Particle.news

Polls Predicted a Democratic Wave but Understated Its Size

Post-2024 changes to weighting, respondent screening, turnout models appear to have shifted many surveys toward undercounting Democratic voters.

Overview

  • Polls correctly signaled a large Democratic victory in the 2026 midterms but were imprecise on magnitudes, missing the scale of the party’s gains.
  • Nationally, Democrats won the House popular vote by about 13 points, roughly four points larger than the final poll average of 8.9 points.
  • Errors were far bigger in several states, with polls underestimating Democratic margins by about nine points in Texas, Georgia and North Carolina, seven points in Mississippi and five points in South Carolina.
  • Analysts link the systematic Democratic undercount to post-2024 methodological shifts — heavier use of past‑vote weighting, new filters for unrepresentative respondents and altered turnout models — which traded one type of bias for another.
  • Performance varied widely by firm: a few pollsters closely matched final results while broad averages still helped blunt extreme misses, and the state-level errors are likely to have changed seat outcomes and renewed scrutiny of polling methods.