Abstract
This paper develops a race-by-race forecast of control of the U.S. House after the 2026 election. Using AI-assisted judgmental probabilities for 68 potential seat changes, each contest is modeled as a Bernoulli event and aggregated into a national baseline forecast. The baseline produces an expected Democratic gain of approximately five seats and an expected House of about 220 Democrats, 214 Republicans, and one Independent, with Democrats having roughly a 77 percent probability of reaching 218 seats under an assumption that race results are independent.
The analysis then examines two factors that could produce substantial departures from the baseline: (1) a larger-than-expected blue wave associated with President Trump’s low approval rating, and (2) Democratic weakness in rural and exurban districts that proves greater than the baseline already incorporates. Regression analysis shows that the baseline itself assigns Democrats substantially lower probabilities of success in the most rural districts; the question is whether actual Democratic performance will be weaker still.
The paper’s principal contribution is therefore not simply its prediction of the final House composition. It provides the race-by-race probabilities underlying that prediction, statistically examines one potential source of systematic forecast error, and identifies four sets of contests that should provide early evidence that the baseline is failing—and in which direction. A complete appendix reports the probability assigned to every race included in the model.
Executive Summary
We model 68 possible seat changes: 60 races identified as key contests and 8 additional redistricting-driven or effective seat changes. Each race is treated as a Bernoulli event—a seat either flips or it does not—with the probability assigned in our race-by-race forecast. Expected flips can be summed without assuming independence. For the variance calculation and the probabilities of House outcomes, however, we make the simplifying assumption that race outcomes are independent.
· Expected R→D flips: 17.62
· Expected D→R flips: 13.23
· Expected Democratic net gain from direct R→D and D→R flips: 4.39 seats. Including Independent-related seat changes, the expected Democratic gain is 5.04 seats.
· After adjusting for possible seats held by Independents, expected total seat changes are 31.70, with a variance of 12.48 and a standard deviation of 3.53 seats.
· The expected new House is approximately 220 Democrats, 214 Republicans, and 1 Independent.
What follows is the part of the forecast that I think is most useful. The national number alone does not tell us where the model could be wrong. Regression analysis shows that the baseline already assigns Democrats substantially weaker prospects in the most rural districts. The paid analysis asks whether that adjustment is large enough, examines the statistical evidence behind it, identifies four election-night tests that could signal either a substantially larger Democratic wave or greater-than-expected Democratic weakness, and provides the complete race-by-race probability ledger underlying the national result.
