When do cyclists crash?

By Steven Vance On July 26, 2012 · 16 Comments

Chart 1. The chart above shows the hourly activity of aggregated reported crashes in Illinois in 2010. It shows the hour of the day, that, throughout the year, saw the most injuries and fatalities.

This post is fourth in a series on crash data sponsored by Jim Freeman, a Chicago lawyer specializing in pedestrian and bicycle crashes. Read the other posts in this series.

The League of Illinois Bicyclists (LIB) recently posted a link on its Facebook page to an event in August called “Designing for Bicycle Safety”, hosted by the Chicago Metropolitan Agency for Planning (CMAP). A person commented,

We can design for bike safety until we are blue in the face, but unless bicyclists come to their senses and buy lights and reflective clothing for riding after dark – there will continue to be needless rider deaths and incapacitating injuries. I believe this needs to be top priority in rider awareness education.

Safer infrastructure should be the top priority in all things bicycling, and when it comes to reducing crashes at night, we agree that encouraging cyclists to use lights at night is important (Get Lit!). I wanted to know just how many crashes occur at each hour of the day. As is usual when it comes to bike crash data crunching, it takes longer than I originally thought or planned to get the full answer. In essence, though, the majority of crashes and injuries occur during “evening rush hour” while the majority of fatalities, while very small, occurred at night.

Update July 27, 2012: New, interactive charts show the same data in different ways.

Notes about this data

The majority of people involved in crashes in Illinois were in Chicago. I ran the numbers for the entire state of Illinois as this in response to a comment on the League of Illinois Bicyclists’s Facebook page. The data only applies to the year 2010, as that is the only year for which I have statewide crash data. The data only includes crashes reported to the police. The frequency of unreported crashes is unknown; hospital records would need to be consulted to gain a better understanding of how many crashes were not reported.

Data and interpretations

The most crash-frequent hour for fatal bike crashes was 8-8:59 PM.
The most crash-frequent hour for anyone receiving an injury (regardless of severity, but disregarding fatal crashes) was 5-5:59 PM.
The most crash-frequent hour for anyone receiving an incapacitating injury was 6-6:59 PM.
The most crash-frequent hour for anyone receiving an incapacitating injury, or who died in the crash, was the same, 6-6:59 PM.

The hour with the most crashes that resulted in fatalities was 8 PM to 8:59 PM, with 5 fatalities in Illinois in 2010. There was one in April, two in July, and one each in August and September. If you look at fatalities in two hour time blocks, you get the following, selected, results (each number in this article represents a person, not a crash):

7-8:59 AM: 2
9-10:59 AM: 3
11 AM-12:59 PM: 0
1-2:59 PM: 1
3-4:59 PM: 2
5-6:59 PM: 3
7-8:59 PM: 6
9-10:59 PM: 3

Only 3.9% of people in reported bicycle crashes in Illinois in 2010 walked away with no injury.

I’ll have to run a separate analysis to match up these times with the months in which they occur, so there’s a better understanding of which occurred before sunset and which occurred after, but generally, the crash rate follows ridership, and ridership follows the seasons.

Hourly ridership data is still lacking in Chicago; the Department of Transportation does monthly counts in the same locations but only for 4 hours each time (7-9 AM and 4-6 PM), one day per month (here’s July’s data). Automated bike count devices, like the EcoTotem from Eco-Counter, should be installed around the city to collect constant and consistent data every day of the year. With that information we can better understand the seriousness of the hourly crash rates and find anomalies in the data (for example, if the majority of crashes were occurring in the 5-5:59 PM period but the majority of cycling in the evening was from 6-6:59 PM – the information isn’t there).

Chart 2. This chart shows the monthly proportions of injury outcomes in reported bike crashes in Illinois in 2010. The outcomes appear relatively the same, except for a decrease in injuries in January and February. March and December have the lowest “no injury” proportions.

We could also find out if more people are crashing or being injured than expected. For example, are there more crashes and injuries at night on a weekend, or on a weekday? Why do injuries peak on Wednesday? What’s different about cycling ridership on Wednesday than Tuesday or Thursday?

Chart 3. This chart shows injury activity of aggregated reported crashes in Illinois in 2010 by day of the week. Wednesday is the day of the week that sees the most injuries with Friday seeing the second most.

A reader commented on the photo when it was posted earlier suggesting that crash rates on Friday and Saturday nights would be “sky high”. Here is the weekend breakdown:

Thursday night, 10 PM to 4 AM: 28 injuries, 1 fatality
Friday night: 10 PM to 4 AM: 55 injuries, 0 fatalities
Saturday night: 10 PM to 4 AM: 53 injuries, 1 fatality
Sunday night: 10 PM to 4 AM: 31 injuries, 0 fatalities

How does knowing when crashes occur help advocates hone their strategies in reducing the number of injuries, a goal in the Bike 2015 Plan?

http://blog.theplannersdreamgonewrong.com jason tinkey

Steven, I’d be curious to see the spatial distribution of crash events by hour, are there certain intersections/streets where rush hour crashes are clustered?

http://www.stevevance.net/ Steven Vance

Do you know how to make a timeline-animated map?

C L

My prediction is that there will be more crashes after sunset when that is factored in. Reduced visibility, with some cyclists not using lights at all, combined with the other things going on at those times.

I imagine alcohol plays a role late at night. Many of my cyclist friends who would never drink and drive think it’s okay to bike after a long night of drinking — and they’ll joke about how they were drunk and plowed into things, hurt themselves, etc. And unfortunately many car drivers have had too much to drink late at night.
http://www.flickr.com/photos/8432336@N08/ BlueFairlane

The weekness of this data for showing what you want to show is that we don’t have corresponding numbers of bicyclists on the road for those hours. Anecdotally, you know it makes sense that more cyclists will get hit during the rush hour, as there are more cyclists available to hit. It seems like the number of potential cyclists lessens as you move later into the evening, which makes the 8-8:50 fatality spike a bit of a puzzle. It’s hard to draw any real conclusion unless we know what percentage of total cyclists get hit, or the liklihood of being involved in a crash … and, as you point out, that data doesn’t exist. I would bet if you adjusted for per capita, You’d get the spikes in somewhat different places.

http://www.stevevance.net/ Steven Vance

Re: “I would bet if you adjusted for per capita, You’d get the spikes in somewhat different places.”
I don’t know what you mean. Or maybe I do, but I don’t know how that would work for hourly. Are you saying “per capita” as a measure of the number of people using the roads at that hour? Or “per capita” as a measure of the people using each mode?
Ridership data is key. 24-hour automated counts are inexpensive and very doable. The Portland Bureau of Transportation places 24-hour automated counters at various places around the city. Some of them might even be permanent. I noticed a pair of them on the Eastbank Esplanade during my trip last weekend. I’m subtly advocating for the city to purchase an EcoTotem counter that would be permanently installed and collects minute-by-minute data on the number of people cycling past it (and can publish the data in real-time or next day).

http://www.flickr.com/photos/8432336@N08/ BlueFairlane

What I mean is how many bicyclists are involved in crashes during a particular hour, compared to the total number of bicyclists riding at that hour. It stands to reason you will have more bicyclists hit during the rush hour, as there are far more people riding at that time. Increase the number of people riding, and you increase the number of people being hit, even if the likelihood of any one person being hit is lower.

Compare it to murder rates. In 2011, Chicago had 440 murders. Memphis had 118. You’d think, then, that you’re more likely to get killed in Memphis. But adjust for the pool of people available to be killed, which in this case is total population (2,707,120 vs 652,050), you see that Memphis has the higher murder rate. (1.8 murders per 10,000 people, vs 1.6 in Chicago.)

So in a world of perfect data, that’s what I would like to see, some comparison of cyclists involved in accidents at 8 p.m. for every, say, 10,000 cyclists riding at 8 p.m. Otherwise, if you just take the unadjusted numbers, you’re not really comparing even things. My guess is that there is a much smaller number of people riding at 8 than at 5. Therefore, one person hit at 8 constitutes a larger percentage of total ridership at 8 than 1 person hit at 5. It’s possible, then, that while more people are hit at 5, a higher percentage of people are hit at 8. The percentage–or likelihood–is what matters here.

But that just hits the problem you run into all the time, as that kind of data’s just not out there. I’d love to see hourly counts over a full year for thoroughfares across the city, but that’s not likely to happen. I will say I think you usually do a pretty good job of showing what you can with the data you have, but sometimes there’s just too big a gap in data to be conclusive.

http://www.stevevance.net/ Steven Vance

I’ve created some new visualizations for this data, so you can see the different injury categories side-by-side: http://gridchicago.com/highcharts/

cw

We’re looking at this data after the fact. Lights are important, but as is true with any attempt at finding a correlation, they do not imply causation. You can always find a third variable. A third variable I see during evening hours is driving speed. People drive faster outside of rush hour.

From personal observations, I see a higher proportion of speed demons out after 9. You know the people, the ones who mistake a long overpass for a NASCAR audition, and seem to think they’re excellent drivers despite their record of speeding tickets.

Vehicle speed is the number one determiner of fatality in auto accidents. Get past 30 mph, and no one outside the vehicle survives. If we had a speed graph for drivers by the hour, I think we’d start to see how fault in a lot of these accidents is shared. Maybe a cyclist wasn’t well lit, but you can usually see them with headlights if you’re staying under posted speed limits. Blast down a surface street at 60 mph without your high beams on, however, and it’s a different story. Red and white blinkies are only part of the solution.
kob22225

Without statistically significant exposure data, this raw data says very little. These raw numbers look like they will end up a very reasonable match to likely amount of bicycling done at those hours.

http://www.stevevance.net/ Steven Vance

Check out these new charts to see if there are other insights you can gain, without exposure data. http://gridchicago.com/highcharts/bikeinjuries.html

kob22225

Insights about what? Without something that indicates with good statistical significance amount of bicycling taking place versus hour of day, not sure how much farther you can take this data.

http://www.stevevance.net/ Steven Vance

For example, the final chart shows the percentage of each bike crash injury outcome within that hour. Turn off all the items except for “non-incapacitating injury”. This injury outcome is relatively steady throughout the day. The majority of people receive a “non-incapactiating injury”.
So even though the number of crashes peaks in the early PM rush hour, 3 PM to 5:59 PM, the type of injury experienced at that time isn’t different than most other other 3-hour windows.

kob22225

OK. I believe I got your point. Night and twilight has some fingerprint of having greater relative amount worse outcomes.
http://www.stevevance.net/ Steven Vance

And of course exposure data would help us know why that is. Are there more people riding at this time, or is there low visibility, or what combination of each?
kob22225

Poor quality/low use of bicyclist nighttime equipment seems best working hypothesis, given my lifetime of observation.