How we read Spanish polls
How we turn Spain's polls into a forecast for the Congreso de los Diputados (350 seats, 176 for a majority): the 48 polls of the last 30 days, corrected for how Spain's final polls have missed before. Today's average: PP 30.8, PSOE 27.6, Vox 19.0, Sumar 6.1.
Our rule everywhere: measure how far final polls missed at past elections, take off half the average miss, and keep the correction only if it would have made past forecasts better when each election is predicted without knowing its result. Then run the shares through the country's own electoral law.
Why we don't correct
Averaged over 2004–2023 the final polls missed by: PP +0.9, PSOE −1.3, Vox −0.4, Sumar +0.3 points (positive = too high). Predicting each election with the correction learned from the others was closer at only 2 of 8 elections (average miss per party 1.15 against 1.13 for the plain polls), so our projection takes the polls as they are. The test →
What decides the seats
- Small provinces. Half the provinces elect five deputies or fewer. There, D'Hondt rewards the largest parties and a point can move the last seat.
- The 3% line. It applies province by province, so it only bites in the big ones (Madrid, Barcelona, Valencia).
- Split left. Sumar and Podemos ran together in 2023 and are polled apart now; split, each wins fewer seats than together.
- If we're wrong. Today's polls replayed with each past election's miss, every province worked out again.
- 140 final polls at eight general elections (2004–2023), and older ones back to 1977 for the long view.
- Official results by province for every election since 2004, from the Junta Electoral Central's publication in the BOE.
- 282 poll averages at set distances before polling day, for the election-day test.
- The CIS's direct vote beside its estimate at each election, for the raw-figures test.
- Average: each pollster's latest poll within 30 days of the newest, equal weight per pollster. Shares are of valid votes including blank votes, as Spanish pollsters publish them.
- Correction: none: tested on past elections, the polls as published did at least as well without one. Sumar and Podemos are one lineage in the history and share any correction in proportion.
- Provinces: each party's 2023 share in every province, scaled by its national change (proportional swing). Sumar's 2023 vote is split between Sumar and Podemos in their national ratio; SALF, which didn't stand in 2023, gets its national share everywhere.
- Seats: in each province, D'Hondt among lists with at least 3% of the valid vote (LOREG arts. 162–163), with the 2023 number of seats per province (Real Decreto 400/2023). Ceuta and Melilla elect one each. Our calculation reproduces all 350 seats of 2023 from the official votes.
- Chances: 5,000 simulations; each party moves by its typical final-poll miss nationally and a little more in each province.
- Polling day: the same, with extra drift for the days left: rmse(days) = 1.35 + 0.56 × ln(days / 30).
How the local model did
Our seats come from spreading the national shares to the 52 provinces by proportional swing, then D'Hondt in each. To test that step on its own, we took each past election, moved every party in every province by its true national change since the election before, and gave out that election's real seats per province. No polls are involved: this is the error the provincial step adds even when the national vote is known exactly.
| Projected from → to | Provinces with every seat right | Seats given to the wrong party | Share error per party, points |
|---|---|---|---|
| 2016 → April 2019 | 25 / 52 | 33 / 350 | 2.7 |
| April 2019 → November 2019 | 36 / 52 | 21 / 350 | 1.1 |
| November 2019 → 2023 | 34 / 52 | 20 / 350 | 1.9 |
Share error: the average gap between projected and real vote share for each party with 1% or more in a province.
November 2019 → 2023: 20 of 350 seats with the wrong party
| Party | Projected | Actual | +/− |
|---|---|---|---|
| PP | 134 | 137 | −3 |
| PSOE | 125 | 121 | +4 |
| Vox | 31 | 33 | −2 |
| Sumar | 31 | 31 | = |
| ERC | 8 | 7 | +1 |
| Junts | 7 | 7 | = |
| EH Bildu | 6 | 6 | = |
| PNV | 5 | 5 | = |
| BNG | 1 | 1 | = |
| UPN | 1 | 1 | = |
| CC | 1 | 1 | = |
| The 18 provinces that came out differently | Seats | Projected minus actual |
|---|---|---|
| Asturias | 7 | PSOE +1, PP −1 |
| Badajoz | 5 | PSOE +1, Vox −1 |
| Barcelona | 32 | ERC +1, Sumar +1, PSOE −2 |
| Cádiz | 9 | Vox +1, PP −1 |
| Cantabria | 5 | PP +1, Vox −1 |
| Castellón/Castelló | 5 | Sumar +1, Vox −1 |
| Ceuta | 1 | PSOE +1, PP −1 |
| Córdoba | 6 | PSOE +1, Sumar −1 |
| Gipuzkoa | 6 | Sumar +1, PSOE −1 |
| Girona | 6 | PP +1, Sumar −1 |
| Granada | 7 | PSOE +1, Sumar −1 |
| Lleida | 4 | PP +1, PSOE −1 |
| Madrid | 37 | PSOE +2, Sumar −1, PP −1 |
| Málaga | 11 | PSOE +1, PP −1 |
| Navarra | 5 | Sumar +1, PP −1 |
| Santa Cruz de Tenerife | 7 | Sumar +1, PP −1 |
| Sevilla | 12 | Vox +1, Sumar −1 |
| Valladolid | 5 | PP +1, Vox −1 |
Largest share errors: Navarra 6.0, Soria 4.3, Santa Cruz de Tenerife 3.8, Las Palmas 3.6, Lleida 3.4
April 2019 → November 2019: 21 of 350 seats with the wrong party
| Party | Projected | Actual | +/− |
|---|---|---|---|
| PSOE | 122 | 120 | +2 |
| PP | 89 | 89 | = |
| Vox | 50 | 52 | −2 |
| Unidas Podemos | 32 | 35 | −3 |
| ERC | 15 | 13 | +2 |
| Cs | 11 | 10 | +1 |
| Junts | 8 | 8 | = |
| Others | 7 | 7 | = |
| PNV | 6 | 6 | = |
| EH Bildu | 5 | 5 | = |
| UPN | 2 | 2 | = |
| CC | 2 | 2 | = |
| BNG | 1 | 1 | = |
| The 16 provinces that came out differently | Seats | Projected minus actual |
|---|---|---|
| Alicante/Alacant | 12 | Others +1, Vox −1 |
| Araba/Álava | 4 | PP +1, Unidas Podemos −1 |
| Barcelona | 32 | ERC +1, PSOE +1, Others −2 |
| Burgos | 4 | Vox +1, PSOE −1 |
| Castellón/Castelló | 5 | Others +1, Unidas Podemos −1 |
| Ceuta | 1 | PSOE +1, Vox −1 |
| Cuenca | 3 | Vox +1, PSOE −1 |
| Girona | 6 | ERC +1, Unidas Podemos −1 |
| Illes Balears | 8 | PSOE +1, Unidas Podemos −1 |
| Las Palmas | 8 | Unidas Podemos +1, CC −1 |
| Madrid | 37 | Unidas Podemos +1, PSOE +1, Vox +1, PP −1, Others −2 |
| Murcia | 10 | Cs +1, PSOE −1 |
| Santa Cruz de Tenerife | 7 | CC +1, Vox −1 |
| Teruel | 3 | PSOE +1, Others −1 |
| Valencia/València | 15 | Others +3, Unidas Podemos −1, PP −1, Vox −1 |
| Zamora | 3 | PP +1, Vox −1 |
Largest share errors: Teruel 5.0, Valencia/València 3.4, Melilla 2.7, Castellón/Castelló 2.2, Ávila 1.6
2016 → April 2019: 33 of 350 seats with the wrong party
| Party | Projected | Actual | +/− |
|---|---|---|---|
| PSOE | 130 | 123 | +7 |
| PP | 64 | 66 | −2 |
| Cs | 53 | 57 | −4 |
| Unidas Podemos | 39 | 42 | −3 |
| Vox | 26 | 24 | +2 |
| ERC | 15 | 15 | = |
| Junts | 7 | 7 | = |
| PNV | 7 | 6 | +1 |
| EH Bildu | 5 | 4 | +1 |
| UPN | 2 | 2 | = |
| CC | 2 | 2 | = |
| Others | 0 | 2 | −2 |
| The 27 provinces that came out differently | Seats | Projected minus actual |
|---|---|---|
| A Coruña | 8 | Vox +1, PP −1 |
| Almería | 6 | PSOE +1, Vox −1 |
| Badajoz | 6 | PSOE +1, Vox −1 |
| Barcelona | 32 | Cs +1, Junts +1, PP +1, Vox −1, PSOE −2 |
| Bizkaia | 8 | PNV +1, PSOE −1 |
| Burgos | 4 | Vox +1, PSOE −1 |
| Cáceres | 4 | PSOE +1, Cs −1 |
| Cádiz | 9 | PSOE +1, Unidas Podemos −1 |
| Cantabria | 5 | Unidas Podemos +1, Others −1 |
| Ciudad Real | 5 | PSOE +1, Vox −1 |
| Córdoba | 6 | PSOE +1, Vox −1 |
| Girona | 6 | Unidas Podemos +1, Junts −1 |
| Guadalajara | 3 | Vox +1, Cs −1 |
| Huelva | 5 | PSOE +1, Unidas Podemos −1 |
| Illes Balears | 8 | Cs +1, PP +1, PSOE −1, Vox −1 |
| Las Palmas | 8 | Vox +1, Unidas Podemos −1 |
| Madrid | 37 | Vox +4, Cs −1, Unidas Podemos −1, PSOE −2 |
| Málaga | 11 | PSOE +1, Unidas Podemos −1 |
| Melilla | 1 | PSOE +1, PP −1 |
| Navarra | 5 | EH Bildu +1, PSOE −1 |
| Palencia | 3 | PSOE +1, Cs −1 |
| Salamanca | 4 | PSOE +1, PP −1 |
| Sevilla | 12 | PSOE +1, Unidas Podemos −1 |
| Teruel | 3 | PSOE +1, Cs −1 |
| Toledo | 6 | PSOE +1, PP −1 |
| Valencia/València | 15 | Unidas Podemos +1, Others −1 |
| Zamora | 3 | PSOE +1, Cs −1 |
Largest share errors: Melilla 10.1, Albacete 5.4, Ciudad Real 5.3, Ávila 5.2, Illes Balears 4.9
Where proportional swing needs care
- New parties. Vox had 0.2% in 2016 and stood in only 34 provinces; Más País (Madrid) and Teruel Existe first stood in November 2019, and the CUP returned then. A party with no earlier vote in a province gets nothing there, and a party new everywhere gets its national share in every province, which is how the Madrid, Barcelona and Teruel seats of November 2019 were missed. Scaling Compromís's April vote by Más País's national gain overshot Valencia.
- Local alliances that come and go. Navarra Suma (UPN, PP and Cs together in 2019) split in 2023, so the PP seat in Navarra was missed; Coalición Canaria and Nueva Canarias ran together in 2019 and apart in 2023. These provinces have the largest share errors.
- Uneven swings. When a party collapses (Ciudadanos in 2019) or surges (the PSOE in April 2019), it does not lose or gain the same fraction everywhere. Many misses are the last seat of a province going to the party next in line.
For this reason the simulations also move each party's share in every province at random (by about 8% of itself), so the 80% ranges allow for errors of this kind.
Correcting final polls
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 Spanish polls was closer at 2 of 8 and further out at 5 (average miss per party 1.15 against 1.13 for the polls as published). So our projection leaves these polls as they are.
| Election | Polls' average miss | Corrected | Verdict |
|---|---|---|---|
| 2004 | 1.55 | 1.50 | Correction closer |
| 2008 | 0.74 | 0.75 | Level |
| 2011 | 0.74 | 0.78 | Correction further out |
| 2015 | 0.87 | 0.90 | Correction further out |
| 2016 | 2.65 | 2.72 | Correction further out |
| April 2019 | 0.93 | 1.02 | Correction further out |
| November 2019 | 0.69 | 0.76 | Correction further out |
| 2023 | 0.85 | 0.77 | Correction closer |
Pulling back towards the last result
Not closer than the polls as they stand, so not used. Details →
- Proportional swing ignores local candidacies and pacts: if Sumar and Podemos run together again, or regional lists change, the provinces move differently.
- Seats per province are re-apportioned by population before each election; we use 2023's until the decree is published.
- Spanish law bans publishing polls in Spain in the five days before polling day (LOREG art. 69.7). We publish from the UK, so the forecast keeps updating in those five days, but fewer new polls appear then and most come out through outlets abroad.
Figures as of . Published by Polling Forecast.



