Headline polling claims are misleading: Recent projections suggesting Reform UK could form the next government are being overstated and should be treated with scepticism.
Too early to predict outcomes: With the next general election likely years away, voting behaviour is still highly fluid and difficult to model accurately.
Misuse of MRP polling: Multilevel regression and post-stratification (MRP) projections are often presented like exit polls, despite being speculative snapshots of current opinion.
How MRPs work:
- They use survey data to identify demographic and local factors linked to voting behaviour.
- These factors are projected onto constituencies based on population characteristics to estimate seat outcomes “if an election were held today.”
Track record is mixed:
- MRPs successfully predicted a hung parliament in 2017.
- Since then, their use has exploded, with 46 forecasts published between 2019 and 2024, producing widely varying and unevenly accurate results.
Media incentives: Journalists eagerly report MRPs, giving pollsters and political actors guaranteed headlines.
Political manipulation risk:
- MRPs can be used as tools for internal party warfare.
- A leaked projection predicting just 14 Conservative seats appeared designed to destabilise Tory leader Kemi Badenoch.
Key limitations:
- MRPs cannot capture future shifts in party fortunes, leadership changes, or campaign dynamics.
Tactical voting at constituency level adds further uncertainty.
Expert caution: Oxford professor Jane Green warns MRPs miss many factors that will change before the next election.
Proper use of MRPs:
- Useful as snapshots of current public opinion.
- Dangerous when treated as firm predictions.
Alternative modelling approach:
- The Economist uses MRP data but focuses on uncertainty rather than headline seat counts.
- Long-range forecasts should emphasise ranges of plausible outcomes, not precise results.
STOP REFORM UK 