The Missing Piece in Crash Reports: Ridership Data

Bike crashes reported by Chicago Police Department
An early version of what became ##

This is the edited text of a speech I gave Monday night for the City of Chicago’s Earth Data Celebration about the use of data in measuring sustainability performance measures in Chicago. 

As deputy editor of Streetsblog Chicago, and co-editor of Grid Chicago before that, one question I field quite often from readers is, “How many people are biking in Chicago?” Frankly, it’s nearly impossible to tell.

In 2006, the City of Chicago released the Bike 2015 Plan with two overarching goals: to increase bicycle use so that 5 percent of all trips less than five miles are by bicycle, and to reduce the number of bicycle injuries by 50 percent from current levels. Injuries are tracked consistently with crash reports and ER admittance records. Last year, the city commissioned the Active Transportation Alliance and the UIC Urban Transportation Center to analyze those; a report should be issued this year. But tracking bicycle use, be it the number of people bicycling on any given day, or the change in bicycling year over year, has been more elusive.

The crash and injury data is easy to obtain from the Illinois Department of Transportation, but it lacks a key complementary dataset: exposure. How many cyclists are out there, for how long, and where? I made a bike crash map in 2011, but without that information, I wasn’t able to tell people which intersection was the most dangerous when they asked about places they should avoid. I could only tell them which places had the most crashes, because I didn’t know how many people were bicycling by. The exposuire information would be factored into crash location analysis to help prioritize which locations need safety improvements the most.

There are many methods to track exposure, and new ones are coming to the forefront.

Traffic counting devices that use pneumatic tubes are much less expensive than they used to be and now cost about $150. Permanently installed counters that can track bikes cost about $2,000. There are video-based counters that, with a single camera, can count all the users — motorists, bicyclists, pedestrians — at one intersection. These can be rented by the minute. A growing trend in cities that consider themselves bike-friendly is to install “eco-totems,” or vertical displays that show the real-time count of cyclists who’ve passed that point. Beyond the data collection, this has the benefit of telling people riding their bikes that “they counted.”

Counting cyclists in Copenhagen
An eco-totem counter in Copenhagen.

If you’ve enabled location services on your smartphone, then your phone is keeping a log of everywhere you’re going. One new app, “Moves,” from Finland, reads your location log and determines which trips you made were by foot, by bike, or by motorized transport. You can change the determination if it’s wrong. An app called Cycletracks was used in San Francisco to track where people biked and when. Then there are devices like, created by students and faculty at the School of the Art Institute of Chicago to be a comprehensive environmental survey device, measuring ambient noise, lighting, and carbon monoxide as well as location.

Traditional methods of counting include travel surveys, in which all members of a household fill out a form about all of their travel activities for a day or a week. The Chicago Metropolitan Agency for Planning conducted a travel survey in 2008 and found that 1.3 percent of all trips in Cook County were by bike. The report didn’t separate Chicago from Cook County, but the data is available and could be crunched again. These surveys involve a lot of manual work on the part of respondents and researchers, however, and suffer from small sample sizes.

Chicago should experiment with some of the electronic counting methods in order to measure the performance of the Bike 2015 Plan’s goals. This information can also be used to measure the performance of the Chicago Climate Action Plan, which sets out to reduce emissions by 80 percent in Chicago by 2050. The plan states that every million trips by bike and by walking reduces emissions by 10,000 metric tons of carbon dioxide, but that in 2010, the number of trips by bike and by walking was only half a million. So there’s a lot of counting to do from now until 2050!

  • Anonymous

    Maybe you could partner with Inrix (or similar) to create an ap that would capture information from willing participant’s cell phones to create a rich database potentially describing rider demographics, 15-minute increment live travel times, routes people are taking, distance traveled, etc. A lot easier and more accurate than the old survey method.

    We need more and better metrics for all forms of transportation decision-making, including biking. Facts are sorely missing from multi-billion investment plans.

    As Einstein noted, we can’t solve the problems of today by using the same kind of thinking we used when we created them.

  • Anonymous

    Missed the paragraph about Moves and CycleTracks, etc . . .

  • BlueFairlane

    The problem with this kind of smart-phone method is that the only things it gives you are the number of riders with smart phones willing to be tracked, the distances they travel, and their preferred routes. It’s impossible to extrapolate anything about ridership as a whole from that, as you have no idea what percentage of riders have smart phones, what percentage of riders with smart phones are willing to be tracked, whether these particular riders are biased toward certain routes, etc. You get a picture of a very narrow demographic without any way of knowing how representative that demographic is as a whole. This makes the picture kind of useless.

  • Yup. I’ve presented a variety of methods that can complement each other, several of which Chicago has experimented. I’ve seen the MioVision camera set up in Chicago (in at least two places) but I don’t know if that was the city trying it out, or a consultant using it for a specific project.

  • Anonymous

    I suppose that depends on how many people one is able to engage. Agreed there may be resistance to “being tracked”, but some may find it fun and worthwhile – perhaps enough to make it meaningful.

    I suspect that the smart-phone approach is a lot better than a small sample of written surveys and extrapolating those findings. I’d also argue that bias toward certain routes may be indicative of travel times, level of comfort on a route, etc. – not necessarily flawed data.

    When I tap my Inrix map I see there are nearly 7,000 users “near me” right now. When I use the ap, it has generally been accurate to within two minutes travel time during peak commute hours (no matter which route I pick), despite questions surrounding type of vehicle, driver demographics, preferred routes, overall representativeness of the broader population of drivers, etc.

    Data usefulness is surely tied to the number of willing participants. As for the balance of data integrity concerns, they’re relatively easy to overcome.

  • BlueFairlane

    I like the camera idea, as I think it gives the most complete picture. I don’t know the specifics of how the pneumatic tube boxes work, but I’ve always wondered how susceptible they are to false information, ie, what else can trip them besides the thing you’re measuring. (There may be some sort of mechanism in the system that prevents that.) The smart phone apps and surveys each could have uses if you’re tracking certain specific things and arrange the study correctly. But the cameras give you the whole deal.

    You say they can rented by the minute. Any idea what that costs?

  • BlueFairlane

    I don’t think it really matters how many people who own smart phones and are willing to be tracked you’re able to engage, as you still are getting no information from people who aren’t willing to be tracked, or, more significantly, people who don’t own smart phones. It’s kind of like the demographic inverse of the political poll numbers you see around elections, which often just tell you what people who still own landlines and answer them at 4 in the afternoon think about a question. Or internet polls on websites that measure the opinion of people who answer internet polls on websites. Call as many people who own landlines and answer them at 4 in the afternoon as you want, and you still get no information about what people who don’t own landlines think. Track as many people with smart phones as you want, and you still see nothing of the travel habits of people who don’t own smart phones. The Huffington Post tells me that’s 54% of Americans. You can’t tell me that 54% of Americans don’t ride bikes.

    Bias toward certain routes, then, suggests the things you suggest, yes, but it’s tainted by the fact that this single demographic–people who own smart phones–are biased toward certain destinations. I think you’d have to agree, the demographic of smart phone owners tends to be younger and more affluent, and likely has preferences along certain racial lines. This group is going to gravitate toward certain destinations, which means you’ll get data skewed toward the preferences of that demographic. In this case, you’ll get exaggerated numbers toward the North Side, and specifically the Milwaukee corridor around Bucktown and Wicker Park, and along Clark up into Lakeview. You’ll get underrepresentation of the South and West sides and of groups that tend to be older or poor. You’ll always have the question of whether a route isn’t traveled because something’s wrong with the route, or whether it is traveled, but not by smart phone owners.

  • With Inrix, is an app even necessary? Can one purchase cell phone location logs based on the phone’s IMEI number that willing participants could give to the researchers?

  • When I worked at the CDOT Bicycle Program, we obtained an Eco-Counter brand pneumatic tube counter that used two tubes to measure wheelbase. With the wheelbase known, the device can tell the difference between bikes and cars. It can also tell speed and it uses that to separate motorized scooters from the “bikes” category (which also means that if you bike faster than the device’s threshold, you will be counted as using something that’s not a bike).

    Back to the tube and accuracy… I was part of a volunteer group to test the accuracy of the Eco-Counter device. I sat under a shade tree at Wells and Locust Streets to manually count cyclists going northbound. This tally was compared with the device’s counter and, if I remember correctly, the accuracy was ~5%.

    As for the cost, I just emailed MioVision asking them about it.

  • Anonymous

    Automated data collection for the travel behaviors of up to 50% or more of the population seems like an easy target for productive and meaningful data collection, though not by any means perfect information. Of course, data limitations need to be recognized, as well.

  • Anonymous

    idk – I needed to install the ap on my i-phone to use it, which also gives them access to my location data.

  • BlueFairlane

    I think in this case, it could useful in some circumstances. I don’t think it’s useful in planning or analyzing safety issues based on crash data.

  • CDOT will be proposing changes to street geometry for Milwaukee Avenue between Elston Avenue and Kinzie Street next Tuesday evening, which includes different bike lanes than currently exist.

    This segment includes the two intersections of Milwaukee/Ogden and Milwaukee/Chicago. These two intersections combined have the first or second highest number of bike crashes. The other location is Milwaukee/North/Damen.

    Given limited resources, where is money best spent (and when) to reach the goals of the Bike 2015 that say we must cut injuries in half and increase cycling (which is hard to do if people feel they’re going to be injured)?

  • Anonymous

    I don’t necessarily agree on planning limitations, but simple travel times and patterns from a group of self-selected participants would reveal little meaningful safety insights, though route volumes may be some form of proxy for perceived safety, and probably nothing about crash data.

    However, Inrix allows and encourages users to “live-mark” with waypoint icons certain types of hazards. If a similar functionality were deployed with bike riders using a similar system then some very detailed safety and accident data could be captured. I’m not how current the datasets are in use now, but imagine one that is in virtual real time. Not a replacement for standard practice and data, but to augment it.

  • Joseph Musco

    I think the cheapest method would be to use cluster sampling with the pneumatic tubes and then extrapolate for the entire city. Put the rigor in coming up with a true random sample each year and then repeat the study annually. Over time I think you would get excellent data about the city as a whole. I’d ask what a reasonable time and monetary commitment would be for a community bike census and then work backwards fitting the best methodology to the people and $$ resources.

  • CDOT has a decent bike count program going on right now, using manual counts at identical locations once a month. However, they have no intention of providing the data in a compiled tabular format to the public. They are currently a series of ~30 PDFs with aggregated/commingled data on Scribd.

    It needs to improve its automated counting.


IDOT Discusses the Future of Protected Bike Lanes on State Routes

The ribbon cutting for the Clybourn Avenue curb-protected bike lanes in Old Town on November 20 was a Kumbaya moment for the city and state transportation departments after years of conflict over protected lanes. Soon after Chicago Department of Transportation opened the city’s first protect lane on Kinzie Street in July 201 the Illinois Department […]

CDOT’s Two-Year Strategic Update Lacks Street Safety Data

The recently released update to the Chicago Department of Transportation’s first-ever two-year plan, the Chicago Forward Action Agenda, fails to give a status report on critical street safety metrics. The two-year update to Chicago Forward [PDF] does say a lot about what CDOT is doing to improve safety, but it doesn’t provide information on traffic crashes […]