Forecast - Spanish general election 2026

The three outcomes of the election

Each of the 20,000 simulations falls into one outcome only, according to the coalition selected by the counting rule described in the Who governs block. The three outcomes cover every simulation.

These shares are not read like the chances of the formulas above: a formula can reach the majority without being the coalition the rule selects.

Key takeaways

Facts drawn from the 20,000 simulations of the forecast of 1 November 2026.

1. PP and Vox together win at least 176 seats in 85% of simulations. Since 17 October 2026, the first run of the current configuration, this share has moved from 86% to 85%; earlier runs followed another configuration and are not directly comparable.
2. The PP wins the most seats in 73% of simulations, the PSOE in 26%.
3. Vox would win 66 seats at the median, 45 to 90 in 90 simulations out of 100, against 33 in 2023. Even the lower bound is above its 2023 result.
4. The bloc that elected Pedro Sánchez in 2023 wins at least 176 seats in 12% of simulations. At the median, it holds 150 seats, 116 to 186 in 90 simulations out of 100.
5. The forecast aggregates 42 polls from 18 pollsters, all checked at their primary source. The most recent fieldwork ended on 25 October 2026.

The parties

Configuration declared on 16 October, detailed in the box Declared lists. Vote share and seats: median across simulations, then the intervals containing 90 and 50 simulations out of 100.

The last two columns are chances in 100. “Largest party”: share of simulations in which the party wins the most seats. Median seats do not add up to a parliament: their sum is not 350.

The two blocs

Share of simulations for each seat total. Beyond the red line, a bloc reaches the absolute majority of 176 seats on its own. PP and Vox bloc: PP and Vox. 2023 investiture bloc: PSOE, Frente Amplio, Podemos, ERC, Junts, EH Bildu, PNV, BNG and Coalición Canaria.

At the median, the PP and Vox bloc wins 198 seats and the 2023 investiture bloc 150. The two curves do not add up: each bar counts simulations. Outside the frame: less than 0.1% of the PP and Vox bloc simulations, less than 0.1% of the 2023 investiture bloc simulations.

Build a coalition

Tick parties: the page counts the share of the 20,000 simulations in which their seats add up to 176. This is arithmetic: it does not say whether these parties would agree to govern together.

Clear

Same simulations as the rest of the page (forecast of 1 November 2026). Example: PP and Vox together win at least 176 seats in 85% of simulations. Declared refusals to govern together are listed in the register; the page flags them when the ticked parties are concerned.

What pollsters project

Several pollsters publish their own seat projection. When they give a range, the table keeps the central value published by the pollster, or failing that the midpoint, rounded up when it falls on a half seat. These projections are not an input to the model: it reads only voting intentions.

For the PP, 11 of 11 projections fall within our 90% interval. For the PSOE, 11 of 11 projections fall within our 90% interval. Our interval is wide: falling within it is a weak test.

Configuration of the lists

Configuration declared on 16 October. All figures on this page refer to this configuration.

Who governs

An arithmetic majority is not enough: its members must agree to govern together. This reading applies the declared vetoes in the register to each of the 20,000 simulations.

Most frequent selected coalitions

These 5 coalitions cover 86% of simulations. The rest is split between other coalitions and simulations with no workable majority. In each simulation, several workable majorities can coexist. To select one, the rule used in our German state models applies in order: workable majority, minimal coalition, presence of the party that came first, fewest parties, smallest surplus, most seats, then alphabetical order. This is a counting convention, not a forecast of negotiations. Compromís is not simulated separately: its voting intentions are counted in Frente Amplio.

Three measures per party

Indispensable: the party is in every workable majority of the simulation. Selected: it is in the coalition designated by the rule. At least one: it is in at least one workable majority. Indispensable and At least one do not depend on the rule. A large gap between them means the rule drives the figure.

If the first vote fails: second-vote scenarios

Spanish Constitution, art. 99.3: in the second vote, 48 hours after the first, a simple majority is enough. There must be more yes votes than no votes. The scenarios set who votes yes, no or abstains. The figure gives the number of simulations in which yes wins if that scenario occurs. It is conditional on the scenario and does not measure the chance that parties vote that way.

All values in this table are conditional on the scenario.

The veto register

25 entries, 24 of them active. 8 carry a quote. A veto is a recorded, dated and sourced statement: it is not a decision of the model, and it expires.

Active vetoes of fragility 0 to 2 all enter the calculation. UPN inherits the vetoes of the PP, both those it makes and those aimed at it. SALF inherits the vetoes of Vox, both those it makes and those aimed at it. Compromís is not simulated separately: its voting intentions are counted in Frente Amplio.

Quoted statements

Stated positions without a quote in the register

Probabilities over time

One value per run date: the last model output of the day, in the central configuration of that date. Source: the history of model outputs, each entry checked by its fingerprint.

Who will govern

• PP and Vox
• Broad right
• 2023 investiture bloc

Chances in 100 that each combination wins at least 176 seats, as a workable majority. The vertical dotted line marks the change of configuration on 17 October 2026. The two periods are not directly comparable. Last run: 1 November 2026.

Median seats

• PP
• PSOE
• Vox

Median seats on each date. The shaded band covers the 90% interval: 9 chances in 10 that the result falls inside it. The horizontal dotted line marks the absolute majority, 176 seats.

On 1 November 2026: PP 131 seats (99 to 165) · PSOE 106 seats (72 to 146) · Vox 66 seats (45 to 90).

How the forecast has changed

Central configuration of each date, one row per run date. When the model ran several times on the same day, only the last output is kept. Source: the history of model outputs, each entry checked against its fingerprint.

The campaign in polls

Vote shares published by each poll since 18 August, in points, for PP, PSOE, Vox and Frente Amplio. This is the model's window: the 75 days before the last run. The line is a moving average. These are the figures published by the polling firms. Only Compromís, when polled separately, is added to Frente Amplio, as in the model.

42 polls from 18 polling firms, from 25 August 2026 to 25 October 2026.

• PP
• PSOE
• Vox
• Frente Amplio
• One poll
• Moving average

Each dot is a poll, placed on the last day of its fieldwork, or on its publication date when the fieldwork is not dated. The line is the simple average of the polls of the previous 14 days, without weighting. It starts once at least 3 polls are available. It is not the model's aggregate, which weights the polls. Frente Amplio is the Sumar column published by the polling firms. When a polling firm gives Compromís separately (11 polls here), that figure is added to Frente Amplio, as in the model. The table gives both published figures. 3 polls publish no figure for Frente Amplio: they have no dot on its line.

Province by province

Each constituency takes the colour of the list most often first in votes across simulations. The stronger the colour, the more likely that first place. Hovering over a province shows its detail, repeated in the table below the map.

The Canary Islands are shown in an inset. Coming first in votes in a province does not say how many seats the list wins there. Base map: INE, seccionado 2026 (provinces built by merging census sections), licence CC BY 4.0.

The 2023 election

General election of 23 July 2023, Congreso. Votes for each list and share of valid votes, blank votes included. Final results published in the BOE (BOE-A-2023-18907). One row per list that won seats, plus the other lists combined. Last column: the 2026 list or party that stems from it.

In 2023, the lists PP and VOX won 170 seats together, 6 short of the absolute majority of 176. The lists of the 2023 investiture bloc (PSOE, SUMAR, PSC, ERC, JUNTS, EH Bildu, EAJ-PNV, BNG and CCa) held 179. Valid votes include 200,673 blank votes (0.81%). The 2026 joint lists are detailed in the Declared lists box.

The electoral rules

52 constituencies, 350 seats, absolute majority at 176. Vote on Sunday 29 November 2026.

The method in five lines

1. Polls from the last 75 days. Each poll is weighted by its age, with a half-life of 30 days, its sample size and the past accuracy of its pollster. Polls from the same pollster add up. Each pollster's house effect is removed.
2. Two adjustments were tested in cross-validation: bias by bloc and weight of the last election. Neither reduces the error, so neither is applied. Pollster weights are applied as a rule, although in cross-validation they do not beat the raw poll average.
3. Uncertainty is measured on past elections, widened with the time remaining before the vote and increased for new lists.
4. The election is simulated 20,000 times, province by province. Seats are allocated by the D'Hondt method with a threshold of 3% per constituency.
5. Coalitions are read in each simulation from the declared vetoes, with no probability assigned to negotiations.

Seed 20261129. Model v0.2. Each parameter and its status are detailed in the method note generated by the engine.

Why trust us

The model was replayed on the elections of April 2019, November 2019 and 2023, each time without the tested election for uncertainty, biases, pollsters and geography. The final widening of the uncertainty was chosen knowing these elections. 38 party-by-election observations.

Strict record, with a widening chosen each time without the tested election: vote shares within the 90% interval for 35 results out of 38 and within the 50% interval for 19. Seats within the 90% interval for 37 and within the 50% interval for 24.

The four figures below use the retained widening, chosen knowing these elections: they are more favourable than the strict record.

Mean absolute error of vote shares. April 2019: 0.79 points against 0.83 for the raw poll average. November 2019: 0.56 points against 0.58 for the raw poll average. 2023: 0.53 points against 0.66 for the raw poll average.

The test far from the vote

For each past election, we take each pollster's latest poll available at a given date before the vote, then check whether the result falls within the vote-share interval given by our uncertainty at that horizon. This test covers vote shares, not seats, and does not run through the whole model chain.

Elections covered: 2016, April 2019, November 2019 and 2023. This test is in sample: the election uncertainty and its widening were measured on these same elections. Its main purpose is to check that uncertainty grows enough far from the vote.

The limits

• Only three elections serve as tests: one miss weighs a third of the record.
• Cross-validation covers polls from the last fourteen days. Far from the vote, drift is only tested in sample. The distribution used for new lists and its increase are not tested.
• The Brier score rests on 6 events. That is too few for a ranking.
• The polls used in this run are all verified at the primary source. Historical polls are transcribed from secondary sources.
• No Spanish forecast by Le Millénaire has yet been published and then checked against the result.

How it works