Would Autonomous Vehicles Eliminate 94% of All Crashes?

Natalie JaroszewskiAugust 09, 2022 7 minutes

A national study on the causes of crashes found that 94% of all crashes have a critical error assigned to the driver of the vehicle. Various organizations and players in the automotive industry have spread this statistic far and wide through advertisements, social media campaigns, blog posts, new outlets, and government websites. If this statistic is true, then autonomous vehicles could save thousands of lives and prevent even more thousands of life-altering injuries by removing the driver from the equation.

But is it really that simple?

How Did it Come to Be that So Many Believe that Drivers Cause 94% of Crashes?

Various automotive organizations have plucked this tantalizing percentage from a National Highway Traffic Safety Administration (NHTSA) research survey, called National Motor Vehicle Crash Causation Survey (NMVCCS), conducted in 2005-07. Yet the statistic removed from the surrounding context has been misunderstood. In reality, according to the 2015 crash stats summary of that survey, the data found that in 94% of the studied collisions, "the critical reason, which is the last event in the crash causal chain, was assigned to the driver."

The wording used makes all the difference. Only a few paragraphs later, NHTSA explains, "[a critical reason] is not intended to be interpreted as the cause of the crash nor as the assignment of the fault." The critical reason is simply the last one, in a chain of many, that led to a collision.

Let's Take a Look at a Hypothetical

Driver inattention, wet roads, a covered road sign, and a child unexpectedly running into the street could all be factors at the same time in a common crash scenario. If in the NMVCCS, that crash's critical reason would be listed as "driver error." Yet, the reality of the crash could be much more complicated.

Imagine this, the day is overcast, and there is only a light drizzle, not really enough for windshield wipers. Our hypothetical driver enters a section of a city street that has a lower speed than the earlier section. The driver is jamming out to some great tunes, happy that there is finally rain during the oppressive summer drought. Because the driver is distracted, they don't see the speed limit sign, which is half covered by an unruly tree. The road doesn't narrow with the speed limit change, so the driver doesn't unconsciously reduce their speed. An excited child runs out from behind a parallel parked car. Our hypothetical driver slams on their brakes, but due to the wrong speed, there is not enough stopping distance. The child is stuck and killed.

The critical reason, driver error, doesn't tell the whole story because, in this instance, extenuating circumstances (like poorly maintained or obstructed signage) didn't give the driver vital information. The statistic is not 94% of all crashes are caused by drivers, as so many outlets lazily report. It is that in 94%, of the studied crashes, the very last reason in a chain of events can be listed as driver error.

Why Is This Distinction so Important?

Despite the caveats, the NHTSA researchers attached to their findings, their careful wording, and in-depth definitions (and even though the research is over 15 years old), the statistic is currently simplified and misused. Whether it's safety advocates or companies pitching autonomous vehicles, there is a common belief that if we could just get drivers to behave better (or remove them from the equation), then 94% of crashes wouldn't happen.

This oversimplification has real-world impacts. The various state departments of transportation spend millions of dollars focused on education campaigns, that are producing minimal results, and the simplified statistic allows NHTSA to ignore the National Transportation Safety Board's (NTSB) concerns and suggestions regarding the testing of autonomous vehicles on public roads.

Additionally, the mythical 94% is used by transportation, road design, and vehicle manufacturing organizations to push the blame for all vehicle-related crashes onto individual drivers; they absolve themselves of any responsibility for the alarming increase in fatalities.

As Vision Zero Network Director, Leah Shahum, said, "We need to debunk the 94 percent myth because it detracts focus from the actual risk factors that are most deadly, such as poorly designed roads and dangerously high-speed limits."

Our own Department of Transportation, TxDOT, has noticed this trend increase and started "a broad social media and word-of-mouth effort" called #EndTheStreakTX. This campaign is "an urgent call to action for all of us behind the wheel" to reach another day (like in November 2007) where zero people die on Texas roads. This is a great example of an organization, that can impact highway deaths through better road design and construction, focusing exclusively on the role that drivers play.

Driver Error Doesn't Even Account for a Majority of Crashes

More than 46,000 people die yearly on US roadways, and Texas alone averages over 10 deaths every day. If these deaths are just the responsibility of citizens making poor decisions, is there hope for the #EndTheStreakTX campaign to be effective? Autonomous vehicle designers certainly don't think so, which is why they want to remove the driver from the equation.

As automotive designers increasingly explore autonomous driving vehicles, their research shows that drivers may not be as responsible for crashes as we initially believed.

A Deep-Dive Into the Numbers

The NHTSA survey that assigned the critical reason for driver error to 94% of crashes, also created subcategories to further analyze that 94%. Additionally, those subcategories were further divided.

National Motor Vehicle Crash Causation Survey Report to Congress

In the report, the researchers state, "About 41 percent of the driver-related critical reasons were recognition errors that include inattention, internal and external distractions, inadequate surveillance, etc." Of that 41%, one-fifth of the critical reasons were in the subcategory called inadequate surveillance, which is defined as "in which a driver failed to look, or looked but did not see, when it was essential to safely complete a vehicle maneuver."

These percentages need some closer examination if the automotive industry is promising to reduce annual crashes by 94%.

What Types of Crashes Could Autonomous Vehicles Eliminate?

Automated vehicles are not susceptible to inattention or distractions, so they can reduce the total percent of recognition errors to 22.9% from the reported 40.6%. However, to successfully eliminate the "inadequate surveillance" subcategory, automated vehicles will need to have sensor coverage that allows the vehicle to perceive more than a human driver's eye, and the automated vehicle's programming will also have to correctly identify and classify the objects and scenarios it encounters. That's likely a tall order.

In 2018, the first fatal collision due to an autonomous vehicle on Mill Avenue in Tempe, Arizona tragically showed the world that those two conditions are difficult to meet. The Uber self-driving vehicle's sensors initially struggled to classify the victim as a pedestrian with a bicycle, failed to predict that the pedestrian would cross the road, and then the vehicle failed to apply the breaks in time.

Unfortunately, as Phil Koopman, an engineering professor at Carnegie Mellon University, mused in a 2018 blog post, "misclassifying an object or being fooled by an unusual scenario could happen with an automated car just as it can happen to a human." He expresses doubts about automated vehicles reducing all crashes by the full 94% saying, that another 21%, of the other subcategories "might or might not be impaired driving, and might be a mistake that could also be made by an automated driver."

Phil Koopman goes on to explain that not only are self-driving vehicles at risk for similar human error, but they are also at risk for vehicle errors, such as a "buggy" computer code.

"Humans aren't perfect. Neither are robots. Robots might be better than humans in the end, but that area is still a work in progress and we do not yet have data to prove that it will turn out in the robot driver's favor any time soon."

A Reality Check on the 94 Percent Human Error Statistic for Automated Cars by Phil Koopman

So, not only is the 94% only addressing the last cause for a crash (critical reason), but it is also only addressing driving imperfections, which will not go away with a switch to autonomous driving.

The IIHS Estimates Autonomous Vehicles Would Only Reduce Crashes by One-Third

The Insurance Institute for Highway Safety (IIHS) is a nonprofit scientific and educational organization dedicated to reducing deaths, injuries, and property damage from motor vehicle crashes. They examined the same crashes (over five thousand) that the NHTSA survey did, but IIHS specifically examined the crash factors in consideration of autonomous vehicles.

While the IIHS and NHTSA results show similar numbers, the IIHS estimates that only a third of the total 94%, could be avoided by using self-driving vehicles.

Additionally, IIHS argues that, because 40% of the studied crashes had "deliberate decisions made by drivers," for autonomous vehicles to be truly effective they must be designed to always prioritize safety over a driver/rider's preference. Essentially, the IIHS recognizes how complex it is to design autonomous vehicles to be as good as a human driver, let alone better than human drivers.

Semi-Autonomous or Fully Autonomous, These Vehicles Offer Us a Limited Solution, Decades Away

When lives are a stake and the masses want an instant solution, we look for a culprit. In this case, the easy scapegoat is human error. If we could all just drive safer and better, if we all could live our lives in consideration of our fellow man, there would be no death. The automotive industry has offered us the panacea, self-driving vehicles.

However, the NHTSA and IIHS analysis of crashes in 2005-2007 show that while every crash has a common denominator of a driver operating a vehicle when examining the details of a crash, the responsibility of a crash rarely falls squarely on one single reason.

"What we need most is a reexamination of how carmakers, traffic engineers, and community members—as well as the traveling public—together bear responsibility for saving some of the thousands of lives lost annually on American roadways. Blaming human error alone is convenient, but it places all Americans in greater danger."

The Deadly Myth That Human Error Causes Most Car Crashes by David Zipper

The reality is that driving is a complex task requiring constant attention, good judgment, and good driving conditions (maintained roads, safe turns, etc). There is not one solution, and our growing yearly death toll would not go away overnight, even if we had fully autonomous vehicles. Beware of people using a number that tries to simplify a complex problem and those pushing even simpler solutions.