 result then moves by its region's change by uniform swing. The paper covers the 151 regional readings, Quebec and the Bloc, a worked riding, seats by province, and the Alberta and Manitoba provincial forecasts.
Summary
Canada elects 343 MPs in 343 separate first-past-the-post contests, so the seats depend on where the votes are. This paper describes how the forecast turns national shares into riding shares: seven regions, each read from the regional figures pollsters print in their releases and blended with national swing, then each riding's 2025 result moved by its region's change. It covers the special case of the Bloc Québécois, whose whole vote is in Quebec, works through one riding, and gives today's figures region by region and province by province.
Two design choices carry most of the weight. First, each pollster's regional figure is measured against its own national figure in the same release, so a pollster that runs high on a party nationally does not push that party up in every region too. Second, the regions are scaled after blending so that they add back up to the corrected national shares; the regional polls redistribute a party's vote between regions but never change its national total. The last section describes the separate forecasts for the Alberta and Manitoba provincial elections.
Why regions
Under first past the post a party wins nothing for second place, and the same national share can give very different numbers of seats. At the 2019 and 2021 elections the Conservatives won the most votes (34.3% and 33.7%) and the Liberals the most seats (157 and 160 against 121 and 119), because Conservative votes piled up in Alberta and Saskatchewan while Liberal votes were spread more efficiently across Ontario, Quebec and Atlantic Canada.
Canadian elections are also regional in their swings. In 2011 the NDP took 59 of Quebec's 75 seats as the Bloc fell to four; in 2015 the Liberals won all 32 seats in Atlantic Canada; in 2025 the NDP lost ten of its thirteen seats in British Columbia and all five in Ontario. A forecast that moved every riding by the same national change would miss these stories, and the replays in Testing the seat engine: the plurality check, 2025 and the swing replays show what that costs: across seven pairs of elections from 2004→2006 to 2021→2025, uniform swing applied nationally got 373 ridings wrong, and applied region by region with the true regional change, 268.
The seven regions
Pollsters report regional figures in six groups, and we use the same six, plus the territories as a seventh:
| Region | Provinces and territories | Ridings | Share of the 2025 national vote |
|---|---|---|---|
| Atlantic Canada | Newfoundland and Labrador, Prince Edward Island, Nova Scotia, New Brunswick | 32 | 7.2% |
| Quebec | Quebec | 78 | 22.8% |
| Ontario | Ontario | 122 | 38.6% |
| Manitoba and Saskatchewan | Manitoba, Saskatchewan | 28 | 6.1% |
| Alberta | Alberta | 37 | 11.5% |
| British Columbia | British Columbia | 43 | 13.5% |
| The territories | Yukon, Northwest Territories, Nunavut | 3 | 0.2% |
Table: The seven regions. Ridings under the 2023 Representation Order; vote shares from the 2025 riding results.
The territories have no regional polls and follow national swing.
Two readings of each region
For each region \(r\) the forecast forms two estimates of each party's share and blends them.
National swing
Let \(N_p\) be the corrected national share today (How we read Canadian polls: the average, the correction and its test), \(N^{25}_p\) the 2025 national share, and \(B_{r,p}\) the party's 2025 share in the region. National swing scales the regional result by the national change and rescales the region to 100:
This is proportional swing at the regional level. It keeps each region's character (the Conservatives far stronger in Alberta than in Quebec) and moves every region in the same proportion.
The regional polls
Many Canadian releases include a regional table. For a regional figure \(x_{j,r,p}\) in pollster \(j\)'s release, and that release's own national figure \(X_{j,p}\), the deviation is
The regional polls' reading of the region is our national share plus the average deviation across the \(n_r\) pollsters with a regional figure for that region:
where \(n_{r,p}\) counts the pollsters that report party \(p\) in the region. Each pollster's latest reading for the region counts, if its fieldwork ended within 90 days of the newest national poll (a setting on the admin page).
Measuring inside the release is the important step. If a pollster has the Conservatives four points higher than the average nationally, its regional figures will usually be high too. Taken at face value they would push the Conservatives up in every region it covers. Measured against the same release's national figure, the house effect cancels, and only the regional pattern remains. That pattern is then applied to our national share, which is already corrected.
Two kinds of reading need extra rules. A stand-alone regional poll (a Quebec-only or Alberta-only poll, say) has no national figure in the same release, so it is measured against our national poll average at the time it was taken. And some releases name only some parties in a region (Abacus Data's posts give the Liberals and Conservatives, and the Bloc in Quebec, in prose): a table with four parties or more is rescaled to 100, but a partial one keeps its printed figures, and a party it does not name takes its national-swing share in equation (3).
The blend
The two readings are blended with a weight on the polls that grows with the number of pollsters:
One pollster gets a third of the weight, two get half, three get 60% and six would get three quarters. A region with no regional polls keeps national swing. A party that had no vote in a region in 2025 stays at zero there: the Bloc everywhere outside Quebec, and the People's Party in the territories, where it had no candidates.
Adding back up to the nation
The blended regions need not add up to the national shares: the regional polls might put the Conservatives above national swing in four regions and below it in none. The forecast therefore scales each party's regional shares so that, weighted by each region's 2025 share of the national vote \(\omega_r\), they reproduce the national share,
then rescales each region to 100, and repeats both steps three times. The result is a set of regional shares that respects the national figure while keeping the regional pattern the polls show.
This step has a consequence worth stating. If the regional polls show a party gaining less than national swing in some regions, equation (5) gives the difference to the other regions. Today the regional polls have the NDP well below national swing in British Columbia (9.8% against 22.9%), and the blend there is 16.4%; scaling to the national NDP share of 11.9% then lifts the NDP in every region, to 19.5% in British Columbia and 10.2% in Ontario, where both readings had it lower. The regional polls decide where a party's vote is relatively strong; the national polls decide how much of it there is.
Where the regional figures come from
Since the 2025 election we hold 151 regional readings: 138 regional tables in national releases and 13 stand-alone regional polls.
| Pollster | Regional readings | Stand-alone polls among them |
|---|---|---|
| Léger | 79 | 7 |
| Abacus Data | 28 | 1 |
| Innovative Research | 18 | 0 |
| Mainstreet Research | 10 | 4 |
| Liaison Strategies | 7 | 1 |
| Pollara | 6 | 0 |
| Research Co. | 3 | 0 |
Table: Regional readings since the 2025 election, by pollster.
By region: Quebec 36, British Columbia 27, Ontario 25, Alberta 23, Atlantic Canada 20, Manitoba and Saskatchewan 20. Léger's monthly report is the backbone: its PDF has a crosstab with a column for each of the six regions, and the poll check reads it automatically (Rating Canadian pollsters: accuracy, the lead and openness). Other pollsters publish less. Nanos keeps its regional ballot figures on a subscriber portal, so none are stored. Angus Reid prints its regional shares in stacked charts rather than tables. EKOS posts its tables mostly as images. Pallas gives regional leads in prose (margins, not shares). Mainstreet's full crosstabs are behind a subscription, though its public PDFs sometimes carry regional tables.
The 90-day window means that on any day only the most recent readings count. On 15 October 2026 there are 13: Abacus Data for five regions (Quebec, Ontario, Manitoba and Saskatchewan and Alberta from 23 September, British Columbia from 9 September), Léger for all six (7 September) and Liaison Strategies for Quebec and Ontario (29 August).
Today's regions
| Region | Pollsters | Weight on polls | Liberal | Conservative | NDP | Bloc | Green | |
|---|---|---|---|---|---|---|---|---|
| Atlantic Canada | 1 | 33% | swing / polls | 60.2 / 58.6 | 27.9 / 32.9 | 8.1 / 5.8 | – | 1.9 / 1.7 |
| estimate (2025) | 58.6 (55.5) | 29.1 (38.0) | 8.8 (4.3) | – | 1.5 (1.1) | |||
| Quebec | 3 | 60% | swing / polls | 45.3 / 47.2 | 16.7 / 16.2 | 8.3 / 6.8 | 26.2 / 28.3 | 1.5 / 1.7 |
| estimate (2025) | 45.2 (42.6) | 16.0 (23.3) | 8.8 (4.5) | 26.8 (27.7) | 1.3 (0.9) | |||
| Ontario | 3 | 60% | swing / polls | 54.1 / 52.8 | 32.6 / 31.8 | 9.3 / 7.8 | – | 1.9 / 3.7 |
| estimate (2025) | 53.1 (49.2) | 32.1 (43.8) | 10.2 (4.8) | – | 2.5 (1.1) | |||
| Manitoba and Saskatchewan | 2 | 50% | swing / polls | 37.8 / 42.5 | 41.2 / 41.1 | 18.2 / 12.8 | – | 1.1 / 2.7 |
| estimate (2025) | 38.7 (34.1) | 39.8 (54.9) | 18.3 (9.4) | – | 1.5 (0.6) | |||
| Alberta | 2 | 50% | swing / polls | 32.3 / 30.6 | 49.9 / 56.4 | 12.7 / 8.9 | – | 0.7 / 1.7 |
| estimate (2025) | 30.4 (27.9) | 51.6 (63.6) | 12.8 (6.3) | – | 1.0 (0.4) | |||
| British Columbia | 2 | 50% | swing / polls | 42.2 / 55.5 | 28.1 / 27.6 | 22.9 / 9.8 | – | 4.7 / 3.7 |
| estimate (2025) | 47.6 (41.8) | 27.3 (41.2) | 19.5 (13.0) | – | 3.4 (3.0) | |||
| The territories | 0 | 0% | swing | 50.8 | 23.5 | 23.6 | – | 2.1 |
| estimate (2025) | 48.3 (50.5) | 22.5 (34.6) | 27.5 (13.4) | – | 1.7 (1.4) |
Table: Regional readings on 15 October 2026, vote share. "Swing" is equation (1), "polls" equation (3), the estimate the result of equations (4) and (5); the 2025 result in brackets.
The largest disagreement between the two readings is in British Columbia. National swing, which spreads the Liberals' national gain since 2025 in proportion to their 2025 vote, gives them 42.2%. The two pollsters with a recent British Columbia figure put the Liberals 3 (Abacus) and 13 (Léger) points above their own national figures, which on our national share gives 55.5%. The blend, after scaling, is 47.6%. In Alberta the regional polls see the Conservatives holding up better than national swing (56.4% against 49.9%), and in Manitoba and Saskatchewan the Liberals doing better (42.5% against 37.8%).
Mapping any national shares onto the regions
Once today's regional estimates are fixed, the forecast keeps, for each region and party, the ratio of the regional share to the national share,
Any other set of national shares \(N'\), a simulated election, a past miss replayed or a reader's own figures in the what-if tool, is mapped onto the regions as \(N'_p \rho_{r,p}\), rescaled to 100 in each region. A national change is spread across the regions in proportion to each party's regional strength today. A point added to the Conservatives nationally adds about 1.7 points in Alberta (where their ratio is 51.6/30.3) and about half a point in Quebec (16.0/30.3).
Quebec and the Bloc
The Bloc Québécois stands only in Quebec's 78 ridings. Its national share of about 6% is a Quebec share of about 27%, a ratio of 4.39 today. Every national point for the Bloc is therefore about four and a half points in Quebec, and the Bloc's seats are unusually sensitive to its national figure. Run through the seat engine, moving one point nationally from the Liberals to the Bloc raises the Bloc in Quebec from 26.8% to 30.1% and its seats from 21 to 26, while the Liberals fall from 241 to 231.
The Bloc's record in the polls helps here. Its national share has been polled within about a point at every election since 1993 (How far Canadian polls have missed, 1988–2025), and its RMS miss since 2000 is 0.54 points, the smallest of any party. The national correction for the Bloc is accordingly tiny (+0.07 points), and its simulated national error is the narrowest (Simulating Canadian elections: where the chances come from).
Quebec is also the one four-party region. Today's estimate has the Liberals on 45.2%, the Bloc 26.8%, the Conservatives 16.0% and the NDP 8.8%. Of the 48 ridings that change hands under at least one past polling miss replayed on today's polls, 15 are in Quebec, most of them Liberal–Bloc contests (If we're wrong: today's polls with each past miss replayed). Three regional pollsters cover Quebec today, as many as Ontario and more than any other region.
Moving the ridings
Each riding \(i\) in region \(r\) starts from its 2025 result \(b_{i,p}\). Under uniform swing, the method in use, it moves by its region's change in points:
and the riding is rescaled to 100. The candidate of the party with the largest share wins, as under s.313 of the Canada Elections Act.
Three rules complete the step. A party with no candidate in a riding in 2025 stays at zero there, which keeps the Bloc out of every riding outside Quebec and keeps a party that skipped a riding from appearing in it. A share that would fall below zero is cut to zero before the rescale. And a party that had no vote anywhere in the region in 2025 would take the regional share in every riding; with today's rules that can only arise for "others".
Uniform swing was chosen by test. Replaying seven pairs of elections with the true regional change, it got 268 ridings wrong in all, against 278 for an even mix of uniform and proportional and 288 for proportional (Testing the seat engine: the plurality check, 2025 and the swing replays). The choice is made by the code each time the forecast is built: the method with the fewest ridings wrong is used.
A worked riding: Repentigny
Repentigny was won by the Bloc in 2025 with 42.2% against the Liberals' 38.8%, the Conservatives' 15.2% and the NDP's 2.7%. Quebec's change since 2025 in today's estimate is Liberal +2.6, Conservative −7.3, NDP +4.3 and Bloc −0.9 points. By equation (7) the riding becomes Liberal 41.4, Bloc 41.3, Conservative 7.9 and NDP 7.0 before the rescale, and after it, with the smaller parties, Liberal 41.5 and Bloc 41.5. It is the closest riding in the country on today's projection: the Liberals lead by less than 0.05 points, and in the simulations the Liberals win it 48% of the time. Five of the nine past misses replayed on today's polls give it to another party.
Seats by province
| Province | Seats | Liberal | Conservative | Bloc | NDP | Green |
|---|---|---|---|---|---|---|
| Newfoundland and Labrador | 7 | 7 (4–7) | 0 (0–3) | |||
| Prince Edward Island | 4 | 4 | ||||
| Nova Scotia | 11 | 11 (9–11) | 0 (0–2) | |||
| New Brunswick | 10 | 8 (6–9) | 2 (1–4) | |||
| Quebec | 78 | 49 (42–60) | 7 (4–9) | 21 (11–27) | 1 (0–2) | |
| Ontario | 122 | 108 (85–115) | 13 (6–35) | 0 (0–2) | 1 (0–1) | |
| Manitoba | 14 | 7 (6–9) | 5 (4–5) | 2 (1–3) | ||
| Saskatchewan | 14 | 5 (3–6) | 9 (8–11) | 0 (0–1) | ||
| Alberta | 37 | 8 (4–13) | 27 (22–31) | 2 (1–2) | ||
| British Columbia | 43 | 32 (27–37) | 4 (2–10) | 6 (3–7) | 1 (0–1) | |
| Yukon | 1 | 1 | ||||
| Northwest Territories | 1 | 1 | ||||
| Nunavut | 1 | 0 (0–1) | 1 (0–1) | |||
| Canada | 343 | 241 | 67 | 21 | 12 | 2 |
Table: Projected seats by province, 15 October 2026, with the 80% range from the simulations in brackets where it is not a single value.
Ontario decides the size of the majority: its range for the Conservatives, 6 to 35 seats, is wider than in any other province, because many suburban ridings around Toronto are Liberal–Conservative contests within a few points. Against the 2025 result, the projection has 79 ridings changing hands: 70 from the Conservatives to the Liberals, 5 from the Conservatives to the NDP, 2 from the Bloc to the Liberals, and one each from the Conservatives to the Bloc and to the Greens.
Provincial elections: Alberta and Manitoba
Beside the federal forecast the site carries nowcasts for two provincial general elections, both on fixed dates: Alberta on 18 October 2027 (Election Act, s.38.1(2): the third Monday in October, four years after the last) and Manitoba on 5 October 2027 (Elections Act, s.49.1(2)(b): the first Tuesday in October). Both legislatures are elected by first past the post: Alberta's 87 electoral districts, with 44 for a majority, and Manitoba's 57 electoral divisions, with 29.
The method is a simpler cousin of the federal one.
- Average. Each firm's latest provincial poll within 90 days of the newest, equal weight, rescaled to 100.
- Districts. Each district's 2023 result moves by each party's change in the polls since 2023, mixing uniform and proportional swing: for a party with at least 3% in the region, the new share is the average of the proportional result and the larger of the uniform and proportional results. For the two main parties (the UCP and NDP in Alberta, the NDP and Progressive Conservatives in Manitoba) the change is half the province's and half the district's region's, where pollsters publish regional figures (Calgary, Edmonton and the rest of Alberta; Winnipeg and the rest of Manitoba). Pollsters draw those regions differently (city proper or metropolitan area), which is why they get only half the weight.
- Uncertainty. 2,000 simulated elections, each with a province-wide polling miss per party (standard deviation 4 points for the two main parties and 2 for the others, a setting), a regional miss for the main parties (1.5 points) and noise in every district with a standard deviation of \(3 \times (1 + 0.05\,v)\) points for a main party on \(v\)% (6 points at 20%, 9 at 40%), and half that for the others.
- No correction. The provinces' own polling record has not been collected, so their polls are taken as published.
On today's figures the UCP would win 45 of Alberta's 87 seats (80% range 39 to 58) and the NDP 42 (29 to 48), with a UCP majority in 76% of runs; the poll average is UCP 46.4%, NDP 41.9%, from four pollsters. In Manitoba, on one pollster (Probe Research, 14 September 2026: NDP 60%, Progressive Conservatives 29%), the NDP would win 43 of 57 (40 to 48). Both projections start from the 2023 general elections; by-elections since then change who holds a seat but not the base. The provincial forecasts do not feed the federal one.
The 2026 Quebec general election, held on 5 October, is shown on the Canada home page with Élections Québec's count; the site made no forecast for it.
Limits
Thin regional data. Two or three pollsters per region, one in Atlantic Canada and none in the territories, with Léger's monthly report the only one covering all six. A region's reading can turn on one release.
Uniform swing within a region. Every riding in a region moves by the same number of points. Within Ontario, Toronto and its suburbs and the north of the province can move differently, and the model cannot see it.
The ratios are today's. Simulated and replayed national errors are spread in proportion to today's regional pattern. A real miss concentrated in one region is represented only by the separate regional noise in the simulations.
No local candidates. Incumbency, retirements and floor-crossings are not in the swing; each riding starts from its 2025 vote. What we don't know yet: limitations of the Canada forecast discusses each of these.
References
Canada Elections Act, S.C. 2000, c. 9, s.313.
Election Act (Alberta), R.S.A. 2000, c. E-1, s.38.1.
The Elections Act (Manitoba), C.C.S.M. c. E30, s.49.1.
Figures in this paper are as of 15 October 2026; the pages update with each poll.
How to cite
Polling Forecast (2026). From national polls to 343 ridings: regions, Quebec and the swing model. Retrieved 17 October 2026 from https://www.pollingforecast.com/research/canada-regions. Live figures change as new polls arrive; the date above is when the text was last revised.