The Spanish data: results, maps, statistics, deputies and polls
Abstract. Where every number on the Spanish pages comes from: the Junta Electoral Central's results in the BOE, the IGN's boundaries, the INE's statistics, the CIS's studies, pollsters' and newspapers' releases, and Wikipedia's tables where nothing else could be read. How each was changed and checked, what is left out and why, the daily poll check, the licences, and the known gaps, including the Congreso photographs we do not use.
Summary
This paper describes every data source behind the Spanish pages: where each comes from, how it was changed, how it was checked and under what terms it is used. Results by province come from the Junta Electoral Central's official summaries in the Boletín Oficial del Estado; boundaries from the Instituto Geográfico Nacional; statistics from the INE. Polls come from the CIS's own studies, from pollsters' own releases (grade A), from the newspapers and broadcasters that commission them (grade B), and, where neither could be read, from Wikipedia's poll tables (grade C), marked on the site. We use no poll aggregator's data and no other forecaster's.
Two things could not be done from where the data was collected. The Ministry of the Interior's download site, which holds official results for every election back to 1977, refused our connections, so four early national results rest on Wikipedia. And the Congreso's own site refused every request, so we could not read the terms on which its deputies' photographs may be reused, and we do not use them.
What has changed since this paper was written
Updated 9 November 2026. The rest of the paper is as first published, except where a sentence below says it was corrected.
- More frequent checks (13 October). After the morning run the poll readers check again every three hours until 11pm.
- Frente Amplio (16 October). Since the left alliance Frente Amplio (Broad Front) registered on 16 October, pollsters publish figures for it. A change in Wikipedia's poll table stopped Spanish polls loading from 11 October; the reader was fixed on 26 October and the missing polls added.
- Pollster grades rescaled (13 October) (Rating Spanish pollsters: accuracy, the lead and openness).
Principles
- Official data first. Every result, boundary and statistic comes from an official source where one could be reached.
- Every poll is cited. Each poll carries a link to where its figures were taken from and a source grade: A for the pollster's own release or an official study, B for the commissioning outlet's report, C for a secondary source. Each poll's page shows the link and what the grade means.
- No reuse of aggregators or other forecasters. Poll averages, reader panels and third-party re-estimates of the CIS's data are not used.
- Changes are stated. Parties are re-coded, figures aggregated, percentages computed and geometry simplified, and the site says so.
- Official summaries over newspaper tallies. Where a figure differs between an official summary and a secondary one, the official one is used and the difference recorded.
Election results, 2004–2023
After each general election the Junta Electoral Central publishes a summary of every province's official count in the BOE. We read these for the eight elections from 2004 to 2023:
| Election | BOE item | Read from |
|---|---|---|
| 23 July 2023 | BOE-A-2023-18907 | HTML tables; census and totals from the PDF |
| 10 November 2019 | BOE-A-2019-17344, corrected by BOE-A-2019-18147 | the correction's tables; totals from the original PDF |
| 28 April 2019 | BOE-A-2019-8070 | HTML tables |
| 26 June 2016 | BOE-A-2016-7133 | |
| 20 December 2015 | BOE-A-2016-867 | PDF (the HTML version omits several tables) |
| 20 November 2011 | BOE-A-2011-19391 | HTML tables |
| 9 March 2008 | BOE-A-2008-6851 | PDF, rotated pages |
| 14 March 2004 | BOE-A-2004-8358 |
Table: Results sources, grade A.
Each summary has three tables: the census, voters and valid, blank and null votes; the lists that won seats; and the lists that won none. Every list is kept, with its label as printed. Four parsers were needed, because the BOE printed each year differently. In 2004 one page of minor lists has no readable figures; those seven lists, 4,275 votes in nine provinces, are carried as a residual row.
The files passed 48 checks. In every province and year, list votes plus blank equal valid votes, and valid plus null equal voters. The sum of the 52 provinces equals the BOE's national total row, for the census and for every list. The strongest check is the seat calculation: D'Hondt with the 3% threshold, run on every province of every election from 2004 to 2023, reproduces the official seats exactly, and the site's own code does the same for 2023 (From votes to seats: D'Hondt in 52 provinces).
The census in these summaries includes Spaniards resident abroad (CERA), and the votes include theirs. Turnout on the site is therefore voters over the whole census, which is lower than the residents-only turnout reported on election night: 66.6% in 2023, nationally.
Earlier elections
For 1986, 1993, 1996 and 2000 the BOE summaries are scans without a text layer. We transcribed the national totals for every list that won seats and grouped the rest as others; all 30 transcribed figures match Wikipedia's national figures for the same lists. For 1977, 1979 and 1982 the BOE has no summary, and the 1989 summary predates the re-run elections of March 1990 in Murcia, Pontevedra and Melilla, so those four national results are taken from Wikipedia and graded C. The Ministry of the Interior's files, which would replace them and give provincial results before 2004, are listed for fetching from another network. These early results are used only in the long view of polling error (Measuring Spanish polling error, 2004–2023).
Parties and lineages
Every list label since 2004 is coded to a party family, with the rules written down and applied by a script. The families used in the forecast are the PP, the PSOE, Vox, Sumar, Podemos, Izquierda Unida, Ciudadanos, ERC, Junts, EH Bildu, the PNV, the BNG, Coalición Canaria and UPN. Some decisions matter:
- The PSOE includes the PSC in Catalonia, and the joint PSOE–Nueva Canarias lists of 2015 and 2016.
- The PP includes its regional joint lists (with the PAR in Aragón, Foro in Asturias). Where UPN and the PP ran together in Navarra (2004 to 2016, and as Navarra Suma in 2019), the list is coded UPN in the results files and the PP did not stand separately; pollsters mostly report the PP including it.
- The left lineage. Izquierda Unida and its regional partners; Podemos with its regional allies in 2015 (En Comú Podem, Compromís–Podemos, En Marea); Unidos Podemos and Unidas Podemos; Más País; and Sumar. For measuring polling error these are one lineage, coded Sumar (Measuring Spanish polling error, 2004–2023). Since December 2023 Sumar and Podemos are separate parties in the polls and in the forecast.
- Junts covers the Convergència line: CiU to 2011, Democràcia i Llibertat in 2015, CDC in 2016, Junts per Catalunya in 2019 and Junts in 2023.
- EH Bildu includes Amaiur in 2011.
Everything else is "others", including local lists such as Teruel Existe and Soria ¡Ya!. The forecast gives others no seats (What we don't know yet: limitations of the Spanish forecast).
Maps
The real map comes from the IGN's INSPIRE service for administrative units, one request per province, under CC BY 4.0. We simplified the 52 provinces so that neighbours keep identical borders, dropped slivers, and placed the Canary Islands in an inset at true scale and Ceuta and Melilla in insets at six times scale. All geometries are valid and no two provinces overlap.
The hex map gives every seat one hexagon, 350 in all, grouped by province, so that Madrid's 37 seats take up 37 times the space of Ceuta's one. Every province is in one piece, and so is every autonomous community except the Canaries, which are two island provinces by design. The layout is deterministic.
Province statistics
Each province page has a section built from the INE: population and age groups (1 January 2025), foreign nationals and the foreign-born, educational attainment of those aged 25 to 64, activity, employment and unemployment rates (the four quarters of 2025) and net income per person and per household. The statistics describe a province; they are not used in the forecast.
Deputies and photographs
The site lists the 408 people who have sat in the Congreso elected in 2023: the 350 elected and 58 who have since taken a seat. 328 of them were matched to a Wikidata item, through their Spanish Wikipedia article or an exact name match checked against Wikidata's record of their office.
Photographs come only from Wikimedia Commons, under open licences (CC0, CC BY, CC BY-SA, public domain, and the Commons attribution templates of La Moncloa, the Generalitat and the Parlament de Catalunya), with the author and licence read from Commons and shown with each photo. 228 of the 408 deputies have one, including 178 of those sitting now; the rest show an outline. Party leaders' photos come from the same source: 21 of the 22 people we show have one.
The Congreso's own photographs are not used. Its site refused every request we made, including for its robots.txt, so we could not read its terms for reusing images. Until those terms have been read, we treat the photos as all rights reserved, as we did for the French National Assembly's.
Polls
What is held
| Polls | A | B | C | Total |
|---|---|---|---|---|
| Final polls, general elections 2004–2023 | 9 | 21 | 120 | 150 |
| National polls since the 2023 election | 56 | 175 | 149 | 380 |
Final polls at the eight general elections from 2004 to 2023 are the evidence for polling error, the backtest, the replays and the pollster ratings (Measuring Spanish polling error, 2004–2023); ten of them report too few parties to be scored. National polls since the 2023 election make today's average. A further series of about 1,200 polls taken in the twelve months before each election from 2011 to 2023 feeds the election-day test (From today to polling day in Spain), and final polls back to 1977 feed the long view.
Where the figures come from
The CIS publishes every study on its own site: the method note, the questionnaire, the marginals, and a document with the estimate and the direct vote. We read the estimate tables from those documents (grade A) for every barómetro since September 2023 and for the pre-election study before every general election since 2004. Its catalogue pages sit behind an anti-bot challenge, so each study was opened at its own address. The 2004 estimate is a scan and was keyed by hand.
Other pollsters mostly publish through the paper or broadcaster that commissions them. The list of polls, and the figures for polls we could not check, come from Wikipedia's tables of Spanish opinion polls, graded C. For every poll since 2007 we then fetched the pages Wikipedia cites and searched them for each figure of 1% or more, written in any of the usual forms. Where the PP and PSOE figures and at least three shares were found, those figures took the grade of the page: A on the pollster's own site (Celeste-Tel, Opina 360, More in Common and others), B on the commissioning outlet's (El País for 40dB., El Mundo for Sigma Dos, ABC for GAD3, El Español for SocioMétrica and so on). 618 polls were checked this way. Paywalled and script-rendered pages usually could not be read, and those polls stay C.
What we leave out
Rows in Wikipedia's tables that are not polls, or not independent ones, are dropped: Electomanía's reader panel and its averages, Key Data's average for Público, third-party re-estimates built from the CIS's raw data (shown on Wikipedia as "CIS" with another firm's name), polls commissioned by parties with no pollster named, seat-only projections, alternative scenarios of the same poll, and exit polls. Polls Wikipedia marks as taken during the five-day ban and published abroad are excluded from every file.
Conventions worth knowing
- Shares are of the valid vote including blank votes, as most Spanish pollsters publish them. Others, for Wikipedia rows, is 100 minus the listed parties, so it includes blank votes and every party not listed. The CIS publishes blank separately. Celeste-Tel's own reports give shares of the vote for lists, blank excluded; we store them as published.
- The CIS's direct vote is stored beside its estimate as a share of everyone asked, including abstainers, the undecided and those who do not answer, so it adds up to well under 100.
- Dates. Wikipedia's tables give fieldwork dates. Where the first cited article appeared within 30 days of fieldwork, its date is stored as the publication date.
- Final polls are each pollster's last poll ending on or before the last legal day of publication (polling day minus six) and no more than 45 days out.
The daily check
Every morning the same background service that reads British pollsters' tables runs a reader for three Spanish pollsters: the CIS's barómetro (its study page and estimate document, grade A), Celeste-Tel's monthly report on its own site (grade A) and 40dB.'s monthly report for El País (grade B). A new poll must include the PP, the PSOE and Vox, add up to between 85 and 101.5, and have ended in the last 45 days; one that fails is logged and not stored. Polls from other pollsters are not read automatically; they come from the build of the Spanish data, from the sources above.
The publication ban
Spanish law forbids publishing polls in the five days before polling day (LOREG art. 69.7). The site has a switch for it, off by default: when it is on, from 00:00 in Madrid five days before polling day until the last polls close in the Canaries, the Spanish pages show the data as it stood when the ban began. Polls keep being collected.
Licences
| Data | Source | Terms | Credit |
|---|---|---|---|
| Results by province, seats | Junta Electoral Central, in the BOE | Official acts, excluded from copyright; BOE reuse conditions (cite, do not distort, give the date) | "Fuente: Junta Electoral Central, BOE" |
| National results 1977, 1979, 1982, 1989 | Wikipedia | CC BY-SA 4.0 (grade C) | per figure |
| Province boundaries | Instituto Geográfico Nacional | CC BY 4.0 | "© Instituto Geográfico Nacional" |
| Province statistics | INE | Reuse with attribution | "Fuente: INE" |
| CIS studies | Centro de Investigaciones Sociológicas | Reuse with attribution, without distortion or implied endorsement | "Fuente: CIS" |
| Poll figures from pollsters and outlets | each publisher | Published figures, cited | link on each poll |
| Poll figures from Wikipedia | Wikipedia contributors | CC BY-SA 4.0 | "via Wikipedia" on each poll |
| Photographs | photographers on Wikimedia Commons | each file's licence | per photo |
Table: Licences.
Known gaps
| Gap | Effect | What would fix it |
|---|---|---|
| Ministry of the Interior files unreachable | Four early national results from Wikipedia; no provincial results before 2004 | Fetching the listed files from another network |
| Most polls before 2015, and many since, grade C | Transcription errors would pass into the average and the record | Pollsters publishing their own series; more pages readable |
| The CIS's April 2019 study not in the raw-figure comparison | Seven elections compared instead of eight | Adding the study |
| Simple Lógica's site and El Periódico's pages unreadable | Their polls stay C and their openness is unscored | Access from another network |
| Congreso photographs' terms unread | 180 deputies without a photo | Reading the terms, or open photos on Commons |
| Local lists in "others" | A seat won by a local list would be missed | Modelling the largest local lists separately |
Table: Known gaps in the Spanish data.
How to cite
Polling Forecast (2026). The Spanish data: results, maps, statistics, deputies and polls. Retrieved 21 November 2026 from https://www.pollingforecast.com/research/spain-data. Live figures change as new polls arrive; the date above is when the text was last revised.



