Research · When the polls missed

Ten points clear, then second: how the polls missed Duda's first-round win in 2015

Abstract. Every final poll had Bronisław Komorowski about ten points ahead of Andrzej Duda; Duda came first by one point and Paweł Kukiz beat his polls by five. Turnout of 49% and a late break away from the incumbent undid polls that assumed a much larger, steadier electorate.

Polling Forecast · 2 November 2026 · 6 min read · Poland, 2015 presidential election, first round, polling error, final polls, inquiry, pollster methods

Summary

In the final week before the first round of the Polish presidential election on 10 May 2015, the last poll from each of five pollsters put the incumbent, Bronisław Komorowski, on 40.1% of decided voters and Andrzej Duda on 30.1%: a ten-point lead. Duda came first with 34.8%, Komorowski took 33.8%, and Paweł Kukiz, whom the polls had on 15.7%, won 20.8%. The error on the lead between the top two was 11.0 points, and every final poll had the wrong man in first place. Turnout of 49.0%, the lowest at any Polish presidential election to that date, and a late break away from the incumbent were exactly what the pollsters' turnout questions and their handling of the undecided failed to capture.

What the polls said

Polish pollsters usually publish shares of all respondents, with the undecided left in. To compare them with the result, we use each poll rescaled to decided voters, as Robert Konieczny of the Jagiellonian University's Centre for Quantitative Research on Politics did for its Pytia contest. The last poll from each pollster before the campaign silence:

Pollster	Fieldwork	Sample	Duda	Komorowski	Kukiz	Undecided in published figure
IBRiS	8 May	1,100	30.6	39.3	19.3	7.5
Millward Brown	6–7 May	3,016	28.7	41.5	13.8	6.0
TNS Polska	6–7 May	1,000	31.0	40.2	17.2	13.0
Estymator	6–7 May	1,293	29.0	40.0	16.0	none
PP/ROBOP	27 Apr–3 May	6,000	31.1	39.3	12.2	none
Average			30.1	40.1	15.7

Millward Brown's 6.0 is 3.0 undecided and 3.0 refusals. As published, TNS Polska had Komorowski on 35% and Duda on 27%; IBRiS's final poll had 36.3% and 28.3%. Every poll in the campaign had Komorowski ahead: Łukasz Wielgosz counted 51 from January to May, and in January Millward Brown had him on 65%.

What happened

The National Electoral Commission (PKW) announced the result on 11 May 2015:

Candidate	Votes	Share	Polls minus result
Andrzej Duda	5,179,092	34.8	−4.7
Bronisław Komorowski	5,031,060	33.8	+6.3
Paweł Kukiz	3,099,079	20.8	−5.1
Janusz Korwin-Mikke	486,084	3.3	+1.0
Magdalena Ogórek	353,883	2.4
Adam Jarubas	238,761	1.6
Janusz Palikot	211,242	1.4

Four minor candidates took the rest. There were 14,898,934 valid votes, and 48.96% of the 30,688,570 registered voters took part. Duda went on to win the runoff on 24 May, when turnout rose to 55.3%.

The Ipsos exit poll on the night had Duda on 34.8%, Komorowski on 32.2% and Kukiz on 20.3%. Interviewing people who had voted got the order right; polling people who said they would vote did not.

Why it happened: what the pollsters found

There was no inquiry. The pollsters spoke to the press in the days after the vote, and CBOS published its post-election surveys in the summer.

Marcin Duma (IBRiS), in Rzeczpospolita on 12 May 2015: "Sondaż to nie szklana kula, która pokazuje przyszłość, ale zdjęcie rzeczywistości wykonane w danym momencie." He placed the cause in turnout estimation: "Tu też należy upatrywać przyczyn rozminięcia się sondaży i prognoz z wynikami PKW."
Urszula Krassowska (TNS Polska), in the same piece: the gap "wynika z życia". She thought Komorowski's voters had stayed at home, and said TNS had tried asking whether respondents would certainly vote, "ale nie zawsze się to sprawdza".
Sławomir Nowotny (Instytut Badań Edukacyjnych) listed three causes: the difficulty of drawing a representative sample, refusals to take part, and preferences shifting before polling day.
CBOS reported in May that 73% of its respondents had said in April they would certainly vote in the first round, against actual turnout of 48.96%. Its post-election survey (report 97/2015, July 2015) found that 23% of first-round voters decided in the final week or later, and 11% on the Saturday or on polling day itself. Among Duda's voters 22% decided in the last week or later; among Komorowski's, 13%; among Kukiz's, 35%.
Robert Konieczny (Jagiellonian University) scored all seven May polls and found every one understated Duda, by 3.7 to 6.0 points, and overstated Komorowski, by 5.5 to 10.2.
Łukasz Wielgosz (Studia Polityczne, 2022) argued that months of comfortable leads lulled Komorowski's voters, while Duda's steady climb mobilised his own.

What our data shows

Every pollster, same direction. All five final polls understated Duda, by 3.7 to 6.1 points, overstated Komorowski, by 5.5 to 7.7, and understated Kukiz, by 1.5 to 8.6. No pollster had Duda within eight points of Komorowski. As in other industry-wide misses, the error was shared, not the fault of one house.

Part of it was late swing. Komorowski's lead had been falling all spring and Kukiz was rising fast in the last fortnight: from 7 to 11% in the last week of April to between 12 and 19% in the first week of May. The last poll in the field, IBRiS on 8 May, was the closest on Kukiz. The CBOS figures on late deciders, who broke for Duda and Kukiz far more than for Komorowski, account for a share of what the polls could not see.

Most of it was turnout and the undecided. A lead of ten points does not vanish in two days through persuasion alone. The pollsters measured intentions among a public of whom 73% said they would vote; half did. Rescaling the undecided in proportion to decided voters assumes they split like everyone else, yet CBOS found that late deciders broke towards Duda and Kukiz, and TNS Polska's published figures left 13% undecided. Those two assumptions, both favouring the incumbent with a large but soft vote, explain most of the gap.

Sejm polls have never missed like this. On the gap between the PiS and KO lines, the final Sejm polls in our database have missed by at most 3.8 points (2019) since 2001. Five months after this election the October 2015 Sejm polls had PiS on 36.9% against a result of 37.6%. An 11-point miss on the lead belongs to a different kind of contest: two personalities, an insurgent third candidate and less than half the electorate voting.

What the pollsters changed

We found no published method change from any of the five firms. Their public line was that polls are snapshots and that turnout had been the problem, which describes the miss without fixing it. What did change came from outside: in 2015 the Jagiellonian University launched its Pytia contest, an independent scoring of published polls against the result after rescaling them to decided voters. That gave Polish polls a public scorecard they had not had before, but it changed how they are judged, not how they are made.

What it means for our forecast

Our Polish forecast is built on the Sejm record, and its correction rule takes off half the average past miss and keeps that only if it would have improved past forecasts, each election predicted without knowing its own result. For the Sejm it fails, so we use the polls as they stand and show each past Sejm miss as a separate replay. The 2015 presidential first round is not one of those replays, because it was a different contest with a different electorate.

It still sets a bound on what to trust. A ten-point lead in a Polish first-round poll is not a safe lead when turnout is low and a third candidate is rising, and a projection should carry an error on the lead several times larger than the Sejm record alone would suggest. The Sejm replays are in our paper on Polish polling lenses, and pollsters' records are on our ratings page.

Sources

How to cite

Polling Forecast (2026). Ten points clear, then second: how the polls missed Duda's first-round win in 2015. Retrieved 2 November 2026 from https://www.pollingforecast.com/research/missed-pl-2015-pres. Live figures change as new polls arrive; the date above is when the text was last revised.