Why does the president's party lose seats in midterm elections?
The short answer
- From 1934 to 2022, the president's party lost seats in the U.S. House of Representatives in every midterm election except 1934, 1998 and 2002. (American Presidency Project, 2026)
- One study testing three explanations together with U.S. survey data from 1952 on finds each explains a sizable part of the president's party's midterm drop in House votes, and a penalty voters place on that party explains the most. (Knight, 2017)
- That study cannot tell whether the penalty is a verdict on the president's performance or a wish for divided government, where each party controls part of the government. (Knight, 2017)
What the studies found
- In a table built from official House counts, the president's party gained House seats in the midterms of 1934, 1998 and 2002, by 9, 4 and 8 seats, and lost seats in every other midterm from 1934 to 2022. (American Presidency Project, 2026)
- The same table shows midterm House losses for the president's party as large as 72 seats in 1938, 63 in 2010 and 54 in 1994. (American Presidency Project, 2026)
- The Senate record has more exceptions: the president's party gained Senate seats in the midterms of 1934, 1962, 1970, 1982, 2002, 2018 and 2022, and broke even in 1998. (American Presidency Project, 2026)
- In the 2022 midterm, President Joe Biden's Democrats lost 9 House seats but gained 1 Senate seat, so 2022 was an exception in the Senate only. (American Presidency Project, 2026)
- The House's own party counts, taken from first election results, show Democrats held 207 seats in the Congress of 1997 to 1999 and 211 in 1999 to 2001, but 222 in 2021 to 2023 and 213 in 2023 to 2025. (U.S. House, Party Divisions)
- In U.S. midterms from 1938 to 1970, a change of 10 percentage points in the president's Gallup approval rating went with a change of 1.3 percentage points in the national congressional vote for the president's party. (Tufte, 1975)
- That study's author concludes that voters try to make the midterm a referendum, or verdict, on how the president's administration is doing, though the way votes turn into House seats mutes that verdict. (Tufte, 1975)
- A study of the 13 U.S. House midterms from 1946 to 1994 finds the president's party should expect to lose 2.5 seats for every extra percentage point of the two-party vote it won in the last presidential election. (Campbell, 1997)
- That study's explanation, called surge and decline, is that a winning president's strong year helps carry the party's House candidates, and without that boost at the midterm, some of them lose. (Campbell, 1997)
- The same study finds midterm seat losses also partly reflect a verdict on the president: each extra point of approval at the midterm saves the president's party one House seat it would otherwise have lost. (Campbell, 1997)
- Another study disputes surge and decline: using U.S. survey data on voters and nonvoters from 1976 to 1998, it finds the changes between presidential and midterm elections do not match what that theory predicts. (Mebane and Sekhon, 2002)
- That study finds some voters do turn against the president's party at the midterm to moderate policy once they know who holds the White House, but this alone does not explain why losses are so large or so frequent. (Mebane and Sekhon, 2002)
- In national polls on which party voters favor for Congress, one study finds early midterm-year results point to a normal outcome, but as the campaign goes on, preferences almost always shift toward the party out of the White House. (Bafumi, Erikson and Wlezien, 2010)
- Its authors find this shift is not a verdict on the president, since presidential popularity does not fall during midterm years, and that it fits balance theory, where some voters back the other party to get more moderate policy. (Bafumi, Erikson and Wlezien, 2010)
- Across 12 U.S. midterms in survey data starting in 1952, one model finds a penalty on the president's party explains 47 percent of its midterm drop in House votes, surge and decline and voters' views drifting back 27 percent each. (Knight, 2017)
- With a different measure of voters' views, the same study puts the penalty's share at 72 percent, surge and decline at 17 percent and the drift at 11 percent, so the split depends on the measure. (Knight, 2017)
- The same study cannot tell whether that penalty comes from voters judging the president's performance or from voters wanting divided government. (Knight, 2017)
What the evidence can't tell us
- Surge and decline did not foresee 1994: it expected the Democrats to lose only about half as many House seats as they actually lost. (Campbell, 1997)
- The referendum study's estimates rest on a fairly short series of elections, the U.S. midterms from 1938 to 1970. (Tufte, 1975)
- The seat counts compare each party's seats right after each election, and the number has sometimes changed later in the session. (American Presidency Project, 2026)
Sources
Every study here was read in a free copy, linked below. A number on this page comes only from a study read in full; one read only in its summary is described in words.
- John Woolley and Gerhard Peters, The American Presidency Project, University of California, Santa Barbara (2026). Seats in Congress Gained/Lost by the President's Party in Mid-Term Elections. The American Presidency Project (UC Santa Barbara) data table, last updated July 13, 2026; seat changes calculated from the House's and Senate's official party division pages. Free copy
- Office of the Historian and Office of Art and Archives, U.S. House of Representatives (no date). Party Divisions of the House of Representatives, 1789 to Present. History, Art and Archives, U.S. House of Representatives (official House history site). Free copy
- Edward R. Tufte (1975). Determinants of the Outcomes of Midterm Congressional Elections. American Political Science Review 69(3); read as the published article PDF posted on the author's own site. Free copy · Publisher's page
- James E. Campbell (1997). The Presidential Pulse and the 1994 Midterm Congressional Election. Journal of Politics 59(3); read as the published article scan posted on the author's University at Buffalo page. Free copy · Publisher's page
- Walter R. Mebane, Jr. and Jasjeet S. Sekhon (2002). Coordination and Policy Moderation at Midterm. American Political Science Review 96(1); read as the published article PDF posted on Sekhon's own site. Free copy · Publisher's page
- Joseph Bafumi, Robert S. Erikson and Christopher Wlezien (2010). Balancing, Generic Polls and Midterm Congressional Elections. Journal of Politics 72(3); abstract read on Dartmouth's open repository page. Free copy · Publisher's page (summary only: no figures taken from it)
- Brian G. Knight (2017). An Econometric Evaluation of Competing Explanations for the Midterm Gap. Quarterly Journal of Political Science 12(2), 2017; read as NBER Working Paper 20311 (July 2014, revised November 2015). Free copy · Publisher's page