Rating Spanish pollsters: accuracy, the lead and openness
Abstract. We rate each Spanish pollster out of 100 on its final general-election polls since 2004, its miss on the gap between the two largest parties, and openness, ranked against Spain's own final polls since 13 October 2026. 40dB. grades A; SW Demoscopia, Cluster17, IMOP Insights, SocioMétrica and Celeste-Tel B; the CIS and most others C.
Summary
We rate each Spanish pollster out of 100 on the same scale as the British ratings (Rating the British pollsters: accuracy, the Labour–Conservative lead and openness): 50 points for the accuracy of its final general-election polls since 2004, 20 for its error on the gap between the two leading parties, and 30 for how much of its working it publishes. The record is pulled towards a deliberately poor pollster, so a pollster with one good election cannot score well by luck. The rating is shown for readers. It does not weight the polls: our Spanish average gives every pollster's latest poll the same weight.
Three things are particular to Spain. The CIS, a public research centre funded by the state, polls alongside private firms and is rated exactly like them. Spanish final polls have been about as accurate as German ones and more accurate than British ones, so the "poor pollster" the ratings pull towards sits between the two, and Spanish grades cannot be compared with other countries'. And the CIS is the one pollster that publishes both what respondents said and its own estimate for every study; its estimate has been the closer of the two at most elections, which is why, unlike in Britain, we do not start from raw figures.
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.
Pollster grades were rescaled on 13 October 2026. The first scale scored accuracy between fixed end points (a miss of 0.75 points or less earned full marks, 2.75 or more nothing) and pulled every short record towards a poor pollster, one that three in four final polls beat. Almost every pollster ended up with a D or an E, which told readers little. The rating now works like this:
- Accuracy (50 points) is ranked against the country's own checked final polls. If a share \(r\) of them missed by less than the pollster's average miss, it scores \(50\,(1-r)^{0.75}\): the typical (median) final poll earns about 30 and one in the best tenth about 46. Ranking within the country means a pollster is not marked down because its country is harder to poll.
- The lead (20 points) is scored the same way on the miss on the gap between the two leaders, \(20\,(1-r)^{0.75}\).
- A short record is pulled towards the typical (median) final poll, as if the pollster had two more elections there, not towards a poor one.
- Openness (30 points) is scored out of the items checked so far; an item not yet checked counts neither for nor against.
- Every pollster with a final poll on record gets a grade. One whose publications we have not reviewed is graded on its record alone (accuracy and lead scaled to 100) and marked "record only".
The grade boundaries are unchanged: A from 75, B 60–74, C 45–59, D 30–44, E below 30. Across the active pollsters of all ten countries the grades went from mostly D and E to A 8, B 68, C 76, D 31 and E 5. Where the sections below give the old end points, the poor-pollster prior or grades from before 13 October, they describe the first version of the scale; the live tables show today's.
In Spain it gives A to 40dB. (76); B to SW Demoscopia, Cluster17, IMOP Insights, SocioMétrica and Celeste-Tel; and C to the CIS, Sondaxe, GESOP, GAD3, Target Point, Sigma Dos, Ipsos and Hamalgama Métrica. These are the figures of 9 November 2026; the table on the pollsters page is current.
What the rating is for
The rating answers a narrow question: on the evidence we can check, how much should a reader trust this pollster's published figure, and how easily can they see how it was made? It does not say which pollster is right today, and it does not feed the forecast. The Spanish projection is the plain average of each pollster's latest poll within 30 days of the newest, equally weighted (Reading Spanish polls: the correction test and the CIS's raw figures). A pollster that publishes every week counts no more than one that publishes every month.
We do not weight by the rating for the same reason we do not correct Spanish polls: the evidence is eight elections, most pollsters have fewer, and their records move from one election to the next. A weighting that could not be shown to help out of sample would be a hidden assumption.
The evidence: final general-election polls, 2004–2023
The accuracy parts of the rating use every final poll in our record for the general elections of 2004, 2008, 2011, 2015, 2016, April 2019, November 2019 and 2023: 140 polls from 50 pollsters. A final poll is a pollster's last poll published before the five-day ban on publishing polls (Measuring Spanish polling error, 2004–2023).
For pollster \(j\) at election \(t\), write \(e_{j,t,p} = \hat{v}_{j,t,p} - r_{t,p}\) for its share for party \(p\) minus the result. The miss is the mean absolute error over the parties measured at that election that the poll reported,
and the lead error is the error on the gap between the two parties that finished first and second,
The pair follows the result: the PSOE and the PP in 2004, 2008 and both elections of 2019; the PP and the PSOE in 2011, 2016 and 2023; and in 2015 the PP and the left lineage, because Podemos and Izquierda Unida together finished second. The left of the PSOE is measured as one lineage throughout (Measuring Spanish polling error, 2004–2023).
Scoring accuracy and the lead
The scales are the British ones, unchanged. With \(\tilde{m}_j\) and \(\tilde{\ell}_j\) the rated miss and lead error defined below,
A rated miss of 0.75 points or better earns the full 50, 2.75 or worse earns nothing; a rated lead error of 1 point or better earns the full 20, 9 or worse nothing.
Shrinkage towards a poor pollster
Each pollster's record is pulled towards a prior as if it had two more elections at the prior's level (Efron and Morris, 1975):
where \(n_j\) is the number of the pollster's final polls and \(\bar{m}_j\), \(\bar{\ell}_j\) its averages of \(m_{j,t}\) and \(|\ell_{j,t}|\). The prior is a poor pollster: \(m^{*}\) is the 75th percentile of the miss across all Spanish final polls and \(\ell^{*}\) the 75th percentile of the absolute lead error. With the current record,
The British values are about 2.50 and 8.3, the German about 1.48 and 3.29. A poor Spanish final poll misses each party by less than a poor British one, and gets the lead much closer. On the shared scales the Spanish prior is worth about 20 accuracy points and 12 lead points, so a pollster with no record starts at about 32 before openness: a D with no openness points, a C with thirteen or more. In Britain the same pollster could not rise above a D, and in Germany it would start at a C. German, British and Spanish grades are therefore not comparable; within Spain the order of pollsters is what the rating is for.
A worked example
Celeste-Tel has a final poll at seven of the eight elections. Its average miss was 1.31 points per party, and its lead errors were −10.7 (2004), −2.1 (2011), +3.7 (2015), −1.2 (2016), −6.6 (April 2019), −1.5 (November 2019) and +6.7 (2023), an average absolute error of 4.6. Then
so \(A = 50 \times (2.75 - 1.45)/2 \approx 32\) and \(L = 20 \times (9 - 4.55)/8 \approx 11\). Its openness points (below) are 28, for a total of 72, a B. Most of its lead error comes from 2004, when every pollster had the PP ahead, and from April 2019, when it had the PSOE only 5.4 points ahead of the PP against a real gap of 12.
Openness
The 30 openness points reward publishing enough that a figure can be checked. Spanish pollsters publish most of their work through the newspapers and broadcasters that commission them, which print a "ficha técnica" (a method box) with each poll. The items are:
| Item | Points | Why it counts |
|---|---|---|
| Raw or unprojected figures ("voto directo" beside the estimate) | 8 | Shows what respondents said before the pollster's own adjustments |
| Full tables published | 8 | Sub-groups and weighting can be checked |
| Method note (ficha técnica) | 4 | How the sample was drawn and weighted |
| Sample size | 3 | The basics for judging a poll |
| Undecided share | 3 | Shows how firm the figures are |
| Fieldwork dates | 2 | When the poll was taken |
| Member of an industry body | 2 | Signs up to its standards |
Table: Spanish openness items.
Each item is Yes, Partial or No, scoring the full points, a third, or nothing:
The review was done in September 2026 from the pages we could read: each pollster's own site and the commissioning outlet's articles. What we found, in brief:
- The CIS publishes everything: for each study a method note, the questionnaire, the marginals, cross-tabulations and the microdata, with the direct vote beside the estimate. As a public body it belongs to no industry association, and that item scores nothing.
- Celeste-Tel publishes a monthly PDF on its own site with a full method note (telephone, half landline and half mobile), the estimate and the direct vote; it states its membership of Insights + Analytics España.
- 40dB. publishes, through El País, a report with the method note (online panel, sample, dates, margin of error) and the estimate series; its raw figures appear only in part.
- Cluster17 and Opina 360 publish their direct vote beside the estimate. Cluster17 has one final poll and Opina 360 none, so both ratings are provisional.
- GAD3, Sigma Dos, Ipsos and DYM have a full ficha técnica printed by the commissioning paper (ABC, El Mundo, La Vanguardia, 20minutos), but no tables and no raw figures. SocioMétrica's fichas in El Español also state that it weights by recalled vote, and part of its raw figures appear.
- NC Report, Invymark and Data10 are published with dates and usually a sample size, but rarely a method note.
An item we could not check scores the same as No. The Simple Lógica site refused our requests, and El Periódico's pages, where GESOP publishes, came back almost empty; their openness scores reflect what we could verify. Membership of an industry body counts only where the pollster or its client states it: the member list of Insights + Analytics España could not be read under its site's rules.
Grades and the current table
The score is \(S_j = \operatorname{round}(A_j + L_j + O_j)\), graded A (75 and above), B (60–74), C (45–59), D (30–44) and E (below 30). A rating is provisional when the pollster has fewer than two final polls. Pollsters whose publications we have not reviewed get record points but no grade.
| Pollster | Grade | Score | Accuracy | Lead | Openness | Elections |
|---|---|---|---|---|---|---|
| Obradoiro de Socioloxía | A | 77 | 41 | 13 | – | 2 |
| 40dB. | A | 76 | 38 | 14 | 24 | 3 |
| Simple Lógica | B | 72 | 28 | 13 | 30 | 7 |
| SW Demoscopia | B | 70 | 36 | 14 | – | 1 |
| Encuestamos | B | 69 | 34 | 14 | – | 2 |
| LAVANGUARDIA | B | 67 | 35 | 12 | – | 1 |
| Noxa Consulting | B | 66 | 35 | 11 | – | 3 |
| TABULAV | B | 65 | 32 | 13 | – | 1 |
| Cluster17 | B | 65 | 33 | 11 | 21 | 1 |
| APPEND | B | 64 | 32 | 12 | – | 1 |
| IMOP Insights | B | 64 | 38 | 12 | 14 | 3 |
| SyM Consulting | B | 63 | 32 | 11 | – | 2 |
| SocioMétrica | B | 63 | 37 | 12 | 15 | 3 |
| Opinòmetre | B | 63 | 32 | 12 | – | 1 |
| Celeste-Tel | B | 62 | 26 | 8 | 28 | 7 |
| OPINION2000 | B | 62 | 32 | 12 | – | 1 |
| GIPEYOP | B | 62 | 32 | 12 | – | 1 |
| Metroscopia | B | 61 | 34 | 14 | 14 | 5 |
| NETQUEST | B | 60 | 28 | 14 | – | 1 |
| GIINTERNACIONAL | B | 60 | 28 | 14 | – | 1 |
| YouGov | B | 60 | 30 | 12 | – | 1 |
| CIS | C | 57 | 19 | 8 | 30 | 9 |
| Sondaxe | C | 57 | 32 | 13 | 12 | 8 |
| GESOP | C | 57 | 27 | 12 | – | 6 |
| ADVICE | C | 57 | 26 | 14 | – | 1 |
| INTERCAMPO | C | 56 | 26 | 13 | – | 1 |
| TNS Demoscopia | C | 56 | 27 | 12 | – | 3 |
| Gallup España | C | 55 | 30 | 9 | – | 1 |
| GAD3 | C | 55 | 32 | 12 | 11 | 6 |
| JM&A | C | 54 | 26 | 12 | – | 2 |
| MyWord | C | 54 | 24 | 14 | – | 2 |
| DEIMOS | C | 53 | 23 | 14 | – | 1 |
| Demoscopia S.A. | C | 52 | 27 | 9 | – | 1 |
| AtlasIntel | C | 52 | 22 | 14 | – | 1 |
| Target Point | C | 51 | 32 | 8 | 11 | 1 |
| Top Position | C | 50 | 27 | 8 | – | 2 |
| IBES | C | 49 | 24 | 11 | – | 1 |
| LLORENTE | C | 49 | 25 | 9 | – | 2 |
| Opina | C | 48 | 26 | 8 | – | 3 |
| Sigma Dos | C | 48 | 28 | 10 | 11 | 9 |
| REDONDO | C | 48 | 24 | 10 | – | 2 |
| Ipsos España | C | 47 | 28 | 8 | 11 | 6 |
| Hamalgama Métrica | C | 47 | 27 | 9 | 11 | 5 |
| Data10 | C | 46 | 32 | 8 | 6 | 1 |
| DYM | C | 45 | 25 | 10 | 11 | 6 |
| Instituto Invymark | C | 45 | 26 | 10 | 9 | 5 |
| VOXPUBLICA | D | 44 | 23 | 8 | – | 2 |
| NC Report | D | 44 | 30 | 8 | 6 | 6 |
| Demoscopia y Servicios | D | 43 | 26 | 9 | 8 | 3 |
| A+M | D | 42 | 20 | 9 | – | 2 |
| INTERECONOMIA | D | 39 | 19 | 8 | – | 1 |
| METRASEIS | E | 25 | 11 | 7 | – | 2 |
| CITIGATE | E | 24 | 10 | 7 | – | 1 |
| More in Common España | – | – | 30 | 12 | 11 | 0 |
| Public First | – | – | 30 | 12 | – | 0 |
| EM-Analytics | – | – | 30 | 12 | – | 0 |
| Winston | – | – | 30 | 12 | – | 0 |
| Ateneo del Dato | – | – | 30 | 12 | – | 0 |
| ODEC | – | – | 30 | 12 | – | 0 |
At the time of writing no Spanish pollster reaches an A. The pollsters at the top combine a long record with a lot of openness (Celeste-Tel, the CIS) or a short, good record with a published report (40dB., SocioMétrica). The CIS's record is middling: its final polls have ranged from the closest in 2023 to the second furthest in November 2019, and its average miss over eight elections, 1.54 points per party, puts it in the lower half of the pollsters with five or more elections. Its openness is complete.
Each pollster's lean now
Beside the rating the site shows how each pollster differs from the others today: its latest poll within the 30-day window minus the average,
A lean is not an error. The pollster furthest from the rest may be the one closest to the result. Each pollster's page shows its lean now next to its lean at past elections, the mean of its final-poll errors. The CIS is the clearest case of both: its recent barómetros have put the PSOE ahead of the PP, while most private pollsters have the PP ahead. In the average it counts once, like every other pollster, so its lean moves the average by a fraction of its size.
The CIS's direct vote against its estimate
Every Spanish headline figure is an estimate. The CIS is the one pollster that also publishes, for every study, the "voto directo": the party each respondent names, as a share of everyone asked. Rescaled to those who name a party or say they will vote blank, it is the Spanish equivalent of Forschungsgruppe Wahlen's "Politische Stimmung" in Germany. For the CIS's last study before each election we compared both with the result:
| Election | Parties | Direct vote | Estimate | Direct vote closer |
|---|---|---|---|---|
| 2004 | 8 | 1.65 | 1.93 | 6/8 |
| 2008 | 8 | 1.75 | 1.01 | 3/8 |
| 2011 | 9 | 0.69 | 0.66 | 2/9 |
| 2015 | 11 | 1.12 | 1.15 | 5/11 |
| 2016 | 9 | 1.68 | 1.28 | 4/9 |
| November 2019 | 12 | 2.03 | 1.77 | 3/12 |
| 2023 | 11 | 0.81 | 0.65 | 3/11 |
| Average | 1.39 | 1.21 |
The estimate was closer at five of the seven elections, and on average. The largest differences favour the estimate: in 2008 the direct vote had the PSOE fifteen points ahead of the PP and the estimate had it 1.5 ahead, against a real lead of 4.5. The direct vote has understated the PP at five of the seven elections, by up to seven points in 2016; the CIS's estimate corrects much of that. The comparison covers seven of the eight elections, because the April 2019 study is not yet in it.
Why we don't start from raw figures, unlike in Britain
In Britain our projection starts from the raw stage of every pollster whose full tables we can read, because the raw answers' misses have been steadier than the headlines' and can be partly corrected out of sample (Why we start from British pollsters' raw answers rather than the published headline, Correcting the raw answers: the method behind our UK projection). In Spain:
There is one raw series. Only the CIS publishes its direct vote for every study. Celeste-Tel, Cluster17 and Opina 360 publish one in some releases, with short records. A method built on raw figures would rest on the CIS's own sample and its house effects.
The published figure is better. At five of seven elections the CIS's estimate was closer than its direct vote.
The steadiness is not there to use. The direct vote's misses have a lean (the PP too low at most elections), but correcting one pollster's raw figures by their average past lean would amount to rebuilding the CIS's estimate from seven observations, which the CIS already does with decades of data.
We show the CIS's direct vote on the polls page beside every estimate it has published since 2023.
What the rating leaves out
As in Britain, the rating leaves out how often a pollster polls, who commissions it, and which way it leans today. It also leaves out mode. Spanish pollsters use telephone interviews, online panels and mixtures of the two. Mode is a means, and the record already shows how well each pollster's means have worked. Regional elections, polled by many of the same firms, are not used: the record is general elections only.
Limitations
Eight elections, two of them shared misses. In 2004 and 2016 every pollster missed the same way. A pollster whose record includes those elections carries their error; one that started in 2019 was spared it.
The sources. Most final polls are grade C, transcribed from Wikipedia's tables. They agree with the checks we could make, but a transcription error would pass straight into a rating.
The scale. The end points were set for Britain. Against Spanish errors they are neither punitive nor generous, but Spanish grades still cannot be compared with other countries'.
Openness is what we could verify. Unknown counts as No; several low scores say as much about the pages we could read as about the pollster.
The first real test of whether higher-rated Spanish pollsters are more accurate will be the next general election.
References
Efron, B. and Morris, C. (1975). Data analysis using Stein's estimator and its generalizations. Journal of the American Statistical Association, 70(350), 311–319.
Jennings, W. and Wlezien, C. (2018). Election polling errors across time and space. Nature Human Behaviour, 2(4), 276–283.
How to cite
Polling Forecast (2026). Rating Spanish pollsters: accuracy, the lead and openness. Retrieved 21 November 2026 from https://www.pollingforecast.com/research/spain-pollsters. Live figures change as new polls arrive; the date above is when the text was last revised.



