THE IDEA IN ONE LINE
Midterms usually punish the party in the White House.
18 losses, 2 gains: the House record since 1946.
Lost seats: 18 Gained seats: 2
18 of 20midterms with House losses
26seats lost on average
63biggest loss (2010)
2exceptions: 1998, 2002
Midterm elections fall halfway through a president's four-year term. The president is not on the ballot, but the president's party usually pays a price anyway. Since the Second World War, that pattern has been one of the most reliable in American politics.
In 18 of the 20 midterms from 1946 to 2022, the president's party lost seats in the House. On average it lost about 26.
Eighty years in one chart
Each bar shows how many House seats the president's party gained or lost. Bars below the line are losses.
The president's party lost House seats in 18 of 20 midterms
House seats gained or lost by the president's party, midterm elections 1946–2022
See the data
| Year | President | House | Senate | Approval (%) |
|---|---|---|---|---|
| 1946 | Truman (D) | -56 | -12 | 27 |
| 1950 | Truman (D) | -28 | -6 | 41 |
| 1954 | Eisenhower (R) | -18 | -1 | 62 |
| 1958 | Eisenhower (R) | -48 | -13 | 57 |
| 1962 | Kennedy (D) | -4 | +3 | 61 |
| 1966 | Johnson (D) | -47 | -4 | 44 |
| 1970 | Nixon (R) | -12 | +2 | 58 |
| 1974 | Ford (R) | -48 | -5 | 53 |
| 1978 | Carter (D) | -15 | -3 | 45 |
| 1982 | Reagan (R) | -26 | +1 | 42 |
| 1986 | Reagan (R) | -5 | -8 | 64 |
| 1990 | G. H. W. Bush (R) | -8 | -1 | 57 |
| 1994 | Clinton (D) | -54 | -8 | 48 |
| 1998 | Clinton (D) | +4 | 0 | 65 |
| 2002 | G. W. Bush (R) | +8 | +2 | 67 |
| 2006 | G. W. Bush (R) | -30 | -6 | 37 |
| 2010 | Obama (D) | -63 | -6 | 45 |
| 2014 | Obama (D) | -13 | -9 | 41 |
| 2018 | Trump (R) | -41 | +2 | 44 |
| 2022 | Biden (D) | -9 | +1 | 40 |
The losses can be huge. Democrats under Barack Obama lost 63 seats in 2010; Democrats under Harry Truman lost 56 in 1946. In 1994, Bill Clinton's party lost 54 seats and control of the House for the first time in 40 years.
Only twice did the president's party gain seats: in 1998 under Clinton (+4) and in 2002 under George W. Bush (+8), the first midterm after the attacks of 11 September 2001.
Optional: how the average is calculated
Add the 20 results, then divide by 20. The 18 losses add up to 525 seats and the two gains to 12 seats.
(−525 + 12) ÷ 20 = −25.65 ≈ −26 seats
The median, the middle result when they are sorted, is a loss of 22 seats. It is a little smaller than the average because a few very large losses pull the average down.
The Senate is less predictable
In the Senate, the president's party lost seats in 13 of those 20 elections. The pattern is weaker because only about a third of the Senate is up for election each time, so the result depends heavily on which seats happen to be on the ballot. In 2018, for example, Republicans lost 41 House seats but gained two in the Senate.
Why does this happen?
Political scientists offer several explanations, and they can all be partly true:
- A referendum on the president. With no presidential race, many voters use the midterm to say how they feel about the person in the White House.
- Turnout changes. Fewer people vote in midterms, and voters who are unhappy are often more motivated to turn out than those who are content.
- Balance. Some voters like the idea of a check on whoever holds the presidency.
Where the economy comes in
In a well-known 1975 study in the American Political Science Review, Edward Tufte found that two things went a long way to explain midterm results: the president's approval rating and the change in people's real disposable income, meaning income after taxes and after adjusting for prices, in the year of the election.
The second chart shows the first of those factors. Presidents with high approval, like Clinton in 1998 and Bush in 2002, avoided losses. Presidents with low approval tended to lose more.
Popular presidents lose fewer seats
Presidential approval before the election against House seats won or lost
The relationship is real but loose. Obama's approval was 45% in 2010, and his party lost 63 seats; Joe Biden's was 40% in 2022, and his party lost only 9. How many seats a party has to defend, how district lines are drawn, the quality of candidates and the big issues of the moment all matter too.
What this means for 2026
History is not destiny, but it sets expectations. The American Presidency Project lists President Trump's approval at 33% in mid-August 2026, below every president in its table except Truman in 1946. The economic signal Tufte looked at, real income, is also weak: average hourly pay fell 0.7% after adjusting for prices over the year to September, according to the Bureau of Labor Statistics.
That does not tell us the result. Republicans start with a narrow House majority, and the Senate map matters more than history. But it explains why forecasters, as of mid-October, see Democrats as favourites to win the House. For the economic picture in detail, read the US economy before the midterms, in five charts. For what a new Congress could actually change, read what the midterms can and cannot change. And for investors: what the S&P 500 has done after every midterm since 1950.
Sources
- The American Presidency Project, UC Santa Barbara: Seats in Congress gained/lost by the President's party in mid-term elections
- Edward R. Tufte (1975), “Determinants of the Outcomes of Midterm Congressional Elections”, American Political Science Review 69(3), 812–826
- Bureau of Labor Statistics: Real Earnings, September 2026
- Silver Bulletin: 2026 midterm forecast
Seat changes count the result on election night as recorded by the source; special elections and party switches between elections are not included.