Polls had Five Star 3.3 points too high at Italian regionals

Final polls put the Five Star Movement (M5S) too high at all 19 Italian regional elections since 2020 where they reported its list, by 3.3 points on average. At 17 the miss was half a point or more; at the other two it was a tenth of a point or less. At the three general elections the party has fought, the same industry had it too low every time: by 10.4 points in 2013, 4.8 in 2018 and 2.0 in 2022.

That split matters now. Our forecast corrects Italian polls for their past misses at general elections, and that correction adds 2.9 points to the M5S. Under the electoral law passed on 8 October, the centre-left alliance that includes the M5S needs 42% to win the majority bonus. Our projection puts it on 44.1%.

What we counted

We took the last poll each pollster published before Italy's 15-day ban at every regional election since 2020 in the ordinary-statute regions, 22 elections and 100 final polls in all, and set each against the official result from the Interior Ministry. List figures are compared with each list's share of valid list votes, the same measure we use at general elections.

The M5S list appears in 19 of those contests and 53 polls. Fifty-one of the 53 had it too high. Taken together, the polls put the party at about one and a half times the share it won. In Abruzzo in March 2024 the four final polls averaged 14.3% for the M5S; it took 7.0%. In Lazio in 2023 the average was 14.8% against a result of 8.5%, and in Lombardy that year 8.3% against 3.9%. The two near misses were Umbria in 2024 and Veneto in 2025.

The party's own candidates for regional president show the same thing. In the nine contests where the M5S ran its own candidate, the polls had that candidate too high every time, by 3.4 points on average. The Democratic Party (PD) runs the other way: its regional list was understated at 14 of the 19 contests.

General elections tell the opposite story

At general elections the pattern reverses for both parties. The M5S was understated at all three it has contested, and in 2013 and 2018 by every pollster. The PD was overstated at each of them, by 4.7, 3.9 and 2.4 points.

The comparison is like for like. Both sets of figures are list shares measured against list results, both sets of polls were taken at least 15 days out under the same ban, and many of the pollsters are the same: Ipsos, SWG, Tecnè and Noto all appear in both records.

Nor is this a split between regional and national votes. At the 2019 European election, the final polls of the 16 pollsters we found averaged 22.5% for the M5S; it won 17.1%. Il Post reported at the time that YouTrend's poll average had the party on 22.6% and called its overstatement the biggest error of that election. At the 2024 European election, the last polls of ten pollsters before the ban averaged 15.4% for the M5S against a result of 10.0%, and 21.1% for the PD against 24.1%. The line runs between general elections and every other kind of vote. Our own accuracy paper lists the explanations offered for the general-election misses, including late movement during the ban and supporters who are hard for pollsters to reach, and says our data cannot separate them. We found no published work that explains why the direction flips at other elections.

Regional winners pull away

The regional record has a second pattern. In the vote for president, the candidate who went on to win beat their polled margin at 17 of the 22 elections, by a median of 5.6 points. The candidate leading in the polls beat their margin at 14 of 22, by a median of 4.2.

Some misses were very large. In Campania last November, Roberto Fico, the former M5S speaker of the Chamber running for the centre-left, led by 10.2 points in the final polls and won by 24.9. All four pollsters had his lead too small. In Lazio in 2023 Francesco Rocca of the centre-right led by 9.3 points and won by 20.4. In Campania in 2020 Vincenzo De Luca's polled lead of 19 points became a win of 51.

This is not a lean towards either side. The centre-right's margin over the centre-left beat its polls at 12 of the 22 elections and fell short at 10, and on average it ended half a point closer to the centre-left than the polls had it. The three poll leaders who lost, in Puglia in 2020 and in Liguria and Umbria in 2024, all led by less than 2 points.

General elections again look different. Adding up the final-poll averages for each coalition's lists, the centre-left led by seven or eight points in 2013, depending on which small allied lists are counted, and finished 0.4 ahead. In 2018 the centre-right led the M5S by about 9 points and won by 4.3. In 2022 the centre-right's polled lead of about 18 points matched its 17.7-point win. At no general election did the leading coalition pull away from the polls.

What it means for the 42% bonus

The new law, called the Stabilicum by the government, gives 70 extra deputies and 35 extra senators to the coalition that comes first with at least 42% of the vote in both chambers, up to 220 of the 400 deputies, Il Post reported. Below 42% the seats are shared in proportion. The Chamber passed it by 227 votes to 164.

Our projection assumes the line-up the centre-left has used in regional elections since 2024, with the PD, the M5S, the Greens and Left Alliance, +Europa and Italia Viva together. On that basis the centre-left takes 44.1% and the centre-right 40.6%, and the centre-left reaches 42% in 81% of our 2,000 simulations. Our average has the Brothers of Italy (FdI) on 26.9%, the PD on 20.1%, the M5S on 12.9% and Roberto Vannacci's Futuro Nazionale on 7.7%. The polls are steady: in October the M5S has ranged from 11.6% to 13.5%, and SWG's latest, completed on 12 October, had it on 12.7%.

The correction adds 2.9 points to the M5S and takes 0.9 off the PD. That M5S adjustment alone is larger than the 2.1 points by which the centre-left clears 42% in our projection. It rests on the three general elections. If the M5S instead fell short of its polls as it did at the regional votes and at both European elections, by 3.3 and 5.4 points on average, the centre-left's cushion above the bonus line would be smaller than that miss.

The correction has earned its place in our model so far: predicting each past general election without knowing its result, it cut the average miss per party from 1.92 to 1.82 points, and most of that gain came from the M5S. Taken at face value, the polls would give the centre-left 221 seats rather than 223, because the bonus goes to the first-placed coalition either way.

The election must be held by 22 December 2027. Avvenire reported that opponents of the law are preparing challenges in about 30 courts and that the Constitutional Court could rule from February. Giuseppe Conte has said a spring vote is fairly likely, according to Open.

The limits

Three general elections is a small record, and the M5S miss has shrunk at each one. The regional miss is shrinking too: it averaged 3.6 points before 2024, 2.8 since, and 1.7 at last year's four contests. Many regional final polls had small samples, and almost all the figures come from secondary transcriptions on Wikipedia rather than the pollsters' own releases. Since 2024 the M5S has fought most regional elections inside the centre-left and was still overstated. At general elections it has always stood alone, so the next one will be the first test of the alliance nationally.