Work Zone Safety Statistics: What the Federal Data Actually Says

Almost every traffic control company has a version of the same safety conversation: someone quotes a statistic in a toolbox talk, nobody knows where it came from, and the number quietly mutates each time it is repeated. Meanwhile the actual federal data is public, free, and more useful than the folklore.

This is a careful read of what the work zone crash and fatality data actually says as of mid-2026 — including where the published sources disagree with each other, which matters more than most safety briefings admit.

One framing note before the numbers. Statistics about work zones are overwhelmingly statistics about road users, not about workers. Both matter. But if you run crews, you need to know which population a given number describes before you decide what it means for your operation.

The headline numbers

The most commonly cited national series comes from the Fatality Analysis Reporting System, compiled and published in accessible tables by the Work Zone Safety Information Clearinghouse. Their figures record 850 work zone fatalities across 763 fatal crashes in 2024, following 905 fatalities across 824 fatal crashes in 2023.

That is a real year-over-year improvement — roughly a six percent decline in deaths. It is also still more than two people killed per day in American work zones.

The Clearinghouse tables break the totals down in ways that are directly relevant to how you plan a setup. In 2024, trucks were involved in 233 of the fatal work zone crashes, accounting for 282 of the deaths, compared with 252 crashes and 301 deaths in 2023. Heavy vehicles are wildly overrepresented relative to their share of traffic, which is the empirical case for buffer space and for positive protection on high-truck corridors.

The same source separates pedestrian deaths into those killed while at work and everyone else. In 2024 there were 37 "at work" pedestrian fatalities and 119 other pedestrian-involved fatalities, against 33 and 145 respectively in 2023 — a split worth internalizing, because the second category is largely people walking through or around a work zone that did not adequately accommodate them.

Where the sources disagree, and why that is fine

Here is the part that trips people up. FHWA's own work zone facts and statistics page reports 891 work zone fatalities in 2022 and 963 in 2021, with 821 and 880 fatal crashes respectively. Other published tables give a somewhat different figure for 2022.

Both are drawn from the same federal crash reporting system. They differ because of revision vintages, extract dates, and small definitional differences in what gets flagged as work zone related. This is normal for federal crash data and is not evidence that anyone is wrong.

The practical rule: cite one source per number, name it inline, and never average two sources together to produce a third number that appears nowhere. A blended figure is a fabricated figure, even when both inputs are real. If you are writing a bid narrative or a safety program document, the auditor's question is never "is your number impressive" — it is "where did this come from."

What kind of crashes these are

Aggregate death counts do not tell you what to change. The crash-type breakdown does.

For 2022, FHWA reports speeding as a factor in 281 of 821 work zone fatal crashes, rear-end collisions in 174, and commercial motor vehicles involved in 246. Those are roughly 34 percent, 21 percent, and 30 percent respectively.

Read those three together and a picture emerges that most field supervisors will recognize immediately. The dominant failure mode is a driver arriving too fast at a queue they did not expect. Rear-end crashes at one in five of all fatal work zone crashes is the signature of inadequate advance warning, insufficient sight distance to the back of a queue, or a taper that appears without enough lead time to react.

This is where the advance warning area earns its keep. It is also why queue length is an operational variable and not just a traffic engineering abstraction — if the queue grows past the point where approaching drivers can see it in time, the geometry of the setup has effectively changed even though nobody moved a cone.

Exposure-adjusted numbers, which almost nobody quotes

Raw counts rise and fall with how much construction is happening. Two figures on the same FHWA page normalize for that, and they are far more useful for arguing about whether the industry is actually getting safer.

FHWA publishes an estimate of approximately one work zone fatality per 4 billion vehicle-miles traveled, and one work zone fatality per $112 million in roadway construction expenditure.

The second one is the number to bring to a budget conversation. It converts work zone risk into the same units the rest of a highway program is measured in. When a agency doubles its construction spend, the exposure-adjusted framing says something specific about what that implies for expected harm — and therefore about what the traffic control line item is actually buying.

The worker-specific numbers

Workers are a minority of work zone deaths and the population your company is directly responsible for.

FHWA, drawing on the Bureau of Labor Statistics Census of Fatal Occupational Injuries, records 94 highway construction worker occupational fatalities in 2022 and 108 in 2021. The underlying occupational fatality program publishes its own tables through BLS.

Roughly a hundred deaths a year is a small number against 891, and an enormous number against the size of the workforce that incurs it. A driver's exposure to a given work zone is measured in seconds. A flagger's is measured in shifts. That asymmetry is the entire argument for why worker-facing controls — positive protection, buffer space, high-visibility apparel, the discipline of never standing in the traffic space — are not interchangeable with driver-facing controls like signing.

What the industry says about the trend

The trade association for this work is the American Traffic Safety Services Association. Commenting on the 2024 decline, ATSSA reported the drop from 905 deaths to 850, with driver and passenger deaths falling from 709 to 673 and pedestrian deaths from 196 to 177.

ATSSA President and CEO Stacy Tetschner framed the improvement without letting anyone rest on it, saying: "While the recent decline in work zone deaths is encouraging, we will never be satisfied until we reach our goal of zero deaths in roadway work zone crashes."

That is the correct posture for a number that is improving from an unacceptable baseline. A six percent decline in a year is real progress and is also entirely compatible with your specific crew having a very bad month.

What the regulatory floor requires

Statistics motivate; standards bind. The federal floor for how work zones are set up runs through two documents that reference each other.

The national manual is Part 6 of the Manual on Uniform Traffic Control Devices, covering temporary traffic control. The current national edition is the MUTCD 11th Edition, published December 2023 and now carrying Revision 1 dated December 2025. Section numbering shifted meaningfully from the long-familiar 2009 edition, so any internal template still citing the old numbers is citing sections that have moved.

On the occupational side, OSHA's construction standards incorporate that manual directly. The standard at 29 CFR 1926.201(a) states that "Signaling by flaggers and the use of flaggers, including warning garments worn by flaggers, shall conform to Part 6 of the MUTCD."

That sentence is worth reading twice, because of what it does structurally. OSHA does not maintain a separate flagging rulebook. It adopts the traffic engineering manual by reference and then enforces it as an occupational safety requirement. The same physical setup is simultaneously a traffic control question and a worker safety question, judged against the same text by two different authorities.

The broader device requirements sit in 29 CFR Part 1926 Subpart G, and OSHA maintains a highway work zones topic hub that collects its guidance in one place.

Standing caveat, and it is not boilerplate: requirements vary by state, county, and city. States adopt their own MUTCD versions, and local agencies layer permit conditions on top. Nothing here substitutes for the requirements of the agency having jurisdiction over your specific roadway. Verify with them, every time.

Turning statistics into something operational

National data cannot tell you whether your setups are safe. It can only tell you what typically goes wrong. Closing that gap requires evidence from your own jobs, which is where most traffic control companies are genuinely blind.

The usual state of affairs: the approved plan lives in a PDF, the actual setup lives in a foreman's memory, and the only durable record is a paper ticket that may or may not survive the truck. When something goes wrong six months later, there is nothing to reconstruct.

Three things are worth capturing on every job, because each maps directly onto a failure mode in the data above.

Where the devices actually went. Not the plan — the placement. A GPS-stamped record of setup, taken at setup time, is the difference between asserting that advance warning was adequate and being able to show it. Given that rear-end crashes are one in five fatal work zone crashes, advance warning placement is the single highest-value thing to document.

When the setup changed. Work zones are not static. Phases shift, lanes reopen, night configurations differ from day. A record that captures only the initial setup describes a work zone that existed for one hour of a two-week job.

Who was on site and when. Not for surveillance — for reconstruction. When a claim surfaces, the question is always who was where, and memory is a poor witness.

This is the reasoning behind how Traffic OS handles daily tickets: signed in the field, GPS-stamped, timestamped, and attached to the job rather than to a truck. The same record that defends you in a dispute is the record that lets you audit your own setups before anyone disputes anything.

It is also why the tooling matters more than it looks like it should. A system nobody uses in the field produces no evidence. Crews document reliably when documenting is faster than not documenting — which is a product design problem before it is a compliance problem.

What this data does not capture

Every number above counts a death. That is a narrow slice of what actually happens in work zones, and the omissions systematically distort how the industry reasons about risk.

Injuries are far less visible than fatalities. Fatal crash reporting is close to a census — a death generates a police report, a medical examiner record, and a federal data submission. Injury crashes are reported inconsistently across jurisdictions, and minor injuries frequently are not reported at all. The result is that the best-quality national work zone data describes the rarest outcome. Any ratio you have heard about injuries per fatality in work zones should be treated with suspicion unless it names its source.

Near misses are not counted anywhere. A vehicle entering the work space at fifty miles an hour and missing a flagger by four feet produces no record in any federal dataset. It is the single most informative event that can happen on a job — a direct measurement of how close the setup is running to failure — and it is invisible to every statistic in this article. The only place a near miss can be captured is your own reporting, which means the quality of your internal near-miss culture determines whether you have any leading indicator at all. Everything federal is a lagging indicator by construction.

Property-damage-only crashes vanish. A struck attenuator or a flattened array of channelizing devices is a control that did its job, and also evidence that something upstream went wrong. Those events cost real money and predict worse ones, and they appear in no national series.

Attribution to a specific setup is nearly impossible. Federal data tells you a crash occurred in a work zone. It rarely tells you whether the taper was the right length, whether the advance warning signs were where the plan put them, or whether the device that should have been there had been knocked down two hours earlier and not replaced. That level of causal detail exists only in whatever the contractor documented at the time — which, for most companies, is nothing.

Put those four together and the honest summary is this: national statistics are excellent for understanding the shape of work zone risk and nearly useless for managing risk on a specific job. The shape tells you to invest in advance warning, buffer space, and heavy-vehicle protection. Managing your own risk requires your own evidence.

How to pull these numbers yourself

Do not take secondhand figures, including the ones in this article, into a bid document or a safety program without checking them at the source. Both primary sources are free and require no account.

The Work Zone Safety Information Clearinghouse data pages publish year-by-year tables of fatal crashes and fatalities with the truck-involved, bus-involved, and pedestrian breakdowns. This is the fastest route to a defensible number, and the tables state which underlying data release they reflect.

FHWA's work zone facts and statistics page carries the crash-type breakdowns, the worker fatality figures, and the exposure-adjusted rates. It is the better source when you need to argue about causes rather than counts.

Two habits make the difference between a number that survives scrutiny and one that does not. Record the access date — these pages update as federal data is revised, and a figure that was correct in March may be superseded by autumn. Quote the year explicitly and never round. "Roughly 900 work zone deaths a year" invites a follow-up question you cannot answer. "850 work zone fatalities in 2024, per the Work Zone Safety Information Clearinghouse" ends the conversation.

If a figure you want does not exist in either source, the correct move is to not use it. The temptation to reach for a competitor's blog post or a vendor white paper for a more striking number is exactly how unsourced statistics enter circulation and then become industry folklore.

What to do with this

A few concrete takeaways.

Fix your safety briefing sources. If your toolbox talks cite numbers with no attribution, replace them with linked federal figures. The Clearinghouse tables and the FHWA facts page are both free, public, and updated. A crew that knows the number came from FARS treats it differently than a crew that knows it came from a supervisor's memory.

Weight your controls toward the actual failure modes. Speeding, rear-end collisions, and heavy vehicles dominate the fatal crash data. Advance warning distance, queue visibility, and buffer space are where the leverage is.

Separate the two populations in your own thinking. Controls that protect drivers and controls that protect workers are not the same controls, and a program that is strong on one can be weak on the other without anyone noticing.

Document setups as they are, not as they were planned. The gap between the approved plan and the field reality is where both the safety risk and the payment risk live.

If you want to see how the ticketing and dispatch side of that works in practice, the pricing page lays out what is included at each tier, and you can book a walkthrough to see the field documentation flow on real screens rather than in screenshots.

None of this makes a work zone safe. The cones do not care what software you run. But the difference between a company that improves and one that repeats the same near-miss for five years is almost always whether anyone can see the pattern — and you cannot see a pattern you never recorded.