David Becker, executive director of the Center for Election Innovation & Research and a former Justice Department civil rights lawyer, has spent years watching the U.S. Postal Service handle election mail. His verdict on the latest inspector general audit, delivered to ProPublica, was that the agency has a long way to go to meet its core obligations of delivering election mail in a timely manner. He said it while the Supreme Court was already weighing an emergency application from the Trump administration to lift an injunction blocking new USPS mail-voting rules.
The audit’s headline number is reassuring. Nearly 99% of the roughly 6.3 million tracked ballots returned to boards of elections during the 2026 primaries were processed on time, up from about 98% in the 2024 primaries.
The rest of the audit is less reassuring.
The USPS Office of Inspector General report, dated Sept. 4, 2026, found no election-related training materials posted at 22 of 73 delivery units visited — about 30%. In 10 of 82 facilities, management was not aware of the proper way to postmark a ballot. In 13 of 82, staff did not accurately complete the daily all-clear certifications meant to confirm election mail had left the building. At the Harrisburg, Pennsylvania, processing plant, auditors found a hand canceller still set to March 19 in use on the state’s May 19 primary election day, which put the wrong date on at least 56 ballots. At the same plant on the same morning, 108 return ballots turned up inside an automated redirection operation that voter-returned ballots are never supposed to enter, and had to be expedited to Pittsburgh to reach election officials before the 8 p.m. cutoff.
Postmark date is not a technicality. As of June 2026, 14 states, Washington, D.C., and several U.S. territories rely on the postmark to determine whether a mail ballot is eligible to be counted. A wrong date can be the difference between a counted vote and a rejected one.
Consider what that means at street level. A voter who hands a completed ballot across a retail counter on primary day is relying on a rubber date stamp having been reset that morning. In Harrisburg it had not been, and the audit notes that while Postal Service policy forbids putting a second postmark on a mailpiece, there is no policy at all for correcting one an employee has misdated. The voter never finds out.
The 99% figure also captures less than it appears to. A footnote in the report states that the score measures mailpieces from first scan to last scan on processing equipment inside mail processing facilities, and does not reflect final delivery. It certainly does not capture whether a ballot, once delivered, was countable. That distinction is doing a great deal of work in the Postal Service’s public response, which emphasized that overall ballot mail performance is strong.
Strong at the aggregate. Uneven at the facility level. This is the gap the audit actually documents, and it is the gap that matters for what comes next.
What comes next is a fight at the Supreme Court. Executive Order 14399, signed March 31, 2026, directed the Postal Service to write stricter controls for mail-in voting. The agency published a final rule on Aug. 21 that requires states to enroll every mail-ballot voter through a new Federal Ballot Mail Portal, mandates new envelope design and barcode standards, and allows USPS to reject an entire ballot mailing that does not comply. A federal district court in Massachusetts blocked the rule, and the administration asked the Supreme Court to stay that injunction. The Center for Election Innovation & Research then led 100 state and local election officials and local jurisdictions — 81 current and former officials, 19 jurisdictions — in urging the Court to leave the block in place. The American Postal Workers Union, whose members would run the new checks at the counter, filed its own brief opposing the stay on Sept. 9.
The logic of the challenge is straightforward. If USPS cannot reliably train its own staff to date a postmark today, layering on portal enrollment, envelope certification and piece-by-piece verification will not produce compliance. The union’s brief puts numbers on it: roughly 1,300 business mail entry unit clerks nationwide, spread unevenly across the states, and no finished training program for a portal whose specifications clerks have not seen.
Chelsey Wininger, executive director of the Democratic Association of Secretaries of State, told ProPublica that USPS struggles to implement even basic changes to postal process consistently. That framing, from an association with an obvious political interest in the outcome, matches the mechanical finding of the audit, which has no such interest.
Top Postal Service employees, and separately a whistleblower who filed a disclosure with Sen. Richard Blumenthal, have warned that the new system could disenfranchise large numbers of eligible voters. The inspector general has confirmed it received whistleblower allegations about the executive order’s implementation and plans a separate review. That the concern is coming from inside USPS, not just from outside election-integrity groups, is the part of the story that changes its shape.
The structural problem is older than the current executive order, and it is not sabotage. USPS runs election mail through a dedicated headquarters team, but what actually happens on a workroom floor depends on local supervision — and the audit exposes exactly where that chain goes slack. Postal Service headquarters told auditors that posting election training materials at a delivery unit is a best practice rather than a requirement. The same unit’s daily all-clear certification nonetheless asks staff to confirm those materials are posted and shared. Thirty percent of the units visited had none up.
That is what happens when a national responsibility rides on a bulletin board somewhere behind a supervisor’s desk. The Postal Service is not unique in it. Any organization asked to bolt a high-stakes political function onto a routine industrial process will produce the same pattern: strong aggregate performance, embarrassing local failures, and a widening gap between what the numbers show and what the exceptions cost. The ballot problem is that pattern in a category where the consequences are irreversible.
The 99% figure is real. So is the 30% figure. Both describe the same postal service. Which one you cite depends on what you are trying to prove.
The audit does put a floor under the exceptions. Of roughly 1.27 million pieces of election and political mail processed at the facilities auditors visited, they identified 214 delayed pieces, 115 of them ballots. Of those 115, ten did not make it: five were never recovered in time to reach boards of elections before the primary deadline. Against 6.3 million tracked return ballots that is a rounding error. It is not a rounding error if it is your ballot, and the auditors tied the failures to training and supervision rather than to volume — which is to say the rate would not necessarily fall as volume rises in November.
In an expert declaration filed in federal district court on Aug. 28, Tammy Patrick, chief programs officer of the National Association of Election Officials and a former Maricopa County federal compliance officer, told the court that election officials, their service providers and USPS do not have sufficient time or resources to implement the rule before the general election. The audit gives that assessment documentary weight. It is one thing for an outside expert to say USPS cannot do more. It is another for the agency’s own inspector general to document that it is not reliably doing what it is already required to do.
The Court will decide the legal question on its own terms — the scope of executive authority over election administration, the role of the states, the standing of the challengers. The operational question sits underneath, unresolved regardless of how the case is decided. If the rule stays blocked, USPS still needs to close the training gaps the audit found. If the block is lifted, USPS is being asked to enforce a system it has not finished building.
Neither outcome resets the hand cancellers at Harrisburg.
The reassurance embedded in a 99% figure is the kind of number a functioning agency cites when it wants to close a conversation. The audit is the conversation the agency did not want to have. What it describes is a mail system that works well enough at scale to protect its reputation and unevenly enough at the edges to lose real votes, in real states, in elections where the margin sometimes comes down to a few thousand ballots.
A ballot that arrives on time but carries the wrong postmark is a failure the aggregate cannot see. There are more of those failures than the Postal Service’s public response acknowledges, and the agency is about to be asked to do considerably more than it is currently doing well.