The morning after the 2026 midterm elections, the most important word in American politics is not a party name or a slogan. It is a status: uncalled. A race that is uncalled has not been declared by a news organisation’s decision desk, and that is not the same as a race that is disputed, tied, or in doubt about its legitimacy. In most cases, it simply means that the count is not yet complete enough, or the margin is not yet wide enough, for a responsible call to be made. Understanding that distinction is the key to reading the days and weeks after Election Day without mistaking a slow count for a broken one.
Election night is only the first count. Ballots cast in person are often reported first, while mail ballots, provisional ballots, military and overseas ballots, and ballots that need a signature check or another verification step can take longer. Some states allow mail ballots postmarked by Election Day to arrive days later. Others cannot begin processing certain ballots until Election Day itself. Add ranked-choice tabulation in some states, a possible runoff where no candidate reaches the required threshold, and the ordinary work of reconciling precinct returns, and it becomes clear why a national map can look unfinished on Wednesday morning even when voting itself went smoothly on Tuesday.
That unfinished look also has a history. In the 2022 midterms, control of the House was not called by the Associated Press until November 16, eight days after Election Day, and in 2024 it again took more than a week. The reason in both years was concentrated in the West, where close seats and mail-ballot rules stretched the timeline. The lesson for 2026 is not that any particular state is unusually slow by design, but that a close national contest will almost always be decided last in the places where the law gives counting the most time.
Why a Race Stays Uncalled After Election Night
News organisations do not call a race because a candidate is ahead. They call it when they judge that the trailing candidate has no realistic path to overtake the leader, given the votes still to be counted and where those votes are likely to come from. That judgment weighs the reported margin, the share of precincts reporting, the estimated number and location of outstanding ballots, and past voting patterns in those areas. A lead of several thousand votes can therefore still be too close to call if tens of thousands of ballots remain uncounted in areas that favour the trailing candidate. Conversely, a smaller lead can be callable if almost all ballots are in and the remaining ballots cannot plausibly change the result.
Five kinds of races most often stay uncalled. First are races with late-arriving mail ballots, where the law itself keeps the count open after Election Day. California is the best-known example: mail ballots postmarked by November 3 can arrive through November 10, and counties have until December 3 to certify, with statewide certification scheduled for December 11. Second are races in states that do not process mail ballots until Election Day or that face a large number of provisional ballots, each of which must be checked for eligibility before it is counted. Third are ranked-choice races, including Alaska’s Senate contest, where a further tabulation may be needed if no candidate wins a majority of first-choice votes. Fourth are races where the margin is so narrow that even a completed count may trigger a recount. Fifth are races in which reporting itself is uneven, with some counties or precincts reporting much later than others, leaving the statewide or district picture incomplete for hours or days.
Early on November 4, the Associated Press’s own results table illustrated that pattern. Most Senate races were listed as called, while Alaska and Mississippi were listed as too early to call. That snapshot does not mean those two races were the only ones that would ever be watched closely, and it does not mean their eventual winners were in serious doubt in the same way. Alaska’s delay reflects, in part, a counting calendar that includes a final regional count of absentee and questioned ballots scheduled for November 18 and a certification target of November 25. Mississippi’s position in the table reflected a count that was still being completed. Different outlets, updating at different hours, can also show different tallies, which is why readers should always check the time stamp on any live results page and treat a number without a time stamp as incomplete information.
The House picture is usually slower than the Senate picture for a structural reason: there are 435 districts, many of them in states with long mail-ballot windows, and a national majority requires 218 seats. A handful of close districts in California, Arizona, Nevada, and other western states can therefore hold up a chamber-wide call even when most of the country is settled. In 2026, as in previous cycles, the question to ask about the House is not only how many seats each party has won so far, but how many seats remain uncalled, where they are, and whether the outstanding ballots in those districts could plausibly move the total across the majority threshold.
Canvass, Certification and the Difference From a Projection
A projection and a certification are different things, and confusing them is the most common source of misunderstanding after an election. A projection, or race call, is a news organisation’s judgment based on partial returns. Certification is the official act by which election officials attest that the counting and canvassing are complete and accurate. No state certifies a winner on election night. Instead, local officials first canvass the results: they reconcile ballots cast with ballots counted, decide provisional ballots, allow voters to cure signature problems where the law permits, and audit the count. State officials then certify the official result.
The calendar for that work varies widely. Compilations based on state law and the National Conference of State Legislatures show states certifying between November 10 and December 18, with several states setting no single fixed date. California’s county certification deadline of December 3 and statewide date of December 11 sit near the later end of that range. Texas illustrates the two-step structure clearly: local canvasses must be held no earlier than November 6 and no later than November 17, while the state canvass must fall between November 21 and December 7. Florida’s state canvass falls on the 14th day after the election. Other states finish much faster. The spread is a feature of a decentralised system, not evidence that one state’s count is more trustworthy than another’s.
Any statement about which party controls a chamber before certification must therefore be framed as projected, not certified. A projected majority can be a sound and widely shared judgment once enough races have been called, and it matters for planning in Washington, because leadership elections, committee assignments, and the legislative calendar all begin to take shape. But a projected majority is still subject to the remaining calls, to any recount that changes a result, and to any runoff that has not yet been held. The new Congress itself does not begin until noon Eastern Time on January 3, 2027, under the Twentieth Amendment, which gives the counting and certification process time to finish before members take their seats.
Recount Triggers: Automatic, Requested and Court-Ordered
A recount does not begin on election night. It generally comes after the canvass, once officials know the certified or near-certified margin. From there, the rules split into three broad paths, and the path available in a given race depends on the state and sometimes on the office.
The first path is an automatic recount. According to compilations from the National Conference of State Legislatures, 26 states and Washington, D.C., trigger a recount automatically when the margin falls below a set threshold, and the most common threshold is 0.5 percent. Florida orders a machine recount at 0.5 percent or less and a manual review of overvotes and undervotes at 0.25 percent or less. Pennsylvania uses 0.5 percent or less for statewide races. Arizona uses 0.5 percent or less between the top two candidates. Ohio uses 0.25 percent for statewide offices, and Oregon uses 0.2 percent. Michigan uses a vote-count test rather than a percentage for statewide races, at 2,000 votes or fewer. Washington requires both tests to be met for a machine recount: under 2,000 votes and under 0.5 percent. Connecticut’s automatic recanvass applies at under 0.5 percent, capped at 2,000 votes, or under 20 votes.
The second path is a requested recount. In states without an automatic trigger, or in races where the margin is outside the automatic range but still close, a candidate may be able to request a recount, often within a short deadline after the canvass and sometimes at the candidate’s expense. North Carolina, for example, permits a requested recount for statewide races when the margin is 0.5 percent or 10,000 votes or less, whichever is smaller, and 1 percent or less for other races. Wisconsin permits a requested recount when the margin is less than 1 percent for larger races, with different rules for very small races. California permits a requested recount with no margin requirement, but the requester generally pays. Texas is stricter: an automatic recount is available only in the rare case of a tie, while a losing candidate may petition for an initial recount if the margin is less than 10 percent of the winner’s vote, if counting errors are certified, or if fewer than 1,000 votes were cast for the office. Deadlines are tight, often measured in business days after the canvass, so a campaign that waits too long can lose the right to ask.
The third path is a court-ordered recount or contest. In some states, including Mississippi, New York, and Tennessee, there is no automatic recount, and the route to a further count runs through a legal contest after certification. A judge can order a recount as part of that process, but the standard is higher than a close margin alone, and the timetable is set by the court and by state contest law. That distinction matters for public debate: the absence of an automatic recount in a state does not mean a close result cannot be examined, but it does mean the examination will look different, and will usually take longer, than a routine machine recount.
What a recount actually does also varies. A machine recount re-tallies ballots through tabulation equipment. A manual recount, or hand-to-eye recount in some states, examines ballots by hand. Some states sample precincts first, and only expand the recount if the sample shows a significant change. A recount can change vote totals without changing the winner, and it can, in rare cases, change the winner. When it changes the result, the petitioner is often refunded any fee. When it does not, the fee is usually not refunded. Those details are why campaigns and election lawyers read the state-specific rulebook before they announce any plan, and why national commentary that assumes one national recount rule is almost always wrong.
Runoffs, Ranked Choice and Other Reasons a Seat Can Wait
Two further mechanisms can keep a seat undecided well beyond the canvass. The first is a runoff. In Georgia, a candidate in an applicable election must win a majority to win outright; if no candidate reaches that threshold on Election Day, the top two advance to a runoff. Pre-election calendars listed December 1, 2026, as a runoff date, with a later certification deadline for the runoff result. Mississippi was also listed in some pre-election calendars with a December 1 date. Whether any particular 2026 race actually goes to a runoff depends on the certified first-round result, so readers should not assume a runoff until officials confirm that no candidate reached the required threshold.
The second mechanism is ranked-choice voting. Alaska uses ranked-choice voting among the top finishers from its primary, and Maine uses ranked-choice voting for federal elections. If no candidate wins a majority of first-choice votes, votes are redistributed from eliminated candidates according to voters’ later preferences. That tabulation takes time, especially when it must wait for late-arriving ballots to be included, and it can produce a winner who did not lead on first choices. It is not a recount, and it is not a sign of a problem. It is the voting system voters in that state chose, working through its own steps.
Other delays are more mundane but just as real. Ballots with missing or mismatched signatures may be cured by voters within a state deadline. Military and overseas ballots, which travel further, have their own receipt windows. Damaged ballots may need to be duplicated so they can be read by tabulators. Audits, including risk-limiting audits in some states, check a sample of ballots against the reported result before certification. Each step protects accuracy, and each step takes time, which is why officials in many states warn before Election Day that close races will not be called that night.
What to Watch Next, in Order
For readers following the remaining races, the most useful sequence is chronological, not partisan. First, watch the outstanding-ballot updates from county election offices in the uncalled districts, and note whether the number of ballots left is large enough to matter. Second, watch for the end of late-mail receipt windows, including November 10 in California and comparable dates elsewhere, because after that date the pool of countable ballots is essentially fixed, apart from provisional and cured ballots still being resolved. Third, watch county and state canvass dates, because recount deadlines run from the canvass, not from Election Day. Fourth, watch for any automatic-recount trigger being met once a margin is known, and for any requested recount filed within its deadline. Fifth, watch for runoff confirmation and, where applicable, the runoff campaign and result. Sixth, watch certification itself, which is the point at which a projected winner becomes the official winner.
Two cautions apply throughout. First, a candidate’s declaration of victory, or a concession, is a political statement, not an official result. A candidate may declare victory while a race remains uncalled, and a trailing candidate may concede while counting continues, but neither act changes the count. Second, a change in the reported leader as later ballots are counted is normal in states where mail ballots are counted after Election Day ballots, and it is not by itself evidence of irregularity. What matters is whether the ballots being counted were cast lawfully, received by the deadline, and verified under the state’s rules. That is the question canvass and audit procedures are designed to answer.
For Washington, the practical consequence of a slow finish is a compressed transition for the incoming Congress. Leadership contests, committee assignments, and staff hiring all begin before the final seats are certified, and they proceed on the basis of projections. That is normal, and it has happened in every recent cycle in which the House took more than a week to call. The difference this time is only the scale of attention: with both chambers projected to change hands, according to early decision-desk projections, every remaining race is read as a clue to the size of the majority, not just to its existence. Size matters for governing, because a narrow majority constrains what a Speaker or a Majority Leader can pass, while a wider majority gives more room for defections. But size, like control itself, remains projected until the last relevant race is called and certified, and the final margin may differ from the early estimate by several seats.
Until then, the most reliable habit is also the simplest: follow the official county and state results, follow the decision desks that explain their calls, and distinguish between a race that is uncalled, a race that is in recount, a race that is headed to a runoff, and a race that is certified. Those are four different stages, with four different timetables, and keeping them separate is the best defence against both premature celebration and premature doubt.
Sources
USN reviewed reporting from multiple outlets for this story. Read more from Associated Press, Brennan Center for Justice and Texas Secretary of State.