Side-by-Sides Roll Over. The Rule Written to Stop That Is Gone.

In fatal crashes from 2017 through 2024, 69.8 percent of side-by-side off-highway vehicles rolled over, against 12.7 percent of passenger cars. That’s 986 of the 1,413 side-by-sides the federal fatality census recorded in a deadly crash over those eight years. Nearly seven in ten.

On Father’s Day, a 13-year-old girl in Kingman was riding on the back of a 750cc quad her 16-year-old sister was driving. It hit a large rut and flipped. Both girls wore helmets. The quad pinned both sisters to the ground.

Twelve days later, on July 3, 2026, her family said goodbye and donated her organs. Her mother told Arizona’s Family what she wanted other parents to hear: “I don’t think there’s enough information out there for other families to realize how easily those quads flip.”

She’s describing rollover, the failure that ends more of these crashes than anything else. The federal government studied it for nearly sixteen years and wrote a rule to address it. That rule would never have covered her daughter’s quad, and in the end it covered nothing at all.

What Flips, and Why the Two Words Matter

Two different machines get talked about as one thing, and the difference decides who’s regulated by what.

A quad is the straddle-seat four-wheeler you ride like a motorcycle. A side-by-side has a steering wheel, bucket seats, belts, and a metal cage over your head. The industry calls it a recreational off-highway vehicle, or ROV.

That cage has a name too: a rollover protective structure, a ROPS. You’ll see both terms on a recall notice and almost nowhere else. Riders say “roll cage,” or they just say the thing flipped.

The distinction isn’t vocabulary. It decides what the law reaches. When the government finally wrote a mandatory stability rule, it reached only side-by-sides with a “maximum speed capability greater than 30 mph,” and the preamble describing its scope named the machines it would leave out: “golf cars,” “all-terrain vehicles,” “fun karts,” “go karts,” and “light utility vehicles.”

The rule text lists the same five. Quads were excluded by name, and so were light utility vehicles. Everything that follows about the rule is about one of these two machines, not both.

Both classes roll far more than anything you drive to work. But they don’t fail the same way.

Source: NHTSA Fatality Analysis Reporting System, fatal crashes, 2017 through 2024.

Copy embed code Copied. Paste into your CMS. Free for editorial reuse. Embed includes a do-follow link to the source story.

One control belongs on that table before anybody quotes it. Most side-by-sides in fatal crashes crash alone: 75.6 percent are single-vehicle, against 35.4 percent of passenger cars. Single-vehicle crashes roll more for every kind of vehicle, so some of the gap in that last column is crash type, not machine.

Hold crash type still and compare single-vehicle crashes only, and side-by-sides rolled in 81.4 percent of them against 26.7 percent of passenger cars. That’s about 3.1 times, not 5.5. The narrower number is the honest one, and it’s still the widest gap on the table.

A quad throws you off. A side-by-side goes over with you inside it, which is why the government’s proposed rule ran past stability into what it called occupant protection: alongside the handling requirements, it would have made side-by-sides carry “a passive means, such as a barrier or structure, to limit further the ejection of a belted occupant in the event of a rollover.”

That difference shows up in who dies. Passengers are 29.7 percent of side-by-side occupant deaths and 11.2 percent of quad deaths. A quad is built around a single straddle seat. A side-by-side is built to carry people, and nearly a third of its dead are the people it was carrying.

The Regulator Said It Out Loud

None of this was a surprise to anybody with a federal badge.

Side-by-sides are consumer products. They sit with the Consumer Product Safety Commission, not with the highway agency, which is why the database most people would think to check barely holds them. The federal auto-defect complaint database has taken more than two million consumer reports since 1995. Search it for these machines and almost nothing comes back, because they were never its job.

The Commission started tracking deaths and injuries on the category in 2003, the year a second manufacturer entered the market. In March 2009 it announced a free repair program and a voluntary suspension of sale on three Yamaha Rhino models, adding a rear wheel spacer and removing the rear anti-sway bar to, in the agency’s words, “help reduce the chance of rollover and improve vehicle handling.”

CPSC called it a free repair program and a voluntary stop-sale, not a recall, while filing it under a recall number. The label matters because a repair program is something a company agrees to, and a recall is something it can be ordered into.

Seven months later, on October 28, 2009, the Commission opened a formal rulemaking. It ran its own rollover testing in 2010. By 2011 it was telling the industry its draft voluntary standard didn’t go far enough: “Proposed changes do not address rollovers,” the agency wrote, urging tests of “a vehicle’s stability while actually driving the vehicles” rather than tipping a parked machine on a table.

Its safety-education timeline records a 2013 update in the same vein, saying data confirmed that rollovers and occupant ejection dominate the hazard.

Then the Commission said the same thing where it counts, in the Federal Register, in the opening summary of the 2014 proposal.

The Rule Somebody Actually Wrote

On November 19, 2014, the Commission published a notice of proposed rulemaking under docket CPSC-2009-0087, printed at 79 FR 68964. It’s titled “Safety Standard for Recreational Off-Highway Vehicles (ROVs),” and it opens like this: “The U.S. Consumer Product Safety Commission has determined preliminarily that there may be an unreasonable risk of injury and death associated with recreational off-highway vehicles (ROVs).”

That’s a determination in a rulemaking record, not a line on a safety-education page, and it’s the government’s own words nearly eleven years before it walked away. What followed was a set of mandatory federal requirements the agency summarized as lateral stability, vehicle handling, and occupant protection, plus information requirements. The stability requirements would have specified “a minimum level of rollover resistance.”

Those three areas map onto the three things the crash data shows. Stability is whether it goes over. Handling is whether you can avoid the thing that puts it over. Occupant protection is whether the cage and the belts keep you alive once it does.

The rule was written. It was published. It never took effect.

Ten Years of Not Yet

What stopped it wasn’t the agency changing its mind. It was a line in a spending bill, and Congress kept putting it back.

From fiscal 2017 on it runs under the appropriations bill’s administrative provisions for the Consumer Product Safety Commission. The operative words don’t move. The fiscal 2024 act puts it this way: “During fiscal year 2024, none of the amounts made available by this Act may be used to finalize or implement the Safety Standard for Recreational Off-Highway Vehicles published by the Consumer Product Safety Commission in the Federal Register on November 19, 2014 (79 Fed. Reg. 68964).”

Only the year changes from act to act.

Then it names the condition. Two of them, actually. The National Academy of Sciences has to finish a study, working with the highway safety agency and the Defense Department, on whether the stability and handling requirements are technically sound, how many rollovers the standard would actually prevent, what the vehicle hangtag ratings rest on, and what the rule would do to military use of these machines.

And a report on that study has to reach four named congressional committees.

It first appears in the 2016 act, at Section 629. From fiscal 2017 the same language moves to Section 501 and stays there, in Public Laws 115-31, 115-141, 116-6, 116-93, 116-260, 117-103 and 117-328, unchanged each time.

Fiscal 2024 carried it again. Fiscal 2025 didn’t rewrite it at all: that year’s law is a stopgap that continued the previous year’s appropriations “under the authority and conditions provided in” the 2024 act, so the restriction rode along without anyone having to type it.

The National Academy of Sciences study was never completed. A search of the Academies’ own project and publication indices turns up no such study, and on the Academies’ own pages listing every congressionally directed study under those acts, the off-highway vehicle provision carries no linked project and no linked publication, unlike other mandates in the same statutes that do.

So the one condition attached to finishing the rule was a study, the study never happened, and the bar renewed anyway, year after year, for ten fiscal years running.

Meanwhile the agency kept circling its own file. Staff submitted a package recommending the rulemaking be terminated on November 22, 2016, and on January 25, 2017 the Commission decided not to terminate it. Staff prepared a second termination package in September 2020. On September 22, 2020, the Commission split 2 to 2. “A majority was not reached and no action is being taken,” the agenda records.

So the rule sat there for ten years. Not withdrawn. Not finished. Funded staff, a live docket, and a standing instruction not to cross the finish line.

The Surrender

Four things happened between April 2025 and February 2026, and they read strangely next to each other.

On April 3, 2025, the Commission announced a recall of about 910 Polaris RZR Pro R and Pro R 4 side-by-sides in the United States, model years 2022 through 2025, plus about 10 in Canada. The stated hazard, verbatim from the recall record: “The fasteners on the vehicle’s rollover protective structure (ROPS) pillar joints can be missing or loose and fail to protect the occupants, if the vehicle rolls over or tips over.”

The company had found two vehicles with missing fasteners and one with a fastener that hadn’t been tightened. No injuries were reported. The recall was about the roll cage itself.

On August 20, 2025, the Commission announced it was withdrawing a list of pending rulemakings, and the Standard for Recreational Off-Highway Vehicles was on it.

Both the release and the legal notice identify it only by the 2009 notice that opened the docket, 74 FR 55495. Neither one cites 79 FR 68964, the 2014 proposal that actually wrote the requirements. For the three other rulemakings withdrawn in the same notice, CPSC named the proposed rule. For this one it named only the opening notice, from a docket, CPSC-2009-0087, that carried both.

The release explains the list two ways: proposals “that have remained on the agency’s rulemaking agenda for years without final action,” and “instances where Congress has provided clear direction to the Commission to not finalize specific rules.”

The agency’s Acting Chairman framed the broader move like this: “Regulations that promote unscientific agendas, impose unnecessary costs, and reduce competition are no longer agency priorities. We will not squander limited resources on rules that diminish consumer choice or hand unfair market advantages to foreign competitors at the expense of American consumers and manufacturers.”

The legal instrument followed on September 29, 2025, at 90 FR 46541, signed by the Commission’s Secretary. It withdraws the rulemaking “as of September 29, 2025,” and it describes what that rulemaking had been for: whether there may be “unreasonable risks of injury and death associated with overturning and collisions” involving these vehicles.

The same notice withdrew a second proposed rule on these machines, the Safety Standard for Debris Penetration Hazards, which the notice files under fire and debris-penetration hazards. Two federal rulemakings on this vehicle class ended on the same page.

Then, on February 3, 2026, Congress passed the next year’s appropriations act and wrote the ban back in, at Section 501, in the same words. The rule it forbids the Commission to finalize had stopped existing four months earlier.

The Bill Nobody Can Add Up

Here’s the question a reader should ask next, and here’s why it has no answer.

How many people were hurt in a side-by-side rollover during the ten years the rule sat blocked? Nobody can say, and that includes CPSC. A number can be produced. It just can’t be trusted, and the agency is the one that says so.

Start with where a side-by-side injury even lands. The national injury system is a sample drawn from hospital emergency departments, covering 2015 through 2020 in this analysis, and it has no code for these machines.

The nearest thing is product 5044, and its title in the federal coding manual is “Utility vehicles.” Not recreational. Not off-highway. The phrase “off-highway” doesn’t appear anywhere in that manual. A side-by-side reaches the code only because the manual’s index redirects it: look up “Side-by-side ATVs” and the instruction reads “Use: Utility vehicles.”

Read the redirect again. The coding manual lists a side-by-side under ATVs, then sends it to the utility-vehicle code.

Answering a comment on that very problem, CPSC set out what separates the classes: “ROVs are a subset of UTVs and are distinguished by having a ROPS, seat belts, and a maximum speed above 30 mph.” That is the agency talking in the proposal’s preamble, not language that would have become law. A cage, belts, and a speed. Not one of the three is a field in the injury system.

So the injuries scatter, and CPSC says where: they’re “usually assigned to either an ATV product code (NEISS product codes 3286-3287) or to the utility vehicle category (NEISS product code 5044).” Three codes, none of which names the vehicle.

When staff re-read 71 cases from 2023 that had been coded as an all-terrain vehicle with the number of wheels not specified, the answer came back 46 quads, 21 side-by-sides or utility vehicles, two unknown and two outside the study. Of the 67 they could place, roughly three in ten were the other class of machine entirely.

That check covers 2023, and staff say they can’t generalize it to years before 2022, which is every year in our own window.

The staff conclusion on their own number is blunt: the review “provided strong evidence that the NEISS injury estimate for ROVs and UTVs are a substantial underestimate,” and they say they can’t generalize that correction backward to earlier years. These figures come from a CPSC staff report that carries a stamp on every page saying it has not been reviewed or approved by the Commission.

Then staff did the thing that ends the argument. For 2019 they suppressed the code 5044 injury estimate outright, the one covering side-by-sides and utility vehicles together, because the sample was too small for the figure to be reliable enough to publish.

We reproduced the Commission’s table from the raw sample to be sure the limit was in the data and not in our method. For 2019 we matched its sample of 87 cases exactly. For 2020 we matched both the sample of 140 and the published estimate of 7,500. Same data, read the same way, and it still can’t answer the question.

Here is what the agency’s own national injury number rested on when it proposed the rule. CPSC pulled 2,018 emergency-department cases coded as an ATV or a UTV, completed 668 interviews, and found that sixteen of them involved the vehicle it was writing the rule about.

From those sixteen it estimated about 3,000 side-by-side injuries nationally in 2010 that were initially treated in hospital emergency departments, and published the confidence interval alongside: 1,100 to 4,900.

The width is the finding. A federal agency proposing a mandatory safety standard could not count the injuries it was standing on within a factor of four, and said so in the proposal.

We can compute a five-year code 5044 injury estimate, side-by-sides and utility vehicles together, for 2016 through 2020 out of that same sample. We’re not publishing it. Four of its five annual components sit at or below the sample count CPSC itself withheld a figure for in 2019.

The agency’s own test is a statistical one, a coefficient of variation above 33 percent; ours is the cruder one of counting cases. Neither of us should be publishing a national number off samples that thin.

The three quotations that follow aren’t our reading of the agency. They’re what CPSC wrote in the rulemaking record itself, in the document proposing the standard. “NEISS does not contain a separate product code for ROVs.”

The reason is structural, and CPSC spelled it out: these vehicles “cannot be identified on a consistent basis through the NEISS case records because NEISS requires knowledge of the make/model of the vehicle (which is not coded in the NEISS for any product).”

The system has no field for the thing you would have to know to use it. The agency’s own summary of the result: “The miscoding rate for UTVs and ROVs is high.”

What that decade cost the people in these machines can’t be counted, because the government never built an instrument that could count it. CPSC said as much in 2014, in the same document that proposed the standard.

The fatality side has the same shape. The federal fatality census had no body-type code for a side-by-side in any year from 1975 through 2016. The first one appears in 2017, when it recorded 54 of these vehicles in fatal crashes. Quads had their own code since 1986.

So the Commission determined there may be an unreasonable risk in 2014, and Congress blocked the rule in fiscal 2016, both before the national fatality census could count the vehicle class as its own category.

Once it could count them, here is what it counted. Combined quad and side-by-side occupant deaths on public roads ran 557 in 2020, then between 477 and 540 in every year from 2021 through 2024.

Treat any comparison back across 2016 with care in both directions. That year has no side-by-side code at all, so its 354 is quads plus whatever got filed as unknown, which makes the rise look cleaner than it was and the later quad decline look steeper than it was.

In 2023 side-by-side deaths passed quad deaths for the first time, 241 to 236. They didn’t stay there. In 2024 quads were back on top, 300 to 240. One crossover year is a crossover year, not a trend.

CPSC staff independently report the share of off-highway deaths involving side-by-sides and utility vehicles rising from 20 percent in 2015 to 34 percent in 2020, so the longer direction isn’t in dispute. The most recent year in the census still cuts the other way.

On July 22, 2026, the Commission announced it’s rebuilding that injury system. It described the old one as “designed in 1972 for a paper-record era,” dependent on manual coding of records from roughly 70 of the country’s more than 5,000 emergency departments, and leaving 14 states with no participating hospital at all.

Those limits, it wrote, “restricted the system’s geographic reach, slowed analysis, and reduced the Commission’s ability to identify rare and rapidly emerging product hazards.” The replacement is expected to be fully effective by the beginning of 2027.

That’s the agency conceding the instrument was broken. The rebuilt system is due to be fully effective about fifteen months after the rule it might one day have measured was withdrawn.

Arizona, Right Now

Arizona isn’t an average state for this.

Pooling 2017 through 2024, Arizona recorded 147 quad and side-by-side occupant deaths on public roads across 58.4 million person-years, a rate of 2.518 per million residents per year. That’s the same federal fatality file we used to test what Arizona’s motorcycle records actually show, read the same way. The national rate over the same window is 1.376. Arizona’s rate runs 1.83 times the country’s.

Inside the state, the risk is rural by a wide margin. Of those 147 deaths, Maricopa and Pima counties account for 53, against a combined 44,426,416 person-years over those eight years: 1.193 deaths per million residents per year. The rest of Arizona accounts for 93, against 13,948,528 person-years: 6.667 per million.

That’s 5.6 times higher outside the two big counties, on the same unit. The remaining death is one the federal file records with no county at all, which is its own small comment on the counting.

Two cautions ride with those figures, and they’re not decoration. Arizona logs between 13 and 22 of these deaths in a year, which is far too few to say anything about a trend, so nothing here is a trend. And the national road census only sees public trafficways, so a death in a wash or on a trail or on private desert never enters it at all.

How far off is that? Compare the two federal counts for Arizona over 2019 through 2021. The Commission counted 48 off-highway deaths in the state. The road census counted 51.

Nationally the road census sees only about 56 percent of what the Commission does, but in Arizona it saw slightly more, which happens in only a handful of states. That doesn’t mean Arizona is better counted. It means the two federal systems disagree with each other, state by state, by amounts no one has reconciled.

And Arizona’s own file logged 277 vehicles it could not assign a body type to in those three years. Any of them could have been one of these machines, so 51 is a floor, not a count.

The state has moved on its own. Since January 1, 2025, Arizona State Parks and Trails requires every off-highway vehicle owner to complete a free safety course before registering or renewing a decal.

It’s a 10-minute video and 20 multiple-choice questions, developed with the transportation department and the game and fish department, and at least one owner per vehicle has to finish it between January 1, 2025 and June 1, 2027.

We can’t tell you whether it’s working. The federal fatality file ends in 2024, and the requirement started in 2025. Anyone who tells you today that Arizona’s course did or didn’t change the number is guessing.

A rollover like these can produce more than one kind of case in Arizona: a wrongful-death action brought by surviving family, a claim by an injured survivor, a product claim against a manufacturer, a claim against somebody else who was out there.

Our legal editor Ron DeBrigida walks through what Arizona’s wrongful-death statute says, and Stephanie Ramirez’s guide for families covers the process from the inside.

This story isn’t about either one. But one line in Arizona’s own law is worth reporting here, because it decides how the defense that killed the Georgia case would be handled in this state.

The Georgia court threw out that verdict by treating assumption of risk as a complete bar, decided as a matter of law. Arizona’s rule on that defense is written into the state constitution. Article 18, Section 5 reads in full: “The defense of contributory negligence or of assumption of risk shall, in all cases whatsoever, be a question of fact and shall, at all times, be left to the jury.”

Read what that does and doesn’t do. It decides who answers the question, and the answer is always the jury, never the judge. It says nothing at all about how the jury has to answer. Whether an Arizona court would have come out differently on the Georgia facts is a question neither the constitution nor this story answers.

We also can’t tell you how many Arizonans get hurt on these machines, because the national injury sample carries no state field. There is no Arizona figure in it, and there’s no way to build one. Wisconsin’s natural resources department publishes an off-highway fatality summary and annual crash reports on its own site. Arizona’s off-highway program publishes an annual legislative report on trail spending and grants.

The class the withdrawn rule covered carries passengers, it rolls in seven of ten fatal crashes, and nationally it lives on the roads least built for it: across the United States, 85.8 percent of side-by-sides in fatal crashes were on rural roads, and 71.1 percent were on minor collectors or local roads, the lowest two rungs of the road network, the county lanes and neighborhood streets nobody engineers for speed.

No shoulder, no lighting, and no engineering that ever assumed a vehicle whose entire occupant protection is a roll cage and a belt, with none of the crumple zones, airbags and restraint systems a car brings to the same ditch.

None of it reaches the two sisters on a quad outside Kingman. The rule named their machine and left it out. The injury system that counts both classes still can’t reliably tell them apart.

The rule is withdrawn. The instrument that could measure its cost was never built.