Thursday 13 August 2020

Roundabout safety for cyclists and why Cambridge's new 'Dutch' roundabout is not what it should have been

This is not the Cambridge roundabout. It's an example of a different
design, the safe roundabout. This design is much more tolerant of
driver mistakes and results in a far lower rate of cyclist injuries.
Please read my blog post about this design and watch the
accompanying video as both of those describe how this
design keeps cyclists safe.

Cambridge in the UK recently opened a "Dutch style" roundabout on Fendon Road to replace another roundabout which had a poor safety record for cyclists in the city. People keep pointing this out to me. I think it's perhaps time for a response.

The design chosen is that which I have consistently recommended against. Why do I recommend against it ? Because it has a poor safety record here in the Netherlands. It has a safety record which is not very different to an uncontrolled crossing or having no cycling facilities at all on a roundabout, which was the situation in Cambridge before this new design was built. I am not convinced that British drivers are better trained or will behave in a safer way on roundabouts than Dutch drivers and therefore I am also not convinced that it will be safer to use this design in the UK than it is here. Dutch drivers have been trained for decades to expect to give way to cyclists under some circumstances and they have also had decades to get used to this type of roundabout, yet it remains unsafe in the Netherlands. Why should we expect this to be different elsewhere ?

Now you might wonder why all my emphasis so far has been on drivers. There's a simple reason. The danger comes from motor vehicles which are faster and heavier than any cyclist. That's why injury rates for cyclists are higher at roundabouts with more cars. Remember always that there is no design of roundabout which is built for cyclists. If it were not for cars then we would have no roundabouts at all and cycling would be all the safer for it.

Dutch research into roundabout safety

Dijkstra's research into roundabout safety in the Netherlands in 2005 is one of the most referenced articles about this subject. Here's a link. It's written in Dutch and having seen the mess that bad machine translations make of this and similar documents I suggest that if you can't read the Dutch then you're best off going no further than the abstract and my explanation below.

The research drew its conclusions from looking at three different earlier studies each of which looked at large numbers of roundabouts. They're well run studies and the people making them attempted to compare like with like. There was no attempt to study only a subset in order to favour a conclusion already drawn by any of the authors. Dijkstra's article is often used by proponents of priority on roundabouts, though I wonder if they actually read it all the way through. It's very clear about the danger caused by giving cyclists priority on roundabouts.

This graph compares the number of serious cyclist and moped injuries per year (Y) at roundabouts of differing types with differing motor traffic intensity (X). The green line shows the effect of a cycle-lane around the roundabout, which has been known to be dangerous for decades and is recommended by no-one  The black line shows a roundabout with no facilities, cycling taking place on the road amongst cars. The blue line shows the effect of cycle-paths around roundabouts, but this is an average as it is not specified which roundabout design or priority rule it refers to, which is a problem as I'll discuss later.

This graph looks a lot like that above, but it includes all serious injuries in all vehicles, not just those which occur to cyclists. However it is clear from comparing this graph and the one before that roundabouts are, generally speaking, far more dangerous for cyclists than they are for drivers. The relatively small number of motor vehicle occupants who are injured at roundabouts pushes each of the lines upward a little. The effect is especially obvious on the lower, blue line as in this case motor vehicle occupant injuries become dominant as traffic intensity increases. But comparing the two graphs does not make a clear-cut case for all roundabouts with cycle-paths because there is no distinction made between two very different roundabout designs discussed later by Dijkstra.

Three different studies are referenced by Dijkstra, that of Van Minnen/CROW (1998), Weijermars (2001) and Gerts (2002). All three of those studies showed that "with priority" roundabouts are substantially more dangerous for cyclists.

Three studies are summarized by Dijkstra. All studies show worsened safety with cyclist priority ('in voorrang') at roundabouts. Crashes (ongevallen) are between 1.75x and 2.9x more common and injuries resulting in death or hospital treatment (slachtoffers) are found to be between 3x and 6x more common. Gerts' study only reported crashes and not injuries.

Some of the data in the document, such as this example, show all injuries and not just cyclist injuries. Overall, a junction with cyclist priority ('in de voorrang') causes about three times as many injuries.  About as many motor vehicle occupants are injured on the two different types of roundabouts, the difference is made by the radically different results for cyclist safety. See the next table.

The same data from Weijemars is used here, but just the figures for cyclists and mopeds are included.

This table shows the number of roundabouts in total, within urban areas, with cycle-paths, with cyclist priority (470), without cyclist priority (314). These numbers come up repeatedly.

Estimated annual hospitalizations of cyclists and moped riders (on the cycle-path - now largely replaced by e-bikes) in the Netherlands due to crashes with motor vehicles on urban roundabouts. There are about 50% more roundabouts with cyclist priority ('in de voorrang') but they cause ten times as many serious injuries as the safe design of roundabout ('uit de voorrang').

Note that Dijkstra used data from Gerts who referred only to roundabouts which strictly followed the CROW design standards. By doing so both Gerts and Dijkstra tried to exclude potentially worse results which could have come from priority roundabouts which didn't follow the standards. This turned out to be less important than was expected because Gerts found that there was no statistically different result for roundabouts which followed the CROW norms vs. those which did not (bottom of page 13). Also read pages 34 and 35 where this is confirmed. i.e. The difference in safety between differing roundabouts with cycle-paths is almost entirely due to the priority rule. That is confirmed on page 36 where the conclusion is made that there is a significant link between priority and safety.
A note about 'mopeds'. Most of the data lumps cyclists and mopeds together because mopeds are not actually very common. They make up around 1% of the total traffic volume. As a result, I don't expect that their presence here makes much difference to any of the results. Low power mopeds which theoretically are limited to around 25 km/h on the cycle-path are losing ground these days to electrically assisted bikes which travel at a similar speed. Many Dutch people really don't like mopeds, but they do seem to like e-bikes which are now far more numerous. The two modes are quite similar from a safety perspective. There does seem to be an indication that 'with priority' roundabouts are more dangerous for faster cyclists (moped, racing bike, e-bike) who are more likely to surprise drivers because their speed makes them more difficult for drivers to spot, hence the idea that they suddenly appeared "from nowhere" over the driver's shoulder. Faster mopeds (limited to 45 km/h) are banned from urban cycle-paths so are not of interest here.

Comparison with a non-signalled crossing
In order to find a point of comparison for the effectiveness of substituting one type of roundabout or another, comparisons were made of 177 locations where non-signalled crossings ('voor, kruispunt') were converted into roundabouts ('na rotonde').

Note that this is not a comparison with traffic light junctions. That's a completely different story which I have addressed elsewhere: well designed traffic light junctions which separate cyclists in time and space from cars create an almost completely safe space for cyclists and as a result have an excellent safety record in the Netherlands. But they are not discussed by Dijkstra. These comparisons are between roundabouts and the straightforward non-signalled crossings without roundabouts or traffic lights which existed in the same location before they were built.

The comparison over time with non-signalled crossings in the same location across 177 junctions gives us a very good idea of the relative danger creating by giving cyclists priority at a roundabout.

Roundabouts are highly effective at preventing motor vehicle occupant injuries, reduced by 80% on average over the previously existing crossing, however for cyclists they are far less effective. Compared with non-signalled (no traffic lights) road crossings, Dutch roundabouts on average only lead to around a halving of cyclist serious injuries. But this statistic is deceptive because both the 'with priority' and safe roundabout designs have been mixed here to create an average.

In this table they've been separated and we if we consider only the safe roundabouts where cyclists do not have priority ('uit de voorrang') then we see a significant 87% reduction in the number of injuries vs. a non-signalled junction.

By comparison, if we consider only the roundabouts where cyclists do have priority ('in de voorrang'), the improvement in injury rate over a non-signalled crossing is just 11%. i.e. priority for cyclists on roundabouts results in nearly eight times as many cyclist hospitalizations as occur at safe roundabouts when replacing non-signalled junctions. But carry on reading...

The simple conclusion to reach here is that 'with priority' roundabouts achieve about an 11% improvement in safety while safe roundabouts achieve an 87% improvement in safety, making them eight times safer. However it's not actually that simple. Over the same period of time, non-signalled crossings across the country became on average about 10% safer (Dijkstra last paragraph of page 10) so we should perhaps reduce both outcomes by this amount. That would imply that there is actually almost no significant difference in safety between an non-signalled crossing and a 'with priority' roundabout while a safe roundabout achieves about a 77% improvement in safety for cyclists.

How safe are 'with priority' roundabouts in reality ?

This blog post began with one of the first graphs from Dijkstra which showed the relative safety of cyclists using cycle-paths and cycle-lanes around roundabouts with cycling on the roadway on the roundabout itself. Unfortunately, that graph did not distinguish between the different types of Dutch roundabout which include cycle-paths. However later in the same document with the discussion about replacement of non-signalled junctions, Dijkstra calculates the relative safety of roundabouts with and without priority, and we see that the 'with priority' cause nearly eight times as many excess cyclist injuries as the safe roundabout design. I find it unhelpful that both roundabout designs are presented as one item in that graph as they really are not the same.

Below you'll find a modified version of the same graph showing the likely real-world safety of 'with priority' roundabouts vs. the average for all cycle-paths on roundabouts, cycle-lanes and cycling on the road:

The red line shows what the probably injury rate is for roundabouts with priority cycle-paths in the Netherlands. The blue line is an average for all roundabouts with cycle-paths. A line which represented only safe roundabouts would be lower than the blue line, especially in the first half of the graph.

You'll note that up to around 10000 vehicles a day there is no great difference between the black line, showing the rate of injury for cyclists riding on the road, the green line showing a cycle-lane painted around the perimeter and the red line showing the injury rate for cyclists on 'with priority' cycle-paths. Therefore swapping between a roundabout design where cyclists ride on the road to a design where they use a priority cycle-path should not be expected to bring an improvement in safety. The only option which always looks good is the blue line showing as an average for all roundabouts with cycle-paths, which is skewed downward by the greatly improved safety of the safe roundabout design over the 'with priority' design.

Explanation of the red line: Over half of all urban roundabouts in the Netherlands gave cyclists priority when Dijkstra's graph was created. He found that on these roundabouts about eight times as many injuries per year could be expected. i.e. the corrected data for with priority roundabouts should be at about four times the level of that for the safe roundabouts. The red line which I have added to the graph is at approximately three times the height on the Y axis of that for both types of roundabout combined. This places it at approximately at the level we would expect to see if the two different types of roundabouts had been  treated as two different cases.

Explanation of blue line shape and stippled red: The blue line, for all roundabouts with cycle-paths, has a rather unusual shape, with a peak at around 10000 vehicles a day before dipping downward and rising again with motor vehicle intensity. I believe that this is caused by priority roundabouts causing the vast majority of crashes and injuries but being built mostly in places with lower motor vehicle intensity. As a result, the blue line is pushed upward in the first half of the graph by crashes on 'with priority' roundabouts, but I think the second half of the blue line is made up to a higher extent of data for safe roundabouts. I used a stippled red line from ~10000 onward because it's likely that it is no longer reliable past this point for the same reason. Unfortunately the separate data is not available so that I can test this hypothesis.


Back in 1998, SWOV's own research already demonstrated a worse safety record for 'with priority' roundabouts. Their research indicated that 52-73 extra cyclists would find themselves in hospital with serious injuries each year if that design was used. Despite this concern, a political decision was made to go along with a standard recommendation that cyclists should always have priority on urban roundabouts. This happend after they were given two re-assurances: first that the new standards advised by CROW would be applied everywhere and second that introduction of these standards would result in improved safety for 'with priority' roundabouts.

Unfortunately, the expected safety improvement was not realised. This was not merely because the guidelines still were not followed in all cases, but because it was discovered that the guidelines didn't actually have a measurable effect on safety even where they were followed. In an appendix on pages 29 to 35 of Dijkstra ("Voorrang voor fietsers: effect van vormgeving?") it is shown statistically using data gathered from many real world roundabouts spread across the country over five years that there is no detectable difference in safety between 'with priority' roundabouts which adhere strictly to the CROW guidelines and those which do not. Therefore it would seem that roundabouts with significant differences from the norms (e.g. different radii, crossings too close to the roundabout, inadequate road markings, allowing bidirectional cycling around the roundabout etc.) were not worse from a safety perspective at all. The research suggests very strongly that exactly following CROW's guidelines is not important for safety at all. Just one thing makes roundabouts with cycle-paths dangerous and that is giving cyclists priority.

Since the decision was made to go along with the "with priority" guideline in urban areas, the number of roundabouts in the Netherlands has tripled. For that reason we can now reasonably expect that instead of 52-73 extra cyclists receiving injuries resulting in hospitalization or death each year that there are now between 150 and 210 more than would have been the case if the safe roundabout design (used all over Assen with extremely low injuries as a result) had been standardized upon across the entire country.

The CROW argument that roundabouts with priority are "slight less safe" than the safe design was based upon the 52-73 extra casualties being only between 1.8 and 2.5% of the total. Unfortunately, with their number having tripled, we can now expect that these dangerous roundabouts account for between 5.4% and 7.5% of cyclist casualties per year, which is certainly not insignificant. Given that after a period of decline, Dutch cyclist deaths are now rising once again, this is something we should be looking at more closely. We already know how to reduce injuries on roundabouts by 87%. We just have to switch to building the proven safe roundabout design.


So now back to Cambridge. Why did that city outside the Netherlands adopt a design has been proven to cause excess injuries and deaths in the country where it came from ? Why are they following a decision which was made in the Netherlands for political reasons and not based on safety ? Personally, I think they were extremely badly advised.

Valkenswaard roundabout visited in 2006. Not a safe design.
See statistics. This roundabout does not conform 100% to
CROW standards, but as you can read above, we now
know that it probably wouldn't make a difference if it did.

Now I take a little responsibility for this because I used to live in Cambridge, I was a member of the Cambridge Cycling Campaign who campaigned for Dutch style cycling infrastructure to be built in that city, and I organised study tours in the Netherlands for members of the campaigning group and people working for the council. Indeed, on the first of these study tours, way back in 2006, I was myself enthusiastic about the cyclist priority design of roundabout and I showed it to people from Cambridge.

However my enthusiasm for priority roundabouts didn't last long after I moved to the Netherlands, experienced near misses on these designs, learnt Dutch and read Dijkstra's and other studies of roundabout safety. I change my mind when presented with contrary facts and it's simply not possible to continue to believe that these designs are safe when they clearly are not. That is the reason why I promote the safe design of roundabout, especially in countries outside the Netherlands.

The advantage that the safe roundabout design has is of complete predictability for the cyclist. As a cyclist you can take control of your own safety. While mistakes made by drivers on priority roundabouts where the driver does not cede priority often result in crashes and cyclist injuries, drivers on safe roundabouts who make errors typically just slow or stop when they need not have done. This can actually be slightly annoying for a cyclist who has adjusted their speed to slip across the road before or after any given car, but it's a safe failure mode which doesn't result in injury.

There is also the potential difference between British and Dutch drivers. I've not only cycled but also driven in both countries and I think the ability to drive safely is quite similar between the two countries. However there is a difference in behaviour in that Dutch drivers typically drive more slowly, at least on minor roads and in towns, and that there is less aggression. This is probably also a function of the unravelling of routes which keeps drivers who are minutes late for their destination from making through journeys near cyclists. They are also used to the idea of ceding priority to cyclists on roundabouts and elsewhere, and most people here do cycle. What effect this difference in driver behaviour will have on a roundabout design which requires drivers to cede priority in order to ensure cyclist safety is currently unknown.

Unfortunately the mistake in Cambridge goes far beyond just choosing a poor and dangerous design of roundabout. Their ambition unfortunately went no further than substituting one roundabout design for another and it is now being heralded as a victory for cycling before we even know whether it turns out to be safe in this location. In reality there is no roundabout design which has ever been built for the benefit of cyclists. They're all just ways of trying to address the problems caused by motor vehicles.

Most road junctions in Dutch residential
areas look something like this. Much
more common than any traffic light
or roundabout design, a non-signalled
junction in a residential area, with 30 km/h
speed limit, raised table, small corner
radii, and most importantly it's not a
through route for motor vehicles

Addressing the problems of motor vehicles could have been done in many other and more effective ways. The roads leading to the location of Cambridge's new roundabout are places where people live. Residential streets which have for years endured high traffic volumes. If they had their usage changed so that they no longer had to support that through motor traffic (already done on a tiny scale in Cambridge) then the roundabout probably wouldn't be required at all. A simple non-signalled junction but with very cars, or if a stream of traffic was required to flow between two arms of the junction perhaps a signalled crossing, would have resulted in a better outcome for cyclists.

You could have asked...

There are few people who have lived, worked and cycled on an everyday basis in Cambridge for more than a decade and who have then done the same in a Dutch city for more than a decade. Fewer still who have written about cycling for decades, about cyclist roundabout safety and made suggestions for how countries outside the Netherlands can best learn from the Dutch experience regarding roundabouts and other infrastructure. Still fewer who have gone to the effort of learning Dutch, reading the research and trying to advise on the basis of that research, and I would guess that also bringing people from the council and the local cycle campaigning group in Cambridge to the Netherlands in order to demonstrate to them the pros and cons of different roundabout designs amongst many other things on top of those other things would make me pretty much unique. I'm also not hard to find.

So do you think anyone from Cambridge ever thought to contact me about their new roundabout ? Of course not. Actually that's a bit unfair. A couple of people on the periphery who were concerned about the direction in which that was going did make contact and I have heard from them how what I'd written was dismissed rather rudely. But why ? Did they have some other local expert who could read the Dutch documents in depth and translate them ? Did they not believe the stats therein ? Did they expect that a design which is relatively dangerous in the Netherlands with Dutch drivers would somehow be less dangerous in the UK ?

I of course don't expect any answers. But it's a shame. I could have helped. I fear that Cambridge will regret what has been done. Not now. Not soon. It'll take a while before you have stats - the Dutch research on which I based my advice was itself based on the stats from many roundabouts over many years. The stats above are based on up to 940 rotonde-jaren, or roundabout years. These results have not been cooked up after just looking at a single location for a few weeks.


I am left with a fear that what Cambridge has done is merely to replace one dangerous roundabout design with another. It's a pity that they have done this because it's been done with such a lot of publicity and this may tarnish the concept and make it more difficult to adopt a better design elsewhere. Britain's cyclists need safe infrastructure. It's high time that they got it instead of ineffective projects like this example. However this is just one junction in one city of a country of many millions of people. It won't in itself make a huge difference one way or the other. No single junction ever could do that.

What is required, in Cambridge and elsewhere, is to start looking at the bigger picture. It is a mistake to think that any real change can result from a piecemeal one junction at a time approach. The UK, and everywhere else, needs a complete grid of effective safe and attractive cycle-paths in order to enable efficient go-everywhere cycling. The Netherlands is fabulous for cycling not merely because some of the roundabouts are safe (as you can see above, some certainly are not). An holistic approach makes the difference. There is nothing new and novel about any of this, it's been known for forty years.

Planning occurs in the Netherlands on a much greater scale than in the UK. There are traffic circulation plans which limit the effect of motor vehicles on residential areas and direct cycle routes. Almost all roundabouts form part of a plan which removes motor traffic, especially through motor traffic, from locations where cyclists need to be to make everyday journeys. This makes even less safe designs safer. But removal of motor traffic can even result in a much bigger prize: the removal of roundabouts and traffic lights altogether from cycling routes. Cyclists don't themselves need either of those types of junctions. They are not actually for us, but exist only to moderate motor traffic. There exist very good traffic light designs which benefit cyclists and roundabouts which do likewise but in both cases these things should only be built where there is no choice. i.e. where it's impossible to get rid of the motor vehicles. Cyclists are always better off without big road junctions. This has happened on a large scale at least around the centre of most Dutch cities, spreading outward from the centre in many cities and into the renovation of old suburbs and the design of new suburbs. The process did not begin by changing the design of a single busy junction which will remain busy, but then making a lot of noise about it as if a great change has been achieved.

The piecemeal approach, resulting in years of little or no real action and the effect of such small forward steps as occur usually being swamped by backward steps elsewhere, is precisely why people in the UK have gone from saying that their country is 20 years behind the Netherlands to 40 or 50 years behind. It's only possible to catch up by doing the hard work required to make things better, consistently over years. This is something which no country can afford not to do: when all things are considered, it's cheaper to build cycling infrastructure than it is not to build it.

Update May 2023
Three years later we now have reported collision data for the Cambridge roundabout. In the three years since the roundabout was rebuilt as "Dutch" style there have been 10 collisions, including three which caused serious injury. In the three years before the change there were just six of which none were serious.

I'd been advising people in Cambridge about roundabout designs for well over a decade before they ignored my advice and chose to build this dangerous design instead. They now have injured cyclists as a result of their choice. I take no pleasure in saying that "I told you so", but I did tell them, repeatedly, and the result that they now have is just what I predicted.