Choosing safer roads: what the crash data says about route planning

Most advice about cycling safety is about the rider: what to wear, how to behave, what to fit to the bike. Far less is said about the decision that comes before any of that, which is where you choose to ride. Route choice is the one safety input a rider fully controls, and the research on it is reasonably good.

This post is about what the crash data actually shows — where collisions happen, what drives their severity, and which route-planning heuristics have evidence behind them. It is not an argument that cycling is dangerous, and it is not an argument that it is safe. It is an attempt to describe the risk accurately enough to plan around.

Where collisions happen is not where riders fear them

Ask a group of cyclists what frightens them and most will describe being hit from behind on a fast road. That fear is not irrational — rear impacts are over-represented in the most serious outcomes, and rural roads with high speed limits are genuinely worse than urban streets on a per-kilometre basis for fatal injury. But it is not where most collisions occur.

In reported casualty statistics and in the research literature, the large majority of cyclist collisions with motor vehicles happen at or very near a junction. A review of the infrastructure literature found intersections consistently over-represented, and a case-crossover study that compared the exact site of each injury with a control site on the same rider's route found that route characteristics changed injury risk by a factor of several.

The planning consequence is direct. A route's exposure to the most common collision types is roughly proportional to how many junctions it passes through, not how long it is. A 40 km loop on lanes with twenty side roads is a different risk profile from a 40 km loop through suburbs with two hundred, even if the second one has a lower speed limit throughout.

Speed is the severity variable

Frequency and severity are separate questions, and they have separate drivers. Junction count predicts how often something happens. Speed predicts what it costs when it does.

The clearest data on this comes from pedestrian impact studies, which are far more numerous than cyclist-specific ones. The relationship between car impact speed and fatality risk is strongly non-linear: risk stays comparatively low at low impact speeds and then climbs steeply through the 40 to 60 km/h band. The precise numbers differ between datasets and populations — older casualties fare much worse at every speed — but the shape is consistent and it is steep.

There is a second, system-level result that points the same way. Across many studies of whole road networks, changes in mean traffic speed produce much larger changes in crash outcomes: injury crashes scale with roughly the square of mean speed, and fatal crashes with a considerably higher power still. Reviews of the speed literature also find that a vehicle travelling well above or well below the surrounding traffic carries elevated risk, which is the practical meaning of speed differential.

For a rider this is the awkward part of the arithmetic. You cycle at 25 km/h. On a residential street where traffic moves at 30, the differential is small and everyone has time. On a national-speed-limit A-road where traffic moves at 100, the differential is enormous, the closing speed on every overtake is large, and the consequence of any misjudgement sits at the top of that curve. Choosing the 40 mph road over the 60 mph road does more for your expected outcome than anything you can buy.

What the infrastructure evidence supports, and what it does not

The route-level studies are consistent on the broad point: purpose-built cycling infrastructure, separated from motor traffic, is associated with lower injury risk than riding on comparable streets without it. A Montreal comparison of cycle tracks against parallel streets found a lower injury rate on the tracks. The case-crossover work found the same ordering, with the highest risk on major streets with parked cars and no cycling provision, and the lowest on separated tracks and quiet residential streets.

Two honest caveats belong with that. First, most of this research is urban and commuting-focused; it says less about the rural lanes where a lot of sport riding happens. Second, the benefit is not uniform along a facility — the junctions where a separated track rejoins the carriageway are where the design has to work hardest, and where poor implementations concentrate their risk.

For route planning, the useful reading is that quiet-lane routing and separated infrastructure are pulling in the same direction: both reduce the number of moments when a driver has to notice you and act correctly. That is the mechanism, and it is worth optimising for directly.

Junction counting as a planning habit

Here is a heuristic that takes two minutes and is easy to apply to any route before you ride it. Trace the line and count the points where a vehicle could cross your path: side roads, roundabouts, signalised junctions, supermarket entrances, farm tracks with sight lines. You do not need precision. You need a ratio.

A rural lane loop might come out at one or two per kilometre. An urban route through a grid can be five or ten times that, which is one reason a daily commute accumulates exposure faster than the distance suggests — making a commute count as training is worth reading with the same junction map open. When you are comparing two candidate routes of similar length, the junction count is usually the single number that separates them best, and it is more informative than the traffic-volume colouring most planners offer, because volume tells you about overtakes and junctions tell you about conflicts.

Moveee's route library keeps the road-type breakdown alongside each saved route, which makes it easier to see how much of a loop actually sits on a main road rather than inferring it from the map's colours. Whatever tool you use, do the comparison before you commit rather than discovering the answer at kilometre 60. Our piece on building a route that works covers the rest of the planning trade-offs.

Conspicuity: real, but smaller than the marketing suggests

Being seen matters, and there is decent experimental evidence that high-contrast clothing, fluorescent material in daylight and retroreflective material at night increase the distance at which drivers detect and recognise a cyclist. What there is not, honestly, is good evidence that wearing them reduces your crash rate. The Cochrane review of visibility interventions found the experimental detection results reasonably consistent and the evidence about actual collisions essentially absent, because nobody has run the trial that would settle it.

That is not a reason to ride in dark grey. It is a reason to keep conspicuity in proportion: it is a cheap, sensible measure with a plausible mechanism and no demonstrated crash benefit, sitting well below junction count and road speed in anything resembling an ordered list. If you ride in the dark, riding at night covers the lighting side, where the physics is more settled.

Road position is a related and better-evidenced lever. An instrumented study of overtaking distance found that how far from the kerb a rider sat changed how much room drivers left. Some of that study's other findings — particularly the helmet result — have been re-analysed and are contested, and it is worth treating them as unresolved rather than settled. The position finding is the robust part.

What this adds up to

If you rank the levers by how much they change expected outcome, the order that the data supports is roughly: the speed of the traffic you share a road with, the number of junctions you pass, whether the road has parked cars, whether there is separated provision, and then everything else. The first two are decided entirely at the planning stage, on a screen, before you have put your shoes on.

That is an unglamorous conclusion and a useful one. There is no single safe road, and route choice does not make risk disappear. But of all the decisions available to a rider, the route is the one with the most evidence attached and the least effort required. Spend the two minutes. Then go and enjoy the lane you found.