The federal analysts, on a basic level, wanted to know how much motorists could be trusted to tap the “Crash” button on their phones when they saw something amiss. And they wanted to see how that compared to the old-school collision reports that have been collected by government agencies for decades.
The early answer, federal officials said, was encouraging. Using a bevy of statistical techniques, they concluded that predictive models backed by data from Waze did a “reasonably good” job of estimating major crashes.
And they hope that modest finding will be the start of something big.
Injuries from crashes on U.S. roads have risen in recent years as driving — and distractions — have increased along with the growing economy. Dangerous habits, primarily speeding and drinking, have continued apace, increasing the death toll. More than 37,000 people were killed on the nation’s roads last year, according to federal estimates.
At the same time, government data, collected from agencies around the country, can be painfully slow, which can hold up efforts to address problems. So federal officials want to get better at gleaning safety insights from the kinds of private and crowdsourced data regular people use everyday.
“You shouldn’t have to always rely on the private sector,” for those capabilities, said Michael Pack, director of the University of Maryland’s Center for Advanced Transportation Technology Laboratory.
“The most dangerous thing people do everyday is get in their car and drive around,” Pack said. “It’s important that we be making investment decisions that are going to improve the safety and security of those roads based on the most current data possible.”
Using Maryland as a test case, federal researchers are finding they “can closely estimate the number of actual crashes on roads from the Waze data,” according to the Transportation Department. Busy roads with many Waze users provide the best results, while less crowded overnight hours aren’t as good.
Federal officials foresee development of a nationwide “crash count tool,” as well as state and local applications. And they envision broader potential breakthroughs if the efforts are successful.
Top U.S. transportation officials cite the nation’s Fatality Analysis Reporting System, or FARS, which gives a detailed look at factors behind each fatal crash, as an example of both the power and limits of federal data-collection efforts.
The data is “robust, and forms the basis for a large number of interventions directed at reducing serious crashes. However, the FARS data may not be finalized for a year,” according to a federal description of the Waze project. “In contrast, the Waze data lacks much of the detail about each crash, but is available at a high frequency (in theory, every 2 minutes), for all roads in the U.S. where Waze users are active.”
And federal officials say they also are hoping to weave together information from ride-share companies, data-heavy self-driving car efforts and numerous other private and government sources.
“The Waze pilot will integrate data from sources that provide severe weather, demographic and highway geometric information to test how well we can estimate crash risk,” Derek Kan, the Transportation Department’s undersecretary for policy, told a group of government researchers and transportation experts last month.
Kan, who was formerly general manager for Lyft in Southern California, also pointed to a project using anonymous GPS data to study the role speed plays in rural crashes. The objective is to get a big, textured look at how people in America drive and the dangers they face.
“Ultimately, our goal is to construct a data infrastructure that will allow the exploration of how safety risks interact, from human behavior to vehicle and infrastructure design, to weather, roadway and city conditions,” Kan said, giving transportation officials and others “advanced tools to create a safer system.”
The big prize would be to “forecast risk at a given place and time and deploy public safety resources” there to save lives, according to the department.
Waze, which is owned by Google, says the Transportation Department is one of more than 500 city, state and national governments and other groups it has worked with on data projects worldwide.
Boston tapped Waze traffic-jam data to identify some of the worst intersections in its Seaport District and to analyze and tweak traffic-light timing. The effort cut congestion by nearly 20 percent in some spots, the company said.
In Budapest, Waze data was used to gauge what happened after authorities dropped the speed limit on a crash-prone road from the airport to downtown to just under 45 mph. Accidents fell by more than a third, according to Waze.
Company spokeswoman Terry Wei said the idea is to provide a “two-way data exchange” that “benefits everyone.” Drivers’ phones, or the drivers themselves, provide slowdown, collision and other data to Waze, which makes it available to governments and other partners. “In exchange, partners provide real-time government-reported construction, crash and road closure data to Waze,” which then can sum it all up in “one of the most succinct, thorough overviews of current road conditions today,” Wei said.
Elena Russo, spokeswoman for the Maryland State Police, said there is a place both for apps such as Waze and the well-vetted incident reports, traffic camera feeds and other data from the state’s official travel information site, md511.org.
Sometimes, Russo said, “it’s hard to rely on information coming from the general public.” But motorists also lean heavily on quick, crowdsourced data to make drive-time decisions. Recently, Russo said, she heard from a motorist who was intensely irked that fellow Waze users failed to report a crash on her route home, leaving her snarled in unexpected traffic.
The commuter’s reaction to her inconvenience? “I don’t think you can print that,” Russo said.