No federal limit on how long a train may block a road crossing — 24,582 reports at 6,738 crossings in 2025 and every state limit tested in court struck down

What is happening?

No federal law and no federal regulation says how long a stopped train may block a public road crossing in the United States. Between 2025-01-01 and 2025-12-08 the Federal Railroad Administration logged 24,582 reports of blocked crossings at 6,738 distinct crossings, and in the same publication the agency states that it has no regulatory authority over the subject and that any regulation is state or local. The comparable earlier window, 2024-01-01 to 2024-10-08, carried 21,108 reports at 5,514 crossings.

These are not all brief delays. In calendar year 2022 the reporting portal recorded 8,450 reports of blockages lasting 31 to 60 minutes, 3,444 lasting two to six hours, 460 lasting 12 to 24 hours and 957 lasting more than 24 hours, out of 30,755 reports that year. The portal opened on 2019-12-19 and had taken 71,282 reports covering 57,810 distinct events as of 2023-06-01.

States wrote limits, and the limits do not survive contact with a federal court. The FRA compilation of state law records a blocked-crossing statute in 38 of the 51 state and District of Columbia jurisdictions, typically five, ten or fifteen minutes. In the brief it filed in the Supreme Court in November 2023 the United States stated that every federal court of appeals and state court of last resort to have addressed the question has held such laws preempted, and that no published decision of either kind holds otherwise. The Supreme Court denied review of the most recent such case on 2024-01-08.

Whose problem is this?

The body that would have to write a national limit says it has no authority to write one. The bodies that did write limits have had them voided. Every institution named in this table can describe the problem and none of them can act on it.

Where does this problem end?

The boundary here is not whether trains stop but whether any government may say how long they stop.

What is the state now, and what should it be?

Now

Needs a new measurementthe target state: no federal source opened here names a duration that is too long. FRA escalates a crossing to its safety management teams when it is reported blocked on three separate days within a month, which is a frequency trigger rather than a duration limit. The only number on paper is the ten-minute maximum in H.R. 6790, introduced 2025-12-17 and still at subcommittee referral as of 2026-08-08, and a bill is a proposal rather than a target. With no target in force, the 957 reports of blockages longer than 24 hours in calendar year 2022 breached no federal standard, because there was none to breach.

How big is it?

What the published record supports is a band measured in crossings rather than in people. The floor is 6,738 crossings actually reported blocked between 2025-01-01 and 2025-12-08. The ceiling is about 130,000 public at-grade crossings as of April 2025, roughly 85 percent of which were not grade-separated as of April 2026, every one of which may lawfully be blocked for any length of time.

The distance between those two numbers is not an uncertainty interval. The floor counts crossings where at least one person completed a federal form that tells them, at the moment of filing, that their report will not be forwarded to any railroad, state or local agency. The ceiling counts crossings where the conduct is permitted. FRA stated as of 2023-06-01 that only about 6 percent of crossings in the country had ever appeared in the portal, without saying whether that share is measured against about 204,000 total crossings or about 130,000 public ones, and that figure is two and a half years older than the window the rest of this block uses.

Needs a new measurementthe affected population is not derivable as a count of people. No source opened here publishes, for any single crossing, how many residents, road users, schools or hospitals sit on the far side without a grade-separated alternative. That per-crossing figure is the one missing input, and H.R. 6790 would make it legally operative by barring penalties where no alternate grade-separated route exists within half a mile — the pending bill assumes a computation that no agency currently performs or publishes.

Under what conditions does it arise?

1. Deregulation was enacted as exclusivity. When Congress abolished the Interstate Commerce Commission in 1995 it declared a policy of minimizing the need for federal regulatory control over rail while granting the Surface Transportation Board exclusive jurisdiction over rail transportation, and the concurrent state jurisdiction that existed before was removed. A field can be occupied by a body that has chosen not to act inside it.

2. Trains stop for reasons the federal government already regulates. Track speed classes, air-brake testing, roadway worker protection and hours of service all shape when a train stops and how long it stays stopped, and none of them mentions how long a crossing may be obstructed. The three blockages at issue in the leading federal appellate case, decided 2022-01-11, ran 38, 80 and 37 minutes, and each occurred while a train sat on a side track waiting for another train to pass.

3. Nobody is required to know. No legal requirement obliges a railroad to track blocked crossings, to investigate their causes or to answer FRA. FRA reported to Congress in January 2024 that railroads sometimes decline to respond and cite the absence of a legal requirement as the reason. A duty that no institution holds is not a duty that anyone can be asked to discharge.

4. The chosen instrument is capital, and capital is rationed one project at a time. Congress appropriated USD 3 billion over five years for grade-crossing elimination and authorized a further USD 500 million a year through fiscal 2026, and a separate program sent at least USD 245 million a year to the states as of April 2025. A grade separation costs several million dollars against about 130,000 public at-grade crossings. Every dollar of it is real and none of it changes what is permitted.

5. The remedy that works is a relationship, and relationships do not scale. Near Raymond, Ohio a railroad voluntarily installed air compressors so a train could be split to open a crossing, and in the Houston East End a railroad rewrote crew and dispatcher instructions for crossings the community had designated as critical. FRA reports that reports there fell from over 400 in September 2022 to under 100 two months later, at the price of two federal inspectors stationed in a railroad dispatch center, 33 site visits, 12 neighborhood meetings and 25 field observations.

A remedy renegotiated district by district is a remedy that a community without political weight cannot obtain.

What has been tried?

What was found?

Why is it still unsolved?

Institutional gap — the power to displace state limits and the mandate to set a federal one sit in different hands, and the hand holding displacement has never been asked to decide.

The first movement is that deregulation was drafted as exclusivity. In 1995 Congress declared a policy of minimizing federal regulatory control over rail and, in the same act, gave the Surface Transportation Board exclusive jurisdiction over rail transportation while ending the concurrent state jurisdiction that had come before. The Indiana Supreme Court put the consequence in a line the United States later quoted with approval — when Congress largely deregulated the railroad industry, it did not invite states to step in and fill the void. Exclusivity plus a policy of not regulating produces a void that is legally protected against being filled.

The second movement is that the preemption test rewards regulatory density rather than regulatory coverage. A state rule survives only until federal regulation covers the subject matter, which the Supreme Court reads as substantially subsuming it. The argument the United States made in November 2023 is that track speed classes, air-brake testing, roadway worker protection and hours of service collectively subsume blocked crossings precisely because they are the reasons a train stops and stays stopped, and it added that a nationwide duration rule may be absent because such a rule would be incompatible with that body of regulation. The more thoroughly the federal government regulates the causes, the more completely it displaces state authority over the effect, without ever addressing the effect.

The third movement is that the harm is filed under the wrong subject heading. The mandate of the federal railroad safety law is safety in railroad operations, and the United States openly doubted that a blocked-crossing statute is even related to railroad safety, because the injury runs to emergency responders and the traveling public rather than to railroad operations. That doubt cuts twice. If the harm is not railroad safety it is not the business of FRA, and it is plainly not economic regulation of rail transportation, so it is not the business of the Surface Transportation Board either. No level of government holds both the power to act and the mandate to act.

This is why the failure type is a gap rather than absent enforcement. State limits are not unenforced for want of will or budget; they are void, and 13 jurisdictions are recorded in the FRA compilation as having no such statute at all, Texas among them, the leading state in every reporting window measured. It is not an inverted incentive either, since nobody is paid to leave a crossing blocked. The clearest evidence is the pair of bills now pending: H.R. 6790 would create the federal rule and H.R. 341 would restore the state power, they are mutually exclusive theories of where the authority belongs, and neither has moved past a subcommittee referral. The gap is not for want of a drafted answer.

What observation would mean it is solved?

Candidates — (a) Congress enacts a duration limit, or restores the power of the states to enact one, and it takes effect (b) the annual count of portal reports and of distinct crossings reported falls over several consecutive years (c) the share of public at-grade crossings that are grade-separated rises materially.

(a) alone is weaker than it looks. H.R. 6790 excepts rail yards and sidings, and every one of the three blockages in the leading federal appellate decision occurred while a train sat on a side track waiting for another train to pass — on the face of the bill text, the fact pattern that produced the leading case would fall outside the fix. The alternative route has the mirror weakness: restoring state authority returns nothing at all in the 13 jurisdictions the FRA compilation records as having no such statute, Texas among them.

(b) alone is worse. The portal is voluntary, its own keeper disclaims it as not a representative sample and not suitable for regulatory proposals, and the form tells each reporter that the report will not be forwarded to anyone. A falling count could mean fewer blockages or fewer people willing to file. The count in fact rose from 21,108 across roughly nine months of 2024 to 24,582 across roughly eleven months of 2025 while nothing in the legal position changed. A number that nobody is required to produce cannot be read as a measure of anything.

(c) alone counts concrete, slowly. About 85 percent of crossings were not grade-separated as of April 2026, each separation costs several million dollars, and the federal commitment is USD 3 billion over five years plus USD 500 million a year authorized. Even a fully spent program removes a small share of about 130,000 crossings and leaves the rest exactly as permissive as before. The three have to be read together, and (b) has to be read against who was willing to file that year.

What is it connected to?

Fills with researchfreight train length and siding capacity, the pending control proceeding before the Surface Transportation Board involving the two railroads that lead the report counts, emergency medical response time standards, and school siting and pedestrian routing across rail corridors. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• Nothing explains the 906 investigations. The same figure appears in the edition dated 2024-10-08 and the edition dated 2025-12-08 for two windows that do not overlap. No opened source defines what an investigation is, what it produces, or how many ended in a change at a crossing, and none says whether the number is annual or cumulative. What can be established is that it is identical across two windows, and nothing further.
• No federal source says how long is too long. The escalation trigger counts days, not minutes. Nothing in the federal record names an acceptable duration, so no measurement can fail against it.
• The only national inventory of state law has a proven hole, and nobody flags it. Chapter 3 of the FRA compilation records Oklahoma as having no applicable statute, while the law struck down by the Tenth Circuit in January 2022 was a state statute setting a ten-minute limit. Counting from a list with one demonstrated omission bounds every figure drawn from it in one direction only, this document included — the count of 38 jurisdictions with a statute is a floor and the count of 13 without one is a ceiling.
• Nobody publishes which state limits are dead. The compilation prints the statutes and warns generally that some have been nullified on preemption grounds, without saying which. No opened source maps jurisdiction by jurisdiction which limits survive, which were struck down and which are simply never charged, so a city attorney reading the compilation cannot tell whether the statute in front of them is law.
• The two federal positions are never reconciled. FRA publishes that it has no regulatory authority. The United States argued to the Supreme Court that preemption does not depend on the authority of FRA to promulgate the same requirement, because existing federal rules already subsume the subject. Both statements are on the record and neither document acknowledges the other.
• The one national dataset is disclaimed by its own keeper and no replacement is proposed. FRA states the collection is not designed to be a representative sample and is not suitable for regulatory proposals, and separately that train control data contradicted many public reports. FRA named that data as the verification method and then told Congress it lacks the authority to require it. No opened source estimates what the true incidence would be if measured, or what measuring it would cost.
• Emergency-response delay has never been quantified nationally. FRA concluded in August 2006 that it is impossible to quantify the delays emergency responders experience at blocked crossings and that no aggregate statistics on crossing delay exist. The only quantification found in this round is a 2025 study of one crossing and one station in Canada. The April 2025 review by the Government Accountability Office of the federal crossings program does not contain the word block anywhere in it.
• The two most consequential portal fields are never published as annual rates. The live portal tallies how often first responders were observed unable to cross and how often pedestrians were observed climbing through stopped cars, for whatever window is selected. Neither edition of the fact sheet nor the report to Congress states them for a year. The two numbers that would put the safety case in national terms are computable from federal data and are not stated.

See the evidence

Primary federal documents carry most of this dossier, which is the inverse of the usual pattern here. The brief of the United States, both bill texts, two Federal Register notices, the report to Congress of January 2024, the report of August 2006, the compilation of state law, the Tenth Circuit opinion, the Ohio opinion, the Supreme Court docket and the accountability office review were each opened and read; most of the dot.gov and justice.gov addresses return HTTP 403 to automated fetch and were reached through an archive copy, a text-extraction proxy or a plain full-text endpoint, and each row records which. Two rows could not be opened at all and keep an empty URL cell with the reason. Three disagreements are left visible rather than resolved. FRA publishes that it has no regulatory authority while the United States argued that preemption does not depend on that authority because existing rules already subsume the subject, and no document reconciles the two. The figure of 906 investigations is printed identically for two windows that do not overlap, and nothing establishes whether it is annual, cumulative or carried forward. The compilation of state law records Oklahoma as having no blocked-crossing statute while the Tenth Circuit was striking down exactly such a state statute, so the count of 38 jurisdictions used here is a floor and the count of 13 without one is a ceiling. Secondary sources appear only twice, for the state train-length record and for the endorsement of the de-preemption bill, and neither carries a legal or statistical claim that a primary source above does not already hold. This is a Path A output (research-based definition), so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range Not derivable

The reason and what is missing are listed under “What is missing” below