Roundabouts keep replacing junctions near me and everyone hates them — are they actually safer?

The gap between the data and the local reaction is worth taking seriously rather than dismissing, because it is not irrational.

People are evaluating a different thing. A driver's experience is about confusion, hesitation, and near misses — all of which genuinely increase while a population learns a new junction. Those are real costs, they are felt daily, and they do not appear in injury statistics.

The benefit is invisible. Nobody experiences the serious collision that did not happen. The costs are salient and the benefits are counterfactual, which is the hardest possible arrangement for public support.

Familiarity effects are large. Studies where roundabouts are new show a learning period. In places where they have always been standard, drivers find them unremarkable and the crossroads confusing.

So both things are true: they are safer, and they are more annoying for the first few years.

On the independence question, which is a fair one to raise: the finding does not rest on the reports of the agencies that built them.

Studies have been done in many countries, with different road cultures, different driver training, and different research institutions, including in places that reintroduced roundabouts after decades without them and therefore had a clean before-and-after with no local advocacy attached. The direction of the result is the same everywhere.

One methodological caution worth knowing about, though, because it applies to all before-and-after road safety work: regression to the mean. Junctions get rebuilt because they had a bad few years, and a bad few years is partly bad luck, so some improvement would have appeared anyway. Good studies control for this with comparison sites. It does not overturn the result, but it does mean the headline figures from any single site should be treated as the optimistic end.

The evidence is unusually consistent across countries and study designs, and the key is that the total number of collisions falls modestly while the severe ones fall dramatically.

That split is the whole story, and it explains the gap between the data and local opinion.

The mechanism is geometric rather than behavioural. A conventional crossroads has a large number of points where vehicle paths cross, including head-on and side-impact conflicts at the full speed of the road. A roundabout removes those crossing points almost entirely — everyone travels in the same rotational direction, so the remaining conflicts are merging and diverging, at an angle, at low speed.

A side impact at speed is the collision type that produces the worst outcomes. Removing that category while adding some low-speed scrapes is exactly what the figures show: minor collisions may not fall much and can even rise, while serious injuries fall by a large margin.

One more benefit that rarely gets mentioned in the safety argument but drives a lot of the decisions: throughput at moderate traffic levels, and behaviour during power failures.

A roundabout keeps working when the electricity is out, needs no maintenance contract for signal equipment, and does not make anybody wait at a red light with no cross traffic at four in the morning.

Whether that outweighs the annoyance is a local judgement. But it means the choice is not being made purely on the injury figures, which is worth knowing when reading the official material.