Caltrans studies bicycle crash risk

Alert: Wonky stuff ahead.

Although the absolute numbers of bicycle crashes in California have long been available through the Statewide Integrated Traffic Records System (SWITRS), evaluating crash risk has been difficult because of poor exposure data — we don’t track the number of miles people ride on a daily basis like we do for automotive traffic. Although statewide data is still limited, researchers from UCLA and UC Berkeley used newly available bicycle count data to better evaluate crash risk. They used this exposure data to better determine what factors (socioeconomic and infrastructure) correlate to higher crash risk.

Bicycle crashes January 2016 - July 2016
Red dots on this map show bicycle crashes as logged through the California Highway Patrol dispatch system, January 1 2016 through August 2 2016. This map is updated every 15 minutes. Click map for the updated zoomable view.

The study selected 500 locations with at least six hours of bicycle counting from the LA County Bicycle Count Data Clearinghouse. The researchers additionally conducted counts at an additional 14 locations in September 2015 where they found very high crash incidence, but no corresponding bicycle volumes data.

Among the findings and other factoids in this report:

  • Weak Data Collection
    • “Crash risk cannot be understood without bicycle count data.”
    • We don’t know how many crashes actually occur that involve people riding bicycles. A previous analysis of trauma center data conducted in San Francisco found significant underreporting in SWITRS. About 26% of bicyclist trauma cases were not reported to SWITRS, and cyclist-only crashes were dramatically underreported, with only 50% of cyclist-only crashes reported to SWITRS.
    • We still don’t have good exposure data to determine relative risk. Many bicycle counts occur only during peak times and where high bike traffic is anticipated.
  • Infrastructure
    • Right-turn-only lanes double your risk of a crash.
    • Riding on roads that are wider than median width double your risk of crashing compared against riding on roads of less than median width in the study.
    • Roads with three lanes (per direction) have more than triple the crash risk over one and two lanes per direction.
    • Truck routes triple your chances of being involved in a crash.
    • Roads with transit stops double your chances of being involved in a crash.
    • There’s no difference in crash rates for roads with allow parallel parking versus those without.
    • High volume roads (those with more than 20,000 vehicles per day) have higher crash rates than those with lower volumes of traffic. For comparison in San Jose, CA: just under 16,000 vehicles travel on Lincoln Avenue north of Minnesota, with just over 16,000 on Hedding. Around 50,000 vehicles travel on El Camino Real through Palo Alto every day.
    • Bikeways (both lanes and signed routes) have lower crash rates than non-bikeways. The study found no difference in crash rates between Class II and Class III bikeways.
  • Socioeconomic factors. The researchers looked at Census data for the neighborhoods they studied, looking at median income, race, and car ownership.
    • People riding through higher income and white neighborhoods with high car ownership are less likely to be involved in a bicycle crash than those riding through lower-income or Latino neighborhoods. The report does not speculate on the reasons this might be.
  • Report summary
    • “Keep building bike lanes”
    • “Be wary of crashes as a prioritization metric.” The report notes that areas with large numbers of crashes also tend to have many people riding bikes, and suggests “the lower-hanging fruit in terms of safety interventions is where ridership is moderate but risk is high.” ( I would personally add that “safety interventions” for locations that already have a high number of cyclists might also be warranted because they can potentially benefit more people. )
    • To best evaluate where safety improvements can benefit the most, “cities can begin by conducting counts at locations with high crash incidence, allowing planners to distinguish between high risk / moderate volume sites and low risk / high volume sites.”
    • “Bicycle boulevards are promising.”
    • “Corridors with high cycling volumes had lower injury risk, lending some credence to the ‘safety in numbers’ hypothesis.” ( And to be contrary and ornery, the higher cycling volumes might be because the corridor itself is safer to ride on. Perhaps a dozen people attempt to cross U.S. Highway 101 annually on their bikes where the Guadalupe River Trail regularly floods. Three of them get hit by cars. That’s a very low cycling volume and a very high crash rate. )

You can read the complete 97-page report here. Principal Researchers were Professor Emeritus of Urban Planning Robin Liggett at UCLA; and Jill Cooper, the Co-Director of SafeTREC at UC Berkeley. Thank you to Jim Baross of the California Association of Bicycle Organizations for forwarding this report, the members of which are engaging in a lively online discussion on the findings and merits of this report.

Related: Biking In LA looks more deeply at the the numbers of cyclist fatalities in the United States and in California. The latest California Household Travel Survey shows 1.5% of trips are taken by bike, while cyclists make up about 4% of fatalities each year. On a per capita basis, California ranks number six among the states in cyclist fatalities. As has been known for a while, Florida is number one in per capita cycling fatalities.

2 Comments

  1. More proof that it is hard to run data when you don’t/won’t collect it. Which is why cities/counties should link up to at least Strava data or some other such information provider. I do know that there are going to be holes for many reasons in Strava or any other sport minded data collection site. Not in the least being socio-economic….

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.