There’s been a lot of pontificating about what this fall’s midterm elections will look like, and the general consensus is that it will be a tough one for Republicans. But judging from the history of midterm elections, can we be any more precise?
Economic forecast models of midterm election are typically not as reliable or accurate as forecasts of presidential elections. But they can give us an idea of what to expect. More importantly, they give us a baseline against which to measure the actual performance of the President’s party.
We know, for example, that the President’s party almost always loses House seats in midterm elections. (See the chart above.) Since World War II, there have only been two midterm elections where the President’s party has gained seats — 1998 and 2002 — when the incumbent presidents were unusually popular and had approval ratings of 65 or higher. (Bill Clinton was buoyed by a strong economy and a pushback against Republicans for the impeachment campaign, and George W. Bush by a residual 9/11 rally effect.)
As I like to do in a midterm year, I’ve put together a simple model forecasting the number of seats the President’s party wins, using three variables to predict it:
Economic growth. (Specifically, the growth in per capita real disposable income between the second quarter of the year before the election and the first quarter of the election year.)
The President’s average approval rating at Labor Day.
The number of House seats the President’s party currently has. (A larger majority is theoretically more vulnerable to a high number of seat losses.)
Just to give an idea of what the relationship between the economy and the vote looks like, note the figure below. Each data point is a midterm election since 1950, labeled by the year and the president at the time.
A few things to note here. First is that the more the economy is growing the fewer seats the President’s party loses. No surprise there.
Second is that 2022 is a huge outlier. (Honestly it screws up the regressions quite a bit. That red line is a lot steeper without 2022.) At least from this particular economic measure (which was heavily influenced by rising inflation), Americans lost a lot of purchasing power in 2022 — more than in any other year in this study. Yet Democrats under Joe Biden lost fewer than 10 seats. That is often attributed to Donald Trump’s outsized influence in the 2022 midterms, recruiting and endorsing many candidates, pushing them to echo his words about election fraud, and making the elections about himself as much as he could. We could also see this as a pushback against the Dobbs decision overturning Roe v. Wade.
Anyway, the most recent numbers lead to this equation:
Seat gains = -49.9 + 5.3*econ growth + 1.2*prez approval - 0.2*House seats held by prez party
So what’s the forecast for this year? I’m putting presidential approval at 39 (that’s what it is now), and the president’s party has 218 seats. And for the economy… well, we don’t actually know the growth in real disposable income yet, since the time period ends at the end of this month, and the numbers won’t be reported for several weeks after that. Income growth has been pretty anemic for the past few quarters, although several forecasts expect it to pick up in early 2026. So here’s a range:
Now, given that the Republicans’ current majority is by just a few seats, any of these forecasts is more than sufficient to put Democrats in charge of the House.
How reliable is this forecast? Who knows? We have a lot of other factors in play this year, including several mid-decade redistrictings, the likelihood that Trump will again play a large (and probably negative) role in Republican candidate selections, the possibility that Trump’s popularity will drop a good deal as this war continues, Trump’s own efforts to reduce voter turnout in Democratic-leaning areas, and more.
We’ll see what comes. It seems fair to say that Republicans are due for a significant rout, which is is in part why the president is so desperate to impose his own new rules on elections (which he can’t do). Also, if Republicans manage to come away losing only 10-15 seats and the majority, they still will have overcome massive historical trends to do that.
So a theory I've always toyed with but never had time to really finish collecting data on to dig through is that the number of resignations and retirements also is a signal for this. Kind of a "rats fleeing a sinking ship" hunch based on some other work I've done on congressional retirements and midterms. But I'm not sure how much that really moves the needle of the "surge and decline" model, or if their is a causal direction issue (does retirement lead to loss of seats? Or are loss of seats (potentially, based on legislators uncertainty) a cause for retirement?).
This is all interesting but kind of irrelevant at the same time. Whether Dems take the House by 20 or 40 seats doesn't matter in terms of restraining T or putting national politics back on a saner course -- a majority is a majority. The only real question that matters is our chance of taking the Senate as well, which would really put a brake on things. Otherwise this looks like T1 where we had a blue wave in the House but it meant nothing but a lot of investigations because the Reps held onto the Senate and he was not restrained at all.