Polymarket puts a Democratic House at 88%. That, not "divided government", is the 2026 number

Hold the combination markets against the two chamber prices and they do not survive the contact. Republicans cannot keep Congress without keeping the House, so a Democratic House price of 88% caps a full Republican hold at 12% ^[8]. The band quoted for that outcome starts at 13% ^[2], entirely above the ceiling the House price allows ^[9]. Take the midpoints of the three bands and they sum to 98.5, leaving about a point and a half for the fourth possible Congress, a Republican House alongside a Democratic Senate ^[10]. Those are not prices in any usable sense. They are the width of a thin book being quoted as a forecast.

So use the chambers. Treated as independent, 88% and 51% ^[1] put a Democratic sweep at 44.9% and a Democratic House with a Republican Senate at 43.1% ^[11]. Divided government leads by under two points, which is no distance at all in a complex where both books together have traded at most $13m ^[12]. An allocator writing "split Congress" into a 2026 base case on that spread is recording a rounding decision and calling it a view.

The number that does hold up is narrower and more useful. With a Republican in the White House through Trump's second term ^[5], a Democratic House ends unified control by itself, and that is the thing priced at 88% ^[13]. One veto point, high confidence, no Senate assumption required.

Note also where the conviction sits. The 88% is mostly restating a pattern the source itself flags, that the president's party almost always loses ground at the midterms ^[5], so the House price is not telling allocators much they could not already have assumed. The genuinely open question is the Senate, and that is the chamber carrying the smaller book at both ends of its volume range ^[15]. Nate Silver's model puts Democrats at 57% there, six points above the market ^[4]. Six points on a few million dollars of turnover is not an edge to trade against; it is a measure of how loosely the Senate is priced.

The reason to look at any of this, per cryptobriefing, is Polymarket's 2024 record, when its markets read the presidential race better than many conventional polls ^[14]. That argues for using the prices as a check on your own priors. It does not argue for importing a 43-versus-45 split as a policy assumption.