I was today days old when I learned that there’s a real, well-documented historical pattern for how a president’s party fares in that president’s first midterm election — and that it’s a different pattern from the one you’ll probably hear cited on cable tonight. Going back through generations of midterms, the president’s party has lost an average of roughly 25 House seats and about 2 Senate seats in the midterm that falls in year two of a term. That’s the baseline. Whatever happens as polls close across the country this evening, that’s the number the results get measured against. Sources: US Polling Data, Yale ISPS.
First, Which Midterm Is This?
This part matters, because it’s easy to get wrong. President Trump began his second term on January 20, 2025, which makes November 2026 year two of that term — his first midterm of this presidency, not his sixth year. He’s in the unusual position of being a president who gets a second “first midterm,” having already had one in 2018 during his first term. So the relevant historical comparison tonight is the year-two pattern (about 25 House seats, about 2 Senate seats), not the more dramatic year-six pattern that gets a catchier nickname.
The “Six-Year Itch” Is Real — It’s Just Not This Election
That catchier nickname is the “six-year itch,” and it’s a legitimate phenomenon: it describes the midterm in a president’s sixth year, the second midterm of a two-term presidency. The term was coined by political strategist Kevin Phillips in a 1973 column, and the numbers back it up:
- Year two (first midterm): the president’s party loses about 25 House seats and about 2 Senate seats on average.
- Year six (second midterm, the “six-year itch”): losses average about 29 House seats and about 6 Senate seats — noticeably worse on the Senate side.
- Ulysses S. Grant, 1874 (year six): a historic landslide loss for Republicans amid corruption scandals.
- Barack Obama, 2014 (year six): Democrats lost 13 House seats and 9 Senate seats, handing Republicans the Senate for the first time since 2006 — almost exactly the shape the six-year itch predicts.
Sources: Yale ISPS, US Polling Data.
The Twist: The Itch Is Mostly a Senate Thing
Here’s the part that turned this from a dry stat into something I wanted to write about. When researchers at Yale’s Institution for Social and Policy Studies actually compared year-two and year-six midterms side by side, the House numbers didn’t meaningfully differ — the ~25-seat and ~29-seat averages are statistically similar to each other. The thing that genuinely gets worse in a president’s sixth year is the Senate, where average losses jump from about 2 seats to about 6. So the “six-year itch,” to the extent it exists as a distinct curse, is really a Senate curse. The House just reliably punishes the president’s party in any midterm, year two or year six, with roughly the same enthusiasm. Source: Yale ISPS.
The Full Historical Record, 1934–2022
Here’s every midterm since FDR’s first, in full — so you can see the ~25-seat House average and the year-two/year-six Senate split in context, instead of as two isolated numbers. Bars show the president’s party’s net seat change; the dagger (†) marks a president’s second-term midterm (a clean “six-year itch” case). Source: The American Presidency Project, UCSB.
House Seats Gained/Lost by the President’s Party
Senate Seats Gained/Lost by the President’s Party
| Year | President | Party | House | Senate |
|---|---|---|---|---|
| 1934 | F. Roosevelt | D | +9 | +9 |
| 1938 | F. Roosevelt | D | -72 | -7 |
| 1942 | F. Roosevelt | D | -45 | -9 |
| 1946 | Truman | D | -56 | -12 |
| 1950 | Truman | D | -28 | -6 |
| 1954 | Eisenhower | R | -18 | -1 |
| 1958 † | Eisenhower | R | -48 | -13 |
| 1962 | Kennedy | D | -4 | +3 |
| 1966 | L. Johnson | D | -47 | -4 |
| 1970 | Nixon | R | -12 | +2 |
| 1974 | Ford | R | -48 | -5 |
| 1978 | Carter | D | -15 | -3 |
| 1982 | Reagan | R | -26 | +1 |
| 1986 † | Reagan | R | -5 | -8 |
| 1990 | Bush | R | -8 | -1 |
| 1994 | Clinton | D | -54 | -8 |
| 1998 † | Clinton | D | +4 | 0 |
| 2002 | Bush | R | +8 | +2 |
| 2006 † | Bush | R | -30 | -6 |
| 2010 | Obama | D | -63 | -6 |
| 2014 † | Obama | D | -13 | -9 |
| 2018 | Trump | R | -41 | +2 |
| 2022 | Biden | D | -9 | +1 |
Why “Average” Is Doing a Lot of Work
The other thing worth holding onto tonight: that 25-seat House average is an average across wildly different individual elections, not a forecast for any one of them. Trump’s own first midterm in 2018 is the best illustration:
- Donald Trump, 2018 (year two): Democrats gained 41 House seats — well above the ~25-seat historical average.
- Same election, Senate: Republicans actually gained 2 seats, because the specific batch of Senate seats up that cycle happened to favor them.
Same night, same president, same national mood, and the two chambers moved in opposite directions. Which Senate seats are on the ballot in a given year can matter as much as anything the president did. Source: US Polling Data.
What to Actually Watch For
I’m not going to guess at tonight’s outcome — this post goes up before the results do, and the whole point is that the historical pattern is a baseline, not a prediction. But it gives you a useful lens. If the president’s party loses somewhere around 25 House seats and a couple of Senate seats, that’s history doing what history does. If the House losses blow well past that, it’s a 2018-style overperformance by the opposition. If the Senate holds or even gains for the president’s party while the House swings hard, that’s the 2018 split happening again, and it’s a reminder that the Senate map is its own animal. And if the losses are unusually small, that’s the genuinely rare outcome — the one that would actually break the pattern. Whatever the number is when you wake up tomorrow, now you know what it’s being compared to. Sources: Yale ISPS, US Polling Data.