The key to improving bike safety is understanding how and why collisions occur.
Which has been almost impossible to figure out here in Los Angeles, where no one was keeping track of such vital statistics until recently. Let alone analyzing them.
I tried digging the data out of the statewide SWITRS traffic collision database before giving up, as have others before and since.
Now long-time LA bike advocate Dennis Hindman has dug through data compiled by the Los Angeles Police Department to uncover the causes of collisions — at least as determined by LAPD traffic investigators — with surprising results.
And makes the commonsense suggestion bicycling infrastructure should be installed first where cyclists ride, and collisions occur. At least until we have a fully built-out bicycling network.
I’m sharing the results of Hindman’s investigation, with his permission.
It’s a must read for anyone who cares about bike safety, and ensuring that everyone who goes out on a bike ride comes back home in one piece.
The Census Bureau’s American Community Survey data results from 2007 through 2013 have a doubling of commuting by bicycle from 0.6% to 1.2%. Los Angeles Police Department reported 1,335 bicycle collisions in 2007 and 2,413 in 2013. That’s a 81% increase. Although the bicycle collisions have significantly increased, the rate of collisions per total number of bicycle riders has no doubt fallen.
I did a totaling of type of collisions in the first 100 pages (about 500 collisions) of the 484 page 2013 bicycle collision report that mentions each collisions individually and found the reported collision type or primary factor in the collisions to be:
- 220 broadside
- 110 wrong side (usually got hit by driver turning right)
- 146 Right of Way auto
- 70 stop sign
- 40 improper turn
- 48 sideswipe
- 30 head on
- 22 rear end
- 10 improper turn
- 8 too close
- 5 improper driving
- 10 lane change
- 29 unsafe speed (usually unclear when that refers to bicycle or motor vehicle)
I haven’t seen anything in the report that mentions hitting a parked car door. There are several reports about hitting a parked vehicle though. I’ll try to figure out how many times that occurred in the total. Its much less frequent than getting broadsided.
Right of Way auto and broadside I assume would mean a bicycle running a stop sign or running a red light and a motor vehicle that had the right-of-way hitting the bicycle. I have yet to see a collision report state ROW bicycle, although it occasionally mentions ROW pedestrian.
The report does mention collisions when a motor vehicle was making a right-turn as a bicycle was going straight. I’ll try to see how frequently that occurred in relation to all types of collisions. This also seems to be a small proportion compared to the number of broadsides.
A Los Angeles Department of Transportation bikeway traffic engineer recently stated that they do not do treatments for bicycles at intersections. The bike lanes are striped where there are no crossing points for motor vehicles such as driveways, freeway on and off ramps, and cross street intersections.
The MIT Media Lab made a great looking map of all the LAPD reported bicycle collisions for 2012:
When I look at that map it seems to me that the bulk of the LADOT resources for bicycling should be concentrated in the areas of the city where the bicycle collisions are densely packed together. That’s also where the most bicycling occurs. If there are few staff members and a very small budget, then why try to install bicycle improvements across the whole city at once. That dilutes the effect by spreading out the improvements so much that they don’t connect into a network of any sort and the quality of the infrastructure won’t be as good because the emphasis is on quantity.
Hindman followed-up with a brief email providing a little additional information and clarification.
When I mentioned 70 crashes involving a stop sign it should be stop sign or traffic signal. I’m getting better at understanding the abbreviations in the crash data and hopefully I can tabulate the primary collisions factors and collisions types for 2013. I counted 16 bicycle fatalities for 2013.* One pedestrian was killed by a bicycle rider in 2007 and in 2012, but none in 2013. Both of these pedestrians were in their 80’s.
Spot checking the MIT Media Lab results of 54 bicycle crashes for Van Nuys Blvd I noticed that any time the LAPD bicycle crash data mentions Van Nuys as the primary or secondary street it was counted by MIT as a crash on Van Nuys Blvd. I have to assume that all the street crashes mentioned were totaled the same way.
*Editor’s note: My records show 18 bicycling fatalities in the City of Los Angeles in 2013. The discrepancy may be due to one rider killed in a train collision, and another who was walking his bike when he was hit by a car; it’s possible neither was classified as a bike collision in the LAPD stats. Two of the cyclists killed in 2013 died as a result of doorings.
This is an excellent summation of cycling accidents in LA. It would even be more powerful if it were displayed as a frequency plot against a graphic depiction of an intersection and a short stretch of road. It would then be obvious where one should ride to reduce risk of collisions. I would be willing to help in effort.
Thanks Dennis, we appreciate the fine work.
What this shows is that bike infrastructure really does jack squat for safety, as pretty much any infra disappears at intersections, which usually leaves the cyclist in a less than ideal position at that intersection.
It also pretty well debunks the LAB’s findings that rear-end collisions are the most frequent type.
Actually, the Bike League never said rear-end was the most common type of collision. They said it was the most common type of fatal bicycling collision, which corresponds with my data.
Rear-end fatalities typically occur on highways and other high-speed roadways, where the speed contributes to the severity of the impact. As a result, while rear-ends may be less common that other types of collisions, they are more likely to result in death to the rider.
The LAB analysis was that rear-end collision was heavily over-represented in fatal wrecks as 40% of fatalities were in this category despite them being such a small proportion of overall wrecks.
Don’t know LAPD’s practices, but in many communities, doorings are routinely not reported as car/bike collisions.
If they make collision statistics at all, they’re reported as cyclists hitting parked cars or fixed objects. But many places don’t routinely include single-bicycle crashes in crash reporting, they only get reported as collisions if they involved a motor vehicle in motion.
The Fed Hwy Administration says* the following: “The money saved by preventing bicycle and pedestrian injuries and fatalities more than offsets the costs of improving our streets and roads. The National Safety Council estimates the comprehensive cost for each traffic death at $4.1 million and $53,000 for injuries. For 2008 bicycle and pedestrian data, this equates to a cost of roughly $26.8 billion in a single year.”
We need to know why bike/ped/car collisions occur, yet wading through collision data is extremely difficult and time consuming. The spreadsheets for a city the size of L.A. are massive and require an enormous personal investment just to be able to comprehend precisely what the data reveals and what it doesn’t. With that in mind, I would like to ask that any analysis of SWITRS or FARS data be scientifically rigorous and disciplined in its approach. Let’s not make assumptions or draw conclusions where the analysis is incomplete or where questions remain unanswered or unanswerable. Please highlight what questions exist and remain unresolved in the data/collision reporting spreadsheets.
Understanding basic terminology is fundamental to collision analysis. For example, the SWITRS glossary defines a “fatal collision” this way: “A motor vehicle traffic collision resulting in the death of one or more persons within 30 days of the collision.” Similarly, a person “killed” according to the same glossary is: “A person who dies, as a result of a motor vehicle collision, within 30 days of the collision.” So, if a person dies on day 31 post-vehicular collision, the statistical representation of that death is different from what lay persons understand. And to add insult to injury, a “victim” is defined as: “Any person killed or injured as the result of a motor vehicle traffic collision.” That means a person harmed by an infrastructure defect without the involvement of a motor vehicle will not be considered a “victim” in the spreadsheets. It’s also important to know that someone who dies as a result of a collision is represented in the FARS database, not necessarily the SWITRS records.
As bicyclists, we all know that it’s f-ing dangerous out there for reasons within and beyond our control. So, careful analysis independent of law enforcement conclusions and public sentiment as to fault or right of way is important. The ultimate goal is to re-engineer the public right of way to make walking and bicycling safe, viable transportation options. That demands we know why collisions occur and how to fix them.
I’ve spent dozens and dozens of hours studying this stuff and only feel I’ve scratched the surface of the evidence in the data. It would be helpful for all of us if state and local officials would hold a series of workshops (a great grant opportunity project!!) for us advocates to teach us how to properly mine the databases and accurately interpret the evidence therein. That would certainly make more transparent the inherent opacity of collision statistics while facilitating a collaborative relationship between advocates, law enforcement, and policy makers.
No one should be able to hide behind obscure statistical manipulation to avoid spending taxpayer funds on roadway safety.
*Source: “Estimating the Cost of Unintentional Injuries” (2007). See: http://www.fhwa.dot.gov/livability/fact_sheets/transandsafety.cfm
“If there are few staff members and a very small budget, then why try to install bicycle improvements across the whole city at once. That dilutes the effect by spreading out the improvements so much that they don’t connect into a network of any sort and the quality of the infrastructure won’t be as good because the emphasis is on quantity.”
This is the exact same thing that David Hembrow suggested over three years ago. Perhaps it’s time to give the idea a good look and go about accomplishing it.
Potholes. Don’t forget about potholes. Lol.
Great job Dennis. If there’s anyone who can turn these stats into meaningful language, it’s you!
[…] Dennis Hindman Looks Into L.A. Bike Crash Data (Biking in L.A.) […]