Open data

The model outputs shown on the site can be downloaded and reused crediting “altnalysis.com” (CC BY 4.0 licence) and the original sources. The model and the rest of the site are not distributed: all rights reserved.

Downloads

• pronostico.jsonFull forecast: seats and votes by party and constituency with ranges, blocs, turnout and scenarios.
• senado.jsonSenate forecast by constituency and for the chamber (267 seats).
• simulador.jsonSimulator inputs: 2023 votes and shares by constituency used by the simulator and the polling-station page.
• nocional_2026.json2023 votes by constituency allocated with the 2026 seat counts, with each seat's cushion and rating.
• encuestas_2026.jsonPublished estimates since July 2023: pollster, client, fieldwork dates, sample, shares and seats.
• ultimo_escano.jsonLast-seat margin by constituency, 2008–2023.

For media: embed the forecast

Copy the code into your page. The widget resizes itself to its content and updates with every model run.

Terms of use

Free to use with visible attribution and a link to altnalysis.com (CC BY 4.0): do not remove or hide the widget's link.

For commercial use or election-night coverage, contact us at altnalysis@altnalysis.com.

Poll blackout (art. 69.7 LOREG): from 24 to 28 November the widgets will stop updating and will not show any new estimate.

What the model combines

altnalysis is not a poll aggregator: it is its own model, combining six kinds of quantitative and qualitative data.

• Voting behaviour. Results for every polling station and constituency at the 2008–2023 general elections (Ministry of the Interior and BOE): the 2023 vote is the starting point for the local split and the record measures how far polls missed.
• Voter sociology. CIS survey microdata (vote intention, recalled 2023 vote, likelihood to vote, ideology, closest party, religiosity, education, employment and income) plus CEO data for Catalonia. Weights are calibrated to INE population figures (sex, age, education and municipality size), and hidden and undecided voters are imputed with a model, never pro rata. Their published estimates are not used.
• Local sociology. Profile of about 36,000 census sections: age, sex, education, employment, municipality size, foreign-born, estimated religiosity and 2024 net household income (INE: 2021 Census and Household Income Atlas). It adjusts the provincial vote split (demographic swing).
• Changing electorate. Deaths, first-time young voters, naturalised citizens and Spaniards abroad (INE and Electoral Census Office), with priors from the academic literature and a separate model for the overseas vote.
• Published polls. Every estimate since July 2023, corrected for each pollster's house effect and the industry-wide error in 2015–2023.
• Qualitative analysis of the ballot. Coalitions and candidate lists in each constituency, agreements between parties (such as the PP–Vox Senate scenarios) and how polls missed at recent regional elections, such as Andalusia's on 17 May.

The vote-transfer matrices, overseas-vote model, house effects and simulations are altnalysis's own work and are not published; their outputs are published here.

Public sources

CIS and CEO microdata are not redistributed; only aggregates are published.