May 28, 2017

Mapping Traffic Fatalities on America's Roads

Aziz Kamoun

The number of traffic fatalities in the United States increased over 7% in 2015 relative to the previous year, the largest such increase in almost half a century. In response, the Department of Transportation, National Highway Traffic Safety Administration, and White House issued an unprecedented call to action for data scientists to unpack the 2015 surge in roadway deaths.

The map below captures all traffic fatalities in the United States between 2001 and 2015. Click on a marker for a 360° view of the roadway where an accident occurred, powered by the Google Maps Street View Service. Play the map through time or draw a polygon to see the factors most closely associated with accidents in a given area. The DataFace hopes this exploratory tool can help public officials identify problem roadways, understand why certain accident types persist over time, and shed light on what future technologies could help eliminate needless loss of life.

All Accidents Drunk Driving Pedestrian Cyclist Rain Frozen Precipitation Cloudy
Accidents by Type in Polygon Region

Here's the good news: despite the recent uptick, traffic fatalities have been steadily decreasing for the past several decades. Deadly traffic accidents have dropped 29% since 1980, and that's not even controlling for the fact that Americans now drive twice as many miles each year. In terms of fatal accidents per vehicle mile traveled (VMT), America's roads are almost three times as safe as they were 35 years ago.

Even as fatal accidents per VMT have declined in every state, there’s still notable regional variation. Some states like Nevada and New Mexico — which were among the most problematic in 1980 — have seen drastic improvement in road safety and have almost converged with the national average. On the other hand, southern states like Louisiana, Mississippi, and Arkansas have maintained roadways consistently more dangerous than the U.S. at large. As of 2015, South Carolina ranks worst in terms of fatal accidents per VMT, with 1.76 accidents per 100 million miles compared to an average of 1.04 nationwide.

So which states are playing a role in the recent surge in fatal traffic accidents? Is the surge concentrated in the South? Not really. In fact, the increase in fatal accidents per VMT has been spread across the country, with 36 states experiencing an uptick in 2015. Oregon and Maine saw the biggest jumps by percentage, at more than 20% each. In Oregon, that equated to 90 more fatal accidents than the year before.

So what’s behind the increase in fatal accidents? We’ve come across a few explanations. The NHTSA attributes the overall accident increase to cheap gas and post-recession job growth, which tend to encourage more driving time. Indeed, Americans drove 3.5% more miles than they did in 2014. Yet we’ve seen fatal accidents decrease over the last few decades despite continual growth in the number of miles Americans were logging behind the wheel.

Another plausible explanation was offered recently by The New York Times. In an article published back in November, Neal Boudette cited drivers distracted by mobile phones as a leading cause of traffic fatalities. NHTSA’s data doesn’t provide an indication of whether any drivers involved in a crash were distracted. But with texts, calls, and smartphone apps now warring for a driver’s attention, distraction seems a likely culprit.

Even so, there are plenty of reasons for hope. Ongoing efforts to curtail distracted driving have provided some positive early signs. And with companies like Ford and BMW pushing to put self-driving cars into commercial use by 2021 — not to mention the strides being made by Google, Uber, Tesla, and a host of others — we could be looking at a world without human drivers in a not-too-distant future. In such a world, the concept of a “distracted driver” would become a thing of the past. For at least the next few years, however, we’ll have to combat rising fatalities on America’s roads with strategies that address the possibility of human error.


Methodology:

  • Reliable global position data is only available from 2001 onwards in the NHTSA’s Fatality Analysis Reporting System (FARS) dataset.
  • To approximate pedestrian and cyclist traffic fatalities, we queried for “Pedestrian” and “Pedalcycle” in the First Harmful Event data element, which the 2015 FARS Analytical User’s Manual defines as “the first injury or damage producing event of the crash.”
  • Frozen precipitation is an aggregation of values from the Atmospheric Conditions data element in the FARS dataset, including “Sleet,” “Hail,” “Snow,” “Blowing Snow,” and “Freezing Rain or Drizzle.”
  • The Atmospheric Conditions data element only has “Cloudy” as a possible value beginning in 2010. As such, prior to this year the only weather data we map are rain and frozen precipitation.