A conventional four-leg intersection with single-lane approaches contains 32 points where vehicle paths cross, merge or diverge. The same four legs arranged as a roundabout contain 8. FHWA’s informational guide puts the figure in exactly those terms, and for a three-leg intersection the drop is smaller and less impressive, from nine conflict points to six.
The count alone would not matter if all conflicts were alike. They are not, and the guide sorts them by severity. Queuing conflicts are the least severe, because the collisions involve the most protected parts of the vehicle and the relative speed difference is small. Merge and diverge conflicts sit in the middle. Crossing conflicts are the most severe of all and the most likely to involve injuries or fatalities, and their typical crash types are right-angle and head-on collisions.
A roundabout removes the crossing conflicts by converting every movement into a right turn. That is the mechanism. Everything else in the safety case is downstream of it, and the guide draws the conclusion that follows: “the ability of single-lane roundabouts to reduce conflicts through physical, geometric features has been demonstrated to be more effective than the reliance on driver obedience of traffic control devices.” A red light works when it is obeyed. Geometry works regardless.
What the published reductions measure
FHWA’s proven safety countermeasure sheet on roundabouts gives two headline figures: 82 percent for converting a two-way stop-controlled intersection, and 78 percent for converting a signalized one. Both are labelled as reductions in severe crashes rather than in all crashes, and a single credit line on the statistics rail attributes both figures to the Highway Safety Manual; the sheet’s two “Source: FHWA” lines sit beneath the roundabout photographs instead and credit the images, not the percentages. Its own summary of the benefit names severe crashes specifically, and it says the treatment works in urban and rural areas across a wide range of traffic conditions.
The older informational guide is where the underlying evidence and its limits are visible. Its American sample was eleven intersections converted to roundabouts. Across all eleven, total crashes fell 37 percent and injury crashes fell 51 percent. Split by size, the eight small and moderate roundabouts showed a 73 percent drop in injury crashes and the three large multilane sites showed 31 percent.
Because of the small size of the data sample, the only result that was statistically significant was the injury crash reduction for small and moderate roundabouts. One number in that table survived a significance test in 2000, and the guide leaned on larger international samples for corroboration: mean all-crash reductions of 41 to 61 percent in Australia, 36 percent in Germany and 47 percent in the Netherlands, with injury-crash reductions of 57 to 78 percent in France and 25 to 39 percent in the United Kingdom.
There is a selection problem underneath all of it, and the guide names it rather than leaving it to critics. These crash reductions are generally for sites where roundabouts were selected to replace problem intersections, so they do not necessarily represent a universal safety comparison with all other intersection types. An agency that converts its worst junction and then reports an 80 percent improvement has measured the junction as much as the treatment. The same caution applies to any before-and-after figure quoted in a project justification, a problem shared with much of the evidence used in highway project planning.
What still happens at a roundabout
Collisions do not stop. They change type. The dominant crash is a driver failing to yield on entry, what the literature calls an entering-circulating crash, and its share varies enormously by country: 36.6 percent of injury crashes in France, 50.8 percent in Queensland, and 71.1 percent in the United Kingdom. Rear-end crashes at entry and single-vehicle crashes follow. The guide notes that the central island presents a particular hazard that may result in over-representation of single-vehicle crashes during periods of low traffic volumes, which is to say the object placed there to slow entering traffic is a fixed obstruction in the middle of an intersection. Treatment of fixed objects near travel lanes is the subject of roadside barrier design.
Injury crash rates that account for volume tell a cleaner story than raw frequencies. Eight single-lane roundabouts in Maryland and Florida produced 0.08 injury crashes per million entering vehicles, against 0.045 in France and 0.275 in the United Kingdom. The British figure is higher because the British stock includes many high-volume multilane roundabouts, which is the next problem.
Where they do not fit
Adding lanes reintroduces conflicts that single-lane roundabouts do not have: drivers using the wrong lane, and drivers making improper turns from the wrong position. The guide’s most useful number on this comes from United Kingdom crash models. At 10,000 entering vehicles per day, flaring an entry from one lane to two is likely to increase injury crashes by 25 percent.
The severity picture stays favorable even so, because the additional multilane conflicts are generally low-speed sideswipes. The guide’s judgment is that although conflict points increase at a multilane roundabout compared with a single-lane one, the overall severity of conflicts is generally less than under alternative intersection control. In practice a roundabout sized for capacity gives back part of the benefit that made it attractive, so an agency weighing a two-lane roundabout against a signal faces a closer call than the 78 percent figure suggests. Most of the studies the guide reviewed also showed crash reduction in rural areas much higher than in urban areas, which is consistent with a treatment whose main effect is on speed differentials at high approach speeds.
How much smaller the pedestrian benefit is
A pedestrian crossing at a typical signalized intersection with four single-lane approaches faces sixteen potential vehicle conflicts, arriving from four different directions and including movements that are legal, movements that are legal but unexpected, and movements that are illegal. At a roundabout a pedestrian crosses one direction of traffic at a time on each approach, with a splitter island allowing the entry and exit conflicts to be resolved separately.
The measured outcomes support that. British crash rates per million pedestrian trips ran 0.31 at mini-roundabouts, 0.45 at conventional roundabouts and 0.33 at flared roundabouts, against 0.67 at signals. A Norwegian analysis of 59 roundabouts and 124 signalized intersections between 1985 and 1989 recorded 33 personal injury crashes at the roundabouts, one of which involved a pedestrian, against pedestrian involvement in 20 percent of the injury crashes at the signalized sites.
A Dutch study of 181 converted intersections broke the benefit down by mode, and the spread is the finding. Injury crashes fell 95 percent for passenger cars, 89 percent for pedestrians, 63 percent for mopeds and 30 percent for bicycles.
While overall crash frequencies have been reduced, the crash reductions are most pronounced for motor vehicles, less pronounced for pedestrians, and equivocal for bicyclists
The problem the data cannot see
For blind and low-vision pedestrians the guide is blunt about the state of knowledge. “Crash data have not been collected to indicate whether a pedestrian has a disability, and no studies have focused specifically on the safety of visually impaired pedestrians at roundabouts.” Twenty-six years on, the evidence base for the group most affected is still the thinnest part of the subject.
The mechanism of the difficulty is well described even where outcome data are missing. The sound of circulating traffic can mask a vehicle approaching the crosswalk. Entering traffic is slower, but it may be impossible to determine by sound alone whether a vehicle has actually stopped or intends to. Sighted pedestrians often rely on eye contact in that situation, and the guide notes plainly that this is not a useful or reliable technique for a pedestrian who is visually impaired. The exit lane is the greater hazard because speeds there are higher, and a multilane crossing compounds all of it, because a car stopped in the near lane can mask other traffic and block the far-lane driver’s view of a cane or guide dog.
The consequence the guide states is legal, not merely operational: unless these issues are addressed by a design, the intersection is inaccessible and may not be permissible under the ADA. That warning predates a formal federal rule by more than two decades. The United States Access Board’s Public Right-of-Way Accessibility Guidelines were published as a final rule on August 8, 2023, which means the design requirements now exist in a form an agency can be held to, and any project citing the 2000 guide as its accessibility authority is citing a document written before the standard.
The distribution is the story
Bicyclists are the case where the direction of the effect flips. British crash rates per million trips were 3.11 at mini-roundabouts, 2.91 at conventional roundabouts and 7.85 at flared roundabouts, against 1.75 at signals. A French comparison across fifteen towns found half as many injury crashes per year at roundabouts as at signalized crossroads, and at the same time found that people were more frequently killed and seriously injured per crash at the roundabouts by 25 percent, with serious crashes running 27.1 per hundred crashes against 21.9 at the signals.
Both results belong in any honest summary. European practice reads them as a design brief rather than an objection, which is why the standard solution there is a separated bicycle facility outside the circulatory roadway, crossing the entries and exits at least one car length back from the circulating lane. Where volumes are low enough, cyclists mix with traffic in the circulatory roadway instead, and the Dutch threshold cited by the guide is up to 8,000 vehicles per day. Tight entry geometry does the rest, by holding motor vehicle speeds near the 12 to 15 mph a commuting cyclist travels at. That entry-speed logic is what sets the roundabout apart from the rest of more on highway safety, and the national trend these treatments are measured against is examined in roadway fatality data.
The honest version of the roundabout safety case is narrower than a percentage and stronger than one. A roundabout is safer because it deletes the conflict geometry that produces right-angle and head-on crashes, and it delivers that benefit in proportion to how much steel the user has around them.