Planet Money reports  (HT: Martin Brock) that New York City is considering increasing its legal taxi fleet from 13,000 to 15,000. The 2,000 new cabs would be wheelchair accessible and the sale of the medallions for the right to drive those cabs are expected to raise $1 BILLION in revenue for the city. And it presumably would make it easier for travelers to get a cab. Presumably.
Charles Komanoff  disagrees. The transport economist has been analyzing the city’s traffic patterns for almost 40 years. He argues that putting more cabs on the streets will actually slow down traffic — so much so that it would cost travelers not just time but also money.
True, it would be easier to find a cab. But Komanoff argues that all those extra cabs would slow down traffic by 12 percent in the city. And they wouldn’t just slow down traffic for their passengers. They would slow it down delivery trucks, buses, private cars — everyone.
Komanoff has been collecting data about New York City’s traffic patterns in a massive spreadsheet. In the data you can find every lane on every road in the heart of Manhattan. He calls his data trove the Balanced Transport Analyzer .
He figures the slowdown due to the new cabs would cost the city $500 million a year in lost time.
You can listen to the full story here .
Robert Smith, the Planet Money reporter, correctly notes that traffic is like an ecosystem, emergent in its nature. One change leads to other changes. In this case, Komanoff argues that adding the 2000 cabs causes the other traffic. He argues that the extra cabs will reduce waiting time for a cab by 60 second (Yay!) but reduce travel speed for everyone by 12% (Boo!). According to Komanoff, the costs outweigh the benefits by that $500 millions. How can so few cabs have such a big impact? Komanoff points out that cabs drive around all day–they don’t park and get off the streets. He argues that each cab is the equivalent of 40 cars commuting into the city.
Komanoff’s results are from a computer model and their precision (12% increase in travel time, one cab=40 commuting cars) gives his analysis an air of science. But I suspect it is scientism not science. Giving 2000 cabs the right to use the streets of Manhattan will not add 2000 extra cars to the streets or the equivalent of 80,000 commuting cars. The added congestion will discourage some or many cars from traveling as often as they did before. Some will be cabs, perhaps, but many (most?) may be non-cab drivers. If Komanoff’s model assumes that existing drivers will behave as they did before, it is meaningless. Either way, I suspect he cannot model this decision with any precision.
Here is my analysis, made without the benefit of a computer model. If the city of New York allows 2000 more cabs to roam the streets looking for fares, traffic levels will stay roughly the same and the beneficiaries will be those seeking a cab. The losers will be existing medallion holders who will find their medallions have lost value. Not surprisingly, those who already own medallions are fighting the expansion of medallions in court. Their allies are the environmentalists making this (as Martin Brock notes to me) a classic bootlegger and baptist coalition .
UPDATE: Komanoff has written saying he is under time-pressure today and unable to respond directly. But he did send me this paper  which is background for an op-ed he wrote (no longer available) and which has the same results discussed in the NPR piece. Here is the key part:
Medallion taxis account for only 3% of all vehicle miles driven in the city as a whole, but they make up more than two-fifths of CBD traffic. Statistically speaking, this high fraction stems from the stark difference between the minutes and miles logged by taxicabs in the CBD vis-à-vis those of private autos that venture into the CBD at all. The typical private car driven into the
Manhattan Central Business District covers less than 3 miles a day there. In marked contrast, medallion taxis are driven within the CBD an average of 110-115 miles per day — 40 times as much as the average auto.
Giving just one more vehicle a license to pick up street hails in Manhattan is thus tantamount to inviting 40 additional cars to drive into the city’s center each day. The 2,000 new medallions authorized in the new legislation therefore equate roughly to 80,000 extra cars — a 10% increase in vehicle volumes in a traffic system that is already stretched beyond capacity for much of the day, including, by definition, during long “peak” periods in which the number of vehicles impacted by traffic is at its highest.
Modeling the Traffic Impact
In projecting a 12% deterioration in CBD travel speeds from adding 2,000 taxis, the BTA incorporates a number of conservatisms, i.e., factors that work to moderate the traffic impact. These include:
• The expectation that the increase in traffic will “crowd out” some car trips, i.e., discourage
people with access to cars from using them for all or some trips;
• The expectation that the increased ease of flagging down a cab (now that there will be more of
them) may lead more New Yorkers to “leave their car at home”;
• The tendency of taxi usage to be concentrated later in the day or night than car and truck traffic,
which would allow the jump in taxi traffic to be more easily absorbed.
The BTA’s inclusion of these factors doesn’t mean that the model perfectly replicates reality; rather, it attests to the model’s sophistication, and to the seriousness of its effort to portray the complex interactions that constitute travel in New York City as realistically as possible. Certainly, the model’s result, that weekday traffic speeds within the CBD during 6am-6pm will decrease by 12% on average, “falls out” of the model rather than being programmed into it.
This leads to the following puzzle. If new taxis crowd out some car trips, how does a 10% increase in traffic volume lead to a 12% increase in travel time? That suggests some sort of effective crowding-in or some non-linear bottleneck effects. Perhaps Komanoff will have time to give us some more info down the road.