The current conventional wisdom is for the midterm to be somewhat better than average for the president’s party in the House. (I set aside the Senate here.) The fundamentals doubt that, as we’ll see.
The “fundamentals” provide a helpful baseline, even if “non-fundamentals” such as polling, candidate quality, unique issues, may modify that baseline. So let’s only look at historical relationships here.
The starting point is that Democrats currently hold 220 seats in the House, Republicans have 212 and 3 seats are vacant. 218 seats are the minimal majority with no vacancies. 2 of the vacancies were held by Democrats, 1 by a Republican, so call it 222-113 now.
The average loss for the presidents party since 1946 is 26.4 seats. That would put Dems at 196 and Reps at 239. (Note there is very little difference in 1st and 2nd midterm losses on average.)
Losses tend to be larger with less popular presidents. Biden average approval is 41.5% at FiveThirtyEight.com today and 42.4% at RealClearPolitics.com. Let’s call it 42%. It is now October. See the orange line for seat loss by October approval. That fit is a 40 seat loss.
Losing 40 of 222 seats would give Dems 182 seats are Reps 253 seats, considerably worse than an “average” loss of 26 seats. So 196 Dem seats if average, 182 Dem seats if as presidential approval would suggest.
Do note the variation around the orange line. It includes far larger losses, as 1994, and far smaller losses, as 2014. While the best estimate is -40 seats, for a president at 42% approval we see a lot of variation in seat loss, hence uncertainty.
A third fundamental approach combines the loss of popular vote for the House candidates of the president’s party in the prior presidential year and in the midterm. In 2020 Dem House candidates won 51.6% nationwide. But that implies they win only 47.8% in the midterm.
Again notice the variation around the blue line, and we haven’t seen a presidential year close to 51.6% since 1946. So more uncertainty here, but best estimate is a drop on nearly 4 percentage points in popular vote.
So how does popular vote translate to share of seats?
DemSeats% = -25.07 + 1.47*DemVote%
At 47.8% of the vote we’d expect Dems to win 45.2% of the seats, or 197 seats. That is back to an “average” loss, not the larger one based on approval.
There are other factors, even fundamentals, not considered here. The size of the current majority is rather small historically, at least for Democrats. So there are fewer seats and risk, and Dems lost rather than gained seats in 2020.
But there are issues pushing one way (inflation) and issues pushing the other way (abortion). Those are fit topics for a “beyond the fundamentals” analysis, but are not my topic here.
The conclusion is that simple fundamentals suggest a loss of 25 to 40 seats for the Democrats, giving them between 182 and 197 seats and the Republicans between 253 and 238 respectively. Anything in that range would be a strong GOP majority.
I stressed twice above the uncertainty in these estimates. For a given approval or a given national vote share there is considerable uncertainty in the share of seats that result. But if you want to consider the fundamentals, that’s what this gives.
For a “non-fundamental” take, consider the latest CBS News model, based on polling but with a sophisticated model for seats from that poll. As of Oct 16, CBS News estimates 211 Dem seats, a loss of just 11, to 224 Rep seats.
That is a much better result for Dems than the fundamentals expect. We’ll know in early November which was closer to the mark.
