Data-Driven Streets: How UVA Darden’s Chris Parker is Reshaping Traffic Safety in D.C.

In the bustling streets of Washington D.C., a city striving for zero traffic fatalities, an unexpected ally has emerged from the halls of academia. Chris Parker, tenured associate professor at the University of Virginia's Darden School of Business, has turned his attention to a critical yet often overlooked aspect of road safety: the very data that informs the city's safety decisions.

D.C.'s Vision Zero Struggle

Washington’s Vision Zero initiative, launched in 2015, set an ambitious goal: to eliminate all traffic-related fatalities and serious injuries by 2024. Part of a global movement, the program focused on redesigning streets, enforcing traffic laws, and educating the public to enhance road safety for pedestrians, cyclists and motorists.

Despite some progress, such as expanded bike lanes and lowered speed limits in certain areas, the initiative has faced significant challenges. Critics point to uneven investments in infrastructure, particularly in lower-income neighborhoods, and a lack of urgency in meeting targets. City officials acknowledge that while strides have been made, much more work is needed to truly transform the safety of the city's streets.

As the 2024 deadline approaches, it’s clear the city will fall short of its goal. Traffic-related fatalities have remained stubbornly high, rising in all but two years since Vision Zero's implementation.

Last year marked a grim 16-year high with 52 traffic deaths. According to recent data, 33 people have already lost their lives in traffic crashes this year. A Washington Post analysis suggests that the nation's capital is on track to match last year's tragic statistic, underscoring the urgent need for more effective strategies to improve road safety in D.C.

The Hidden Patterns in Crash Data

Parker's journey into traffic safety began with a startling realization: the official crash data in D.C. was significantly underreporting accidents, potentially skewing critical safety decisions. His research, conducted in collaboration Howard University’s Karthik Balasubramanian and data scientist Charlotte Jackson, revealed a disconnect between reported crashes and the reality on the streets.

“We're trying to de-bias the police-reported crash data,” says Parker. His team's analysis uncovered a significant discrepancy: “They are under-reporting crashes by about 30% or 40%.” This gap in data isn't just a statistical anomaly — it has real-world implications for safety and equity across the city.

Unveiling the Underreported Crashes

The process of how crashes are recorded in D.C. is at the heart of the problem. As Parker describes, “Somebody calls 911, police get dispatched to the scene. The happy path is that the police record what happened.” However, reality often deviates from this ideal scenario. Minor fender benders often go unrecorded, and if no one is present at the scene, no report can be filed. There’s also the issue that not everyone involved in a crash is comfortable calling law enforcement.

“And so, what we end up seeing is D.C. using the official report to decide where the dangerous places in our city are based on the crash count,” says Parker. But the data is flawed.

Parker's research revealed a stark disparity in reporting across different areas of the city.

“We have places like Georgetown, where you have relatively wealthy Caucasians who are very happy to interact with the police, and then you have Anacostia, lower income, people of color, where crashes are not being reported as much,” Parker notes.

This bias in reporting leads to a troubling consequence: “The city is throwing money in places where the wheel is squeaky,” says Parker, “and it's not investing in the infrastructure in southeast D.C.”

Data-Driven Solutions

To combat these issues, Parker and the rest of the Data-Driven Streets team, backed by substantial funding from the D.C. Highway Safety Office, aims to provide better decision-making tools for city planners:

1. 911 call analysis: They transcribe 911 dispatch call audio into text, using it to locate incidents and match them back to official records.
2. User-generated crash reports: They find crashes reported through apps such as Waze that may never reach 911 dispatch to build a fuller picture of crashes throughout the city.
3. Interactive dashboard: The team is creating a dashboard to visualize unreported crashes and identify areas with excess crashes relative to traffic volume.

Beyond Crash Counts

Parker's research goes beyond merely counting crashes. “Where are segments where we're seeing more crashes than you might expect? That's a place where you want to focus your attention,” he says.

The team is exploring various safety measures:

• Daylighting expansion: “Daylighting is massive,” Parker says, referring to a practice that restricts parking near intersections to make pedestrians easier to see.
• Traffic calming measures: Such as bulge-outs at intersections to slow right-turning vehicles and protect pedestrians.
• Harsh braking analysis: Using data on sudden braking events to identify risky areas.
• Automated traffic enforcement: Using speed data to determine how new traffic cameras impact driver speeds.

A Holistic Approach to Urban Safety

The project even extends to school districting. Parker's team is developing a risk score for every block in D.C. and creating an optimization model to balance walking distance with safety for elementary school students.

“We're building optimization models for weighing the two main concerns: walking farther versus walking over shorter distances, but across a busy street,” Parker says.

The Power of Data for Social Good

This research exemplifies how data science can be applied to critical urban issues. As Parker puts it, “As data people, we need to be able to use our skills in ways that have an impact.”

As D.C. continues its journey towards safer streets, Parker's data-driven approach offers a promising path to more equitable and effective traffic safety measures, potentially saving lives and reshaping the urban landscape for the better.