Rating Italian pollsters: accuracy, the lead and openness
Abstract. Each Italian pollster is rated out of 100 on accuracy at its final Camera polls, its error on the gap between the two largest lists, and openness. Since the rescale of 13 October 2026, accuracy is ranked against Italy's own final polls, so Italy's larger misses no longer push every pollster to E: EMG Different and Cluster17 grade B, most others C or D.
Summary
We rate each Italian 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 Camera polls since 2008, 20 for its error on the gap between the two leading lists, and 30 for how much of its working it publishes. The record is pulled towards a deliberately poor pollster, so that a pollster with one good election cannot score well by luck. The rating is shown for readers; it does not weight the polls.
Every Italian pollster we have reviewed is graded E. That is less a verdict on any one pollster than a consequence of two facts. Italian final polls have missed by more than those of any other country on the site, partly because the fifteen-day publication ban makes them early, so the "poor pollster" the ratings pull towards misses by more than the British scale gives any points for. And the official register where Italian pollsters must file the full document of every published poll could not be reached, so we could credit only the openness we could see elsewhere. The order of pollsters within Italy still carries some information; the grades carry little.
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 Italy, where every pollster we reviewed was graded E on the first scale, it gives B to EMG Different (67) and Cluster17 (66); C to SWG, Tecnè, Ipsos, Demopolis, Quorum / YouTrend and Noto; and D to Winpoll, Ixè, Demos & Pi, BiDiMedia, Lab21.01 and Termometro Politico. 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 Italian average gives every pollster's latest poll within 21 days of the newest the same weight (Reading Italian polls: the correction and its test). A pollster that publishes weekly, as several do, counts no more than one that publishes monthly.
The evidence: final Camera polls, 2008–2022
The accuracy parts of the rating use the 82 final polls in our record for the Camera elections of 2008, 2013, 2018 and 2022, each taken at least fifteen days before the vote (Measuring Italian polling error, 2008–2022). Earlier final polls give coalition totals only and cannot be scored. No pollster has more than four final polls.
For pollster \(j\) at election \(t\), the miss \(m_{j,t}\) is its mean absolute error over the lists measured that election, and the lead error \(\ell_{j,t}\) its error on the gap between the two lists that finished first and second: the PdL and the PD in 2008, the M5S and the PD in 2013 and 2018, FdI and the PD in 2022.
Scoring, and a poor Italian pollster
The scales are the British ones:
where the rated miss and lead error are pulled towards a poor pollster as if it had two more elections at the poor pollster's level (Efron and Morris, 1975):
The poor pollster is the 75th percentile of all Italian final polls:
Both are beyond the end of the scales: a miss of 2.75 points or worse earns no accuracy points, and a lead error of 9 or worse no lead points. A poor Italian final poll earns nothing on either. The consequences are stark:
- a pollster with no record starts at zero accuracy and zero lead points, so its grade is at most its openness score, which cannot exceed 30: an E;
- a pollster with four final polls at the Italian average miss, about 2.3 points, has a rated miss of about 2.6 and earns about four accuracy points;
- even a pollster with four perfect final polls would earn only about 42 accuracy points and 12 lead points.
In Germany the same rule makes grades read high; in Italy it makes them read low. We keep the British scale in every country so that the same record earns the same points, and we say plainly that Italian grades cannot be compared with other countries'. An Italian scale with its own end points is an open item (What we don't know yet: limitations of the Italian forecast).
A worked example
SWG has final polls at three elections, with an average miss of 2.39 points per list. Its rated miss is
which earns \(50 \times (2.75 - 2.69)/2 \approx 1\) accuracy point. Its rated lead error is above 9, which earns none. Its openness (below) is about 17 points, the highest we could verify for an Italian pollster, for a total of 18: an E.
Openness
Italian law already requires more disclosure than most. Under the AGCOM rules (delibera 256/10/CSP), every poll published in the media must carry a nota informativa (who ran it and for whom, the sample and its size, the number of non-respondents, the fieldwork dates and the address of the full document), and the full document, with the method, the weighting, the margin of error and the text of every question, must be filed within 48 hours on the government's register, sondaggipoliticoelettorali.it. On paper, every Italian pollster should score well on the method, sample and date items.
We could not verify it. The register reset every connection from the network our data was collected from, and several pollsters' own sites did the same or refused automated requests. An item we could not check scores as No, as in every country. The items and points are:
| Item | Points | Why it counts |
|---|---|---|
| Raw or unweighted figures | 8 | Shows what respondents said before the pollster's own adjustments |
| Full tables published | 8 | Sub-groups and weighting can be checked |
| Method note | 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 ASSIRM, the industry body | 2 | Signs up to its standards |
Table: Italian openness items.
What we found, in brief:
- SWG's weekly results, shown on TG La7 and in SWG's own Radar report, carry the full AGCOM note (client, mode, sample, non-respondents, dates, margin of error) and the share of respondents who give no party. SWG is a member of ASSIRM.
- Tecnè's weekly report for Agenzia Dire has a method note with population, mode, dates and margin of error. Tecnè is a member of ASSIRM.
- Demopolis and Quorum/YouTrend publish a note with sample and dates on their own pages.
- BiDiMedia states the undecided share in its releases.
- Ipsos's figures appear in the Corriere della Sera with prose on those who do not express a vote.
- EMG, Eumetra and Izi are ASSIRM members, but their releases could not be read.
- Termometro Politico, Ixè, Demos, Euromedia, Index, Noto, Piepoli, Lab21.01, Only Numbers and Winpoll could not be checked beyond their membership of ASSIRM, which none of them holds.
No Italian pollster publishes raw figures: what is published is the share of those expressing a vote, after weighting. That item scores nothing for everyone.
Grades and the current table
| Pollster | Grade | Score | Accuracy | Lead | Openness | Elections |
|---|---|---|---|---|---|---|
| Izi | A | 79 | 36 | 13 | 30 | 1 |
| Dinamiche | B | 72 | 36 | 14 | – | 1 |
| Quaeris | B | 72 | 36 | 14 | – | 1 |
| Crespi Ricerche | B | 70 | 35 | 15 | – | 1 |
| Hokuto | B | 70 | 35 | 14 | – | 1 |
| EMG Different | B | 67 | 26 | 11 | 30 | 3 |
| Cluster17 | B | 66 | 33 | 13 | – | 1 |
| Datamedia Ricerche | B | 65 | 32 | 14 | – | 1 |
| NET Ricerche | B | 65 | 33 | 13 | – | 1 |
| Ekma Ricerche | B | 65 | 32 | 14 | – | 1 |
| ClandestinoWeb | B | 65 | 31 | 14 | – | 1 |
| Unicab | B | 64 | 32 | 13 | – | 1 |
| GfK Eurisko | B | 63 | 32 | 13 | – | 1 |
| Demoskopea | B | 62 | 30 | 14 | – | 1 |
| Ferrari Nasi & Associati | B | 62 | 31 | 12 | – | 1 |
| GDC | B | 62 | 31 | 12 | – | 1 |
| GPF | B | 62 | 31 | 12 | – | 1 |
| CIRM | B | 61 | 29 | 14 | – | 1 |
| Digis | B | 61 | 29 | 14 | – | 1 |
| Demetra opinioni.net | C | 57 | 29 | 11 | – | 1 |
| IPR Marketing | C | 56 | 28 | 11 | – | 4 |
| SWG S.p.A. | C | 55 | 27 | 11 | 17 | 4 |
| Lorien Consulting | C | 53 | 26 | 11 | – | 4 |
| Tecnè srl | C | 52 | 26 | 11 | 15 | 3 |
| ISPO Ricerche | C | 50 | 24 | 11 | – | 2 |
| Ipsos Italia | C | 48 | 29 | 11 | 8 | 5 |
| Istituto Demopolis | C | 46 | 28 | 11 | 7 | 4 |
| CISE | C | 46 | 23 | 8 | – | 2 |
| Quorum / YouTrend | C | 45 | 26 | 10 | 10 | 3 |
| Noto Sondaggi | C | 45 | 32 | 13 | 0 | 1 |
| Datamonitor | D | 44 | 22 | 9 | – | 1 |
| Scenari Politici | D | 44 | 21 | 10 | – | 1 |
| SpinCon | D | 43 | 21 | 9 | – | 1 |
| Winpoll | D | 42 | 30 | 12 | 0 | 1 |
| Index Research | D | 41 | 30 | 12 | 0 | 2 |
| Euromedia Research | D | 40 | 29 | 11 | 0 | 5 |
| Ixè | D | 39 | 28 | 11 | 0 | 2 |
| Demos & Pi | D | 38 | 27 | 11 | 0 | 4 |
| BiDiMedia | D | 38 | 21 | 9 | 8 | 3 |
| Lab21.01 | D | 38 | 27 | 11 | 0 | 1 |
| Termometro Politico | D | 35 | 25 | 10 | 0 | 3 |
| Istituto Piepoli | D | 30 | 22 | 9 | 0 | 4 |
| Eumetra MR | – | – | 30 | 12 | 30 | 0 |
| Lab21 | – | – | 30 | 12 | – | 0 |
| Pieopoli | – | – | 30 | 12 | – | 0 |
| Sigma Piave | – | – | 30 | 12 | – | 0 |
| Only Numbers | – | – | 30 | 12 | 0 | 0 |
| Excellera Advisory Group | – | – | 30 | 12 | – | 0 |
The order within the table says something: SWG and Tecnè, which publish most of what the law requires where we could see it, are at the top, and pollsters we could not check are at the bottom with no openness points. The accuracy and lead columns are close to zero for almost everyone, for the reason above. Pollsters with one final poll and no recent polls (such as Izi, whose 2022 final poll was the closest) are pulled so far towards the poor pollster that their record barely shows.
Each pollster's lean now
Beside the rating the site shows each pollster's latest poll within the 21-day window minus the average. Italian leans are small: most pollsters are within a point or two of the average on every list. A lean is not an error. Each pollster's page shows its lean now beside its errors at past elections, which for most pollsters are dominated by the shared misses on the M5S and the PD.
What the rating leaves out
As elsewhere, the rating leaves out how often a pollster polls, who commissions it, and mode. It also leaves out regional elections and by-elections, which are polled by fewer pollsters and less often.
Limitations
The scale. The British end points give Italian pollsters almost no record points. An Italian scale would spread them out; it would not make them more accurate.
Four elections, one of them 2013. Every pollster with a 2013 final poll carries the industry's ten-point miss on the M5S.
The register. Openness is scored on what we could read elsewhere. Once the register can be read, the method, sample and date items can be checked for every pollster, and most scores should rise.
Sources. Almost every final poll is a Wikipedia transcription.
The first real test of whether any rating predicts accuracy in Italy will be the next Camera election.
References
AGCOM, delibera n. 256/10/CSP, Regolamento in materia di pubblicazione e diffusione dei sondaggi sui mezzi di comunicazione di massa, Allegato A.
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.
Legge 22 febbraio 2000, n. 28, article 8.
How to cite
Polling Forecast (2026). Rating Italian pollsters: accuracy, the lead and openness. Retrieved 21 November 2026 from https://www.pollingforecast.com/research/italy-pollsters. Live figures change as new polls arrive; the date above is when the text was last revised.



