Lessons Learned From 2016
U.S. Fatal Car Crash Data

Motor vehicle travel is a major means of transportation in the United States, yet for all its advantages, each year fatal motor vehicle crashes in the U.S. lead to an estimated societal burden of more than $230 billion from medical and other costs [1]. Motor vehicle crashes are also the leading cause of death for persons every age from 5 to 32 years old [2]. In this project, we deep dived into the fatal car crash records in the U.S. in year 2016, collected by National Highway Traffic Safety Administration (NHSTA) [3] and encoded using the government's Fatality Analysis Reporting System. By wielding this dataset and related research, we developed this report trying to thoroughly explore the top risk factors that are highly correlated to fatal motor vehicle crashes.

Anatomy of a Car Crash

Time of Crash
Weather
Condition
Road
Condition
Vehicle
Condition
Driver
Behavior
Passenger
Behavior
Pedestrian
Behavior
Rescue
Delay
Fatalities
& Injuries

A motor vehicle crash results from a combination of factors related to the components of the system comprising roads, the environment, vehicles and road users, and the way they interact. Some factors, such as speeding and alcohol or drug impaired driving, contribute to the occurrence of a collision and are therefore part of crash causation. Other factors, such as old vehicles and unused restraints systems, aggravate the effects of the collision and thus contribute to trauma severity. Some causes are immediate, but they may be underpinned by medium-term and long-term structural causes. Identifying the top risk factors that contribute to fatal motor vehicle crashes is important in developing interventions that can reduce the risks associated with those factors.

Time of Crash

Hover on circles to view number of crashes by month and hour.

Methodology: The size and color indicates crash frequency during the given hour in the given month.

By plotting the number of fatal crashes in 2016 over hour and month, we observe that deadly crashes tend to happen more often in the afternoon, particularly during rush hours, e.g. from 17:00 to 19:00. The risk of an accident still increases during rush hour due to increased traffic, tension among commuters, and an increase in drivers heading home from "happy hour" and driving under the influence of alcohol.

It can be seen that fatal crashes also tend to happen more frequently in winter seasons, e.g. November and Decemeber, where adverse weather conditions (e.g. snow on the East Coast and rain on the West Coast) increase risks. Click on UNDER ADVERSE WEATHER tab to see how adverse weathers impacts fatal crash count.

Note that we do not observe the same high frequency during rush hours in the morning. We further explored the light condition that existed at the time of the crash and see how it might have impacted fatalities. Click on UNDER BAD LIGHTING tab to see more details.
Due to the changing length of daylight, we see a bell shaped curve: the "dangerous" period starts early and lasts longer in winter, but starts late and lasts shorter in summer.

Hover on circles to view number of crashes occurred under bad lighting conditions by month and hour.

Methodology: This plot highlights fatal crashes that happened in a dark and unlighted environment.

Bad light condition appears to be a risk factor highly correlated with fatal car crashes: what account for the darker colors in evening and early morning hours also emerge in the ALL CRASHES chart.

Hover on circles to view number of crashes under adverse weather by month and hour.

Methodology: Weathers other than clear or cloudy (i.e. rain, sleet, snow, fog, severe crosswinds, or blowing snow/sand/debris) are considered as adverse weathers.

Weather-related accidents prevail particularly in December, which aligns with the pattern we see in the ALL CRASHES chart.

Weather Condition

Approximately 21% of all vehicle crashes are weather-related and on average, nearly 5,000 people are killed in weather-related crashes each year (source: Ten-year averages from 2007 to 2016 analyzed by Booz Allen Hamilton, based on NHTSA data). Weather acts through visibility impairments, precipitation, high winds, and temperature extremes to affect driver capabilities, vehicle performance (i.e. traction, stability and maneuverability), pavement friction, roadway infrastructure, crash risk, traffic flow, and agency productivity. See NHSTA's summary of how weather impacts various weather events on roadways, traffic flow, and operational decisions.

Hover on circles to view fatal crash counts by states.

Methodology: Occurrence of various adverse weather conditions that existed at the time of the crash is counted for each state. Weathers other than clear or cloudy (i.e., rain, sleet, snow, fog, severe crosswinds, or blowing snow/sand/debris) are considered as adverse weathers.

According to NHSTA's survey, the vast majority of most weather-related crashes happen on wet pavement (70%) and during rainfall (46%). The chart above supports this finding: Rain - which occurs more frequently than snow, ice, and fog — leads to greater number of fatal crashes. Recent news stories here, here and here.

Road Condition

The Most Dangerous Highway

Hover on map to view details about county, population and fatalities per 100,000 population.

Methodology: The fatality rate is computed as {[total fatalities in car accidents in year 2016] / [county population]} × 100%. County population data source.


One notable outlier from this map is Loving, Texas, a sparsely populated county with a soaring fatality rate of 3.158%. The culprit is one West Texas highway that is notorious for being dangerous - Highway 285, which runs from New Mexico right on through the small town of Fort Stockton. According to the New Mexico Department of Transportation, there were another three fatal crashes from the Texas state line to Loving in 2017. "Because there are so many trucks coming off the side roads and you know they don't see everybody and they will just pull off," said Sherrif Cliff Harris,"Or pull right in front of people sometimes." It was said the construction zones in Fort Stockton "are not going anywhere anytime soon."

Driver Behavior

Driver Circumstances
Alcohol and drug stand out as the most common risk factor.

Hover on bars to view crash counts by circumstances.

Methodology: Circumstances that may have contributed to a crash include distraction, driver's physical impairment, alcohol/drug influence, and impediments to the driver's visual field.

Alcohol and drugs usage is highly risky for both drivers and pedestrains (click on tabs to see how they correlate to fatality risk). Avoid alcohol and drugs when walking and driving; they impair your judgment and coordination.

Alcohol-Impaired Driving

Hover on bars to view injury counts.

Drug-Impaired Driving

Hover on bars to view injury counts.

Speeding

Speeding endangers everyone on the road: According to NHSTA, for more than two decades, speeding has been involved in approximately one-third of all motor vehicle fatalities. In 2016,

  • Speeding killed 10,111 people, accounting for more than a quarter (27%) of all traffic fatalities.
  • 37% percent of all speeding drivers were alcohol-impaired in fatal crashes.

How much does speeding have an impact on fatality risk of the vehicle occupants?

Hover on circles to view crash counts and average overspeeding amount (MPH) for the selected overspeed range.

Methodology: The overspeeding amount of a vehicle is computed as {[vehicle travel speed] - [road speed limit]} at the time of crash. The average fatality percentage for a overspeeding range is computed as {[fatality in vehicle] / [all occupants in vehicle]} × 100% averaged over all vehicles involved in fatal crashes in that overspeeding range at the time of crash.

There is a clear trend that as vehicle's overspeeding amount goes up, there is also a higher fatality risk associated for vehicle occupants.

Speeding tickets are the most commonly used tool to deter speeders, but do they actually have an impact on driver's behaviors (and reducing car crashes related to speeding)?

It is a Class 3 Misdemeanor to operate a motor vehicle either greater than 15 MPH above the posted speed limit. § 20-141
A person will be charged with reckless driving when the speeding exceeds 25 MPH above the speed limit. § 20-141
Still 50 fatal crashes involved reckless drivers who have previously received 3 speeding citations.

Hover on bars to view details.

Methodology: This visualization relates the vehicle's overspeeding amount at the time of crash to the driver’s past speeding convictions. We trace the change in ratio of drivers involved in fatal crashes by the number of previous (within five years from the crash date) speeding convictions of the driver.

Overall, it can be observed from the visualization above that speeding citations are effective at reducing the likelihood of receiving subsequent speeding tickets, as significant drops in number of fatal crashes for drivers who have previously received speeding citations. However, while the total speeding-related crash count sees a downward trend, the absolute crash count remains significant, especially for reckless speeders, suggesting that speeding citations have limited effects on deterrence in the context of the current traffic enforcement system.

Passenger Behavior

Restraint System Use

Of the 23,714 passenger vehicle occupants killed in 2016, there were 11,282 (48%) who were restrained and 10,428 (44%) who were unrestrained at the time of the crashes. Restraint use was not known for the remaining 2,004 (8%) of the occupants. For passenger vehicle occupants involved in fatal crashes in 2016, nearly half of those who were killed were unrestrained in the crash, compared to only 14 percent of those that survived.

Hover on bars to view injury counts.

Methodology: The data describes restraint usage of passengers of a motor vehicle in-transport.

Hover on bars to view injury counts.

Methodology: The ejection scenario describes the ejection status and degree of ejection for this person, excluding motorcycle occupants.

Ejection refers to occupants being totally or partially thrown from the vehicle as a result of an impact or rollover. Ejection from the vehicle is one of the most injurious events that can happen to a person in a crash. Seat belts are very effective in preventing total ejections; in 2016 only 1 percent of all passenger vehicle occupants (those killed as well as survivors) in fatal crashes reported to have been using restraints were totally ejected, compared to 29% of those unrestrained.
Seating Position

Among all vehicle occupants, people riding on vehicle exterior have the highest fatality rate, followed by passengers sitting in cargo areas or trailing unit, a result from lacking safety protection measures. Inside the vehicle, seats in the front row (Front Seat, Second Seat positions) are shown to be more dangerous than seats in the back row (Third Seat, Fourth Seat positions).

Hover on bars to view injury counts.

Pedestrian Behavior

In 2016,
• 5,987 pedestrians were killed in traffic crashes, accounting for 16% of all traffic fatalities.
• On average, a pedestrian was killed in traffic crashes nearly every 1.5 hours.

Pedestrians, along with cyclists and persons on personal conveyances, are one of the road user groups with the highest fatality risks in road traffic crashes.

Hover on bars to view injury counts.

Below we show how the number of fatal crashes correlate to pedestrian actions immediately prior to the crash and how the actions relate to the orientation of their collision with respect to the striking vehicle.

Beware the dangers of pulling over on the side of the road when your car get disabled!

Hover on circles to view details.

Data show that far more people involved in fatal crashes when dealing with disabled vehicles (see recounts of such incidents here, here and here). Do you know what to do if your vehicle becomes disabled on a highway? Experts offered advise on how to remain safe during a roadside emergency. "You are better protected in the car than anywhere else," said Cathleen Lewis, director of Public Affairs and Government Relations for the Northeast division of AAA.

"If possible, get off the highway before stopping," said Lt. Theodore Schafer of the New Jersey State Police. "Is it worth you protecting the rim on your car and pulling over in a dangerous spot versus possibly damaging the rim and making yourself safe? Try to get far enough out of the way and off the main roads so that you won’t be an obstacle for other cars."

Both experts agree that the best thing to do is call 911. Don’t be shy about it. "There seems to be a misconception that we will be upset with you if you call 911," Schafer said, "If your car is broken down on the side of the road to us that constitutes an emergency that requires assistance, so absolutely call 911."

Vehicle Condition

We investigated how passenger vehicle occupant fatality correlates to various vehicle conditions in fatal crashes.

Vehicle Model Year

Hover on circles to view details.

Methodology: The average fatality percentage for a vehicle model year is computed as {[fatality in vehicle] / [all occupants in vehicle]} × 100% averaged over all vehicles of that model year involved in fatal car crashes.

A higher proportion of the occupants of older model-year vehicles suffered fatality or a fatal injury.
Vehicle Age

Hover on circles to view details.

Methodology: The age of a vehicle is measured by subtracting the vehicle model year from the calendar year at the time of the crash (vehicle whose age was calculated to be -1 was recoded to be age 0).
The average fatality percentage for a vehicle age is computed as {[fatality in vehicle] / [all occupants in vehicle]} × 100% averaged over all vehicles of that vehicle age involved in fatal car crashes.

There is a trend that the proportion who were fatally injured increases with vehicle age.

The analysis shows that among all passenger vehicle (passenger cars, SUVs, pickup trucks or vans) occupants involved in a fatal crash, the proportion who were fatally injured increases with vehicle age, i.e., the proportion was higher among occupants of older vehicles as compared to the occupants of newer vehicles. Also, the proportion of occupants who were fatally injured was higher among occupants of older model year vehicles as compared to the occupants of newer model year vehicles.

The improved quality of cars is a bit of a double-edged sword when it comes to the age issue. As cars get better on multiple fronts, such as safety technology and crashworthiness, they also become more reliable and longer lasting, leading to Americans driving cars until they're much older and pushing the average age of a car in the U.S. up to 11.6 years at last count. Despite these gains in car quality, the older a car gets, the less reliable and less safe it becomes.

NHTSA has released a shopping guide to help consumers make sense of the myriad vehicle safety and driver assistance systems available on modern cars.

Airbag Deployment

Hover on bars to view details.

On the surface, the data deviates from our expectation that airbags deployed are associated with lower fatality rate. Our hypothesis for this pattern is that airbags are generally designed to deploy only in moderate to severe crashes, therefore when the airbags are observed to be deployed, we can assume that the crash is much more severe than when airbags are not deployed. If factoring in crash severity, airbags may have been effective in lowering the fatality risk of vehicle occupants in severe crashes. In fact, NHTSA estimated that 2,756 lives were saved by frontal air bags in 2016.

It is important, however, to follow guidance on how to safely position yourself and your passengers, as well as young ones in car seats and booster seats to prevent injury from air bags in a crash. Generally, when there is a moderate to severe crash, a signal is sent from the air bag system's electronic control unit to inflate the air bag within the blink of an eye (less than 1/20th of a second). Because air bags deploy very rapidly, serious or sometimes fatal injuries can occur if the driver or passenger is too close to – or comes in direct contact with – the air bag when it first begins to deploy. Side-impact air bags inflate even more quickly since there is less space between the driver or passengers and the striking object, whether the interior of the vehicle, another vehicle, a tree, or a pole. See more info on airbag-related fatality here, here, and here.

Vehicle Occupant Satety Tips:
  • Sit as far back from the steering wheel or dashboard as possible and using seat belts help prevent drivers and passengers from being "too close" to a deploying frontal air bag. This is why rear-facing car seats should not be placed in front of an active air bag, and children under 13 should be seated in the back seat.
  • Always buckle your seat belt. The seat belt is the single most effective vehicle safety technology. Air bags are supplemental protection and are designed to work best in combination with seat belts.
Vehicle Body Type

We have also studied which vehicle body types render their occupants vulnerable to fatality risk in road traffic crashes. It was observed that fatality risk varies widely based on vehicle body types, and notably, motorcycles (of all types) and snowmobiles occupants suffer higher fatality risk by a large margin. Trucks, vans, and buses have some of the lowest fatalities.
Click here to compare how different vehicle body types relate to fatality risk below.

Rescue Delay

Response Time in Urban Area
Long delays in Washington, Vermont and Virginia.

Hover on bars to view average response time in urban area of each state.

Methodology: "crash2notification" tracks the time from the crash occurred to the emergency medical service was notified. "notification2arrival" tracks how long it took for the emergence medical service to arrive on the crash scene. "arrival2hospital" tracks how long it took for the emergence medical service to transport victims of the crash to the treatment facility. The time is averaged over all crashes that happened in the urban area of the given state.

In Virginia, Kansas and Vermont, there existed noticeable delays between crashes and notification. In Oklahoma, the emergency service reacted relatively slowly. In Washington, transferring victims from the scene to the treatment facility was a long trip. Sadly, the state also has the longest total response time.
Response Time in Rural Area

Hover on bars to view average response time in rural area of each state.

Methodology: The time is averaged over all crashes that happened in the rural area of the given state.

Overall, response time was longer in rural areas, which might be explained by the limited access to emergency services.
Response Time v.s. Emergency Medical Service (EMS) Facilities
Fewer EMS stations per capita in states with long delay: Washington, Vermont and Virginia (red circles).

Hover on circles to view details.

Methodology: By combining Homeland Infrastructure Foundation-Level Data (HIFLD) and U.S. Census Bureau data, we further relate response time to the per-capita number of emergency medical service stations which provide ambulances.

In general, the response time decreases as the number of EMS stations increases and we immediately see that Washington has low number of EMS stations per capita. The HIFLD data also includes details on the facility conditions, such as the number of ambulances, the number of EMS members, level of professionalism etc., which are all factors that potentially contributed to the delay. We also found out that Othello, the city that reacted the slowest in Washington has no EMS station records according to the HIFLD data.

Further Exploration

Analyzing and visualizing the NHSTA traffic record dataset has been a challenging task: the dataset consists of 20 data tables which collectively describe over 400 known attributes of 34,619 fatal motor vehicle crashes. Furthermore, as a motor vehicle crash usually results from a combination of factors including vehicles, road users, environment, and the way they interact, it possess challenges to underpin their individual influences in road traffic fatalities.

To overcome this shortcoming, we have built a predictive model using machine learning, based on our analysis, for predicting a person's fatality risk in a road traffic crash scenario given input features about this person, such as alcohol usage, seating position, person type and so on. We encourage interested readers to play with our model and see how much changing a factor while keeping other variables controlled will affect fatality risk quantitatively.

Due to the sheer volume of data present in the NHSTA dataset, our analysis is likely to have overlooked some risk factors contributing to fatal car crashes. Interested readers are encouraged to use the interactive map of the U.S. below to learn about case details of the fatal car crashes occurred in 2016 and filter records by selected attributes (explore this map in fullscreen mode ).

There were 34619 fatal car accidents in 2016.
Active map filters: clear filter
Case Report