In D.C., the District Department of Transportation (DDOT) is responsible for planning and building the city’s transportation infrastructure, including where bicycle lanes, crosswalks, and safety features are installed. When making decisions about public infrastructure investments, DDOT relies on public crash data provided by the Metropolitan Police Department (MPD) to understand where crashes happen in the District.1 However, multiple sources of crash data exist, and the reported number of collisions varies wildly between sources, indicating potential issues with how the city documents collisions, which could lead to underreporting on public safety.
Comparing data from various sources show that crash reports systematically miss certain incidents. Sometimes this is the result of the fragmented nature of enforcement in D.C., sometimes because of how 911 calls are coded, and sometimes because parties involved in the crash refuse help. This prompted us to ask: How big is this problem? What is DDOT missing when relying on crash reports? And what should the agency be concerned about going forward?
About MPD crash reports and other crash data
An MPD PD-10 report detailing a crash in the city is required every time there is an injury of any kind, whenever there is any serious traffic violation or suspected criminal activity, or if there is any major property damage. This means that fender benders and other low-speed collisions are likely excluded from the data, but that this dataset should encompass all major crashes and all injuries caused by vehicle collisions: both incidents of a car hitting another car, and incidents of a car hitting a bicyclist or pedestrian.
Over the years, bicyclist and pedestrian advocates have speculated that the public crash reports are not fully capturing the extent of serious incidents and injuries in the city, pointing to anecdotal evidence of crashes that didn’t have corresponding reports in the public data.
The reported number of collisions involving cars in the District also varies wildly between sources. For example, two annual reports on the city’s trauma centers, released by the District Department of Health (DC Health) in 2016 and 2017, also report on motor vehicle crashes. These reports show that in 2017, 252 pedestrians were admitted to D.C. trauma centers due to motor vehicle crashes. In addition, 240 bicyclists were admitted, 74 of whom were tied to motor vehicle crashes. The 2016 trauma report numbers closely resemble those of the 2017 report —and are distinctly higher than numbers report by MPD in following years. Specifically, pedestrian crashes reported by trauma centers are more than double the numbers released by MPD (122 pedestrian injuries in 2019), and cyclist injuries are four times as high (52 bicycle injuries reported by MPD in 2019). Unfortunately, no direct comparison of data can be made, as trauma reports only exist for 2016 and 2017, and publicly-available MPD crash data are not available data for those years.
To test whether the public crash data was capturing all crashes and injuries called into 911, we scraped data on 911 calls in progress to D.C. Fire and Emergency Medical Service (DCFEMS) from Pulsepoint2 for the six weeks between March 18 and April 30 of 2021. We then matched the data to the MPD crash reports available on the District’s Open Data portal.
The DCFEMS data report whether collisions were between vehicles or if someone was struck outside of a vehicle, but do not distinguish whether the “pedestrian struck” outside of the vehicle was a pedestrian, a bicyclist, or someone on a scooter. We also listened to available EMS scanner audio for collisions in the six-week period (and were able to gather scanner audio for 85 percent of incidents). In scanner audio, the dispatcher sometimes identified the type of crash involving someone outside of a car, but often did not. Due to these limitations, we have combined data on collisions for pedestrians and bicyclists who have been hit by vehicles into a single “pedestrians” category.
We matched collisions reported to DCFEMS with MPD crash reports based on the date of the incident, the type of collision, and its location. To account for errors in the data and make sure all crashes reported by MPD were matched, we expanded the location of each crash to a 500-meter (approximately 550 yards) radius around each reported location. As such, a 911 call is marked as not having a matching MPD report if there was no reported crash of the same type (person outside vehicle, or vehicle-only), on the same date within 500 meters of the 911 call location. When one MPD-reported crash matched multiple 911 calls, the report was matched to the closest call location. The 911 call data was cleaned to ensure that there was only one 911 call per incident, ensuring that crashes marked unreported are separate crashes and not duplicate calls.
A significant percentage of 911 calls are unreported in crash data
In the six-week study period, there were a total of 2,802 collisions called into DCFEMS and/or reported by MPD. Out of 2,566 total vehicle-only collisions, 259 (10 percent) were called into 911 but did not have corresponding MPD crash reports— the majority of which occurred in Wards 2, 7, and 8. Importantly, for the 236 collisions involving a person outside of a vehicle (most commonly a pedestrian or a cyclist), 71 (30 percent) were called into 911 but did not have corresponding MPD crash reports—a rate three times as high.
Of the total crashes called into DCFEMS or reported by MPD, over a third of the crashes occurring in the District took place east of the Anacostia River in Wards 7 and 8. Ward 3 had the fewest crashes, at approximately four percent. This pattern likely reflects the fact that the I-295 freeway runs through Ward 7 and 8, and many commuters coming into D.C. pass through those wards to get downtown.
Unreported crashes (those with a DCFEMS call that could not be matched to an MPD crash report) followed similar patterns, with the most unreported crashes occurring east of Rock Creek Park. Wards 7 and 8—along with Ward 2—had the highest percentages of unreported crashes. However, because Wards 7 and 8 had a much higher volume of crashes, this means that the plurality of unreported crashes are occurring in Wards 7 and 8.
Crashes in which a car hit a pedestrian or bicyclist occurred disproportionately in Wards 6, 5, and 2
Out of all the crashes in which a car hit a person outside of the vehicle, such as a pedestrian or bicyclist, the most crashes occurred in Ward 6 (17.8 percent), Ward 2 (17.2 percent), and Ward 5 (16.9 percent). Within each ward, Ward 1 has the highest percentage of crashes in which a car struck a person outside the vehicle (16.4 percent of the crashes that occurred in Ward 1), versus a car hitting another car. However, because Ward 1 has less than half the volume of crashes (201 crashes in Ward 1) than Wards 5 and 6 (428 crashes and 438 crashes, respectively), Wards 5 and 6 still have the highest number of crashes in which a person was struck outside the vehicle.
A car crash involving a pedestrian or bicyclist is three times less likely to be reported in the crash data
While 90 percent of the car-only crashes had a corresponding police report, one third of the crashes in which a car hit a pedestrian or bicyclist went unreported in the public crash data. While there might be legitimate reasons why these crashes go unreported, the missing information on how many crashes happen and where they happen makes MPD reports an incomplete source for DDOT’s planning purposes. For example, incomplete data may conceal a pressing need for bicycle lanes, sidewalks, stop signs, and other measures necessary to protect people walking and bicycling in the District.
Why might there be no MPD crash report for a 911 call?
We can think of a variety of reasons why it was not possible for us to find a corresponding MPD crash report for all DCFEMS calls involving a serious crash. These include lack of MPD jurisdiction at the location where the incident happened and parties leaving the scene of the crash or refusing emergency medical attention:
Crashes could have occurred in places in the city not under MPD jurisdiction. There are several law enforcement agencies in the District including the United States Marshals Service, the U.S. Park Police, U.S. Capitol Police, Federal Bureau of Investigation Police, and the National Guard. D.C. has small geographic borders and a significant amount of park land that is managed by the National Park Service–as such, collisions that happen in parks are likely to be under the jurisdiction of United States Park Police, and so MPD would have no reason to create a report. Additionally, some crashes that occur near D.C.’s borders with Maryland might have been handled by neighboring jurisdictions of Montgomery County or Prince George’s County. While data on crashes occurring in other jurisdictions within the District is necessary for a complete safety profile, it is clear from the map of unreported incidents that crash locations occurring in other jurisdictions is not the main driver of missing crash reports.3
Finally, car crashes might be underreported if collision-involved individuals fled the scene or refused EMS treatment. Of the 71 incidents for which there is a 911 call but no MPD report, there are at least 21 instances in which the person injured at the scene refuses EMS care or left the scene.4 In such a situation, MPD policy is to not issue a report if there is no cooperating victim.
Regardless of the reason for people refusing to cooperate or leaving the scene, this discrepancy causes official crash report data to be incomplete and not accurately reflect the true number of pedestrians and bicyclists being hit by a car in D.C.
Why might reported crashes not have a corresponding 911 dispatch?
We also found that some incidents have MPD crash reports without corresponding 911 calls. This occurs when there is no 911 call made but MPD was involved, or when there is a 911 call but it was not dispatched as traffic collision (and thus call data was not collected for this report):
There could have been no 911 call for EMS services. MPD, per their own standards, are required to write a report for any crash that resulted in any type of injury, even a minor one. It’s possible that for the 71 pedestrian crashes without a corresponding 911 dispatch, one or more involved parties experienced an injury that met the threshold for an MPD report, but was not severe enough to merit a call for emergency medical services. For vehicle-only crashes, it is possible that people are more likely to file police reports (even if there is no injury) as they are often required for insurance claims.
Some incidents are dispatched by DCFEMS as “medical emergencies” instead of “traffic collisions.” Medical emergencies could be any acute medical issue such as a heart attack or a serious injury, and some of the crashes, particularly ones in which a person was struck outside of the vehicle, could have been dispatched as medical emergencies rather than traffic collisions.
Of the 236 total collisions involving a person outside of a vehicle for this period, 71 of them did not have a corresponding DCFEMS traffic collision incident.5 To be clear, we do not consider the lack of 911 calls in this case to be a safety issue.
However, this points to another problem: there are possibly car crashes that have no 911 calls and no police reports. We have no way of knowing how often this happens, but we can only say that the numbers presented in this report are a low estimate of unreported crashes, given that the data here only represent crashes that were coded by DCFEMS as “traffic collision”. Including medical emergencies in which a car hit a person walking or bicycling, or another car, is necessary in order to gain a complete picture of the safety incidents in the District, and would likely reveal a larger number of 911 calls without corresponding MPD reports.
What is needed to have a complete picture of crash data?
At present, the public crash data reports are the only source of data that DDOT uses to make decisions about transportation infrastructure and safety. For DDOT to have a complete picture of the crashes that happen in the District, the agency needs a dataset that incorporates traffic incidents coded as medical emergencies, crashes where people have left the scene, and collisions that were not reported by MPD for any other reason. For this to happen, DDOT should coordinate with DC Health and DCFEMS to learn where people are being injured. Without a complete dataset on where people are being injured, especially where people who are walking, bicycling, or scooting are being injured, DDOT cannot make truly informed decisions about transportation infrastructure. Many locations will continue to remain unsafe, particularly for these most vulnerable road users.
- For cases when MPD is not required to generate a report, there is a form to report crashes directly to DDOT so that can be captured in the data. Self-reported data, however, is often unreliable, as not all people file reports and many people do not know of this form or do not see the value in filling it out.
- Pulsepoint is an online website and application that shows all 911 calls in progress, the type of incident, the location of the reported incident, and other details such as if an ambulance was dispatched.
- Additionally, it is MPD policy not to write crash reports for crashes occurring on private property unless there is a reported injury, a diplomat was involved, or an MPD vehicle was involved. While this may include some cases, almost all crashes for which 911 calls were made occurred in intersections and streets (public property) and it is likely that crashes that merit a 911 call include an injury of some kind. See: https://go.mpdconline.com/GO/GO_401_03a.pdf
- It is possible that this occurred in more than 21 of the crashes, for example, in crashes for which we do not have scanner audio or in crashes in which the person refused treatment but it was not announced in the audio.
- In the collisions for which a car hit someone outside of the vehicle, there were 71 incidents (30 percent) for which there was an MPD report but no DCFEMS (911) traffic collision dispatch, as well as 71 incidents (30 percent) for which there was a DCFEMS (911) call but no MPD report.