What is Crash Count Chicago?

What is the current state of this site?

I'm treating Crash Count Chicago as a side-project, so I describe its current functionality as: well, it's useful to me. It downloads data daily from the city, and I can look up serious crashes near me and get a count of how many pedestrians and cyclists have been hit and how many people have died — and I can see how bad it is now compared to in the past.

The city's data is accurate enough when you look at it zoomed out. But the details can be confusing or plain wrong when reading the individual records. I've tried my best to remove or redact data that seems to be obviously incorrect:

• The city data classifies 20 serious crashes with first_crash_type=ANIMAL. But it appears that the incapacitated/killed "person" might be an animal (e.g. the data says they were taken to veterinarian hospitals). It's impossible to tell if a person was actually hurt in these crashes, so for now I've removed them from the "serious crash" tally (which purports to be about crashes where "people" were seriously injured or killed)
• Some crash reports have duplicated victim and vehicle records, ostensibly because old records weren't removed after amended ones were added. I've removed those duplicate records where I've identified them in 20 fatal crash records.

Unfortunately, there's a lot more cleaning work needed — for example, there's at least 100+ serious (non-fatal) crash records have mismatched injury tallies, which I haven't had time to manually research and fix. I don't have a short-term solution, so for now, I can only advise you to treat every number and data point on this site as being provisional.

What's so complicated about the data?

Data is the plural of anecdote, so let's look at how one story of a crash is currently recorded in the city data.

On the Monday afternoon of February 28, 2022, 41-year-old Gerardo Marciales was riding a Divvy bike in the Loop when a BMW sedan struck him as he crossed DuSable Lake Shore Drive. According to reporting by Streetsblog Chicago, Marciales was at a crosswalk and had the signal to cross. The 26-year-old driver of the BMW had the green signal to turn left, but apparently instead drove straight through and hit Marciales.

Marciales was taken to Northwestern Hospital in critical condition, and he died 2 days later. According to a police statement to Streetsblog, the driver was not cited.

Marciales is currently listed twice in the city's data. In his initial record, he's assigned blame for the crash: driver_action="DISREGARDED CONTROL DEVICES", pedpedal_action="CROSSING - AGAINST SIGNAL" However, in the newer record, his reported action has been updated to say "CROSSING - WITH SIGNAL".

Typically, when a record is amended, the city's system removes the obsolete version, and apparently that hasn't happened in this case. This is an example of a clear cut error that I can easily "fix" by compiling and maintaining my own "ignore this record where id=[old record id]" list, as I mentioned earlier.

Here's how Marciales' crash record looks like now on this site, versus how it looks like had I not manually fixed the records after downloading them from the city's data portal.

But that's the easiest of the data problems to programmatically identify and deal with. The clearest outstanding error is how the city's record classifies the 41-year-old male cyclist's injury as an "INCAPACITATING INJURY", when it should be "FATAL". Since Marciales died 2 days after the crash, it's understandable his death wasn't initially recorded in the city's data. But it should have been updated by now, 4 years later.

Other possible errors and omissions are harder to classify. The city data records the primary cause of the crash as "DISREGARDING TRAFFIC SIGNALS", but doesn't specifically assign blame to the driver: driver_action=NONE. Is that a data entry problem? Or a misjudgment by the on-scene police officer? Or did the driver, contrary to Streetsblog's reporting, actually have the right of way when he drove through that intersection? That's one of the many things that can't be known from the published data alone.

Right now the crashcount.org page for Marciales' crash reproduces the city's data and its apparent errors. Unfortunately I don't have a wide scale programmatic query that can flag and fix all the discrepancies between reality and reality as converted into a database record. I'm still thinking how to filter and improve upon the city's data that doesn't require me to manually research and edit hundreds of records. But I also hope that making the city's crash data more visible and accessible will spur interest in making that data as accurate and up-to-date as possible.

The future of Crash Count Chicago

An incomplete list of improvements I hope to add:

Increasing the scope of the crash data

There are a lot of crashes which didn't kill or incapacitate someone, but are obviously "serious". So I intend to include more of the 915,000+ crash records from the city's data, such as the 81,000 crashes that recorded at least one non-incapacitating injury.

The streets and intersections in my mapping database are compiled from the city's street center line shapefiles. I plan to include streets by segments since it makes much more sense to view crash activity by blocks instead of entire streets.

Supplementing the crash report data with news reports

As I've mentioned above, the crash data is limited in how the city documents and publishes crashes. Beyond basic facts of what, where and when, we don't know what led to crash, or even whether anyone was even cited, nor anything about the victims beyond age and sex. While ~130 traffic fatalities in a year is far too many compared to the city's Vision Zero aspirations, it's not too many to research them and augment their reports with links to related stories.

More analysis and reporting

I'd like to make this site more than just a database of anecdotes by building some analytical tools and making it easy to compare specific locations and time periods, e.g. if you want to see the impact of a new street improvement.