Ensemble Models of For-Hire Vehicle Trips

Recently published:

Wu, Hao and Levinson, David (2022) Ensemble Models of For-Hire Vehicle Trips. Frontiers in Future Transportation. 3 [DOI]

Ensemble forecasting is class of modeling approaches that combines different data sources, models of different types, with different assumptions, and/or pattern recognition methods. By comprehensively pooling information from multiple sources, analyzed with different techniques, ensemble models can be more accurate, and can better account for different sources of real-world uncertainties. The share of for-hire vehicle (FHV) trips increased rapidly in recent years. This paper applies ensemble models to predicting for-hire vehicle (FHV) trips in Chicago and New York City, showing that properly applied ensemble models can improve forecast accuracy beyond the best single model.

A twitter conversation about Uber

The End of Traffic and the Future of Access: A Roadmap to the New Transport Landscape. By David M. Levinson and Kevin J. Krizek.
The End of Traffic and the Future of Access: A Roadmap to the New Transport Landscape. By David M. Levinson and Kevin J. Krizek.

Gabe Klein said he was talking on Bloomberg about Uber. I tweeted him my article.

@trnsprttnst good piece. What do u think of Instagram vs kodak? Same issue, tech vs people + reach via internet and lower cust aquis. cost?

@gabe_klein Social networks mostly don’t stick around (Friendster, Myspace). We’ll see whether Twitter or FB will be around.

@gabe_klein Kodak lasted more than a century and earned great profits. It was undone not by Instagram but by digital cameras and smartphones

@gabe_klein Nothing stops taxis from (or anyone) from emulating Uber. E.g. Lyft. An app does not get you a monopoly position.

@gabe_klein Taxi industry is trying to fight back with TaxiMagic etc.

@trnsprttnst great points. I think Uber can be displaced for sure. People  said same things about Amazon though when just selling books.

@gabe_klein Amazon has an actual distribution system. (And cumulative profits to date are only $2B.  Uber doesn’t even own cars

@trnsprttnst platform to have multiple offerings can be very effective

@trnsprttnst true but that’s why making $ too.  I think @gett is going to be good competitor, and @lyft as well.

@gabe_klein @gett @lyft It’s great people are interested, and it should of course be made legal (& safety regulated) but $10B is a lot of $.

Is Uber worth $10 B?

The End of Traffic and the Future of Access: A Roadmap to the New Transport Landscape. By David M. Levinson and Kevin J. Krizek.
The End of Traffic and the Future of Access: A Roadmap to the New Transport Landscape. By David M. Levinson and Kevin J. Krizek.

So Bloomberg reports that Uber, a so-called Transportation Network Company is worth $10 B. Clearly I am in the wrong business.

This implies that the net present value of all future revenue from Uber minus the costs of that revenue will amount to $10B (assuming they invest the money they raise).

Nationally there are under 250,000 taxi cab drivers who earning a salary of $23,000 apiece, giving a total labor cost of $5.7 B. New York City has about  20% of US taxi cabs and NYC alone serves 241 million annual passengers. Let’s assume 1 billion taxi passengers per year nationally in the US, and $15 revenue per trip. So if Uber captures 100% of the  market, there will be $15 B in revenue per year. But that is before labor costs, which are $5.7 B, leaving $9.3B per year. (I am assuming no international profits, since urban surface transportation is highly localized, and why would any other country let a US company dominate?)

If Uber’s gets  100% of this market keeping 10% of remaining revenue for profit margin, they are plausibly worth $10B. This leaves the remaining 90% of that $9.3B to pay for the costs of fuel, maintaining the vehicle, owning the vehicle, and operating a back end app.

Currently Uber has 900 employees. I assume most of them are in the field.

Let’s just say there is a lot of “hope” and “if” in this valuation.

The Financial Times lists companies by Market Value (March 2013). The smallest company with over $10B is Chipotle. Chipotle uses 37000 employees to be worth $10B. $10B is more than Southwest Airlines, an actually profitable airline, was worth. It is about the same size as United Continental.

We have other comparisons – Mastercard is worth $63B. Which would you rather own 1/6 of Mastercard or all of Uber? UPS was worth $62B with 400,000 employees. Do you think Uber with less than 1000 employees will be more valuable than 1.6 of  UPS? FedEx was worth $31B with 278,000 employees.

Is this a bad investment for the current investors? That depends. Are there greater fools, or are they at the bottom of this pyramid?

up Lyft

As a somewhat famous person in the field of transportation within Minneapolis, from time to time companies give you products in exchange for reviews. In this case, Lyft gave me a free Lift in exchange for a review. Since I was planning to try Lyft anyway, this seemed opportune. And since I was participating in the Theater of Public Policy, I actually needed a ride from point A to point B, a trip that transit planners agreed was not optimally served by transit, (i.e. there was a 1/2 mile walk or a transfer) and was too far to walk.

Step 1. Download the App. I use iOS, this is easy.

Step 2. Enter the required info

Step 3. Enter the promo code. They said it was for $50, but the app only applied $25 apparently, OOPS. I did however get an email following upgrading me to a “Lyft Pioneer” giving 5 $25 rides in the next two weeks. I guess that’s how I will get home too.

Step 4. Summon a Ride.

Step 5. Get in the Car

Step 6. Tell Driver Destination

Step 7. Ride

Step 8. Get out of the Car.

Step 9. Check App to “Pay Driver”

Repeat 4-9 to go home.


Step 10. Write this review.

The app is straight-forward to use. The only thing missing was for me to enter the expected destinations in the app before drivers “bid” on me, which would seem to better match drivers with where they want to go. On the other hand, that might lead to too few drivers “bidding” on prospective customers. (I am not entirely clear how it works from the drivers perspective, and don’t really plan to become a driver, so am not sure exactly how customers are matched to drivers. I assume some of this is proprietary).

Lyft ride #1, in a Mercedes
Lyft ride #1, in a Mercedes

The rides went well. The first driver picked me up within 11 minutes. He didn’t have the pink mustache on his car, but called me on arrival and then I saw which car it was. He took me to my destination with a minimum of confusion (he had 3 GPS systems in the car, two on smart-phones and one built in). The bill was $17 (UMN to Bryant Lake Bowl), though as noted, I had Lyft Credits already. He was doing this while (temporarily?) unemployed. His vehicle was quite nice, including a back massager in the front seat. (With Lyft it seems customary to get in the front, in contrast with Taxi 1.0, where the paying passenger sits in the second row). We chatted a bit. He says he started recently and does this part time, for about 4 passengers a day, though it is not very profitable for him. (And given his car and previous line of work, I imagine not).

At the end of T2_P2, I summoned a second driver from Lyft, who arrived within 2 minutes. I was also texted when he arrived, though I identified his car, which did have the mustache. He does this a few hours a day between his main gig which is in the catering field and thus sporadic. He has been doing this since last fall when Lyft started in St. Paul. From Bryant Lake Bowl to Prospect Park was $14. I received receipts in my inbox almost immediately. My donations to the drivers were  automatic.

Lyft ride #2, in a Corrolla.
Lyft ride #2, in a Corrolla.

Without making any broad social statement (though readers are free to infer), the racial composition of Lyft drivers is noticeably different from the racial composition of Taxi 1.0 drivers in Minneapolis. Since moving to Minneapolis I believe I have had 1 white (immigrant) taxi driver, while both Lyft drivers were white.

Lyft is just an app with a back-end dispatch service. They claim to be a ” transportation network company whose mobile-phone application facilitates peer-to-peer ridesharing by enabling passengers who need a ride to request one from drivers who have a car.” They insist the drivers are independent (as are the riders). The difference between this and a taxi dispatcher is thin. Google says a taxicab is:

a car licensed to transport passengers in return for payment of a fare, usually fitted with a taximeter.

So for taxicabs, the arrangement between the rider and the passenger is mediated by the government (which licenses the vehicles).

Are Lyft drivers licensed to transport passengers for payment? No, they are licensed drivers, and any licensed driver (above a certain age, level of experience) is eligible to carry passengers. The cars are private cars (at least sometimes) though that is little different than how taxis operate in other parts of the world. I am told many Singaporean taxi drivers will take fares when going between where they are going anyway, but otherwise treat the taxis as a personal vehicle. Lyft passengers “donate” to drivers (like a tip, although it is more or less automatic). I guess voluntary payment differs from contractual payment legally, sort of, kind of. The courts will decide. Lyft does insure its drivers, which makes this even murkier.

Also,  here is a $25 discount code for you if you want to try it yourself:  “TRANSPORT” (valid only for new users)

They also gave me this information:

I don’t think Lyft yet does jitney (shared taxi, dollar van, informal transport) type services (which undoubtedly they would spell Jytney … register that domain now kids).  These would serve either one pickup going to multiple destinations, or multiple pickups going to one destination, or multiple origins to multiple destinations, and would begin to compete with public transit. (Though it would not be exactly fixed routes, one could imagine regular runs with a known coterie of passengers). Presumably this would be at a lower rate.

I have not yet tried Uber. I may use Lyft again from time-to-time. At $30 or so round-trip within the city, it is more expensive than driving and parking when I have a car at my disposal, and certainly more expensive than transit, or car2go and may be comparable to taxi. But if I am without a vehicle, or someone else is paying, it is a nice option to have.

No, Ride-Sharing Is Not the Death of Public Transportation

Eric Jaffe at Atlantic Cities, commenting on the BART strike, writes: No, Ride-Sharing Is Not the Death of Public Transportation: “Work from Columbia University scholar David King and others has supported the role of taxis as part of a multi-modal travel network that does not involve a private automobile. Their data of taxi movement in Manhattan neighborhoods found that a majority of these trips were late-night origins — meaning riders must have arrived there by some other mode earlier in the day, with this mode obviously not being a single-occupancy car. King and colleagues write:

In many cases taxi trips are part of journeys that began with transit trips, yet planning and expanding taxi service as an extension of transit networks is rarely undertaken in practice.”

One taxi Über alles

I am quoted in this article Smartphone apps are changing the way riders hail taxis by Katie Humphrey

“When you go to most cities, there are many different taxi providers and it’s hit-or-miss how reliable any of them are,” said David Levinson, a professor at the University of Minnesota who studies transportation. “It’s easier to download an app than to look up a telephone number.”




Via BS, Betabeat: You Were Warned: Uber Says E-Hailing Fares Will Skyrocket Amid New Year’s Eve Demand :

“NYE pricing is not for the faint of heart. The average surge multiple will likely be 2x normal prices, but during extreme spikes it could cost you $100 MINIMUM before time and mileage charges! So be careful with those Uber ride requests. Uber rides will be reliable on New Year’s Eve, but they’ll also be pretty pricey.”

The more of this and *not* having road prices vary by time of day will seem strange. We should just use the words “surge pricing” rather than “road pricing”.

London’s black cabs to get free high-speed WiFi hotspots from early 2013 – The Next Web

If anyone was wondering why Google is interested in self-driving vehicles … imagine the future as robot black cabs. The Next Web: London’s black cabs to get free high-speed WiFi hotspots from early 2013

Who will raise my Taxis | streets.mn

Cross-posted from streets.mn: Who will raise my Taxis?

Who will raise my Taxis


Taxis are an essential feature of an urban transportation system where travelers must go from point to point but lack automobiles. This mainly occurs in two situations: first where travelers do not possess a car, and fixed route transit is nonexistent or inadequate; second, where travelers are coming from out of town, and do not wish to rent a car, or did not want to store one at the airport or train station. Taxis complement transit in an urban environment, and a fleet of taxis is indicative that one can survive without private vehicle ownership. The convergence of (driverless) taxis and carsharing with autonomous vehicles is my vision of the future of urban transportation in the mid 21st century.

Taxi drivers in some cultures are held in high esteem. London taxi drivers must learn “The Knowledge”, before getting licensed, which allows them to deliver travelers safely to any location in London without consulting a map. The Knowledge has been required since 1865, and typical drivers practice for 34 months and take the test 12 times before passing. It has been shown the hippocampus, the part of the brain associated with navigation, is larger for London cabbies than the average person.

In Minneapolis-St. Paul, it is clear that airport taxi cab drivers are required to lack The Knowledge about the street network. Actually knowing where things are must be disqualifying. The only qualification is that one arrived at the airport yesterday oneself, without transportation, and in exchange for volunteering to drive a cab, they let you. The only locations airport cabbies know are downtown Minneapolis, (which I can reach by transit), and perhaps St. Paul. The only roads they know are limited access.

I have yet to meet a cab driver who knows where Southeast Minneapolis or Prospect Park is. I have yet to meet a cab driver who can find my house without a long set of directions or resorting to a GPS system that requires the use of freeways (doubling distance, and greatly increasing fare) without saving any time. I have yet to meet a cab driver who has been on more than half the streets I use to get home. The most recent 2 out of 3 didn’t know where Cleveland Avenue was. (One intelligently asked which city, as Cleveland discontinues into Roseville, though I am not sure that’s why he asked.)

If I land during the daytime, I try to take transit, where at least the extra distance comes without extra cost and without my having to give directions. But when I land at night, transit is not a practical option (off-peak, the nearest operating bus stop is about 1/2 mile away, despite living in the city).

I do have a regular cab driver, who is quite good, who takes me to the airport at odd hours. He is not allowed to pick people up at the taxi stand, that is a separate class of taxi. (He can pick people in the regular pickup area). I don’t want to hire him from the airport since I don’t know in advance when I will land, given the vagaries of air travel.

I don’t envy the life of a cab-driver, waiting in a car all day, for a few round trip fares. But that is the life they chose.

I question the licensing of taxis in such a way that (1) Knowledge is not required, or (2) Knowledge is not rewarded (i.e. it cannot be used as a competitive feature). If you want to allow fresh-off-the-airplane immigrants to drive cabs, fine, so long as they don’t kill people. However if someone actually knows the network, they should be allowed to use that as an advantage in recruiting customers. One imagines they take a test like The Knowledge, customized for the Twin Cities, and can post their scores in their window. I can go for the smart cabbie and pay a higher per mile rate, but have a shorter time, or go with the new cabbie at a discount. Instead, the airport taxi whip routes travelers to taxis in a first-come, first-serve process where taxis are assumed to be identical commodities (differentiated only by the size of the vehicle).

If we hope to have a world-class city, we need world-class taxis.