Sign up for the Talking Points newsletter, a recap of the most important business news, delivered fresh each afternoon, Monday through Friday.
About 14,000 taxicabs swarm the streets of New York City. That’s 11,000 too many, according to research from the Massachusetts Institute of Technology.
A team led by Daniela Rus, director of MIT’s Computer Science and Artificial Intelligence Laboratory, found that if more riders were willing to use carpooling services offered by ride-hailing companies like Uber and Lyft, instead of riding alone, New York would need only about 3,000 taxis.
Rus said the team has improved the routing software behind such services to more efficiently deploy vehicles for carpooling.
Making it work, of course, is based on a big premise: that many riders would be willing to share their trips with a car full of strangers.
With carpooling, “we have fewer vehicles, we have less pollution and we have a better travel situation for everyone,” Rus said. The same approach could be applied globally, leading to less crowded streets and cleaner air in cities around the world.
A research paper describing the method is being published Monday in Proceedings of the National Academy of Sciences.
Ride-hailing companies such as Uber offer a carpooling option that allows two or more passengers heading in the same direction to share a vehicle and pay a lower fare. A previous MIT study from 2014 looked at data from 150 million New York City cab rides and found that 95 percent could have been handled via carpooling, while adding just five minutes to each passenger’s trip.
Rus said her team’s improved routing systems allow any driver to be redirected in real time to the nearest next passenger, and calculates a new and better route to the passengers’ destinations.
The MIT system hasn’t been tested on the streets. Instead, the researchers ran a simulation using a week’s worth of actual trips by New York City taxis — three million fares. The database includes pickup and drop-off times and locations, trip distances, and fares. Rus said her team did not use the figures on the number of passengers from the database in its calculations out of concern they were unreliable.
Rus said the MIT simulation was able to efficiently carry nearly all of those trips, or 98 percent, by using just 3,000 four-passenger cars. On average, customers would have had to wait just 2.7 minutes to get a ride and would have arrived 2.3 minutes later than if they’d used a standard cab.
Steve Goldberg, president of the Boston Taxi Owners Association, said he was more concerned about cabs facing tougher regulations than about ride-hailing companies.
“That’s a bigger issue” than carpooling, Goldberg said, “and unfortunately we’re going to have to keep fighting in court.”
But Donna Blythe-Shaw, former staff representative for Boston’s cab drivers for the United Steelworkers union, said carpooling is yet another threat to taxis. However, she questioned how many customers would be willing to share rides.
“It’s like being on a train with a bunch of people, only with closer quarters,” she said.
Still, Blythe-Shaw called the MIT research “pretty fascinating” and said Boston taxi regulators ought to prepare for a future in which carpooling is common.
Rus said that carpooling would not completely replace single-passenger trips, because many customers would always prefer to have the back seat to themselves.
Still, competition from ride-hailing companies has caused a collapse in the value of taxi licenses, and a better carpooling system would probably further erode the now-fragile business.