Ride-sharing Doesn't Solve Traffic Problems, Israeli Research Finds

A simulation of the Tel Aviv area showed the system to be very inefficient

Send in e-mailSend in e-mail
Send in e-mailSend in e-mail
Cars in traffic in central Tel Aviv, December 4, 2019.
Cars in traffic in central Tel Aviv, December 4, 2019.Credit: Moti Milrod

Imagine your ideal work commute. Instead of getting in your car alone and fighting traffic, you use an app to find a nearby driver who picks you up within a few minutes and drops you off at your office. Someone else does the driving, and the removal of cars from the road reduces congestion.

Ride-sharing has been tried by services such as Uberpool overseas as a convenient, low-cost transportation alternative. But in practice it was found that shared transportation actually increased traffic as drivers circled around awaiting riders and going from pickup to pickup.

In Israel, a service called Bubble Dan is experimenting with the same formula. Looking into the future, many see self-driving cars as finally fulfilling the dream.

Not so fast. A recent study by a group of researchers from Tel Aviv University and Ben-Gurion University of the Negev throws cold water on the whole idea. Using a simulator developed in Israel, they created an ideal world of self-driving cars and ride-sharing serving Greater Tel Aviv.

Even though the model didn’t account for factors such as traffic accidents, stalled cars, inclement weather and the like, the simulation showed that the feasibility of mass ride-sharing was very limited.

Prof. Eran Ben-Elia of Ben-Gurion University, who helped supervise the study, said the simulation showed that conventional mass transportation solutions were more promising.

“The fact that even under laboratory conditions flexible public transportation isn’t economical or efficient is forcing us to go back to many classic mobility solutions – trains and buses, operating on regular routes and from fixed stations to which passengers come by foot and bicycle.”

The researchers set standards for shared-rising success. Users had to be able to reach their destination in no more than one and a half times what it would take them using their own car, point to point. Waiting time for a ride must not exceed 20 minutes, and pickup points must be no more than 500 meters from the user’s starting point.

The simulation involved 300,000 “passengers,” representing about 10% of the total number of daily commuters in the area bound by Ashdod to the south, Modi’in to the east and Netanya to the north (and the Mediterranean Sea to the west). It’s Israel’s most congested area, and the simulation looked at peak commuting hours.

The team tested the feasibility of shared rides using various numbers of vehicles, starting with 50,000 and increasing gradually to 100,000.

A passenger boards a Bubble Dan shared taxi in Tel Aviv, April 15, 2019.Credit: Moti Milrod

One of the most disappointing findings from the simulation, which was conducted by doctoral candidate Golan Ben-David under the supervision of Ben-Elia and Prof. Itzhak Benenson of Tel Aviv University, was how inefficient shared transportation is. Even at peak demand times, vehicles on average could only transport two riders at a time. Most of the day, vehicles carried just one passenger.

With 50,000 vehicles on the road at peak hours, each was carrying just 2.02 passengers on average; with 100,000, the average edged up only slightly to 2.12. At peak hours only 16% of the vehicles on the road had an average of three riders and no more than 10% had four.

When the simulator set as a standard that vehicles had to carry as average four riders during peak commuting hours, the waiting time for a ride jumped to three times what it would take for the commuter to drive themselves.

The simulation also found that the bigger the shared-riding fleet, the less it was utilized. With 50,000 shared vehicles on the road, a third of the time they were parked and 10% of the time they were carrying no passengers. At 100,000, they were parked 70% of the time. They were travelling without riders only 2% of the time, but twice as many cars were needed to achieve that goal.

The study also showed that at rush hour, one out of every five riders had to wait an unreasonable amount of time to be picked up. With 50,000 vehicles, the wait averaged 22 minutes and could reach up to 42; at 100,000, it was still an average of 12 minutes and could be up to 21.

Worse still, throughout the day, 6% of all requests for rides were not fulfilled at all when there were 50,000 vehicles available. Boosting the number to 1000,000 only cut the nonresponse rate to 2%. It also found that the longer the requested ride, the more likely it would not find a driver.

Ben-Elia said that was a particular vexing problem for the shared-transportation mode. “Most of the rides that were hard to provide because of long waiting times involved long trips from the outskirts (of the region) to the center. Assuming those are the commuters that will give up on the service and travel by private car instead, it will mean less of a reduction in traffic,” he said.