Grillo's tsunami: how Italy's 2013 polls missed the Five Star surge by ten points
Abstract. The last polls before the blackout had Bersani's centre-left about 5 points clear and the M5S on 15%; the centre-left won the Chamber by 0.4 points and the M5S took 25.6%, the largest single list. Every pollster understated the M5S and overstated the PD, through a new party with no past vote to weight on, reluctant centre-left switchers and a late surge inside the 16-day blackout.
Summary
When the legal blackout on publishing polls began, 16 days before Italy voted on 24 and 25 February 2013, the polls gave Pier Luigi Bersani's centre-left coalition a lead of about 5 points over Silvio Berlusconi's centre-right and had Beppe Grillo's Movimento 5 Stelle on about 15%. In the Chamber the centre-left won by 0.4 points, 29.6% to 29.2%, and the M5S took 25.6%, more than any single list. The miss on the lead between the two coalitions was 4.9 points; on the M5S alone it was 10.2 points, the largest single-party miss in our Italian series. The pollsters' explanation: a new party with no past vote to weight against, centre-left voters who would not admit they were switching to Grillo, and a late surge inside the blackout.
What the polls said
The last polls that could legally be published came out on 8 February. All are party-list figures for the Chamber.
| Pollster | Fieldwork end | PD | PdL | M5S | Scelta Civica | Lega |
|---|---|---|---|---|---|---|
| Euromedia | 3 Feb | 29.8 | 22.0 | 14.6 | 7.5 | 5.2 |
| ISPO | 5 Feb | 32.2 | 22.0 | 14.3 | 9.3 | 5.4 |
| Piepoli | 5 Feb | 32.0 | 22.0 | 13.0 | 9.5 | 6.0 |
| Tecnè | 5 Feb | 28.9 | 20.7 | 16.3 | 8.5 | 5.5 |
| Demos & Pi | 6 Feb | 29.9 | 20.4 | 16.0 | 11.5 | 5.0 |
| Ipsos | 6 Feb | 30.6 | 20.3 | 15.9 | 10.8 | 4.4 |
| SWG | 6 Feb | 29.5 | 19.5 | 18.8 | 9.0 | 5.2 |
| EMG | 7 Feb | 30.0 | 19.9 | 16.0 | 10.0 | 4.9 |
| IPR | 8 Feb | 29.5 | 20.0 | 16.5 | 10.0 | 4.2 |
| Termometro Politico | 8 Feb | 29.3 | 18.5 | 15.1 | 10.0 | 5.1 |
| Average, 16 pollsters | 29.8 | 20.0 | 15.3 | 9.3 | 5.1 | |
| Result (Chamber, Italy) | 24–25 Feb | 25.4 | 21.6 | 25.6 | 8.3 | 4.1 |
The average covers each of the 16 pollsters' last release between 3 and 8 February. Italian coverage focused on the coalitions, which decide the Chamber's majority bonus. Leggioggi's average of the final polls, published on 8 February, had the centre-left on 34.6%, the centre-right on 29.3%, the M5S on 15.3% and Mario Monti's centrist coalition on 13.4%.
What happened
In the Chamber (Italy, excluding Valle d'Aosta and the overseas constituency) the centre-left coalition took 29.55% (PD 25.43%, SEL 3.2%, Centro Democratico 0.49%, SVP 0.43%) and the centre-right 29.18% (PdL 21.56%, Lega Nord 4.09%, Fratelli d'Italia 1.96%, La Destra 0.65%). The M5S took 25.56% and Monti's coalition 10.56% (Scelta Civica 8.3%, UDC 1.79%, FLI 0.47%). Rivoluzione Civile won 2.25% and Fare per fermare il declino 1.12%, both below the threshold. Turnout was about 75%.
The 0.37-point lead was enough for the centre-left to collect the majority bonus: 340 of the 617 seats elected in Italy outside Valle d'Aosta, against 124 for the centre-right, 108 for the M5S and 45 for Monti's coalition. The Senate, with regional bonuses, gave no one a majority.
Against the average poll, the M5S was 10.2 points short and the PD 4.3 points too high. The centre-left coalition was 5.0 points too high, the centre-right almost exactly right (29.3 against 29.2), Monti's coalition 2.8 points too high and the PdL 1.6 points too low.
Why it happened: what the pollsters found
Italy held no official inquiry. In April 2013 lavoce.info published a dossier in which five heads of polling firms gave their own accounts and answered the same questions.
- Nando Pagnoncelli, Ipsos (8 April 2013), "the perfect storm". He listed an unusually large grey area of abstainers and undecided voters, 169 lists on the ballot, new parties (M5S, Scelta Civica, Rivoluzione Civile) with no electoral history, and "an unprecedented acceleration of voting choices in the last ten days". The main method failure was that there was no past vote for the M5S or Scelta Civica to weight against. Ipsos caught voters moving from the centre-right to Grillo but not the full flow from the centre-left, which grew after the Monte dei Paschi scandal and Grillo's "Tsunami Tour" rally in Rome, and which centre-left voters were reluctant to declare.
- Renato Mannheimer, ISPO (8 April 2013). 35% of voters said they had decided in the last few days and another 30% in the last two weeks. ISPO's own polls in the final days, which could not be published, had the M5S rising to 23%, against about 16% a month earlier. He added that weighting by recalled 2008 vote fails when millions of votes move: before the election only about 24% of respondents recalled voting PdL in 2008, against its real 38%.
- Alessandra Ghisleri, Euromedia (5 April 2013). Her firm's estimates were within half a point for every party except the M5S and PD, where the error was 5 points each way. About ten days before the vote, inside the blackout, 29.5% of centre-left voters told Euromedia they would welcome the M5S in parliament. Her reading: PD voters with a strong party loyalty were embarrassed to "confess" a vote for Grillo, and there was no historical anchor for the new party.
- Roberto Weber, SWG (5 April 2013). Samples systematically over-represent the highly educated, the non-religious and centre-left voters, and the share of people who will not say how they vote rarely falls below 15%. He also blamed herding: pollsters doubt their own numbers, look at their rivals' and soften the trend.
- Fabrizio Masia, EMG (2 April 2013). EMG's series from January showed the M5S rising steadily, and he argued that a result near 25% was realistic for anyone reading the trend rather than a single poll.
Roberto D'Alimonte, writing in Il Sole 24 Ore on 26 February 2013 (reprinted in CISE's election dossier), said Italian methods had simply been unable to capture the change, misjudging both Grillo and the South.
What our data shows
Every one of the 16 final polls in our database understated the M5S, from SWG's 6.8 points to SpinCon's 13.1, and every one overstated the PD, from 2.1 points (SpinCon) to 6.8 (ISPO). This was not one method failing: the polls were close on the PdL, the Lega and Scelta Civica, and almost the whole error was a transfer of about ten points from the M5S to the PD and the centre.
Two things are structural. The final published polls were taken 16 to 21 days before the vote, so any late surge is counted as poll error. And the M5S error persisted after 2013 as the party matured: our data has it understated by 5.1 points in 2018 and 2.0 in 2022, while the PD was overstated by 4.0 in 2018 and 2.4 in 2022. The 2013 M5S miss is the largest single-party error in our Italian series, ahead of the 5.6 points by which the Margherita was understated in 2001.
What the pollsters changed
The concrete change was to the anchor for past-vote weighting, the method Italian pollsters lean on most. Before 2013 they weighted against recalled votes from 2008 and 2009, which had no M5S in them; afterwards they had a recent election with the M5S at 25.6%.
- ISPO said it could now weight on recalled 2013 vote, which no longer suffered from the distance of the 2008 and 2009 elections and described the new party system, and that this removed "a large part" of the distortion.
- SWG said it already corrected for non-response with propensity-score weighting on demographic, behavioural and attitudinal questions, and that after each election it re-contacts its online panel to measure the gap between stated intention and actual vote and updates its correction coefficients.
- Euromedia said it compares its final estimates with the official results to set new weighting standards across phone, web and face-to-face interviews.
- EMG and Ipsos mostly defended their methods: EMG argued for a 70/30 split of landline and mobile-only interviews, Ipsos for dual phone samples and discriminant analysis to assign a probable vote to respondents who would not declare one.
None of this touched the blackout, which still stops publication more than two weeks before polling day, and the M5S was understated again in 2018.
What it means for our forecast
Our correction rule adds half of each party's average past miss to its polling, and keeps the correction only if doing so would have improved forecasts of past elections made without knowing their results. For Italy it passes narrowly, cutting the average error from 2.29 to 2.21 points a party, and the gain comes mostly from the M5S, understated at every election since 2013; so it is applied. Because Italy's last publishable polls are more than two weeks old on polling day, the simulated polling error also has to cover a late swing of the 2013 kind. The method and the backtests are in our paper on Italian polling error, and each company's record is in our Italian pollster ratings.
Sources
- Il Post, 4 March 2018: Come andò a finire nel 2013 (Chamber results and seats)
- Leggioggi, 8 February 2013: Elezioni politiche 2013, sondaggi finali: medie coalizioni e seggi
- Il Fatto Quotidiano, 8 February 2013: ultimo giorno per i sondaggi
- lavoce.info, April 2013: Dossier sondaggi (Pagnoncelli, Mannheimer, Masia, Ghisleri, Weber)
- CISE, Dossier 4: Le Elezioni Politiche 2013 (D'Alimonte, Il Sole 24 Ore, 26 February 2013)
- Pagella Politica, 1 August 2022: Quanto sono affidabili i sondaggi politici in Italia?
- Wikipedia: Opinion polling for the 2013 Italian general election (cross-check)
How to cite
Polling Forecast (2026). Grillo's tsunami: how Italy's 2013 polls missed the Five Star surge by ten points. Retrieved 2 November 2026 from https://www.pollingforecast.com/research/missed-it-2013. Live figures change as new polls arrive; the date above is when the text was last revised.