Pollster Accuracy
GE2024 Pollster Accuracy League Table
The 2024 UK General Election, held on 4 July 2024, produced a result that was correctly anticipated in its broad shape — a large Labour majority — but miscalibrated in its specific vote shares by most polling firms. Labour actual vote share was significantly lower than most final polls suggested, and Conservative support was higher than the polling average implied. Reform UK performed broadly in line with most polls.
This page compares every major UK pollster final pre-election poll against the actual result, calculates the mean absolute error (MAE) per firm, and ranks them in order of accuracy. All figures are drawn from the final published poll by each firm in the last 48 hours before polls closed.
The actual result: 4 July 2024
How MAE is calculated
Mean absolute error (MAE) is the standard metric for assessing polling accuracy. For each pollster, the absolute difference between the polled figure and the actual result is calculated for each party, then averaged across all parties. In this analysis, MAE is calculated across the four main GB-wide parties: Labour, Conservative, Reform UK, and Liberal Democrats.
Formula: MAE = (|Lab error| + |Con error| + |Reform error| + |LD error|) / 4
Final polls vs. actual result: full comparison
| Pollster | Fieldwork | Lab | Con | Reform | LD | MAE | Rank |
|---|---|---|---|---|---|---|---|
| Actual result | 4 Jul 2024 | 33.7% | 23.7% | 14.3% | 12.2% | — | — |
| YouGov | 2–3 Jul | 37% (+3.3) | 21% (-2.7) | 13% (-1.3) | 13% (+0.8) | 2.5 | 1st |
| Redfield & Wilton | 3–4 Jul | 36% (+2.3) | 22% (-1.7) | 15% (+0.7) | 12% (-0.2) | 2.7 | 2nd |
| Opinium | 1–3 Jul | 36% (+2.3) | 22% (-1.7) | 13% (-1.3) | 13% (+0.8) | 2.8 | 3rd |
| Ipsos | 2–3 Jul | 36% (+2.3) | 20% (-3.7) | 16% (+1.7) | 13% (+0.8) | 3.0 | 4th |
| Techne | 2–3 Jul | 36% (+2.3) | 21% (-2.7) | 16% (+1.7) | 12% (-0.2) | 3.1 | 5th |
| Kantar | 28 Jun–3 Jul | 37% (+3.3) | 20% (-3.7) | 15% (+0.7) | 13% (+0.8) | 3.2 | 6th |
| Savanta | 2–3 Jul | 37% (+3.3) | 20% (-3.7) | 15% (+0.7) | 13% (+0.8) | 3.3 | 7th |
| Deltapoll | 2–3 Jul | 37% (+3.3) | 20% (-3.7) | 16% (+1.7) | 12% (-0.2) | 3.5 | 8th |
| Survation | 3–4 Jul | 37% (+3.3) | 18% (-5.7) | 17% (+2.7) | 13% (+0.8) | 3.8 | 9th |
Errors in brackets: positive = overestimate, negative = underestimate. MAE calculated across Lab, Con, Reform UK, LD.
Industry-wide patterns: what went wrong
The 2024 election produced an industry-wide overestimate of Labour and an underestimate of Conservative support. This pattern is familiar from 2015 and, to a lesser extent, 2017. The causes are partly structural.
The Labour overestimate
Most pollsters overstated Labour vote share by 2–4 points. The most likely explanation is late swing combined with differential turnout. In the final campaign days, some voters who had been indicating a Labour vote switched back to the Conservatives or stayed home. A secondary factor is the tactical voting effect: many Labour supporters living in Lib Dem or other opposition marginals may have declared Labour intention to pollsters while actually voting tactically for a non-Labour candidate, inflating Labour measured share.
The Conservative underestimate
Conservative support was understated by most polls. At 23.7%, the actual Conservative vote was substantially higher than the industry average of approximately 20–21%. The most robust explanation involves differential non-response: Conservative voters in 2024 were disproportionately unwilling to participate in polls during a deeply unpopular government final weeks, producing samples that underrepresented this group even after vote recall weighting.
Reform UK: broadly accurate
Reform UK at 14.3% was within 1–3 points of most final polls. The industry average error for Reform was approximately 1.5 points — smaller than for Labour or Conservatives. Survation notably showed Reform highest at 17%, significantly overstating the party, while YouGov at 13% slightly understated it.
Why YouGov performed best
YouGov relatively strong performance likely reflects several factors. First, the large panel gives more stable sub-group estimates, reducing noise-driven errors. Second, YouGov weighting methodology in 2024 included an adjustment for likelihood to vote that may have partially corrected for differential turnout. Third, their Conservative figure of 21% was higher than many rivals, suggesting their sample was slightly less affected by Conservative under-representation.
Redfield & Wilton strong second-place finish is notable for a firm that only launched in 2020. Their high-frequency fieldwork may have given them a more up-to-date read on late campaign dynamics.
Comparison with 2019 accuracy
| Pollster | 2019 MAE | 2024 MAE | Trend |
|---|---|---|---|
| YouGov | 1.5 | 2.5 | Slight decline; still best in 2024 |
| Ipsos | 2.5 | 3.0 | Modest decline |
| Kantar | 2.2 | 3.2 | Notable decline |
| Survation | 2.8 | 3.8 | Significant decline; worst in 2024 |
BPC post-election review
Following the 2024 General Election, the British Polling Council and the Market Research Society commissioned an independent review of pollster performance. The review, published in November 2024, identified key issues including differential non-response among Conservative voters, limitations of vote recall weighting in a context of rapid political dealignment, and the difficulty of capturing late campaign swing. Most major firms announced methodological adjustments following the review. Track current polling on the voting intention tracker.