clock menu more-arrow no yes mobile

Filed under:

All of SF cabs could be replaced with just 650 ride-share vehicles, new report says

New, 1 comment

Massachusetts Institute of Technology’s numbers provide possible solutions to your everyday commute

A Lyft car with pink mustache driving at sunset with the Transmamerica Pyramid in the background. Pkg203

In December, SFMTA complained that while ridesharing apps are generating huge revenue for San Francisco companies like Lyft and Uber, they’re also generating monstrous traffic, with 45,000 drivers circling our streets.

Comparing this to the mere 2,026 legal taxi medallions citywide, it’s not hard to see why officials might grouse about the number of wheels on blacktop since ridesharing took off.

Now, possibly, along comes the miracle of MIT-brewed math to the rescue.

A study conducted by the university’s Computer Science and Artificial Intelligence Laboratory claims that ridesharing could cut traffic in cities as much as 75 percent—if applied properly.

Researchers used an algorithm of their own design to crunch how a small fleet of carpooling vehicles operating at prime efficiency would affect traffic patterns.

Their conclusion: A mere 3,000 app-hailed, four-seater vehicles could do the same work as almost the entirety of New York City’s taxi fleet.

Automated vehicles would be particularly well suited to the task, they note, as similar algorithms could inform them which byways see the most demand which hours.

What would that look like in San Francisco? Well, New York fields about 14,000 cabs, running about 485,000 trips per day.

(Although the MIT study assumed 440,000.)

If MIT’s math translates into the real world efficiently, that means one of those 3,000 perfect vehicles would have to run between 146 and 161 trips per day.

In 2014, the average San Francisco cab ran about 500 trips per month, or about 16 per day.

If all 2,026 cab medallions are operating daily (which they probably aren’t, but this is back of the envelope stuff here), that comes out to roughly 32,416 daily trips.

Note that the city’s largest cab company with over 500 medallions, Yellow Cab, is bankrupt but still operating.

If the same algorithm takes to San Francisco the way MIT says it would to New York, it would only take about 430 optimized, app-hailed vehicles to do the job of all of those cabs.

We should note, though, those 500 rides per month are down from the 1,424 that a San Francisco taxi used to average as recently as 2012.

Factoring in the past demand that rideshare companies are presumably already meeting, the ideal number of vehicles on the street bumps up to as much 650.

Either way, it would theoretically meet the same demand with only 32 percent as many vehicles.


SFMTA ought to love that. Of course, the big problem is that this is a mathematical model, not a business one.

A few hundred vehicles operating in an ideal way is great for the city, but the companies themselves may not take to the idea.

And many Lyft and Uber drivers presumably favor breaking up those futuristically efficient riding models in favor of the less efficient, non-automated, non-carpool system that supports a larger workforce.

In fact, it used to be that one of the big selling points of rideshare companies was that they worked around the city’s stingy taxi medallion system to get more people doing pickups.

Which is why instead of a few hundred cars humming along under the guidance of pure, powerful, MIT-developed math, we presently have tens of thousands filling lanes and driving up SFMTA’s blood pressure.