Shortly before 4 a.m. Eastern on Wednesday, the Mississippi count had Scott Colom ahead of Cindy Hyde-Smith by 43 votes: 462,821 to 462,778, each at 48.9 percent, with about 96 percent of the expected vote counted (HuffPost, 4 November 2026; Dark Horse Press, 4 November 2026). Counties would resume counting that morning, and Mississippi gives five business days for absentee and affidavit ballots to arrive or to be cured.
There is nothing to ask for after that. Mississippi provides no automatic recount and no requested recount; both answers are no, in the state’s own law (Ballotpedia, recount laws in Mississippi). More striking, as Derek Muller pointed out within hours of the count closing, state law supplies no mechanism even for a contested Senate election. The election code addresses county-wide contests and legislative races, and for a Senate seat the forum it leaves is the Senate itself, which under Article I, Section 5 judges the elections of its own members (Election Law Blog, 4 November 2026). Muller expects the remaining ballots to moot the question by Friday. They may well. But the shape of the thing is visible either way: a 43-vote margin in a state that has deliberately written no recount into its law, deciding whether a majority reaches 55.
That is the strangest shape of this election. The wave was not close, and the last two seats are not being decided by voters. They are being decided by a calendar and by an absence in the statute book.
The floor moved
The Senate is at 54 to 44 with two seats outstanding (Electoral-vote.com, 4 November 2026), and the majority was settled before Alaska’s polls closed at all (AP). What matters is not the size of the win but the places it happened, because several of them are not places where a Democratic win should be a rounding error away from a loss.
James Talarico led Ken Paxton 58 to 40 in Texas, with 77 percent counted and the race called. Jon Husted, the Republican incumbent in Ohio’s special election, lost to Sherrod Brown 49 to 46 with 89 percent counted. Troy Jackson unseated Susan Collins in Maine, 52 to 48, at 67 percent. Roy Cooper took North Carolina 59 to 39. Josh Turek won Iowa 51 to 46. Annie Andrews won South Carolina 51 to 47. Abdul El-Sayed won Michigan 52 to 46, and Jon Ossoff held Georgia 59 to 41 (Electoral-vote.com).
Those are not the margins of an election decided by turnout in the last few precincts. An eighteen-point win in Texas and a twenty-point win in North Carolina are what a map looks like when the structure underneath it has shifted, not when a campaign found a few thousand votes it did not expect.
The governorships and the House say the same thing in a different register. Gina Hinojosa defeated Greg Abbott in Texas, Amy Acton won Ohio, and Rob Sand won Iowa (Ballotpedia, 2026 gubernatorial elections). NBC and The New York Times both projected a House of 258 Democrats to 177 Republicans (Electoral-vote.com).
The control group
There is one result that keeps this from being a story about incumbency in general, and it is in Nebraska.
Pete Ricketts beat Dan Osborn 50.4 to 43.7 with 99 percent of precincts reporting (watch.vote, read 9:25 a.m. ET, 4 November 2026). Osborn ran as an independent, deliberately, on the theory that the Democratic label was the liability rather than the policies. That theory did not survive the night. A seven-point loss, in a state where Osborn’s whole theory was that the party label was the liability, is the evidence that voters were not punishing officeholders as such. They were punishing one party, and in Nebraska, where no Democrat was on the ballot to punish, the incumbent walked through.
The distinction is the whole argument. A wave that removes incumbents is a mood, and moods revert. A wave that removes a party is a realignment, and realignments outlive the candidates who happen to be standing when they arrive. Ricketts winning by seven is what separates the second from the first, and it is the single most important number in this election that nobody will put in a headline.
Counting is a policy choice
Which brings the two undecided seats back into focus, because their delay is not a mystery. It is a rule that someone wrote down.
Alaska accepts ballots postmarked by Election Day for fifteen days after the election from overseas voters, which puts the deadline for final receipt at 18 November, and ranked-choice tabulation begins only after that (AP, 4 November 2026). Both Dan Sullivan and Mary Peltola finished the first count below the fifty percent threshold, so the second choices decide it. That is a defensible system, and it is slow by design. Peltola said as much on election night: “the results are not going to be fast.”
The Alaska count also carries a second, uglier weight. A third candidate, a retired teacher from Petersburg named Daniel J. Sullivan Jr. on the ballot, ran with no party affiliation and took 2.5 percent in the August open primary. National Republicans argued he was a sham candidate recruited to confuse voters, and the President ordered the Justice Department to sue to block certification if the incumbent lost. A federal judge dismissed the suit less than two weeks before the election, and the department appealed. It has since confirmed a criminal investigation, without saying what crimes or whom it concerns. The senator called the affair “the biggest election cheating scandal in the history of the state of Alaska” (AP).
So one of the last two seats will be settled by ranked-choice arithmetic in Anchorage while a federal appeal and a criminal probe run alongside it, and the other will be settled by 43 votes in a state where the law offers no recount at all. Neither outcome depends on persuasion. Both depend on procedure, and procedure was written years ago by people who did not imagine it would ever decide a majority.
What would make it stick
The temptation now is to call this a realignment on the strength of one night, and the honest answer is that one night cannot tell you. What one night can tell you is where the floor sits, and where it sits is not where it sat on Monday: an eighteen-point margin in Texas, a twenty-point margin in North Carolina, governorships changing hands in Texas, Ohio and Iowa, and Nebraska as the exception that proves the rule about which party the voters were punishing.
The test is not 2026. The test is whether the same seats are competitive in 2028 when the wave is not there, because a realignment is a floor, and a floor is only a floor if it holds in a bad year. Mississippi being separable by 43 votes is a curiosity. Texas being won by eighteen points is a data point. Texas being won by eight points in 2028, with a Republican at the top of the ticket who is not this one, would be the thing that makes tonight structural rather than meteorological.
Until then the correct posture is the one the count itself recommends: patience, and an insistence on counting every ballot from every far-flung community, because the two seats still outstanding are exactly the ones where the mechanics matter more than the message.
PRH | huffmanwrites.org | © Philip Huffman