What we don't know yet: limitations of the Polish forecast
Abstract. Every known weakness of the Polish forecast in one place: the assumed line-up and the thresholds, a correction that fails while the polls still miss, old misses applied to new parties, proportional swing for the splinters, the pull-back for Trzecia Droga, independent errors, don't-knows, the seats per okręg, the President and second-hand polls. For each, how much it matters and what would change our mind.
Summary
The Polish forecast has one part that is exact and several that are not. The seat law reproduces every Sejm election since 2005 to the seat. The rest depends on things the polls cannot settle: who runs with whom, which parties clear their thresholds, how parties that split from the 2023 lists are spread across the okręgi, and how today's polls, taken more than a year out, relate to the result. The polls themselves are not corrected, because a correction would not have helped, but they have missed by more than in most countries on the site.
This paper sets out each weakness, how much we think it matters, and what evidence would change our mind. The British counterpart is What we don't know yet: limitations and open questions in our UK forecast; the method is in Reading Polish polls: why they are not corrected.
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 Poland the 1993 to 1997 seat check is still shown but no longer counts.
- Pollster grades rescaled (13 October) (Rating Polish pollsters: accuracy, the lead and openness).
The line-up and the thresholds
The largest uncertainty about the next Sejm is not in the polls. Whether the PSL and Polska 2050 run alone, together, or on KO's list decides whether four points of the vote count at all:
| Line-up | Governing coalition | PiS + R+ + Konf. + KKP | Lists under their threshold |
|---|---|---|---|
| Every party on its own list (our projection) | 215 | 245 | Rozwój Plus 4.7, Razem 4.3, PSL 3.4, Polska 2050 1.8 |
| PSL and Polska 2050 together again as Trzecia Droga | 215 | 245 | TD 5.2, Rozwój Plus 4.7, Razem 4.3 |
| Two government lists: KO with PSL and Polska 2050; Lewica alone | 232 | 228 | Rozwój Plus 4.7, Razem 4.3 |
| One joint list of the governing coalition | 244 | 216 | Rozwój Plus 4.7, Razem 4.3 |
Our default, every party on its own list, is the line-up the parties are in today, and with today's shares one of the two least favourable to the governing coalition (Line-ups and thresholds: 5%, 8% and the Trzecia Droga question). The simulation, the replays and the polling-day forecast all hold it fixed, so none of their chances includes the chance that it changes. And a joint list is given the sum of its parties' shares, which is generous.
What would change our mind. The registration of the lists, a few weeks before polling day.
A correction that fails, and polls that miss
The correction fails its test, so the forecast takes the polls as they are:
| Election | Polls' average miss | Corrected, leave one out | Verdict |
|---|---|---|---|
| 2001 | 2.37 | 2.69 | further out |
| 2005 | 4.27 | 4.31 | further out |
| 2007 | 0.53 | 0.47 | closer |
| 2011 | 0.81 | 0.70 | closer |
| 2015 | 0.43 | 0.73 | further out |
| 2019 | 1.63 | 1.65 | further out |
| 2023 | 1.86 | 1.81 | closer |
| Average | 1.70 | 1.77 |
That is not a finding that Polish polls are accurate. They have missed by 1.7 points per list on average, and by more than 4 in 2005. It says only that the misses have not kept their direction, so an average of them does not help. One line has: the lists that became Trzecia Droga were understated at five of seven elections, by every pollster in 2023. The forecast does not correct it, because the rule is applied to all lines together.
What would change our mind. The next election. The rule is checked every time the forecast is built, and if a correction came to beat the polls on average it would switch on without any change to the method.
Old misses, new parties
The record's largest misses come from 2001 and 2005, from a party system that no longer exists, and the lines that join past lists to today's parties are rough: Lewica's joins the SLD–UP alliance of 2001 to an 8% party today. Since 2023 the system has split again: Rozwój Plus from PiS, KKP from Konfederacja, Razem from Lewica, and Trzecia Droga into three. None of these parties has a record of its own. The simulation gives them spreads that grow with their size, and the replays move them with their parents.
What would change our mind. Elections, local or European, in which the new parties stand on their own.
Proportional swing and the splinters
Each 2023 list's okręg vote is scaled by its parties' national change and shared among them in their national ratio (From votes to seats: the Kodeks wyborczy). So Rozwój Plus is strongest where PiS was strongest, KKP where Konfederacja was, and the PSL where Trzecia Droga was. That is a guess, and for Braun's KKP, whose vote may have a different geography from Konfederacja's, it may be a poor one. Proportional swing has not been tested on Polish okręgi, and the party system has changed too much since 2023 for a test on 2019 to 2023 to say much.
What would change our mind. Okręg-level results for the new parties, which only an election will give.
Pulling back towards 2023
The polling-day forecast pulls each line about half-way back towards its 2023 result, because that was closer than the polls at most distances in the test (From today to polling day in Poland). For Trzecia Droga's parties, which have fallen from 14.4% to about 4%, it adds several points and lifts the PSL towards 5% on polling day. If Trzecia Droga's break-up is lasting, the pull-back overstates its parties.
What would change our mind. The PSL or Polska 2050 recovering in the polls, or not, as the election nears.
Independent errors
The simulation draws each party's error independently (Simulating Polish polling error: where the chances come from). For the right, PiS, Rozwój Plus, Konfederacja and KKP, that gives a total more uncertain than the record of PiS's and Konfederacja's lines together, so the right's chance of 231 is spread wider than the evidence supports.
What would change our mind. Drawing linked errors from past elections' patterns, once there are enough elections to make the patterns more than anecdotes.
Don't-knows
Most pollsters publish a don't-know share of five to twelve points, which the forecast takes out, assuming the undecided split like the decided. The same was done for every final poll in the record, so any lean it causes is in the measured misses. But if the undecided lean differently this time, the forecast will not see it.
The seats per okręg
The forecast uses the seats per okręg in force since 2011. A bill passed by the Sejm and Senate in September 2026 would let the PKW reset them each year; by its calculation of January 2026, 13 okręgi would lose a seat and 8 gain one. It awaits the President's signature or veto. If it becomes law, the same votes will give somewhat different seats, with more elected around Warsaw and Wrocław and fewer in smaller towns.
What would change our mind. The President's decision, and the PKW's published figures.
The President
The forecast reports the chance of 231 and of 276, the votes needed to override a presidential veto (The President's veto and the 276 line). It does not model what the President does, or which laws a government could pass. A majority below 276 may govern very differently depending on whether the President is on its side.
Second-hand polls
About a third of this term's polls reach us only through Wikipedia's tables, every OGB poll among them, and the election-day series entirely. A transcription error would pass into the average.
What would change our mind. Pollsters and outlets publishing their figures in text we can read.
Summary
| Limitation | How much it matters | What would address it |
|---|---|---|
| The line-up and the thresholds | More than any polling error | The registration of the lists |
| No correction, sizeable misses | Polls taken as they are, one steady line uncorrected | The next election; the test reruns itself |
| Old misses, new parties | Spreads and replays rest on a changed party system | Elections with the new parties |
| Proportional swing for splinters | Okręg projections may be biased | Okręg results for the new parties |
| The pull-back for Trzecia Droga | May overstate its parties on polling day | Their polls as the election nears |
| Independent errors | The right's range too wide | Linked errors from past patterns |
| Don't-knows | An uneven split would go unseen | – |
| Seats per okręg | A reapportionment would move a few seats | The President's decision |
| The President | 231 and 276 mean different things | – |
| Second-hand polls | Transcription errors would pass into the average | Readable releases |
Table: Known limitations of the Polish forecast.
The forecast's claim is narrow: Polish polls, taken as they are because correcting them would not have helped, carried through a seat law that reproduces every election since 2005 exactly, under a line-up we state. The next Sejm election will be the first real test, and we will publish how it went.
References
Kodeks wyborczy (Dz.U. 2025 poz. 365).
Jennings, W. and Wlezien, C. (2018). Election polling errors across time and space. Nature Human Behaviour, 2(4), 276–283.
Konstytucja Rzeczypospolitej Polskiej, article 122.
How to cite
Polling Forecast (2026). What we don't know yet: limitations of the Polish forecast. Retrieved 21 November 2026 from https://www.pollingforecast.com/research/poland-limitations. Live figures change as new polls arrive; the date above is when the text was last revised.


