Urban planners and some bicycle and pedestrian advocates tell me that “roundabouts are safer than traditional signalized intersections.” It comes up in the context of conversations about pedestrian-friendly urban street design. I hear this line a lot and see it on websites and in social media posts. In most cases, folks are parroting various viral videos and news articles they’ve seen or they’re just repeating what some traffic engineer told them. Empirical evidence shows and I will not deny that roundabouts are safer and more efficient for motorized vehicles, however pedestrians and bicyclists are another matter.

At the end of October, The Minnesota Department of Transportation (MnDOT) released “A study of the Traffic Safety at Roundabouts in Minnesota.” It got a lot of press because, like most DOTs, MnDOT is pushing roundabouts, even in urban areas. Currently there are around 200 roundabouts in the state, most in suburbs, exurbs and rural locations but, over the years, MnDOT has also proposed building them in the Twin Cities, including at the Dale Street I-94 off-ramps and as part of the “35 Access project” back in 2004. DOTs like roundabouts because they reduce serious and fatal crashes for motor vehicles. More importantly, they can handle more motor-vehicle thru-put (“Level of Service” for cars) and they have a longer “service life” and don’t require electricity or traffic control devices. So they are cheaper to maintain than traditional signalized intersections. The only problem is there’s almost no good data showing that roundabouts are safer or better for pedestrians. In fact there is some data showing they are more dangerous and plenty of evidence that they greatly reduce “Level of Service” for pedestrians. As such, roundabouts should be resisted in urban areas, particularly if your city wants to create density, walkability and pedestrian safety.

First, what do we mean by “roundabouts?” We’re not talking about little neighborhood traffic circles on bicycle boulevards. The photo at the beginning of this article, taken from the cover of the MnDOT report depicts a typical roundabout as envisioned by most state DOTs. It is telling that the example they choose for their cover has no sidewalks and no pedestrian crossings at all.

Having lived in New York City, Boston and other parts of New England where automobile roundabouts have existed since the 1920s, I have experienced urban roundabouts as both a driver and a pedestrian. Roundabouts waste huge amounts of space versus a normal signalized intersection. The space in the middle is completely useless. You can’t put a building in it, because there’s no way to safely reach it and you can’t walk into it as a pedestrian without getting trapped or hit by a car. Instead, you have to walk around them. This greatly increases walking distances and travel times for pedestrians. As such, roundabouts decrease pedestrian “Level of Service” and waste valuable urban space.

Additionally, as a pedestrian crossing an arm of the circle, you have no idea whether a vehicle in the roundabout is going to exit the circle in front of you or continue around it. Because drivers are either looking for their exits or trying to merge with vehicles in the roundabout, they aren’t paying attention to pedestrians and behave more unpredictably.

In contrast, at a traditional signalized intersection, there are turn lanes and/or vehicles occasionally use their turn signals. If I see a vehicle in a turn lane or signaling, I have some idea that it’s turning and whether it’s safe to cross. If the intersection’s signals feature a “Lead Pedestrian Interval” walk signal, it’s even safer because I am given a head-start across the intersection, so motorists are more likely to see me before turning. If the intersection or the city also bans “right turns on red”, it’s the safest of all, because it all but eliminates “right-hook” pedestrian crashes. If a vehicle is going straight across the intersection, my crossing signal is red and theirs is green. People run red lights but, if I’m paying attention, I can see this unfolding and react accordingly.

None of this predictability exists with a roundabout. For all these reasons– unpredictability, safety, distance and travel time– crossing a roundabout as a pedestrian can be awful. The statistics and photos that follow bear this out.

 

MnDOT’s Roundabout Study

Reading the recent MnDOT roundabout study, their method of analysis is solid. First they gathered crash data for several years at signalized or 4-way-stop intersections before a roundabout was installed and divided this by the total number of vehicles entering the intersections (traffic counts) to get an overall before-conversion “crash rate”. Then they collected crash data for the same number of years after roundabouts were installed and divided this by the total number of vehicles entering the new roundabouts to get an overall after-conversion crash rate. Then they compared the before and after data to determine whether roundabouts were safer. They broke this data down by crash types and severity. Theirs is not the first study to do this. There have been dozens.

The problem is they didn’t apply the same level of analysis to pedestrian crashes. They included pedestrian crash data but they didn’t do actual pedestrian counts to determine a “crash rate.” Most of the current roundabouts are in locations that have very low or non-existent pedestrian traffic and there is currently no reliable way for DOTs to automate pedestrian counts. Counts have to be done by human volunteers with clipboards and count sheets. MnDOT almost never does these and only a few U.S. cities do them in a comprehensive way. None have done comprehensive pedestrian count and crash studies before and after the installation of roundabouts at intersections, particularly intersections with significant pedestrian volumes.

But, even with just crash data and no count data, MnDOT’s study shows that roundabouts dramatically increased pedestrian crashes. Pedestrian crashes went from 10 to 18. Single lane crashes went up 17.3% (page 22) and dual lane crashes went up 500% (page 26) …for a total increase in pedestrian crashes of 70.2%. To be fair, these are statistically insignificant numbers because most roundabouts are in locations with little or no pedestrian traffic …but, combined with other evidence discussed later, these numbers suggest that roundabouts are actually less safe for pedestrians than traditional intersections. Nowhere in the report is this pedestrian data highlighted or summarized. The only data summarized is the benefits of roundabouts for cars and car crashes.

According to the study, bike crashes went down slightly but, again, the overall numbers are tiny and without bike count data and more detailed crash information, it’s hard to say why. The very design of the roundabout may have deterred cyclists from entering it. A very cursory look at “fatal and serious injury crashes exploratory analysis” (table 11) shows one of the 6 “severe or fatal” crashes was a bicyclist hit by a vehicle exiting the roundabout. Since their “Crash Rate” analysis formula (on page 21) doesn’t include pedestrian or bike counts (but just vehicle ADT), it’s useless.

What’s fascinating about the report (and others) is that the overall number of crashes, even for motor vehicles, actually increased 15.6%! (page 28). It’s just that the overall severity of these crashes (for cars) decreased. So certain types of crashes like “Sideswipe Same Direction” increased 683% but the more severe “Right Angle” crashes declined 35.8%, and everyone was moving slower. So, despite the increased number of crashes, the overall severity was lower. This is what roundabout proponents mean by “safety” for cars. In urban areas, proponents will often cite this “safety” but I could find no statistical data that show reductions in pedestrian injuries or fatalities.

 

The NACTO Report

The lack of pedestrian data and anecdotal evidence of roundabout danger and inconvenience to pedestrians is corroborated by a National Association of City Transportation Officials (NACTO) study. Here’s a few quotations:

“Current international research shows that modern roundabouts improve vehicular and pedestrian safety compared to conventional intersections. However, their effects on pedestrian safety in the U.S. remain unsubstantiated. Complicating this problem is a scarcity of pedestrian accident data at roundabouts, especially at intersection locations that were reconstructed as roundabouts and could potentially provide critical before/after accident statistics.”

“Relatively little pedestrian data exist compared to vehicle crash data; even less pedestrian crash data exist for roundabout treatments. Yet, roundabouts are often proposed for traffic calming in high pedestrian areas”

“Indeed, anecdotal evidence leads to concerns that pedestrians may minimize their walk distance by taking short cuts across the central island and cause impatient drivers to challenge pedestrians.”

Note that even the authors of this study acknowledge here and in table 1.1 that traffic circles increase crossing distances for pedestrians, to an annoying enough degree that pedestrians try to jaywalk on them. The report cannot find a single statistical analysis case study where:

1. someone did regular pedestrian counts at a traditional, signalized intersection to determine average daily counts for pedestrians and;
2. collected at least 5-10 years of pedestrian accident data at that intersection;
3. Then did the same thing after the intersection was converted to a roundabout (pedestrian counts and accident data for 5-10 years).

Because no such statistical analysis appears to exist, the authors had to resort to computer simulations for their proposed roundabout in Hillsboro, North Carolina. Anyone who is being honest knows computer models like SYNCHRO and VISSIM are fraught with problems and are only as good as the data put into them (which the authors acknowledge is non-existent). So any results obtained are “theoretical”. Again, I quote the document–

“Statistical analysis in the context of roundabout safety evaluation would ideally have pedestrian crash data before and after an intersection is reconstructed as a roundabout. Better yet, to account for other possible causal factors, it is desirable to have multiple reconstructions sites. Persaud had nearly 30 intersections in his vehicle safety analysis. Yet, as discussed previously, there is very little data for pedestrian accidents.”

The report concludes: “The advantage of statistical analysis is that the results can be very pragmatic and statements about the results of the study can be authoritatively made” …but it has no statistical analysis. So nothing about this study is “authoritative.” It’s just a literature survey and computer simulation for a proposed roundabout.

 

The FHWA Report

While offering all sorts of guidelines, The Federal Highway Administration Informational Guide on Roundabouts similarly lacks any real pedestrian or bicycle data. In Chapter 2, page 23, it cites one U.S. study of just 11 intersections to support its contention that roundabouts are “safer” but says that “This study yielded insufficient data to draw conclusions regarding the safety of bicyclists and pedestrians.” On page 24, it makes the inference that, because low vehicle speeds “reduce crash severity for pedestrians and bicyclists” and roundabouts lower speeds, then roundabouts must be better for pedestrians and bicyclists. This is like saying “correlation proves causality”, which it doesn’t, and it ignores many facets of roundabouts that are potentially hazardous to pedestrians. On page 25 and 105-110, it says roundabouts have “fewer conflict points, which means fewer opportunities for collisions” …but, as the more rigorous MnDOT study showed, the number of collisions actually INCREASED. Only their severity decreased. So, once again, the report’s statements read like theoretical philosophy, not science. With double lane roundabouts, even the FHWA acknowledges that:

“Pedestrians crossing double-lane roundabouts are exposed for a longer time and to faster vehicles. They can also be obscured from, or not see, approaching vehicles in adjacent lanes if vehicles in the nearest lane yield to them. Children, wheelchair users, and visually impaired pedestrians face particular risks. Bicycles are also more exposed to severe conflicts when choosing to circulate with motor vehicles.”

With all roundabouts (on page 33) it acknowledges that they

“allow pedestrians to cross one direction of traffic at a time; however, traffic may be moving (albeit at a slow speed), thus making it more challenging to judge gaps, especially for visually impaired users, children, and the elderly. …Roundabouts pose problems at several points of the crossing experience, from the perspective of information access. When crossing a roundabout, there are several areas of difficulty for the blind and or visually impaired pedestrian. Unless these issues are addressed by a design, the intersection is ‘inaccessible’ and may not be permissible under the ADA (Americans with Disabilities Act).” (…And later, on page 40), ” Current guidelines do not specifically address ways to make roundabouts accessible.”

With bicyclists, the FHWA report states that “Roundabouts may not provide safety benefits to bicyclists. …The complexity of vehicle interactions within a roundabout leaves a cyclist vulnerable, and for this reason, bike lanes within the circulatory roadway should never be used.” A lot of the foreign studies in the report (discussed below) showed increased risk for cyclists.

Chapter 5 of the report summarizes some foreign data from various countries. I’ve tried to access and sift through some of these. A French study of 15 towns showed statistically insignificant (0.7%) risk reduction for pedestrians in roundabouts over “All Crossroads”, and a 3.6% increase in risk for cyclists but the methodology of the study is unclear.

A British study showed more improvement but the study’s methodology is also unclear. Same for a Norwegian study that may or may not have included pedestrian counts and crash data from before/after conversions.

An oft-cited 1993 Dutch study by Schoon and van Minnen of 181 intersections converted to roundabouts showed huge improvements but the FHWA report didn’t spell out the study’s methodology and I couldn’t find a copy of the study online, only references to it.

Ultimately, when it comes to pedestrians and cyclists, the entire 277 page FHWA report is mostly based on theory and has scant data to show roundabouts are safer for pedestrians than standard, signalized intersections– something it more or less acknowledges. The report is overwhelmingly focused on cars and driver safety, just like the FHWA.

 

IIHS Studies and Conclusion

Finally consider two articles from the Insurance Institute for Highway Safety (IIHS), once given to me by an MnDOT employee as evidence of roundabout safety. The first, from 2013 says that so called “non-injury crashes” actually increased at the two discussed Bellingham intersections when they were redesigned with traffic circles. This is similar to what MnDOT found. The author wants to blame “driver confusion” but they have no polling or scientific evidence for this and the cause is irrelevant. The article has zero pedestrian statistics. Everything is about moving more cars, faster. Everyone should look at the aerial picture of one of the roundabouts. I include it below. It doubles or triples the crossing distances for pedestrians because the crosswalks have to be set back so far from the circle to be even marginally safe. Pedestrians are forced to cross “double jeopardy” two-lane highways with no traffic light. The intersection now wastes double its original space. Were this in an urban area or town, you could fit four additional properties on this corner in the extra, wasted space. What it does accomplish is increasing the average volume and speed of traffic that can traverse the intersection, something that appeals to car-centric DOTs. This isn’t even an area where people walk. Note there are no sidewalks on the east-west legs of the roundabout. This is a car-oriented strip-mall area.

 

The second IIHS article is even more revealing. It references experiences of Carmel, Indiana from 2010. There are no pedestrian counts. The sum total of its statistics are as follows:

“Brainard credits roundabouts with keeping the number of traffic injuries from growing along with the city’s road network. In 2003, there were 252 crashes causing injury on 220 miles of roads, according to Carmel officials. By 2008, the city had 395 road miles, but injury crashes went down slightly”

“A 2001 Institute study of 23 intersections in the United States found that converting intersections from traffic signals or stop signs to roundabouts reduced injury crashes by 80 percent and all crashes by 40 percent.”

The first stat tells us nothing except some Indianapolis suburban community is building a ton of new roads (a red flag for being a “pedestrian-friendly community”) with a goal of “managing high traffic volumes”– also a red flag, since it’s impossible to build your way out of traffic congestion. We don’t know what “Vehicle Miles Traveled” on those roads were and, most importantly, we don’t know what the number of pedestrian trips or intersection crossings on those roads were because Carmel, like most towns and cities in the USA, doesn’t collect that data in any comprehensive or scientific way, if at all.

The second, older, national statistic doesn’t tell us what pedestrian volumes were before and after on those 23 intersections so we don’t know the crash-rate-per-intersection-crossing before and after traffic circle installation. We don’t know the type of intersections or much of anything …and there’s no link to the original study even though it’s often cited as a source on various state DOT websites. I highly doubt they did any pedestrian counts. All the quotes from people in the article are completely car-oriented– how it’s improved things for cars.

 

I urge everyone to look at the photograph of their Keystone “Double Teardrop” roundabout. It is one of the worst pedestrian nightmares I’ve ever seen. It makes the Maryland Avenue overpass at I-35E look like a pedestrian paradise. Stop for a moment and imagine walking across this overpass as it’s currently designed versus walking across it as a straight overpass with parallel ramps ending in a pair of synced traffic signals (like Dale Street at I-94 in Saint Paul). This Keystone roundabout design adds at least an eighth of a mile to pedestrian crossing distances when crossing the highway, and even more when crossing parallel with the highway because the crosswalks have to be set back so far from the roundabout. Pedestrians are forced to make two crossings of double-jeopardy, high-speed, blind-corner slip turns, all without signals. Slip turns and unsignalized two-lane crossings are something that even engineers will admit increase pedestrian crashes and we’ve already seen multiple people hit on Snelling Avenue by Macalaster, on Lexington Parkway and other 4-lane streets with medians and no signals… and these are on straight roads! Medians on 4-lane boulevards reduce crashes versus no median …but both are much less safe than signals.

On top of this, look how much extra concrete and wasted space the Keystone design adds to the intersection. You could put in 4-8 additional buildings (houses, retail or apartments) if this was a traditional overpass with tight, parallel ramps and traffic lights, where the ramps intersect the overpass road at right angles. You can see this traditional design where Dale Street passes over Interstate 94, an area where double roundabouts were once proposed and (fortunately) rejected.

What the “Keystone Project” roundabout design is good at is moving more cars at a faster average speed, with less severe car collisions. It even says the roundabout will “shorten commute times.” If your goal is moving more cars, faster, then roundabouts are for you. If your goals are to control traffic for pedestrian crossings, shorten pedestrian trip distances and make a pedestrian-friendly city or town, then roundabouts are completely counterproductive. So please stop saying they are “safer” or “better” for pedestrians. They’re not.