What will the national House popular-vote margin be in the November 3, 2026 midterms (the generic congressional ballot outcome)?

Polymarket's House Popular Vote Margin market has real money behind it — $112.8k traded across thirteen buckets — and when you sum the D+4-to-D+10 buckets they carry 58 cents of every dollar. The wave buckets carry 26 cents, the toss-up-and-worse buckets carry the rest. That's not a coin flip dressed up as a forecast; that's a book that's actually settled. I read the tape as priced and don't fight it without a reason the market doesn't already have.

Center the model on the polling average — D+4.8 to D+5.3 across every tracker I pulled — and apply the actual historical error, not a guessed one: the final-poll-to-certified-result gap has averaged about 3.3 points in absolute value across the last five midterm cycles, split roughly 3-of-5 toward overstating the leader and 2-of-5 toward understating the swing. Fold a normal-ish spread of that width around a D+5 center and most of the mass — but not all of it — lands inside D+4 to D+10. The market's 58% is close to where the decomposition actually lands, maybe a touch high once you weight in the shy-Republican risk from 2024.

I count the ledger of every midterm since 2006, and the ledger warns against trusting an August number whole. In three of the last five cycles — 2006, 2010, 2022 — the final generic-ballot polls overstated the leading party's eventual margin; only twice did the polls understate the coming swing. And the specific mechanism that compresses a lead — the shift from registered to likely voters as pollsters tighten their screens in the closing months — has not even happened yet in most of these August numbers. The market prices the toss-up at twelve cents; the ledger says the true number is closer to a third.

Follow the actual mechanism, not the mood. Trump's net approval sits near minus-eighteen to minus-twenty-one, a genuine drag, and that's the engine behind the Democratic lead holding through the summer. But the NRCC is sitting on $92.7M in cash against an $18.5M DNC deficit, and Republicans are winning the economy argument in Washington Post/Ipsos data — that's the counter-mechanism capping the lead below wave territory rather than letting it run to D+10-plus. Bad environment, funded opposition. That combination lands you mid-band, not at the extremes.

I read the man at the center, not the average. In October 2018, Trump's approval sat near 40% and the country still handed Democrats an 8.6-point win. Today he is measured at 33 to 38%, worse than that same point in his first term, by pollsters ranging from AP-NORC to the Economist. A president this underwater, at the head of his own party's ballot, has historically bought bigger verdicts than this one — the wave scenario is not the edge case here, it's the pattern repeating with a weaker man at the center of it.

Everyone's anchored on a spring number that's already stale. The RCP average has shed 2.1 points since June 1 — D+7.4 down to D+5.3 in eight weeks — and national polling undercounted the actual 2024 Republican House margin by about three points. Run that trend forward and the current D+5 average is a snapshot of a number still falling, not a floor. The market prices the toss-up bucket at twelve cents; I think the tightening trend plus the shy-Republican pattern makes that a live underpriced path, not a tail.

Look at the shape of the story so far: a first-term president, a cost-of-living grievance that will not quiet, an approval line worse than the one that produced 2018's reckoning. That earlier midterm did not end at a modest lean; it ended at 8.6 points and forty seats, because the story that was writing itself refused to stop at 'close.' I read the current arc as the same tale told with a harsher opening chapter, and a tale like that does not resolve itself in the toss-up column. It runs to the wave.