Gallery

A Look at U.S. Traffic Fatalities

I was talking with a friend yesterday when he mentioned he came across an article that looked at the number to accidents across the United States.  He said they had a map showing the per capita rates.  He described it as being darker in the south and in the Rocky Mountain states (with the exception of Utah), and lighter on the coasts.  I didn’t find the article he was referencing so I decided to pull the latest summary data of the U.S. Department of Transportation’s Fatality Analysis Reporting System (FARS) data from the IIHS (insurance institute for highway safety) and make the map myself.

iihs-map1
Fig 1. – Deaths per 100,000 Population (2013)

This didn’t exactly match what he described (or what I thought he described) so I broke the states out into quartiles.

iihs-map2
Fig 2. – Deaths per 100,000 Population Quartiles (2013)

Now that seemed closer.  He expressed his objections to the measurements.  He said using a per capita rate because it makes New York’s rate look good because New York City makes up a large part of the population and a lot of the people who live there don’t drive.  He thought that it would be better to use highway miles traveled as a denominator.  He thought rural states would not be as dark as they would using a per capita measure.  Well I wanted to explore the data and see if there was something to his objection.  The data existed in the IIHS data set so here’s what I produced side by side to the above maps for easy comparison:

iihs-map3
Fig 3. – Deaths per 100,000 Population & per 100 Million Miles Traveled (2013)

By changing the rate from a person based to a mileage based measure we see some impact.  For example we see changes in the quartiles (i.e. Virginia is considered a safer state and Texas a more risky one).  We don’t see much of a change for New York (the shade of orange is slightly darker).  Wyoming seems to be a much safer state when viewed through a per 100 million miles lens. For those interested you can see my code in the project’s GitHub repository.