Nate Silver's Silver Bulletin put its generic congressional ballot average at D+6.2 on 20 June, down from the D+6.9 it reached on 28 May1. The generic ballot, the poll asking which party a voter would back for the US House, no longer agrees with itself across pollsters. Institutional pollsters tell a quieter story: NBC, Marquette and Ipsos cluster at D+3 to D+5, with Reuters/Ipsos at D+3, while Emerson sits out at D+112. The readings disagree by eight points on the same question in the same fortnight.
That spread is what campaigns have to work with. Harry Enten of CNN put the stakes plainly on 17 June: Democrats need roughly D+3 to D+4 to overcome the Republican redistricting edge and win the House, which leaves them, on the institutional numbers, "right on the border"3. Aggregators place them in comfortable wave water; the high-quality polls place them at the waterline. The gap is a fight about who turns out, decided by how each pollster screens for likely voters and weights party identification. Midterm electorates are smaller and more partisan than presidential ones, and the eight points reflect competing bets on whether 2026 turnout looks like a base-mobilisation year or a broad-anger year.
Intra-party defection supplies the evidence for the pessimistic read. Donald Trump's economy approval hit a record-low 33% in a Marist Poll fielded 8 to 11 June and published over 18 to 20 June, his worst since Marist began asking in 20194. The figure inside his own coalition matters most: 22% of Republicans disapprove of his handling of the economy5, the home-team erosion that has preceded past waves. A Republican strategist would counter that headline approval near 36% still holds in an electorate that votes its tribe regardless of grocery bills.
