2026 House midterm · November 3 · updated daily until election day
The 2026 House midterm
Republicans hold the White House and won 220 seats in 2024; 218 is a majority.
The first 2026 midterm forecast is written by the next daily run.
How much to trust each line
Tested on the 19 midterms since 1950, each predicted by a model that hadn’t seen it, against a no-model guess: “the president’s party loses the average number of seats”.
| Forecast | Seats | National vote | Verdict |
|---|---|---|---|
| Economy (income growth) | ±24 | ±2.7 pts | No better than guessing |
| Mood (approval) | ±19 | ±2.1 pts | Better than guessing |
| Guess the average midterm | ±22 | ±2.6 pts | The yardstick |
The economy line has not beaten the guess: on past midterms, income growth told you nothing that the usual midterm loss didn’t. The mood line has. It is shown anyway, because the site shows the economy and the mood side by side and every miss, and because that finding is the point: midterms have followed the mood, not the economy.
These are national models: they can’t see redistricting, retirements or individual races, so the ranges are wide on purpose. No House polls are used. Seats are counted as first elected, before any later vacancies.
The national House vote
Winning the national House vote and winning a majority of seats usually go together, but not always.
What goes in
Every midterm since 1950
| Year | In power | Seats | Economy said | Mood said | Approval | Income growth |
|---|---|---|---|---|---|---|
| 1950 | Democrats | −28 (235) | −22 | −39 | −10 | 7.5% |
| 1954 | Republicans | −18 (203) | −25 | −12 | 38 | 0.0% |
| 1958 | Republicans | −50 (153) | −18 | −13 | 23 | −0.6% |
| 1962 | Democrats | −6 (258) | −34 | −15 | 52 | 2.8% |
| 1966 | Democrats | −47 (248) | −35 | −30 | 9 | 2.8% |
| 1970 | Republicans | −12 (180) | −16 | −15 | 24 | 1.7% |
| 1974 | Republicans | −48 (144) | −16 | −33 | −35 | −2.7% |
| 1978 | Democrats | −14 (278) | −45 | −44 | 0 | 2.3% |
| 1982 | Republicans | −26 (166) | −16 | −25 | −2 | 0.1% |
| 1986 | Republicans | −4 (177) | −14 | −8 | 38 | 2.7% |
| 1990 | Republicans | −8 (167) | −16 | +1 | 52 | −0.7% |
| 1994 | Democrats | −54 (204) | −28 ✗ | −32 ✗ | −3 | 1.8% |
| 1998 | Democrats | +4 (211) | −22 | −17 | 26 | 1.0% |
| 2002 | Republicans | +8 (229) | −24 ✗ | −9 ✗ | 57 | 2.0% |
| 2006 | Republicans | −30 (202) | −23 | −40 | −23 | 2.4% |
| 2010 | Democrats | −64 (193) | −28 ✗ | −30 ✗ | −1 | 0.9% |
| 2014 | Democrats | −13 (188) | −18 | −31 | −10 | 1.5% |
| 2018 | Republicans | −42 (199) | −26 | −35 | −14 | 1.9% |
| 2022 | Democrats | −9 (213) | −63 | −37 | −16 | −6.5% |
✗ = the forecast got House control wrong. Income figures are as first published before each election from 1982; earlier years use the earliest available or revised figures (tagged in the data). Out-of-sample: each year predicted by models fitted on the other 18.
How it works
Four small models, fitted by ordinary least squares on the 19 midterms from 1950 to 2022. The economy line uses income growth; the mood line uses presidential approval. Each predicts the seat change (it also counts the seats the party holds going in, since more seats means more to lose) and the swing in the national House vote.
Chances come from each model’s out-of-sample errors (a Student-t spread with 16 degrees of freedom for the seat models). Coefficients are fixed (version 1, fitted October 17, 2026); they change only when a new midterm is added. The forecast updates daily until election day and then freezes.
Coefficients: econ seats: const 27.424, income 1.303, seats_before −0.234 (LOO ±24.3); mood seats: const 1.792, approval 0.399, seats_before −0.134 (LOO ±18.9); econ vote: const −3.663, income 0.033 (LOO ±2.7); mood vote: const −4.289, approval 0.061 (LOO ±2.1).
House only: no Senate or individual races. Not betting or investment advice.