DataScience is supporting Los Angeles's efforts to eliminate traffic deaths by 2025 through pro-bono work with Vision Zero, a road-safety initiative that began in Sweden more than a decade ago. The initiative, which has been implemented in other U.S. cities including New York, San Francisco, Seattle, and Portland, takes a data-driven approach to identifying problem areas on a city's roads and making much-needed safety improvements.
We've downloaded and geocoded data from the Statewide Integrated Traffic Records System (SWITRS) and will be integrating data from additional sources. Over the next year, we will be mashing up the gigabytes to identify the most effective mechanisms for reducing traffic fatalities. Ben Van Dyke, one of our data analysts, has experience analyzing National Highway Traffic Safety Administration (NHTSA) data in Python, and he's been the first to take a crack at the data and generate the initial insights below. Ben's national data engineering work will soon be featured on the fantastic y-hat blog. We built on his LA-based work to generate lists of hazardous intersections which we'll be cross-referencing with LA's "High Injury Network" data.
In this first preview of our LA-based analysis, we've compiled below:
  • Insights from our initial exploration
  • Top 10 Worst Intersections for Collisions in LA
  • Top 5 Worst Intersections for Pedestrian Injury in LA
  • Top 5 Worst Intersections for Hit and Runs in LA
Early Insights

According to data from 2015:

  • Pedestrian fatality rates are much higher than motorist rates. In fact, pedestrians are nearly 5 times as likely to receive life-threatening injuries in a collision than non-pedestrians.
  • Pedestrians fatalities, surprisingly, frequently occur in crosswalks. Over half of pedestrian collisions occur in the crosswalk when the pedestrian has the right of way.
  • Motorists are most frequently at fault for pedestrian injuries. Motorists are at fault (as determined by law enforcement) 41% of the time. Pedestrians are at fault in 31% of cases, while bicyclists are at fault in less than 1% of cases.
  • More than 1 in 4 pedestrian injuries are the result of hit-and-runs. Hit-and-runs account for 27% of all pedestrian collisions. Fault in these cases is unknown, as law enforcement does not record fault in most hit-and-runs.

Questions to Pursue:

Data exploration is the search for interesting questions. Here's a few, based on the collision data alone, that we want to answer in upcoming analyses:

  • Why are crosswalks at intersections in Los Angeles so dangerous? Motorists should expect to see pedestrians in these areas.
  • For collisions not in a crosswalk, can we establish distance from the nearest crosswalk?
  • Why are there so many hit-and-runs and where do they happen?
  • How are collision totals and types changing over time?
  • Can we compare Los Angeles to other metro areas (NY, SF, Chicago) in pedestrian fatalities per capita and hit-and-run rates?

LA's Worst Intersections

Many of the locations where collisions occurred in 2015 had multiple accidents some as high as 20, in the case of Imperial Highway and Vista Del Mar. We've compiled some lists of the worst intersections for collisions, pedestrian injury and hit and runs below.

Here's a map of the most dangerous intersections:


And in ranked order:


 Subset by Pedestrian Injuries:


 And, by hit and runs:


We'll be building out interactive maps as part of our work. As an example, here's the cumulative growth of alcohol related incidents this year:



We'd love to hear your questions and comments as we work towards the overarching goal of Vision Zero: to identify and implement interventions most likely to save lives. There's many directions we can go in and we'd love requests and suggestions as to which angles we should pursue. We're also interested in collaborating and welcome interest from coders, designers, statisticians and all data enthusiasts. For more information, please contact us at with "vision-zero" in the subject line.

Dave Goodsmith
Dave Goodsmith

Data Scientist. Fan of neuroeconomics, kayaking, algorithmic music and the zen of data wrangling. @thegoodsmith