So when Trump said the country took a "giant step back" after the 2026 results, he was reacting to something close to an American political tradition. The scale is less routine. CBS News estimated that Democrats will hold around 256 House seats and at least 54 Senate seats in January after flipping more than 40 Republican districts. More than 20 House races and two Senate contests were still undecided.

How often does the president's party lose?

Almost every time. From 1946 to 2022, the president's party lost House seats in 18 of 20 midterms, about 26 seats on average. House control changed hands in seven of those elections.

The longer record from the American Presidency Project at UC Santa Barbara starts in 1934. That series has only three exceptions:

  • 1934: Franklin Roosevelt's Democrats gained 9 seats.
  • 1998: Bill Clinton's Democrats gained 4.
  • 2002: George W. Bush's Republicans gained 8.

The president's party lost House seats in every midterm from 1938 through 1994.

Some of the losses were huge. In the same table, Roosevelt lost 72 seats in 1938, Truman 56 in 1946, Clinton 54 in 1994, and Obama 63 in 2010. Even Biden's mild 2022 result was a loss of 9.

The Senate tells a messier story. The president's party lost Senate seats in 13 of the 20 midterms from 1946 to 2022 but lost the majority only six times. Because only about a third of Senate seats are on the ballot in a given year, the president's party sometimes gains in the Senate while losing the House, as in 1982, 2018 and 2022. Over 1934–2022, political scientist John Woolley counts average losses of 27 House seats and 3 Senate seats. About two-thirds of the time, the party lost ground in both chambers.

What are the main explanations?

University of Georgia political scientist Jamie Carson lists several forces that work together.

The presidential-year boost fades. A presidential election draws occasional voters who mostly back the winner, and some of them stay home two years later. This is Angus Campbell's 1966 "surge and decline" theory, which once dominated the field. It describes a "high-stimulus" presidential race followed by a "low-stimulus" midterm.

The midterm becomes a referendum. Governing disappoints some voters, and unhappy voters on the other side are motivated to turn out. Woolley finds that the lower a president's approval, the bigger the losses.

Voters want a check. Some people vote for the out-party to keep the president from having a free hand. Political scientists call this "balancing."

More seats are exposed. A party that just won big holds competitive seats it won under better conditions. Woolley names the number of seats the president's party must defend as the second factor that consistently tracks midterm swings.

Do political scientists agree on which one matters most?

No. Researchers disagree about how much each one matters.

In a working paper, Joseph Bafumi, Robert Erikson and Christopher Wlezien argue that surge-and-decline can't explain the full pattern. The president's party doesn't just fall back to its normal level of support at midterm. It drops below it. By their estimate, a party gains about 2.6 points of the House vote in a presidential year when it wins the White House. At midterms, it does about 4.1 points better when it is out of power than when it holds the presidency. Their conclusion: "to do well in midterm House elections, it is best to lose the presidency."

They also see a weak link in the referendum theory. Presidents' approval is not especially low at midterm, and majorities usually approve, yet the president's party still loses. They favor balancing instead. Their polling shows voters drifting toward the out-party as the midterm campaign goes on.

The data partly back them up. Woolley finds approval does correlate with seat losses, but only moderately: r=0.58 for the House and r=0.36 for the Senate.

2022 shows the limits of a pure referendum. Despite Biden's low approval, Democrats lost the House only narrowly and gained a Senate seat. Carson credits abortion, worries about Trump, and weak Republican candidates. The most likely answer is that all of these mechanisms contribute, but no agreed formula says how much each one counts.

Where does 2026 fit?

Several of the classic warning signs were present at once. The American Presidency Project lists Trump's Gallup approval at 35% in early August 2026 and 33% in late August. For comparison, Truman stood at 27% in late October 1946 and lost 56 seats. Republicans were also defending 220 House seats and 20 regular Senate seats.

Before the election, Woolley's model treated GOP losses of fewer than 25 House seats or 4 Senate seats as an unexpectedly good result for Republicans. It saw losses of more than 40 House seats and up to 12 Senate seats as more likely. He also warned that mid-decade redistricting and other 2026 novelties made history a shakier guide than usual. Carson and Max Whitford put Democrats' chances of retaking the House at about 80% and the Senate at roughly 50-50.

One factor seems to have mattered less than you might expect. Trump is a lame duck, since he can't run again. But by our count from the American Presidency Project table, the six lame-duck midterms from 1946 to 2022 cost the president's party about 20 House seats on average. That is no worse than other midterms.

Woolley's closest historical match is Grover Cleveland, who also served two non-consecutive terms. In the 1894 midterm, Cleveland's Democrats fell from 218 House seats to 93 and lost the Senate majority.

Is this the biggest midterm loss on record?

It's too early to say. Obama's 2010 loss is the biggest since 1946, but Roosevelt's 72-seat loss in 1938 was larger. Lists that start in 1946 simply leave it out.

Even the familiar numbers depend on how seats are counted. The Business Standard puts Obama's 2010 loss at 64 seats and Trump's 2018 loss at 42. The American Presidency Project gives 63 and 41. Exact rankings for 2026 will have to wait for the final count, and for an agreement on which count to use.

What we don't know yet

  • The final tally. More than 20 House races and two Senate races were unresolved as of Nov. 8.
  • Whether redistricting changed the size of the swing. Before the election, Carson expected Texas-style redistricting to reduce, but not prevent, Republican losses. Nobody has measured its actual effect yet.
  • Who changed their behavior. Did 2024 Trump voters stay home, or did they switch to Democrats? That's the kind of question that separates surge-and-decline from referendum and balancing explanations. Answering it will take post-election voter data.