Poland Parliamentary electionby 11 Nov 2027in 355 days Government majority 10% Right-wing majority 86% Calendar Track record Research
Research · Poland
EN PL

Simulating Polish polling error: where the chances come from

Abstract. Seat ranges, each list's chance of clearing its threshold and each combination's chance of 231 and 276 come from 2,000 simulated elections under the default line-up, each party moved by its typical final-poll miss scaled to its size today, the thresholds checked on the national result, and each okręg moved by a local factor. The thresholds make the results lopsided. Drawing the right-wing parties' errors independently makes their total more uncertain than the record suggests.

Polling Forecast · updated 9 November 2026 · 6 min read · Poland, simulation, uncertainty, thresholds, majority, 231, 276, okręgi, independent errors

Summary

The seat ranges, the chance that each list clears its threshold, and the chance that each combination of parties reaches 231 and 276 all come from 2,000 simulated elections. In each one, every party's national share is moved by a random error the size of its typical final-poll miss, scaled to its size today; the thresholds are checked on the resulting national shares; every party's vote in every okręg is moved again by a small local factor; and D'Hondt is run in each okręg under the default line-up. The outputs are counts across the 2,000 runs.

The thresholds make the Polish simulation more lopsided than most. A list near 5% either wins a dozen or more seats or none, and when it wins none its votes go to nobody, raising every other list's seats. So the ranges of the large lists depend on the small ones, and the chance of a majority can move several points with a small change in a small party's share.

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.
  • Every outcome (3 October). The seats in every simulated election are shown as histograms at /pl/outcomes.

One run, step by step

  1. Move each party's national share. Add to each party's share a random error drawn from a normal distribution with the standard deviation below. A share cannot fall below 0.05%.
  2. Check the thresholds. Sum the parties into lists under the line-up and compare each list's national share with its threshold: 5% for a party list, 8% for a coalition list, none for the German minority (Line-ups and thresholds: 5%, 8% and the Trzecia Droga question).
  3. Project the okręgi. Scale each 2023 list's okręg vote by the national change of the parties descended from it, exactly as for the projection (From votes to seats: the Kodeks wyborczy), multiplying each by its own random local factor.
  4. Count the seats. D'Hondt in each okręg among the qualifying lists.
  5. Record each list's seats, whether it cleared its threshold, the leading list in each okręg, who holds each seat on the hex map, and whether each combination of parties reached 231 and 276.

The random number generator has a fixed seed, so the same polls give the same chances.

National error

For the main party of each 2023 line (PiS, KO, Lewica, Konfederacja), the standard deviation is the line's typical final-poll miss, scaled to the party's size today:

\[ \sigma_p = \max\!\left(0.3,\ \text{RMS}_L \sqrt{\operatorname{clip}\!\left(\frac{S_p}{\bar{v}_L},\ 0.25,\ 1.5\right)}\right), \tag{1} \]

where \(\text{RMS}_L\) is the root mean square of line \(L\)'s final-poll errors (Measuring Polish polling error, 2001–2023), \(S_p\) the party's share today and \(\bar{v}_L\) the line's average result at past elections. The scaling matters because the lines have changed size: Lewica's line won 41% in 2001 and polls at 8% now, and a miss of several points on a 41% list says little about the miss to expect on an 8% one. Every other party, including those that split from a 2023 list, gets

\[ \sigma_p = \max(0.3,\ 0.3 + 0.12\,S_p). \tag{2} \]

In run \(r\),

\[ S^{(r)}_p = \max\!\left(0.05,\ S_p + \sigma_p\, z_{r,p}\right), \qquad z_{r,p} \sim \mathcal{N}(0,1) \text{ independently for each } p. \tag{3} \]

Today's values:

PartyShareElections measuredTypical final-poll miss (RMS)Standard deviation used
KO32.672.732.83
PiS22.071.641.37
Konfederacja14.031.291.57
KKP8.80–1.36
Lewica7.473.382.21
Rozwój Plus4.70–0.87
Razem4.30–0.82
PSL3.40–0.71
Polska 20501.80–0.51
Live The simulation's national error: standard deviation drawn for each party (points) Computed from the database, polls to 19 November 2026.

KO's spread, the widest, comes from its line's miss of 6.5 points in 2005. Lewica's line has the largest typical miss (6.8 points in 2001, 5.6 in 2005), but at less than half its average size the spread used is smaller. Konfederacja is polling at more than twice its line's average result, so its spread is scaled up to the limit. The parties with no record of their own get a spread that grows with their share: about a point and a third for KKP, a point for Rozwój Plus.

What the independent draw leaves out

Each party's error is drawn independently. For the governing coalition's lines the record and the draw roughly agree: the combined miss on KO, Lewica and Trzecia Droga had a root mean square of about 3.6 points over the seven elections, against about 3.7 for the independent draws, though since 2005 the record is smaller, about 1.9. For the right, the draw is wider than the record: PiS and Konfederacja's lines together were missed by about 1.9 points, while the independent draws for PiS, Rozwój Plus, Konfederacja and KKP give their total a standard deviation of about 2.6. So the right's total, and its chance of 231, are more uncertain in the simulation than the record suggests.

The British simulation avoids this by replaying whole past misses (Simulating UK polling error: where the chances come from). The replays on the what-if page do the same for Poland (If we're wrong: replaying past Polish polling misses on today's polls); the simulation does not.

Local error

Each 2023 line's projected vote in each okręg is multiplied by its own factor,

\[ \hat{u}^{(r)}_{k,L} \leftarrow \hat{u}^{(r)}_{k,L} \times \max\!\left(0.2,\ 1 + 0.08\, u_{r,k,L}\right), \qquad u_{r,k,L} \sim \mathcal{N}(0,1), \tag{4} \]

and, where a line has split into several parties, each party's part is multiplied by a further factor of the same size, so that Rozwój Plus is not always strongest exactly where PiS is. The German minority's vote is held at 2023. The factor's size, 8%, was chosen, not estimated. The thresholds use national shares and are not affected by it.

What comes out

Seat ranges and thresholds

For each list, the 80% range is the band within which its seats fell in the middle 80% of runs; the threshold chance is the share of runs in which its national share cleared its threshold.

ListPollsThresholdSeats80% range2023Clears the threshold
KO32.65%191168–206157over 99%
PiS22.05%131117–143194over 99%
Konfederacja14.05%7359–8518over 99%
KKP8.85%4127–49–over 99%
Lewica7.45%240–432686%
Rozwój Plus4.75%00–15–38%
Razem4.35%00–10–21%
PSL3.45%00–0–1%
Polska 20501.85%00–0–0%
Live Our projection by list (every party on its own list) Computed from the database, polls to 19 November 2026.

A list whose threshold chance is between a fifth and four fifths has a range that starts at 0: in some runs it wins a dozen seats, in others none. Those lists' uncertainty spills into everyone else's ranges.

231 and 276

For each combination, the chance of 231 is the share of runs in which the lists of its parties won 231 seats or more, and likewise for 276 (The President's veto and the 276 line).

CombinationSeats231+276+
The governing coalition21510%0%
The governing coalition + Razem21513%0%
PiS + Konfederacja2042%0%
PiS + Rozwój Plus + Konfederacja2045%0%
PiS + Rozwój Plus + Konfederacja + KKP24586%2%
Live Combinations: seats in our projection and the chance of 231 (a majority) and 276 (three fifths, enough to override a veto) Computed from the database, polls to 19 November 2026.

The combinations overlap, and their chances do not add up to anything: in one run both the governing coalition with Razem and the right can fall short, and never can both reach 231. A combination's seats are those of every list with one of its parties on it, so under a joint list the combination gets the whole list.

Okręgi and seats

For each okręg the simulation records the range of each list's seats there and how often each list leads; for each hexagon on the seat map, how often each list holds that seat. The okręg pages and the map show them.

What we know about the simulation

We can say how it was built, not how well it performs: seven elections are too few to check how often results fall inside the ranges. Four points matter.

The spreads come from seven elections, and the largest misses come from 2001 and 2005, when the party system was different.

The scaling to today's size is a choice. It stops Lewica's 2001 miss from giving an 8% party a spread of more than three points, and it lets Konfederacja's spread grow with its vote. Both are reasonable; neither is tested.

Independent draws make the right's total too uncertain, for the reason above.

Everything holds the line-up fixed. The largest source of uncertainty about the next Sejm, who runs with whom, is shown by the line-ups side by side, not in the simulation (Line-ups and thresholds: 5%, 8% and the Trzecia Droga question).

Limitations

The simulation turns one projection into chances with the least machinery the evidence allows: a normal draw per party, a local factor per line and okręg, the thresholds and D'Hondt under one line-up. It does not model linked errors, turnout, candidates or the line-up. The polling-day version adds drift for the time left (From today to polling day in Poland). What we don't know yet: limitations of the Polish forecast collects the rest.

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). Simulating Polish polling error: where the chances come from. Retrieved 21 November 2026 from https://www.pollingforecast.com/research/poland-simulation. Live figures change as new polls arrive; the date above is when the text was last revised.

You may also like