How we read French polls
How we turn France's polls into a forecast for the Assemblée nationale (577 seats, 289 for a majority): the 12 polls of the last 30 days, corrected for how France's final polls have missed before. Today's average: Le Pen 32.8, Mélenchon 18.3, Philippe 12.2, Glucksmann 11.6.
The rule is the one we use in Britain: measure how far the final polls have missed at past elections, take off half the average miss, and keep that correction only if it would have made past forecasts better when each election is predicted without knowing its result. In France the answer differs by contest.
Presidential round one: taken as they are
At the 5 presidential elections from 2000 on (the model uses none before), final round-one polls missed by 1.0 points per candidate family on average, but not in a steady direction. Correcting would have been further out at 4 of 5 elections, so we don't.
Round two: the RN is overstated
In 2002, 2017 and 2022 the RN's candidate reached round two, and the final polls put them too high each time. The correction is −1.8 points on the RN's share, with the opponent up by the same; it was closer at 4 of 5 past runoffs.
Legislative polls: corrected
At the Assembly elections of 2002, 2007, 2012, 2017, 2022, 2024, final round-one polls put the RN and its allies too high each time (+2.1, +1.0, +1.8, +4.3, +0.5, +2.9). Correcting was closer at 6 of 6 elections, so our Assembly projection uses it.
Polling day: polls as they stand, with wider uncertainty
Months out, British polls have tended to drift back towards the previous result. We tested the same idea on French presidential polls since 2002 and it did not help, so our polling-day forecast keeps today's polls and only adds the drift that polls this far out have shown. Election day →
Every figure behind the choices above, with its source, is in the track record. The main evidence:
- 40 final round-one polls at five presidential elections (2002–2022), from the pollsters' releases and Commission des sondages notices.
- 37 final round-two polls at the same five elections.
- 41 final legislative polls (2002, 2007, 2012, 2017, 2022, 2024) and the seat ranges pollsters published between the rounds.
- 198 poll averages at set distances before polling day, for the election-day test.
- Final polls of every presidential election since 1965, for the long view only: elections before 2000 are not used by the model.
The presidential forecast, step by step
- Pick the line-up. French polls test several line-ups at once. For each poll we take the scenario closest to our main line-up (Le Pen (RN), Mélenchon (LFI), Philippe (Ensemble), Glucksmann (PS)), and separately the one with Jordan Bardella for the RN.
- Average. Each pollster's latest poll within 45 days of the newest counts once.
- Round two. For each pair that could meet, the head-to-head polls are averaged the same way, then corrected as above.
- Chances. 20,000 simulated first rounds, each candidate moved by a random miss the size of their family's typical final-poll miss (at least 1.5 points). The top two meet; the winner comes from their corrected head-to-head, plus a random round-two miss of about 4.4 points. Pairs never polled count as a coin toss.
- Polling day. The same simulation with extra drift for the days left: rmse(days) = 3.01 + 1.16 × ln(days / 30).
The Assembly projection, step by step
- National shares by bloc from the latest legislative polls, corrected.
- Each circonscription starts from its 2024 round one and moves by the national change in each bloc.
- Qualifying for round two: the top two and anyone with 12.5% of registered voters; over half the votes and a quarter of registered voters wins outright.
- Withdrawals: where the RN and two or more opponents qualify, the weaker opponents stand down at the 2024 rate.
- Duels: a model of round-two shares fitted on the 2024 duels (it picks 370 of 403 2024 winners from the actual round one).
- Chances: 5000 simulated elections with national and local random misses.
- If we're wrong: today's polls replayed with each past election's miss, every seat worked out again.
correction = −½ × mean(final-poll average − result), if ≥ 2 elections and the leave-one-out test passes
How the local model did
Our seats come from moving every circonscription's last round one by each bloc's national change (uniform swing), deciding who qualifies and who withdraws, and settling each duel with a model of how voters transfer. To test those steps on their own, we took the 2022 round one in every circonscription, moved each bloc by its true national change to 2024, and ran the rest of the model unchanged. No polls are involved: this is the error the local steps add even when the national vote is known exactly. The boundaries were the same in 2022 and 2024. The same test on the earlier pairs is further down.
| 2024 seats | Projected from 2022 | Actual | +/− |
|---|---|---|---|
| Left (NFP and other left) | 249 | 192 | +57 |
| Ensemble and centre | 148 | 162 | −14 |
| RN and allies | 115 | 142 | −27 |
| LR and other right | 57 | 69 | −12 |
| Others and regionalists | 8 | 12 | −4 |
| Circonscriptions with the right winner | 423 / 577 |
| Round-one share error per bloc, points | 4.2 |
| Won outright in round one: projected / actual | 17 / 76 |
| Three or more qualified: projected / actual | 70 / 312 |
| Round-two seats with the right finalists | 305 / 501 |
| … of which the duel model called right | 267 / 305 |
Share error: the average gap between projected and real round-one share for each bloc with 1% or more in a circonscription (others left out). The duel model was fitted on the 2024 duels, so the round-two step is tested on the data it was fitted on; the round-one steps are not.
The 154 circonscriptions that came out differently (116 with the wrong finalists, 38 with the right finalists but the wrong winner)
| Circonscription | Projected | Actual | Where it went wrong |
|---|---|---|---|
| Ain 1re 001-01 | RN and allies | LR and other right | finalists |
| Ain 2e 001-02 | RN and allies | Ensemble and centre | finalists |
| Ain 5e 001-05 | LR and other right | RN and allies | finalists |
| Aisne 3e 002-03 | Left (NFP and other left) | Others and regionalists | finalists |
| Allier 2e 003-02 | Left (NFP and other left) | RN and allies | finalists |
| Allier 3e 003-03 | Left (NFP and other left) | LR and other right | finalists |
| Alpes-de-Haute-Provence 2e 004-02 | Left (NFP and other left) | RN and allies | duel |
| Alpes-Maritimes 1re 006-01 | LR and other right | RN and allies | finalists |
| Alpes-Maritimes 3e 006-03 | Left (NFP and other left) | RN and allies | duel |
| Alpes-Maritimes 5e 006-05 | Ensemble and centre | RN and allies | finalists |
| Ardèche 2e 007-02 | Ensemble and centre | RN and allies | finalists |
| Ardèche 3e 007-03 | Left (NFP and other left) | LR and other right | finalists |
| Aveyron 3e 012-03 | Left (NFP and other left) | Ensemble and centre | finalists |
| Bouches-du-Rhône 6e 013-06 | Ensemble and centre | RN and allies | finalists |
| Bouches-du-Rhône 8e 013-08 | Ensemble and centre | RN and allies | duel |
| Bouches-du-Rhône 11e 013-11 | RN and allies | Left (NFP and other left) | duel |
| Bouches-du-Rhône 14e 013-14 | Left (NFP and other left) | RN and allies | finalists |
| Calvados 1re 014-01 | Left (NFP and other left) | LR and other right | finalists |
| Calvados 3e 014-03 | RN and allies | Ensemble and centre | finalists |
| Calvados 5e 014-05 | Left (NFP and other left) | Ensemble and centre | finalists |
| Charente 2e 016-02 | Left (NFP and other left) | Ensemble and centre | finalists |
| Charente 3e 016-03 | Left (NFP and other left) | RN and allies | duel |
| Charente-Maritime 4e 017-04 | Ensemble and centre | RN and allies | duel |
| Charente-Maritime 5e 017-05 | RN and allies | Ensemble and centre | finalists |
| Corrèze 2e 019-02 | Left (NFP and other left) | LR and other right | finalists |
| Côte-d'Or 5e 021-05 | Ensemble and centre | RN and allies | duel |
| Côtes-d'Armor 1re 022-01 | Left (NFP and other left) | Ensemble and centre | finalists |
| Côtes-d'Armor 5e 022-05 | Left (NFP and other left) | Ensemble and centre | finalists |
| Creuse 1re 023-01 | Left (NFP and other left) | RN and allies | finalists |
| Dordogne 1re 024-01 | Left (NFP and other left) | RN and allies | duel |
| Dordogne 2e 024-02 | Left (NFP and other left) | RN and allies | duel |
| Dordogne 3e 024-03 | Left (NFP and other left) | RN and allies | duel |
| Doubs 1re 025-01 | Left (NFP and other left) | Ensemble and centre | finalists |
| Doubs 3e 025-03 | LR and other right | RN and allies | finalists |
| Drôme 4e 026-04 | Left (NFP and other left) | RN and allies | finalists |
| Eure 4e 027-04 | RN and allies | Left (NFP and other left) | duel |
| Finistère 1re 029-01 | Left (NFP and other left) | Ensemble and centre | finalists |
| Finistère 5e 029-05 | Left (NFP and other left) | Ensemble and centre | finalists |
| Finistère 7e 029-07 | Left (NFP and other left) | Ensemble and centre | finalists |
| Finistère 8e 029-08 | Left (NFP and other left) | Ensemble and centre | finalists |
| Haute-Corse 2e 02B-02 | Others and regionalists | LR and other right | duel |
| Gard 5e 030-05 | Left (NFP and other left) | RN and allies | duel |
| Haute-Garonne 3e 031-03 | Left (NFP and other left) | Ensemble and centre | finalists |
| Haute-Garonne 5e 031-05 | Left (NFP and other left) | Ensemble and centre | finalists |
| Gers 1re 032-01 | Left (NFP and other left) | Ensemble and centre | finalists |
| Gironde 5e 033-05 | RN and allies | Left (NFP and other left) | duel |
| Gironde 9e 033-09 | Left (NFP and other left) | Ensemble and centre | finalists |
| Gironde 10e 033-10 | RN and allies | Ensemble and centre | finalists |
| Hérault 4e 034-04 | Left (NFP and other left) | RN and allies | duel |
| Hérault 5e 034-05 | Left (NFP and other left) | RN and allies | duel |
| Hérault 6e 034-06 | LR and other right | RN and allies | finalists |
| Ille-et-Vilaine 1re 035-01 | Ensemble and centre | Left (NFP and other left) | finalists |
| Ille-et-Vilaine 2e 035-02 | Ensemble and centre | Left (NFP and other left) | finalists |
| Ille-et-Vilaine 7e 035-07 | Ensemble and centre | LR and other right | finalists |
| Indre 1re 036-01 | RN and allies | Ensemble and centre | finalists |
| Indre-et-Loire 3e 037-03 | Left (NFP and other left) | Ensemble and centre | finalists |
| Isère 1re 038-01 | Ensemble and centre | Left (NFP and other left) | finalists |
| Jura 3e 039-03 | Left (NFP and other left) | LR and other right | finalists |
| Loir-et-Cher 1re 041-01 | Left (NFP and other left) | Ensemble and centre | finalists |
| Loir-et-Cher 2e 041-02 | Ensemble and centre | RN and allies | finalists |
| Loire 3e 042-03 | Left (NFP and other left) | Ensemble and centre | finalists |
| Loire 4e 042-04 | Left (NFP and other left) | LR and other right | finalists |
| Loire-Atlantique 1re 044-01 | Ensemble and centre | Left (NFP and other left) | finalists |
| Loire-Atlantique 5e 044-05 | Ensemble and centre | Left (NFP and other left) | finalists |
| Loire-Atlantique 9e 044-09 | Ensemble and centre | LR and other right | finalists |
| Loire-Atlantique 10e 044-10 | Left (NFP and other left) | Ensemble and centre | finalists |
| Loiret 3e 045-03 | RN and allies | LR and other right | finalists |
| Loiret 5e 045-05 | LR and other right | Ensemble and centre | finalists |
| Lot-et-Garonne 1re 047-01 | RN and allies | Ensemble and centre | finalists |
| Lot-et-Garonne 3e 047-03 | RN and allies | LR and other right | finalists |
| Lozère 1re 048-01 | LR and other right | Left (NFP and other left) | finalists |
| Maine-et-Loire 1re 049-01 | Left (NFP and other left) | Ensemble and centre | finalists |
| Maine-et-Loire 3e 049-03 | Left (NFP and other left) | LR and other right | finalists |
| Marne 1re 051-01 | Left (NFP and other left) | Ensemble and centre | finalists |
| Marne 3e 051-03 | Ensemble and centre | RN and allies | duel |
| Marne 5e 051-05 | LR and other right | Ensemble and centre | finalists |
| Meurthe-et-Moselle 3e 054-03 | Left (NFP and other left) | RN and allies | duel |
| Meuse 1re 055-01 | LR and other right | RN and allies | duel |
| Moselle 1re 057-01 | RN and allies | Ensemble and centre | finalists |
| Moselle 2e 057-02 | Left (NFP and other left) | Ensemble and centre | finalists |
| Moselle 3e 057-03 | Left (NFP and other left) | LR and other right | finalists |
| Moselle 5e 057-05 | LR and other right | RN and allies | duel |
| Nièvre 1re 058-01 | RN and allies | Ensemble and centre | finalists |
| Nord 4e 059-04 | Left (NFP and other left) | Ensemble and centre | finalists |
| Nord 5e 059-05 | RN and allies | LR and other right | finalists |
| Nord 7e 059-07 | Left (NFP and other left) | Ensemble and centre | finalists |
| Nord 13e 059-13 | Ensemble and centre | Others and regionalists | finalists |
| Nord 14e 059-14 | Ensemble and centre | LR and other right | finalists |
| Nord 18e 059-18 | LR and other right | RN and allies | finalists |
| Oise 1re 060-01 | LR and other right | RN and allies | duel |
| Pas-de-Calais 2e 062-02 | RN and allies | Ensemble and centre | finalists |
| Pas-de-Calais 5e 062-05 | Left (NFP and other left) | RN and allies | duel |
| Pas-de-Calais 6e 062-06 | Ensemble and centre | RN and allies | finalists |
| Pas-de-Calais 7e 062-07 | LR and other right | RN and allies | duel |
| Pas-de-Calais 8e 062-08 | Left (NFP and other left) | RN and allies | duel |
| Puy-de-Dôme 4e 063-04 | Left (NFP and other left) | Ensemble and centre | finalists |
| Pyrénées-Atlantiques 1re 064-01 | Left (NFP and other left) | Ensemble and centre | finalists |
| Pyrénées-Atlantiques 4e 064-04 | Others and regionalists | Left (NFP and other left) | finalists |
| Pyrénées-Atlantiques 5e 064-05 | Ensemble and centre | Left (NFP and other left) | finalists |
| Pyrénées-Atlantiques 6e 064-06 | Ensemble and centre | Left (NFP and other left) | finalists |
| Rhône 8e 069-08 | Ensemble and centre | RN and allies | finalists |
| Rhône 12e 069-12 | Left (NFP and other left) | Ensemble and centre | finalists |
| Rhône 13e 069-13 | Ensemble and centre | RN and allies | finalists |
| Saône-et-Loire 3e 071-03 | Ensemble and centre | RN and allies | duel |
| Saône-et-Loire 4e 071-04 | Left (NFP and other left) | RN and allies | duel |
| Saône-et-Loire 5e 071-05 | Left (NFP and other left) | RN and allies | finalists |
| Sarthe 1re 072-01 | Left (NFP and other left) | Ensemble and centre | finalists |
| Sarthe 3e 072-03 | RN and allies | Ensemble and centre | finalists |
| Sarthe 5e 072-05 | Left (NFP and other left) | Ensemble and centre | finalists |
| Haute-Savoie 2e 074-02 | Left (NFP and other left) | Ensemble and centre | finalists |
| Haute-Savoie 3e 074-03 | Left (NFP and other left) | LR and other right | finalists |
| Paris 11e 075-11 | Ensemble and centre | Left (NFP and other left) | duel |
| Seine-Maritime 5e 076-05 | RN and allies | Left (NFP and other left) | duel |
| Seine-Maritime 6e 076-06 | Left (NFP and other left) | RN and allies | duel |
| Seine-Maritime 10e 076-10 | Ensemble and centre | RN and allies | duel |
| Seine-et-Marne 2e 077-02 | Left (NFP and other left) | Ensemble and centre | finalists |
| Seine-et-Marne 3e 077-03 | Left (NFP and other left) | LR and other right | finalists |
| Seine-et-Marne 5e 077-05 | RN and allies | Ensemble and centre | finalists |
| Yvelines 9e 078-09 | RN and allies | Left (NFP and other left) | duel |
| Yvelines 11e 078-11 | Left (NFP and other left) | LR and other right | finalists |
| Deux-Sèvres 2e 079-02 | Left (NFP and other left) | Others and regionalists | finalists |
| Tarn 1re 081-01 | Left (NFP and other left) | Ensemble and centre | finalists |
| Tarn 3e 081-03 | Left (NFP and other left) | Ensemble and centre | finalists |
| Tarn-et-Garonne 1re 082-01 | Left (NFP and other left) | RN and allies | duel |
| Vaucluse 1re 084-01 | RN and allies | Left (NFP and other left) | duel |
| Vienne 2e 086-02 | Left (NFP and other left) | Ensemble and centre | finalists |
| Vienne 3e 086-03 | Left (NFP and other left) | Ensemble and centre | finalists |
| Vosges 2e 088-02 | Ensemble and centre | RN and allies | duel |
| Vosges 4e 088-04 | Left (NFP and other left) | RN and allies | finalists |
| Territoire de Belfort 2e 090-02 | Left (NFP and other left) | RN and allies | duel |
| Essonne 8e 091-08 | LR and other right | Left (NFP and other left) | duel |
| Hauts-de-Seine 2e 092-02 | Ensemble and centre | LR and other right | finalists |
| Hauts-de-Seine 3e 092-03 | Ensemble and centre | LR and other right | finalists |
| Hauts-de-Seine 12e 092-12 | Ensemble and centre | LR and other right | finalists |
| Seine-Saint-Denis 5e 093-05 | LR and other right | Left (NFP and other left) | duel |
| Val-de-Marne 1re 094-01 | Ensemble and centre | LR and other right | finalists |
| Val-de-Marne 4e 094-04 | Left (NFP and other left) | Ensemble and centre | finalists |
| Val-de-Marne 7e 094-07 | Left (NFP and other left) | LR and other right | finalists |
| Val-d'Oise 1re 095-01 | Left (NFP and other left) | RN and allies | finalists |
| Val-d'Oise 4e 095-04 | Left (NFP and other left) | Ensemble and centre | finalists |
| Français établis hors de France 9e 099-09 | Ensemble and centre | Left (NFP and other left) | duel |
| Martinique 1re 972-01 | Left (NFP and other left) | Others and regionalists | finalists |
| Martinique 2e 972-02 | Left (NFP and other left) | Others and regionalists | finalists |
| Martinique 4e 972-04 | Left (NFP and other left) | Others and regionalists | finalists |
| Guyane 1re 973-01 | Left (NFP and other left) | Others and regionalists | finalists |
| Guyane 2e 973-02 | Ensemble and centre | Others and regionalists | finalists |
| La Réunion 3e 974-03 | LR and other right | RN and allies | finalists |
| La Réunion 4e 974-04 | LR and other right | Left (NFP and other left) | finalists |
| Mayotte 1re 976-01 | Left (NFP and other left) | LR and other right | finalists |
| Mayotte 2e 976-02 | Ensemble and centre | RN and allies | finalists |
| Polynésie française 1re 987-01 | Ensemble and centre | LR and other right | finalists |
| Polynésie française 2e 987-02 | Others and regionalists | LR and other right | finalists |
| Polynésie française 3e 987-03 | Others and regionalists | Left (NFP and other left) | finalists |
| Nouvelle-Calédonie 1re 988-01 | Ensemble and centre | LR and other right | finalists |
Largest round-one errors: Polynésie française 1re 25.8, Polynésie française 3e 22.7, Marne 5e 22.0, Mayotte 1re 21.6, Polynésie française 2e 20.6
Where uniform swing needs care
- Incumbents. Ensemble lost 5 points nationally, but only about 1.4 points in the seats its deputies held and about 8 elsewhere. Uniform swing takes 5 everywhere, so it pushed Ensemble into third place in seats it held on to; 36 of the misses are seats we gave to the left that Ensemble kept.
- The RN's surge was not uniform either: from 2022 to 2024 it gained about 10 points where it was weakest and about 17 where it was strongest. Uniform swing gives it 14.5 everywhere, so it misses RN seats won outright or in close duels in its heartlands.
- Turnout. Valid votes rose from 46% of registered voters in 2022 to 65% in 2024, which lowers the 12.5%-of-registered bar in votes and makes three-way run-offs far more common. Run on 2022 turnout, the model finds 70 circonscriptions with three or more qualifiers against 312 in reality. With each circonscription's real 2024 turnout it finds 349, but the seats barely change (153 wrong instead of 154), because the weaker anti-RN candidates withdraw either way.
- Changing alliances. In 2024 part of LR stood with the RN (the UDR, counted with the RN), and many circonscriptions had no Ensemble or LR candidate. A 2022 base cannot know who stands where.
Most of the error is in round one: the wrong pair reaches round two. This is why the simulations also move every bloc in every circonscription at random (by about 2.5 points for the larger blocs) and withdrawals happen only at the 2024 rate.
Every pair of elections since 2002
The same replay on each pair of legislative elections fought on the same map: 2002 → 2007 on the 1986 circonscriptions, then 2012 → 2017, 2017 → 2022 and 2022 → 2024 on today's. Before 2024 the duel model is tested out of sample (it was fitted on 2024). Uniform swing, which the model uses, named 1476 of 2308 winners right; proportional swing 1428, and it was never better.
| Pair | Right winner, uniform | … proportional | Right finalists | Round-one error, uniform | … proportional |
|---|---|---|---|---|---|
| 2022 → 2024 | 423 / 577 | 421 | 305 / 501 | 4.2 | 4.0 |
| 2017 → 2022 | 338 / 577 | 335 | 343 / 572 | 3.9 | 3.7 |
| 2012 → 2017 | 344 / 577 | 311 | 240 / 573 | 4.8 | 8.8 |
| 2002 → 2007 | 371 / 577 | 361 | 239 / 467 | 4.4 | 5.1 |
Round-one error: average gap in points per bloc with 1% or more. Results: Ministry of the Interior files on data.gouv.fr. In 2002–2012 the UDF and MoDem count with Ensemble and the centre, the Nouveau Centre with the right; the blocs are otherwise the ministry's nuances grouped as for 2024.
Round one
Each election is predicted with the correction worked out from the other elections only, so it never sees the result it is tested on. Correcting round-one polls was closer at 1 of 5 and further out at 4 (average miss per party 1.06 against 0.98 for the polls as published). So our projection leaves these polls as they are.
Elections before 2000 (1965, 1969, 1974, 1981, 1988, 1995) are not in this test: they are shown in the track record but not used by the model.
| Election | Polls' average miss | Corrected | Verdict |
|---|---|---|---|
| 2002 | 1.02 | 1.07 | Correction further out |
| 2007 | 1.10 | 1.15 | Correction further out |
| 2012 | 0.91 | 1.21 | Correction further out |
| 2017 | 0.60 | 0.57 | Correction closer |
| 2022 | 1.26 | 1.31 | Correction further out |
Round two
Each election is predicted with the correction worked out from the other elections only, so it never sees the result it is tested on. Correcting round-two polls was closer at 4 of 5 and further out at 1 (average miss per party 1.86 against 2.57 for the polls as published). So our projection uses the correction.
Elections before 2000 (1965 round two, 1969 round two, 1974 round two, 1981 round two, 1988 round two, 1995 round two) are not in this test: they are shown in the track record but not used by the model.
| Election | Polls' average miss | Corrected | Verdict |
|---|---|---|---|
| 2002 round two | 3.21 | 1.62 | Correction closer |
| 2007 round two | 0.73 | 1.22 | Correction further out |
| 2012 round two | 1.11 | 0.78 | Correction closer |
| 2017 round two | 4.48 | 3.45 | Correction closer |
| 2022 round two | 3.32 | 2.24 | Correction closer |
Legislative
Each election is predicted with the correction worked out from the other elections only, so it never sees the result it is tested on. Correcting legislative polls was closer at 6 of 6 and further out at 0 (average miss per party 0.86 against 1.03 for the polls as published). So our projection uses the correction.
| Election | Polls' average miss | Corrected | Verdict |
|---|---|---|---|
| 2002 | 1.18 | 0.99 | Correction closer |
| 2007 | 0.68 | 0.40 | Correction closer |
| 2012 | 0.70 | 0.62 | Correction closer |
| 2017 | 1.57 | 1.35 | Correction closer |
| 2022 | 1.14 | 1.04 | Correction closer |
| 2024 | 0.93 | 0.73 | Correction closer |
The duel model
Fitted on 2024 and given 2024's real round one, it picks 370 of 403 winners. Fitted on 2022 and applied to 2024, it gets about 64% right, because voters switched differently between the two elections; this is the step we watch most closely.
- The presidential model learns from five elections, and the round-two correction from three RN runoffs.
- Candidacies are not settled; we follow the line-ups pollsters test.
- Legislative polls are old: none since autumn 2025.
- Uniform swing ignores local candidates and incumbency, which matter a lot in French legislative elections.
- Withdrawal and transfer behaviour in round two can change sharply from one election to the next.
- Raw figures: only OpinionWay and Cluster17 print them, and they pull in opposite directions, so we don't build on them (in Britain our projection starts from raw figures). See them.
Figures as of . Published by Polling Forecast.




