Midterms Appear In The Middle Distance

Eeeeeek, midterms. I cannot goddamn believe that the endless cycle of elections has brought us back around to caring about midterms. I am very tired.

Midterms are, almost universally, a rough time to be the party holding the presidency. Thermostatic backlash tends to mean that the country has become a little unhappy with you, and that your electoral results are somewhat worse than however you did in the presidential election year. Presidential approval ratings tend to dip over time as you have to “do policy” and “take actions” rather than promising stuff and making statements. Voters hate when stuff happens. If you’re the out-party, this means that Midterms are a big opportunity to gain (or keep) control of the House and Senate, and turn the last 2 years of the presidential term into a period where your opponent can’t enact any policy. Generally I say that parties should plan to get everything done in the first 2 years of their Presidential term, because the odds of their situation improving after that point are extremely bad. 

The folks at The American Presidency Project have some quick data on how many seats each party was defending and then gained or lost in midterm years. The House is fairly clear cut since all the seats are up every time. I did also look at gains/losses as % of seats currently held by the party, which looked exactly the same as this lazier graph. At a glance, while the magnitude of the loss varies a ton, the vast majority of the time, midterms are not great for whoever holds the presidency. 

There are radically different baselines each year, so whether or not this shift puts the presidential party into a House minority varies. 

Note I am doing some math simplification here and ignoring independents who caucus with particular parties because I don’t know anything about the varying caucuses in like, 1980, and I did not care to look it up. 

It is pretty hard to infer anything from presidential elections over time, because there are just not that many of them and the conditions wildly change year to year. What I feel most confident in is: midterms will not be great for Republicans. The current House majority is extremely narrow, 5 seats is well within the people retiring/dying margin (close enough that anxiety prevails among Republicans). I would be very surprised if Democrats didn’t take control of the House. However, it’s quite hard to know how many seats specifically they’ll take. 

The other data point we’ve got available is:

Approval Ratings

We have these for the incumbent President going quite far back, also sourced from the American Presidency Project. I’m using the last approval ratings before the election, late October. When I played around with it, early October was slightly more predictive, but not much. With this few datapoints, it’s hard to know what’s an artifact and what’s real, so I’m sticking with late October. 

If you limit this to modern ish presidents (Bush 1 and after, because I asked around and “post cold war” seems like a decent benchmark), there’s a maybe sort of clearer trend. This is obviously worse from a number of data points perspective. Please do not treat the line as a real prediction of future outcomes.

Using a purely vibes and squinting analysis, doesn’t look super great for Trump! He’s in line to roughly match his previous term’s approval trajectory, which was not good. 2018 was pretty famously a Democratic wave year, in which Republicans netted -40 House seats. If half that happens this year, Democrats would be firmly in control of the House. How much would that help them stop Trump’s terrible policies? Kind of unclear, because of DOGE’s commitment to fully ignoring congressionally allocated spending and because holding the House alone gives you the power to say “no” but not to really make policy. For that, you need....

👻The Senate 👻

My first cut at writing about the Senate map was in a piece on tipping point seat analysis, and my initial conclusion was “ouch, not a great map”. I stand by that: not a great map for Democrats!

The statistics on Senate seats gained and lost by incumbent parties in midterms are much messier than the House ones, because the map matters so much for the Senate. 

 In 2018, that Democratic wave year, Republicans gained 2 senate seats. This is because they were only defending 9 seats, while Democrats were defending *26*. The Senate is pretty much never a good time for Democrats, since the apportionment means that low-population states get inflated individual representation, but that map was worse than usual. 

2026’s Senate map is less bad than 2018, but not great. 

From wikipedia

This is 22 Republican seats and 13 Democratic seats. Crucially, Democrats have several incumbents in contested seats retiring, which makes everything worse. Incumbency is incredibly helpful to your electoral prospects, and losing that makes Michigan and Minnesota in particular far more in play. 

I included the special elections in OH and FL here as “retirements”, which they sort of are. The Florida retirement doesn’t matter at all, since it’s not plausibly in play. Ohio might maybe matter if Democrats have a good year and get a good candidate. Dems still need NC, OH, IA, as well as to hold all their existing seats, in order to take the Senate. That’s plausible but difficult. It’s early days yet, so we might see more retirements and maybe some strong candidates announcing challenges. 

Back of the envelope, in 2022 the House popular vote was D 43.7% vs R 50.5%, compared to the 2020 Presidential election which was D 51.3% - R 46.8%.  A shift of that size (-7.2 in margin of victory) away from the President’s party would put Iowa, Ohio, and North Carolina all over the line for Democrats and make Maine a tossup. It’s unlikely that would be precisely what happens, not least because swing by Senate seat tends to vary off a uniform swing by a decent margin, but it gives you an idea of how close things might be. 

Reading The Tea Leaves

For the next year or so, there’s not gonna be a lot you can use to predict the midterm outcomes other than what I’ve described above. The thing you’ll hear about *constantly* is special elections. I can tell you right now: Democrats are going to do a lot better in special elections than they did in the general election. This was true in PA, this will keep being true. Will it be enough to win Florida 1 and 6? No, certainly not, those are R+32/33 seats. Will it be a big shift from past vote in favor of Democrats? Yes. 

Right now, Democrats do far better in low turnout off year elections because the sort of very educated high income voters who show up for those elections are overwhelmingly Democrats. This has been a consistent pattern for the last few years. If you’re on twitter a lot, it might feel like the whole entire county is breathlessly following special elections, but most people barely know that they’re happening. It’s a quirk of the Democratic coalition that the sort of people who *do* know they’re happening are heavily Democratic.

It’s quite hard to know how much this overperformance is predictive of larger elections. If you’ve got access to a voter file, you can look at how the turnout for the special compares to general election turnout, and specifically look for signs that new or infrequent voters are showing up. Even that is tricky and uncertain, especially with this much time before the midterms. I would treat special election analysis as marginally better than informed guessing. 

The elections in NJ and VA are slightly better than specials information-wise, since they’re at least standard elections, if weirdly timed. VA’s 2021 election saw surprising Republican overperformance, which was blamed on voter dissatisfaction with Joe Biden. This was a decent indicator of how the 2022 midterms would pan out. Perhaps more predictive of long term trends was the NJ election in which the D Gov results went from a margin of 14 in 2017 to a margin of 3 in 2021. NJ has continued to be on a negative trend for Democrats since then, including in the 2024 general, forecasting some discontent in deep blue states. 

Will all of this stop me from following special elections? Absolutely not, are you kidding. If you care about Democrats winning, it’s still worthwhile to donate to the closer elections where you can (especially Susan Crawford). Just be careful about how much you infer from these unusual elections.