Election pollsElection polls
Analysis lab
Explore how methodological assumptions affect results. These experimental calculations are not an official forecast or a ranking of pollsters.
90-day method comparison · Experimental
This analysis does not change the main monthly average. It includes source-checked records with sample size, fieldwork and publication dates, method and treatment of undecided voters. Records with the same method and undecided treatment are compared in separate groups. Editorial checks are still needed to establish whether the questions and target populations are comparable.
14 Aug 2026–11 Nov 2026 · 4 eligible records · 74 records excluded for coverage or missing information.
Equal: each poll has weight 1. Pollster-balanced: each pollster's average has equal weight. Sample/recency: √sample × 2^(−days since fieldwork ended / 30). These are not quality scores or margins of error and do not guarantee a better forecast. Historical views use records in the archive today, rather than reconstructing the archive as it stood then.
Yüz yüze:Kararsız dağıtılmış · 1 polls / 1 pollsters
Results require at least 3 polls and 2 pollsters.
Telefon:Kararsız hariç · 1 polls / 1 pollsters
Results require at least 3 polls and 2 pollsters.
Telefon:Kararsız ayrı · 1 polls / 1 pollsters
Results require at least 3 polls and 2 pollsters.
Karma:Kararsız dağıtılmış · 1 polls / 1 pollsters
Results require at least 3 polls and 2 pollsters.
Historical accuracy assessment
This archive does not contain a comparable set of polls conducted before the 2023 election, so we do not assign accuracy scores to pollsters. An assessment requires the same election, vote type, target population and treatment of undecided voters; a window specified before the election; complete sources; and comparable final results. Differing from other pollsters' averages does not establish an error.