It Don’t Mean a Thing if it Ain’t Got That Swing: The Outlook for Electoral Accountability in 2026

KEY POINTS FROM THIS ARTICLE

— Midterm elections often act as a brake on presidential power, but extreme partisan polarization calls into question whether the 2026 midterm will serve its usual role.

— However, despite growing partisan polarization, large partisan swings in the House and Senate still appear to be possible.

— This is truer of the House than the Senate, though, as nationalization has helped Republicans realize an advantage on the overall Senate map.

— States with Senate seats up for election in 2026, in aggregate, vote several points to the right of the nation, so even a substantial Democratic lead in generic ballot polling would not necessarily translate to a Democratic Senate majority, even though it very likely would in the House.

Electoral accountability in 2026

The 2026 midterm elections could well be the most consequential in modern history. Whether Democrats or Republicans control the U.S. House and Senate in the next Congress will determine whether the Trump administration can continue the drastic changes in the direction of domestic and foreign policy that it has been pursuing with almost no resistance from Congress, or whether one or both chambers of Congress will begin to act as a check on presidential power. However, despite the overwhelming importance of these elections and the growing unpopularity of many of the policies being pursued by the Trump administration, there are serious questions about whether electoral accountability is even possible in an era of extreme partisan polarization and declining electoral competition.

Midterm elections have frequently acted as a brake on presidential power—almost always resulting in losses for the president’s party in the House of Representatives and usually in the Senate as well. When the president and his policies are unpopular, those losses can be dramatic. In 1966, voters unhappy with President Lyndon Johnson’s handling of the presidency handed Republicans 47 additional seats in the House and 4 additional seats in the Senate. Eight years later, in the aftermath of the Watergate scandal and President Nixon’s resignation, Democrats gained 49 seats in the House and 4 seats in the Senate.

Over the past several decades, however, the American electorate has become increasingly divided along party lines, and the number of swing districts and states has declined sharply. This raises a question about whether there is still a potential for the midterm elections to deliver a clear rebuke to an unpopular president. In this article, I will examine the implications of these trends for electoral accountability in congressional elections, and in 2026 in particular. I will show that despite growing nationalization of congressional elections and the declining number of House districts and states that are closely divided in presidential elections, large swings in House and Senate seats are still possible, and these seat swings remain highly responsive to the voting preferences of the electorate. However, these conclusions apply more clearly to House elections than to Senate elections due to the drastic overrepresentation of sparsely populated rural states in the Senate and the peculiar makeup of the Senate seats at stake in a given election.[1]

Trends in congressional elections: nationalization and polarization

Two trends have profoundly changed the nature of congressional elections in the United States over the past three decades: nationalization and polarization. Nationalization refers to the growing connection between congressional and presidential elections. Evidence of this trend can be seen in Table 1, which displays the percentage of House and Senate races won by the party that carried the district or state in the previous presidential election and the correlation between the Democratic vote margin in the House or Senate election and the Democratic margin in the previous presidential election. The correlation data are based on districts and states with two major party candidates.

Table 1: Nationalization of midterm House and Senate elections, 1978-2022

Source: Data compiled by Gary Jacobson and author. Based on major party vote in contests with two major party candidates.

The data in Table 1 demonstrate a drastic increase in the nationalization of both House and Senate elections over these four decades. During the 1970s and 1980s, House and Senate elections were largely detached from national politics. There was only a modest relationship between the results of both House and Senate races and presidential elections. In fact, in some years, a majority of Senate races were won by the opposite party from the one winning the previous presidential election in the state. This was an era in which the personal appeal of candidates and incumbency were dominant factors in elections. During these years, voters did not hesitate to cross party lines to support congressional and presidential candidates from different parties.

However, since the 1990s and especially since 2010, both House and Senate elections have become highly nationalized. The overwhelming majority of winning House and Senate candidates come from the party winning the previous presidential election in the state or district, and there is almost a one-to-one correspondence between vote margins in congressional and presidential elections. These voting patterns reflect the emergence of a highly partisan electorate in which very few voters are willing to cross party lines to support candidates from the opposite party from their presidential choice. Candidate personality and incumbency are far less important in this new age of partisanship. Incumbents still win reelection at a very high rate, but that is not because of their personal appeal to the voters—it is because they come from the dominant party in the district or state.

Table 2: Geographic polarization of House districts and states by decade

Source: Data compiled by Gary Jacobson and author

Along with a dramatic increase in the nationalization of House and Senate elections, another key trend affecting these contests has been growing partisan polarization. Both House districts and states have become increasingly divided along party lines. This can be seen very clearly in Table 2, which displays the margins of victory for the winning presidential candidate in each district or state relative to that candidate’s margin in the entire nation. The data show that over time, the percentage of districts and states with outcomes close to that of the nation has declined, while the percentage won by a much larger margin than the nation has increased. In other words, there are fewer marginal districts and states and more blowout districts and states in presidential elections. During the 1970s, about a third of all House districts and more than 40% of all states were highly competitive at the presidential level. In the current decade, however, only 7% of all House districts and only 16% of all states have been highly competitive at the presidential level. Likewise, during the 1970s, less than a third of House districts and only about one-sixth of states deviated by 10 points or more from the national presidential vote margin. In the current decade, however, more than 80% of House districts and more than 40% of states deviated by 10 points or more from the national presidential vote margin. This trend is clearly unrelated to partisan gerrymandering, which has no impact on Senate elections.

Table 3: Competitiveness of House and Senate elections by decade

Source: Data compiled by Gary Jacobson and author

Given the increasing nationalization of House and Senate elections and growing one-sidedness of both districts and states at the presidential level, one might expect to observe a marked decline in the competitiveness of House and Senate elections. However, that has not been the case, as Table 3 shows. In fact, there has been little or no decline in the competitiveness of House elections and only a modest decline in the competitiveness of Senate elections since the 1970s. House elections were very uncompetitive during the 1970s, and they have remained very uncompetitive. The percentages of highly competitive and blowout races have changed very little. Senate elections have always been more competitive than House elections, and they have remained so, but the percentage of highly competitive races has declined modestly. At the same time, the percentages of very one-sided and blowout races have not increased that much.

The explanation for why competition has not decreased markedly in House and Senate elections despite the increasing nationalization of these contests and the increasing partisan polarization of House districts and states is very simple. These trends have been offset by a sharp decline in the personal advantage of incumbency. The large majority of House and, to a lesser extent, Senate elections were uncompetitive in the past because of the personal electoral advantage enjoyed by incumbents in these races. But as congressional elections scholar Gary Jacobson has documented, that personal advantage has shrunk dramatically over time in U.S. House elections and is now fairly small.[2] My analysis of Senate election data shows a similar though more modest decline in the personal advantage of incumbency. The vast majority of House races and most Senate races remain uncompetitive, but that is not due to the advantage of incumbency. It is because the vast majority of House districts and most states clearly favor one party at both the presidential level and at the district or state level.

Trends in seat swing and electoral accountability in midterm elections

The key to electoral accountability in midterm congressional elections is the responsiveness of seat swing in House and Senate elections to the preferences expressed by voters at the ballot box. The first question that we need to examine in that regard is the trends in the magnitude of seat swing in midterm House and Senate elections. Figure 1 displays the trend for the 12 House and Senate midterm elections between 1978 and 2022. In almost every case, House seat swing was in the direction of the opposition party, but in both 1998 and 2002, the president’s party made small gains in the House elections. There were several midterm elections in this series in which the president’s party made small gains in the Senate elections.

Figure 1: Party seat swing in midterm elections, 1978-2022

A. House

B. Senate

Source: Vital Statistics on Congress, Table 2-3

The data displayed in Figure 1 indicate that there has been no decline in the magnitude of seat swing in midterm House or Senate elections over these four decades. Relatively large seat swings are still possible despite growing nationalization and partisan polarization. Thus, in 2018, Democrats gained 40 seats in the House of Representatives, replacing the Republicans as the majority party. Likewise, in 2014, Republicans gained nine seats in the Senate, becoming the majority party in the upper chamber.

The results in Figure 1 suggest that both House and Senate seat swing may still be responsive to the preferences expressed by voters. We can measure responsiveness more directly, though, by conducting regression analyses of House and Senate seat swing in midterm elections since 1946. The dependent variable in this analysis is the change in Democratic House or Senate seats in the election. The independent variables are the national Democratic popular vote margin in the House or Senate elections and the number of seats held by Democrats in each chamber going into the election. The latter predictor is the exposure factor. In general, the larger the Democratic popular vote margin, the more seats Democrats should gain; on the other hand, the more seats Democrats held going into the election, the fewer seats they should gain or the more they should lose. The results of the two regression analyses are displayed in Table 4.

Table 4: Regression analyses of House and Senate seat swing in midterm elections, 1946-2022

Source: Data compiled by author

The results displayed in Table 4 provide strong support for our expectations about the effects of popular vote margin and seat exposure on seat swing in midterm elections. Together, these two predictors explain a remarkable 97% of the variance in seat swing in House elections and 77% of the variance in seat swing in Senate elections. The disparity between House and Senate elections reflects both the much smaller number of Senate contests in a given election and the fact that, in Senate elections, there are vast disparities in state populations, so small states like Wyoming and North Dakota have a very disproportionate impact on the outcome compared with large states like California and Texas. As a result, Senate seat swing is much less responsive than House seat swing to the results of the national popular vote.

The partial correlation between House Democratic popular vote margin and House Democratic seat swing, controlling for previous Democratic House seats, is .98. In contrast, the partial correlation between Senate Democratic popular vote margin and Senate Democratic seat swing, controlling for previous Democratic Senate seats, is only .77. After controlling for seat exposure going into the election, the national popular vote margin explains almost all of the variation in seat swing in House elections but only about three-fourths of the variation in seat swing in Senate elections.

Figure 2: Scatterplot of seat swing by predicted seat swing in midterm elections, 1946-2022

A. House Seat Swing

B. Senate Seat Swing

Source: Data compiled by author

The much greater predictability of House vs. Senate seat swing is very evident in Figure 2, which displays scatterplots of the relationship between actual seat swing and the seat swing predicted by a combination of seat exposure and popular vote margin in House and Senate midterm elections. In the scatterplot for House elections, the fit between predicted and actual seat swing is almost perfect, and there is no apparent trend in the accuracy of the predictions: more recent elections are predicted just as accurately as earlier elections. In the scatterplot for Senate elections, however, the fit between predicted and actual seat swing is not as good, and there appears to be a partisan trend in the accuracy of the predictions: Democratic seat swing has fallen below the predicted level in recent elections. In every midterm election since 2010, Democrats have won fewer seats than expected based on seat exposure and the national Senate popular vote. Democrats won three fewer seats than expected in 2010, two fewer seats than expected in 2014, four fewer seats than expected in 2018, and three fewer seats than expected in 2022. These are substantial differences in a chamber with only 100 members and typically only 33-34 seats in play in a single election.

There is a highly significant partial correlation of -.76 between the passage of time, measured by years elapsed since 1946, and the change in Senate Democratic seats after controlling for Senate Democratic popular vote margin and Senate Democratic seat exposure. In contrast, there is a very small and statistically insignificant partial correlation of -.08 between years elapsed since 1946 and change in House Democratic seats after controlling for House Democratic popular vote margin and House Democratic seat exposure. The significant underperformance of Democrats in recent Senate elections suggests that Democrats have been penalized in recent Senate elections by the growing association between state population and partisanship.

Republicans’ overperformance in Senate elections due to their domination of most of the less populous states is a relatively recent phenomenon that reflects the growing association of urbanism with Democratic voting. In the 1960s and 1970s, sparsely populated rural states like Idaho, the Dakotas, and Utah regularly elected Democrats to the Senate while Republicans often won Senate elections in heavily populated urban states like New York and California. In the past decade, however, almost no Democrats have won Senate elections in these sparsely populated rural states. The result is that Democrats are underrepresented in the Senate relative to their support in the national electorate.

What to expect in the 2026 midterm elections

We can use the results of the regression equations in Table 4 to make conditional predictions of seat swing in the 2026 midterm elections. Since we already know that Democrats hold 215 House seats and 13 Senate seats going into the election, the only unknown factor is the national popular vote margin in the House and Senate elections. Table 5 displays conditional predictions of the change in Democratic House and Senate seats depending on the national popular vote margin. I have included a range of possible national popular vote margins from a Republican margin of 4 points to a Democratic margin of 10 points. However, one can easily project outcomes based on smaller or larger Democratic vote margins from this table.

Table 5: Predicted change in Democratic House and Senate seats in 2026

Source: Data compiled by author.

The results in Table 5 suggest that a large Democratic seat swing is possible in both the House and Senate, although we have much more confidence in the House predictions. The national popular vote results will not be known until after the election, but we can use recent generic ballot polling, averaging the results of pre-election polls asking voters which party they would support in the congressional election, to estimate the likely outcome of the national popular vote. However, the generic ballot is a much more accurate predictor of the national House popular vote than of the national Senate popular vote since not all states hold Senate elections in a given year. The correlation between the generic ballot and the national House popular vote is .89 while the correlation between the generic ballot and the national Senate popular vote margin is a much weaker .75.

A Democratic vote margin of 5-6 points, roughly what recent generic ballot polls have been averaging, would result in an expected Democratic gain of about 20-25 seats in the House and 6-7 seats in the Senate, more than enough to shift control of both chambers. However, the group of states holding Senate elections in 2026 is considerably more Republican than the nation. In 2020, the last year in which this Senate class was up for election, Democrats won the national popular vote for the House by 3.1 percentage points but lost the national popular vote for the Senate by 2.3 percentage points. A 5-6 point popular vote margin in the House elections might predict a less than one-point popular vote margin in the Senate elections and a gain of only about four seats—barely enough to shift control of the upper chamber. These results suggest that the battle for control of the Senate is likely to go down to the wire even if Democrats easily win a majority of seats in the House. Once again, we see that the peculiar make-up of the Senate, with its vast overrepresentation of sparsely populated rural states, poses a serious potential obstacle to electoral accountability in the 21st century.

Endnotes

[1] These conclusions are based on the assumption that, despite threats from Donald Trump to interfere in the midterm elections, there will be an election in 2026 and that the votes in this election will be counted fairly and accurately.

[2] Gary Jacobson, “It’s Nothing Personal: The Decline of the Incumbency Advantage in U.S. House Elections,” Journal of Politics, 77 ( 2015).

See Other Commentary by Dr. Alan Abramowitz.

See Other Political Commentary.

Views expressed in this column are those of the author, not those of Rasmussen Reports. Comments about this content should be directed to the author or syndicate.