If we're wrong: replaying past Italian polling misses on today's polls
Abstract. Six readings of today's Italian polls: as they stand, our corrected projection, and each Camera election's final-poll miss since 2008 replayed in full, all through the same seat engine and line-up. Under the default line-up the M5S's large misses move seats mostly inside the centre-left; the Lega's move the centre-right. We explain the count of collegi at risk.
Summary
A lens is one stated assumption about how wrong today's polls are. For Italy the site shows six readings of the same polls: the polls as they stand; our projection, which takes off half the average past miss because that passed its test (Reading Italian polls: the correction and its test); and four replays that take each Camera election's final-poll miss since 2008 and apply it in full to today's average. Every reading goes through the same seat engine under the same line-up (Who stands with whom: the coalition line-up assumption), and we show them side by side without averaging them.
Two things make the Italian readings different from the German and Spanish ones. The projection is not the polls: it moves points from the PD to the M5S and a few tenths elsewhere. And under the default line-up, in which the PD and the M5S stand together, most of what the misses move stays inside the centre-left. The replays change the M5S's own seats a great deal, from about sixty to more than a hundred, and the coalition totals much less. 21 of the 147 collegi change hands under at least one replay.
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.
- Only elections from 2000 shape the forecast (28 September). Earlier elections stay in the long view and on the history pages but no longer feed any test, correction or error size. In Italy the tests were already on elections since 2001, so nothing changed.
- The lenses on the map (November). The replays are also on the map as "The polls miss again", beside "You set the vote".
- When the polls missed (2 November): Grillo's tsunami: how Italy's 2013 polls missed the Five Star surge by ten points.
What a lens is
Let \(M\) be the eleven lists the forecast models, \(v_l\) today's average for list \(l\) (each pollster's latest poll within 21 days of the newest, equally weighted) and \(S = \sum_{l \in M} v_l\). For a past election \(t\) the error on list \(l\) is \(e_{l,t}\), final-poll average minus result. The replay of election \(t\) subtracts that error from every list it measured, with a floor of 0.1, and rescales the modelled lists to their original combined share:
The lists measured change from one election to the next. The PD, the Lega and FI (as the PdL in 2008 and 2013) are measured at all four; FdI, the M5S and AVS from 2013. Futuro Nazionale, Azione, Italia Viva, +Europa and Noi Moderati are measured at none, and move only through the rescale. So the 2008 replay, which moves only the PD, the Lega and FI, says nothing about the M5S.
Each replay applies a whole past miss to the polls as they stand, not half of it to our projection. The projection is the rule's reading; the replays are what the Camera would look like if this election's polls missed exactly as a past one's did.
After the replay, every reading goes into ItalyForecast.Run under the default line-up: proportional swing to the collegi, the coalition winners, the proportional seats and the seats abroad (From votes to seats: the Rosatellum).
The readings
The polls as they stand. The average with no adjustment.
Our projection. Half the average past miss taken off each list measured at two or more elections, then rescaled. Of the four countries outside Britain, Italy is the one where this row differs from the first.
Four past misses. One replay for each of 2008, 2013, 2018 and 2022, built from 82 final polls.
The reader's own. The what-if tool lets a reader set an error for each list from −6 to +6 points in half-point steps, and runs the same engine. Its presets are the readings above and the 2022 result. The prediction tool does the same from shares the reader types in.
Every reading through the same seat engine
| Reading | Centre-right | Centre-left | of which M5S | Others |
|---|---|---|---|---|
| The polls as they stand | 144 | 222 | 69 | 34 |
| Our projection | 143 | 224 | 82 | 33 |
| If the polls miss like 2001 | 144 | 223 | 67 | 33 |
| If the polls miss like 2006 | 222 | 145 | 46 | 33 |
| If the polls miss like 2008 | 149 | 220 | 68 | 31 |
| If the polls miss like 2013 | 143 | 224 | 112 | 33 |
| If the polls miss like 2018 | 220 | 148 | 65 | 32 |
| If the polls miss like 2022 | 141 | 224 | 81 | 35 |
The table is recomputed whenever a poll is added. Three features follow from the shape of the record and should survive most changes in the polls.
The projection and the polls differ inside the centre-left. The correction moves between two and three points, most of it from the PD and AVS to the M5S. Under the default line-up both are in the centre-left, so the coalition totals barely change, while the M5S's own seats, the "of which M5S" column, rise by a dozen or more. Under the 2022 line-up, with the M5S standing alone, the same correction would move seats between coalitions.
2013 is the M5S's replay. The 2013 replay adds 10.4 points to the M5S and takes 4.7 off the PD. It is the reading under which the M5S wins most seats, and, because the M5S's gain is larger than the PD's loss, the one under which the centre-left is largest.
The Lega moves the centre-right. The Lega was understated by 3.0 points in 2008 and 3.8 in 2018; the replays of those elections add that to it, and they are the readings under which the centre-right does best. The 2013 and 2022 replays, in which the Lega was overstated, take points off it.
Why the collegi jump
A collegio is won by plurality, so a coalition's collegi are a step function of its vote. When the two main coalitions are within a few points of each other nationally, a shift of a point or two moves many collegi at once, and the "Others" column moves with Futuro Nazionale and Azione, which win only proportional seats. The M5S's large replayed errors move the South more than the North, because that is where its 2022 vote is largest and proportional swing scales it most there.
Collegi at risk
For each collegio the site counts how many of the four replays give a different winning coalition from the projection:
A collegio is "at risk" if \(F_k \geq 1\); today 21 are. This is a count of historical scenarios, not a probability, and the replays are not independent: 2008 and 2018 both understated the Lega, and 2013 and 2018 both understated the M5S. The simulation's chance for each collegio is shown beside the count on the collegio pages.
Limitations
Four past misses are not a distribution. The replays show what known misses would do, not how likely each is.
One list dominates. Most of the spread between readings is the M5S's record, and one election, 2013, carries most of that.
The line-up is fixed. Every reading uses the default line-up. The difference between line-ups (Who stands with whom: the coalition line-up assumption) is larger than the difference between readings.
Lists without a record move only through the rescale. Futuro Nazionale is the largest of them, and nothing in the record says how well its support is measured.
References
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). If we're wrong: replaying past Italian polling misses on today's polls. Retrieved 21 November 2026 from https://www.pollingforecast.com/research/italy-lenses. Live figures change as new polls arrive; the date above is when the text was last revised.



