AN EXPLICIT METHOD
How 127QC works
127QC turns polls into estimates for Quebec’s 127 ridings. The projection describes the race based on the data available; it does not predict the election result with certainty.
From polls to seats
- Voting intentions. The latest eligible poll from each firm is kept within a 28-day window. Recent polls and large samples carry more weight, with a cap to limit the influence of any single poll. The same rule applies to every firm.
- Local distribution. The 2022 results, transposed onto the 2026 map, are the starting point. They are adjusted according to how support for each party has changed.
- Regions. Recent regional polls are used to adjust support in the Montreal and Quebec City metropolitan areas and in the rest of Quebec. Because they rest on small samples, they count for 75% of the regional adjustment; the remaining 25% comes from the historical distribution. A riding that straddles a regional boundary is split between the regions concerned. The results stay consistent with the provincial average.
- Uncertainty. The model computes 5,000 scenarios by varying voting intentions and local results. These scenarios produce the averages, intervals and probabilities shown.
How to read the numbers
The central scenario
The large numbers on the home page, the chart, the map and the party filter count the ridings where each party leads the central estimate. A narrow lead can easily flip.
Averages and intervals
The simulation average can differ from the number of ridings in the lead. The 90% interval covers the results between the 5th and 95th percentiles of the simulations; it depends on the model’s assumptions.
Probabilities
A displayed probability of 70% means the party wins about 7 simulations out of 10. This figure is indicative and is not an electoral guarantee.
Limits of the projection
- Personal effects. Two rules common to all parties are applied. If the party that won a riding in 2022 is running its incumbent MNA, an initial adjustment of +1.3 points applies; otherwise, the adjustment is −1.0 point. If a party’s 2022 leader is not running again in their riding, the initial adjustment for that party is −4.9 points (L’Assomption, Saint-Henri–Sainte-Anne, Gouin, Sainte-Marie–Saint-Jacques and Chauveau). The leader rule replaces the incumbent rule when both concern the same party. Regional and provincial recalibration then changes the net effect in each riding. These values come from previous elections: 2008, 2014 and 2022 for incumbents, and five leader departures for the second rule. The votes are redistributed among the other parties, and regional and provincial totals stay those of the polls. Other candidates’ profile and by-elections since 2022 are not included. Candidate names are listed for information.
- Changes in the electorate. The model does not directly forecast the effects of demographics, turnout or late-campaign shifts.
- Data limits. Regional polls rest on smaller samples. The geographic matching and the transposition onto the new map involve some estimation.
- Small parties and independents. Their vote is counted with the other parties, but their wins are not modelled.
No predictive advantage over other projection models has been demonstrated.
Going further
Main calculation choices
Provincial weights use the square root of the total sample, capped at 1,500 respondents, and halve every five days. Age is measured from the exact fieldwork midpoint. A poll must have been published by the projection date. No reputation bonus is assigned to any firm.
The initial distribution combines, in equal parts, a proportional change and a change in points, with 5% smoothing toward the provincial average. It is then adjusted to match regional and provincial support at the same time.
For the regions, the region-to-province ratios from recent Léger polls are combined according to sample size and recency, with a five-day half-life. Only regional tables eligible within the 28-day window enter this adjustment. Historical local differences remain in the starting data.
The regional profiles are then blended 75% with the regional polls and 25% with the distribution from the model without regional polls. This weight was selected after comparisons on 2022 alone. The best weight depends on the measure used; 75% is not an optimum demonstrated across several elections.
Statistics Canada metropolitan boundaries and Élections Québec polling divisions are used to estimate each riding’s share in each region. Within a shared polling division, voters are allocated by area: this is an approximation.
The simulations use errors observed in previous elections. The sampling error of the regional tables is not estimated separately. Turnout weights are based on valid votes in 2022.
The historical transposition shows a gap of 44,154 votes, about 1.07% of the official 2022 total, in changed or new ridings. The provincial recalibration does not guarantee that each local transfer is exact.
Model documentation
The detailed method, code and checks are kept in a separate archive, available on request at info@127qc.com. The technical documentation is in French.
Independence and data
127QC is an independent project, with no affiliation to any political party, to Qc125 or to Élections Québec. Other sites’ projections are not used in the calculation. Qc125’s poll directory was used for the initial survey of polls; links to the original publications are preferred.
Data credits
Election results and boundaries: Élections Québec. Metropolitan areas: Statistics Canada. Provincial polls: Mainstreet Research; Research Co.; Noovo/Crave/Nanos; Segma Recherche/Radio-Canada/Les Coops de l’information; Léger/Le Journal/TVA; Pallas Data; Synopsis Recherche/La Presse. Regional polls used: Léger, September 21 and Léger, September 28 (in French). Historical transposition to the 2026 map: Kyle Hutton (Blunt Objects).