On June 4, 2016, the Washington Metropolitan Area Transit Authority, or WMATA, began its badly needed subway maintenance surge. Overnight, thousands of riders scrambled to find alternatives, and already congested roadways faced even greater demands. With so much news coverage of the advances in autonomous vehicle research, it is tempting to dream that one day soon, fully autonomous vehicles, or AVs—also known as self-driving or driverless cars—will eliminate congestion and make public transportation obsolete.
However, this enticing line of thinking overlooks the limitations of AVs and the benefits that transit provides, especially during peak commuting hours.
A recent report from the Organisation for Economic Co-operation and Development, or OECD, modeled the effects of AVs on travel in the municipality of Lisbon, Portugal. The report concluded that if nearly everyone in Lisbon sold their cars, a comparatively smaller fleet of AVs could provide an equivalent level of personal mobility.
This result is intriguing but requires some unpacking with respect to travel in the United States.
Increased vehicle miles of travel
For starters, fewer vehicles do not necessarily mean less congestion. One overlooked downside of AVs is the additional miles of travel required to reposition vehicles to areas that generate a high number of trips or to drive empty to the next booking. Unfortunately, the numbers increase quickly. In the case of Lisbon, total vehicle miles of travel, or VMT, nearly doubled when researchers removed public transportation and assumed most AV trips would be taken by single riders.
Increased VMT is problematic because traffic congestion grows exponentially. As roadways become full, the negative impact of each additional vehicle or mile of travel is much greater than when roadways are mostly empty. Notably, the results of a second Lisbon simulation in which researchers included public transit and assumed that nearly all AV trips would involve ride-sharing showed total vehicle travel increasing by only 6 percent.
However, U.S. policymakers should remain skeptical of this second simulation for three reasons. First, travel behaviors are slow to change. Seventy-six percent of U.S. commuters drive to work alone, while only 9 percent carpool—typically with family members. In the past 25 years, as urban area congestion has grown by 145 percent, the share of commuters who drive alone has remained constant, while the share who carpool has fallen. Most of the gains in the OECD study resulted from ride-sharing rather than from vehicle autonomy.
Second, Lisbon has an average population density of approximately 17,300 people per square mile. In the United States, only two incorporated cities—excluding less dense suburban communities—have an equivalent or greater density: New York City and San Francisco. Even in large U.S. cities, densities are far lower. For example, Washington, D.C., at 9,850 people per square mile; Houston at 3,500; and Atlanta at 3,154 all fall well below the population density of Lisbon.
Why does this matter? Lower residential densities make it more difficult to match people successfully for ride-sharing trips.
The ability of AVs to help reduce congestion comes down to the difference between ride-hailing and ride-sharing. Ride-hailing typically involves a single traveler matching with a nearby vehicle for one pickup and one drop-off. In effect, ride-hailing is a taxi service with a more sophisticated technology platform for booking and payment.
By comparison, ride-sharing involves bundling multiple occupants into one extended, multistop trip. This requires a far more complex form of matching. Rather than simply finding a nearby car, a ride-sharing trip requires multiple travelers to have approximately the same origin and destination pair at the same time. Regardless of whether the vehicle is driven by a computer or a person, the matching challenges are formidable—especially as urban density gives way to suburban and exurban land use patterns.
The challenge of matching riders in low-density areas is important when considering the potential impacts of AVs on commuting. On average, Americans drive 19 miles each way to and from work, which is more similar to a trip to the airport than running a weekend errand or meeting a friend for dinner. Therefore, SuperShuttle vans can offer some potential insights.
Unlike the ride-sharing the OECD report envisioned, airport shuttles enjoy the substantial advantage of having a single origin or destination—the airport. This simplifies matching. Additionally, people budget significant time for airport runs, allowing SuperShuttle to have latitude in overall trip times. Finally, shuttles offer a price advantage over an equivalent taxi ride.* Yet even with these built-in advantages, shuttles serve only a modest number of total travelers. This suggests a relatively low top-end limit in consumer demand for commuting by ride-sharing.
Another limitation facing AVs is capacity. Today, most passenger vehicles sit idle more than 90 percent of the time. However, these vehicles are almost all in use during the same core commuting times, meaning that they have little to no ability to serve multiple users—again, assuming those users mostly prefer traveling alone. A single-occupant driver and single-occupant rider in an AV both add to roadway congestion. By comparison, WMATA’s Orange Line carries 15,400 passengers per hour during peak commuting times.
The coming AV age will deliver important benefits, including reduced fatalities due to driver error, improved mobility for people who cannot drive, and increased roadway productivity. These welcome benefits will require policymakers to rethink how to plan, build, and manage cities. Yet one thing is certain: Metropolitan regions will always need high-capacity public transportation, with AVs complementing and strengthening that system.
* Author’s note: According to Dulles International Airport, for example, a taxi ride into the city is about $68 dollars, significantly more than SuperShuttle, estimated to cost $28 dollars. On marketing materials, SuperShuttle notes that it tries to price service between the cost of transit and taxis.
Kevin DeGood is the Director of Infrastructure Policy at the Center for American Progress.