The historical pattern is clear, and ominous for Joe Biden and Democrats this year: The president’s party usually does poorly in midterm elections.
“It’s not quite a law of physics, but it’s probably as close as you’re going to get in the social sciences,” says Carlos Algara, a professor of political science at Claremont Graduate University.
Despite the persistence of this pattern, we know surprisingly little about why it happens — and why, every so often, it’s somehow averted.
Political scientists have argued about it for decades — there’s a 1988 paper called “The Puzzle of Midterm Loss.” Some theories focus on lower turnout among the president’s supporters. Others emphasize the public’s tendency to sour on an incumbent president. They may both be correct to some extent.
Other theories focus on why some presidents tend to do worse than others in midterms. Maybe the results are mainly about presidential approval these days. Or maybe they’re about the economy or, more specifically, real personal income growth. Some national crises, like 9/11, are associated with unexpectedly strong midterm performances for the president’s party — but others are associated with blowout defeats.
None of these signs are looking great for President Biden right now. His approval rating is the second-lowest of any president’s at this point in their presidency since modern polling came into use. The economy is booming by some metrics, but inflation is at a 40-year high and eating into voters’ spending power. The country is still in the midst of the pandemic, but Biden hasn’t unified the country around his leadership.
There’s no one weird trick that can guarantee midterm success, or one theory to perfectly explain every midterm result. But there are several that, considered together, go a long way toward helping explain why this so often happens — and what November’s midterms might herald for Biden.
Midterm election outcomes, by the numbers
The general story of midterm outcomes is clear: The president’s party usually has a bad time. And even in the unlikely event that it has a good time, it isn’t that good.
The most commonly used shorthand is a simple count of the number of House seats lost (or won) by the president’s party. Every House seat is up in every midterm and presidential year, so this makes a good comparison to the election cycle two years prior. And it shows a stark pattern — the president’s party has lost House seats in 17 of 19 midterms since World War II.
The losses have been big, while the gains are both rarer and smaller: Most of these seat losses are double digits, with some rising remarkably high (54 seats in 1994, 63 seats in 2010). But in the only two midterms where the president’s party gained House seats, the gains were small (five seats in 1998 and eight seats in 2002). That imbalance is a clear sign there’s something structural at work here.
An issue with the seat loss metric, though, is that it’s partly dependent on other factors. A party with a bigger majority going in has more to lose; shifting district lines over the decades affect how many districts are competitive (better gerrymandering makes fewer competitive districts).
So to assess the overall opinion of the electorate, looking at the nationwide House popular vote — the percentage of voters who voted for Democratic candidates versus Republicans — might be better. Conveniently, that’s been a perfect match for which party ends up with House control (the winner of the popular vote ended up with House control in all of these postwar midterms, though this hasn’t always been the case in presidential years):
The president’s party got fewer votes in 14 of these 19 contests, but presidents Truman, Kennedy, Johnson, and Carter did all manage to win more votes and hold the House despite losing seats. That was a different era of US politics, though — one that was less nationalized and less polarized, when voters were more willing to split their tickets, when incumbents had more of an advantage, and in which Democrats held the House for 40 years straight. Biden can no longer rely on any of that.
If we look beyond the House, the same general trend is evident, but because there are fewer contests, results are more affected by which seats happen to be up for election and the dynamics of individual races. The president’s party lost Senate seats in 13 of the 19 postwar midterms — most, but hardly all (for instance, Trump’s GOP benefited from a favorable map and picked up seats in 2018). There’s a similar story in governor’s contests, too.
Overall, waves against the president’s party are common and often quite large, while the best the president has been able to achieve in midterms anytime recently is something close to a draw.
The trend: Why does the president’s party usually lose?
So why does this happen? There are a few clues that can rule out possible explanations. The trend predates World War II, so it’s not about recent developments. It happens in states (the governor’s party usually loses seats in off-year legislature elections), so it’s not just about the presidency. It’s not just an American phenomenon, either. “It also occurs internationally in systems where there is a chief executive election separate from a midterm,” says Matt Grossmann, a professor of political science at Michigan State University.
And the trend usually reverses itself — at least partly — by the time the next presidential election rolls around, since most presidents get reelected and their party’s down-ballot performance usually improves relative to the midterms.
Any explanation must take all this into account. It should also grapple with the two main factors that explain why election results can differ: turnout and persuasion.
Turnout means a change in the composition of the electorate: different people voting or not voting, as compared to last cycle. (There’s a lot of that going on in the midterms, where turnout is always significantly lower than in presidential years.) Persuasion, on the other hand, involves voters who show up both times but change their minds, switching between parties.
One turnout-centric explanation is that the president not being on the ballot is inherently demobilizing to many of his supporters. Perhaps they realize, correctly, that he’ll still be in power after the midterms — the stakes are simply lower for them; there is no prospect of total defeat. Some might also grow disappointed in the new administration’s performance and feel less urgency about voting (though many will show up again when reelection rolls around and the presidency is on the line). Meanwhile, the president’s critics are quite likely to vote — they are fired up in opposition, and motivated to try to check his power.
A persuasion-centric explanation, in contrast, emphasizes public opinion’s tendency to shift against the president, as seen in a common pattern of sinking approval ratings in the first two years of a presidency. This theory holds that the public functions essentially like a thermostat, kicking in when it’s either too hot or too cold to restore the preferred temperature. Voters could conclude things are too conservative under a Republican regime and elect a new Democratic president. But then they could quickly conclude things have become too liberal, and swing back toward Republicans in the midterms. Then after Republicans regain some power, perhaps opinion will swing back toward Democrats and get the president reelected. (The Obama and Clinton presidencies both followed this trajectory.)
In political science, this is referred to as the “thermostatic model.” Some political scientists posit these swings happen in response to policy changes — say, Democrats move policy further to the left than voters prefer. Others argue that it’s not really about policy, but rather who’s in charge — that, for instance, the mere fact that Democrats hold power gives voters the sense that things are too liberal in the country now, swinging many to the right.
Which is correct: turnout and demobilization, or persuasion and thermostatic shifts? It’s a trick question because both are happening, though their relative impacts can vary. The Democratic firm Catalist analyzed voter file data (official government records of who turned out to vote) and concluded that the party’s 2014 midterm woes were more due to plummeting turnout, as Obama voters stayed home; its 2018 successes were more due to persuasion, as Trump voters changed their minds and voted for Democrats.
However, both effects often work in the same direction — supporters can be demobilized and swing voters can be lost for similar reasons. “You tend to get higher turnout on the side that public opinion is moving toward,” says Grossmann. And if that happens, that’s a powerful one-two combination against the president’s party.
The variation: Why are some midterm outcomes worse than others?
So the midterms usually go badly for the president, and they often go very badly, with something like a draw being the best-case plausible outcome. But Democrats would surely love to end up with a draw this year. Why have a few past presidents managed to get there, while others have suffered disastrous blowout defeats?
In recent years, a president’s approval rating has been the clearest predictor of their midterm fate. Over the past seven midterms, we’ve had two presidents with over 60 percent approval by midterm time, and they were the two who defied the trend of midterm defeat: Bill Clinton in 1998 and George W. Bush in 2002. (In 1998, Democrats lost the national House vote, but they did pick up seats.)
Those two midterms differ dramatically from the other five, which featured presidents with approval ratings in the 40s or below, all of whom fell victim to midterm wave defeats. (Biden’s current approval rating of 41 percent would place him in that latter batch.)
But presidential approval isn’t a perfect predictor, especially if you look further back in the data. Popular presidents like Dwight Eisenhower, Richard Nixon, Ronald Reagan (in his second term), and George H.W. Bush all suffered midterm defeats, as their approval ratings didn’t translate to popularity for Republican candidates down the ballot. That may be an artifact of a time when politics was less nationalized, but it’s a useful caution on the limits of presidential popularity.
Still, how did second-term Bill Clinton and first-term George W. Bush end up becoming so popular? The standard explanations are that Clinton benefited from a strong economy and the overreach of Republicans’ effort to impeach him, while Bush was still riding high off his massive popularity boost after the 9/11 attacks. But they had weaknesses, too — Clinton had just admitted publicly lying about his affair with Monica Lewinsky, while Bush’s economy still wasn’t strong.
Like Bush, earlier presidents with solid midterm performances may have benefited from being seen as competent leaders in national crises, such as Franklin D. Roosevelt in the Great Depression midterms of 1934 and John F. Kennedy in the post-Cuban missile crisis midterms of 1962. But Herbert Hoover, Jimmy Carter, second-term George W. Bush, Donald Trump, and now Joe Biden can tell you that not every crisis makes the president popular.
It’s the president’s response to that crisis, and whether it appears to produce results, that seems to matter. By 1934 GDP growth was booming, by November 1962 the standoff with the Soviets over Cuba had ended, and in November 2002 there hadn’t been another terror attack comparable to 9/11 on US soil.
The influence of the economy on midterms is not so simple to suss out, and scholars and commentators have argued about it over the years. Suffice to say things aren’t nearly as simple as that a good economy will lead to a good midterm — there’s disagreement about how much economic statistics like GDP tell us, or whether it’s voters’ perceptions of the economy (rather than any objective reality) that are really most important.
Is there a metric that can somehow capture both? University of Denver political scientist Seth Masket has zeroed in on one: real per capita disposable income growth. Rather than, say, GDP or the unemployment rate, this may come closer to reflecting the economy as the average voter experiences it — it measures whether pay is going up, when adjusted for inflation.
Masket found that, on average, higher income growth tends to be associated with better midterm outcomes for the president’s party. Specifically, he measured the growth between the second quarter of the year before the midterms and the second quarter of the midterm year. Here I’ve plotted it against the national House vote margin (Masket used the net change in seats in his own chart).
It does look like there’s a relationship here. Notably, those four Democratic presidents who held the House — Truman in 1950, Kennedy in 1962, Johnson in 1966, and Carter in 1978 — all had incomes growing strongly, and so did Clinton in 1998. Two recent anti-incumbent landslides, 2010 and 1994, are distinguished by weak income growth of around 1 percent. In fact, every cycle with below 3 percent income growth is a bad midterm year, except for George W. Bush in 2002.
But clearly there are other things going on, too. Nixon’s GOP did poorly in 1970 despite 4 percent income growth. And income growth of 2 to 3 percent was associated with outcomes ranging from a 12-point popular vote loss to 5-point popular vote victories. So a strong economy may well help, but it isn’t solely “the economy, stupid.”
What does it mean for Biden and Democrats?
If presidential approval, perceived crisis response, and real per capita disposable income growth are what we should be looking for, Biden is currently flopping on all three.
His approval is currently around 41 percent, and assessments of his handling of the pandemic and the economy have darkened. The income metric is particularly ominous for Biden — some forecasts project real income growth over the past year will be negligible or even negative, because of the effects of inflation. (The most recent data, for the fourth quarter of 2021, showed real incomes decreasing by 0.5 percentage points compared to the previous year, which I’ve added as a red line to the chart here.)
The only presidents to preside over negative income growth by the midterms — Eisenhower in 1954 and 1958, and Nixon in 1974 — did quite poorly. Even Obama in 2010 and Clinton in 1994, with their historic midterm losses, saw some income growth, though it was notably weak.
But all is not necessarily lost just yet. The midterms are nearly nine months away, and there is still time for things to change.
As Roll Call’s Nathan Gonzales writes, most presidents do not see their approval ratings significantly improve in their second year. There is one notable exception, though. Donald Trump’s approval was mired at 37 percent in December 2017, but by November 2018 it had recovered to about 42 percent, according to FiveThirtyEight’s tracker. Those midterms still went poorly for the GOP, but they probably would have been significantly worse if Trump’s approval stayed so low.
Feeding into that, there are questions about what will happen with the pandemic and the economy now that the omicron wave is subsiding. Masket’s model specifically focused on income growth in the second quarter of the midterm election year, and that second quarter is fast approaching. No one economic statistic or model is a perfect predictor, but the broader point is that we are heading into the period in which the economy will be fresher on voters’ minds.
Considering how historically difficult it is for a president to even get a draw in the midterms, Biden will probably need not just one but several things to break his way — an improving approval rating, growing real incomes, and an improving pandemic. And he likely needs both turnout and persuasion to break in his favor: He has to give sporadic Democratic voters a reason to cast ballots this year, and to win back some voters who initially approved of him but who have since soured on his presidency.