|

Vehicle Accidents in the USA

Driving is dangerous.  Many sources claim that driving is the “most dangerous thing most people ever do,” it is easy to find many sources which support this idea. I wondered, how dangerous is driving really? Does the level of danger vary based on any meaningful factors? Are there any interesting trends that we can see related to the risk of driving a vehicle? If driving is one of the things most likely to kill me it seemed worthwhile to investigate it more deeply. The result of my inquiry became this analysis, an attempt to learn more about the mundane but murderous practice of driving in America. 

Github

Notebook Part 1

Notebook Part 2

The data source for this analysis is from the National Highway Transportation Safety Administration, a government organization focused on road safety. The NHTSA produces a variety of fascinating, high-quality datasets including those from the Fatality Analysis Reporting System (FARS) which is available on the NHTSA website

Many of the findings from this analysis are not surprising, namely that drunk driving is a huge problem in the United States and that the chance of being involved in a fatal vehicle accident increases on the weekend. It may also not be a surprise to learn that motorcycles are much more dangerous to operate than other vehicle types and have a higher chance of death given an accident.  

Not all insights were so conventional: when trying to answer the question “which state has the worst drivers” it turns out there are many ways to approach the question and it quickly becomes difficult to compare states. How can a state the size of New Hampshire be compared to Texas? It quickly becomes a question of apples and oranges. 

By adjusting for the number of registered vehicles and population size this analysis found a middle ground that prevents misrepresentation. This is done by creating casualty rates, the number of people killed in accidents per 100,000 registered vehicles. This type of metric prevents over representation of states like California which have a high population and many registered vehicles. This methodology also prevents underrepresentation from states like New York, which have a large population but many fewer drivers due to the presence of urban centers, or North Dakota, which has very few people but many vehicles given the population size. 

The conventional wisdom that New York has superior drivers to New Jersey is visible in this analysis, though to be fair the states have casualty rates that are close. Yet some states seem to have a larger problem, specifically Mississippi and South Carolina, in which the accident rate is 2-3x the median. If asked which states are most dangerous to drive in these would not readily come to mind, making this is an interesting observation.

When looking at fatal accidents by vehicle manufacture year there is a large spike in fatalities from vehicles that were manufactured in 2006 – 2007.  This trend then abruptly reverses in subsequent years, does that mean that cars manufactured after that time are safer in some way?  Keep in mind this data is from accidents which occurred in 2017 and the manufacture year of all vehicles on the road will not be distributed evenly. 

Though it looks like cars from the 2007 era may be more dangerous, perhaps there are simply more of them on the road. Could there be an economic explanation instead? There  may be more cars on the road in 2017 that were built in 2005-2007 because more people bought cars during a time when the economy was thriving.  It is also plausible that after the 2008 recession there were fewer people buying cars, making casualties from vehicles of that era less likely in 2017 because there were fewer on the road!

Look instead at the probability of fatality, given an accident, by vehicle manufacture year. This trend shows a downward trend that makes sense given the continuous technological improvements in vehicle design and safety. From this visual it is clear that newer vehicles are more survivable, casting doubt on the claims that older vehicles are safer based on claims they are heavier or “have a steel frame”. It turns out crumple zones, curtain air-bags and other modern innovations matter more. 

In conclusion, a good way to improve your life expectancy is to avoid drunk driving, driving a motorcycle and driving a very old car. Be extra vigilant in Mississippi or South Carolina. Remember that driving is the most dangerous thing you will probably ever do, so give that risk the respect it deserves.

Similar Posts