Most people who use Uber know they may pay more for a ride on a rainy winter evening than on a sunny summer’s day. Uber’s use of dynamic surge pricing–its practice of charging more when demand for rides is higher than the supply of cars–is now famous (or infamous, if you are someone who paid hundreds of dollars for a New Year’s Eve lift.) Other industries, even Disney’s theme parks, are copying the model. And the entire selling point of would-be competitor ride hailing companies is that they’re surcharge-free.
Uber says it uses surge pricing to create the most efficient market and maximize the number of rides it can provide (Lyft also uses a similar system). But no one outside the company knows how Uber’s surge pricing actually works–it isn’t public with details of how it sets prices moment by moment in a given neighborhood or city. And because no one can see the prices others are getting, neither riders nor drivers know whether they’re actually getting a fair deal or whether Uber is manipulating the system to pad profits. It’s basically the “trust us” approach to pricing.
“Uber has all the information, and it’s making decisions about prices. Whether its interest is the same as the customers and the drivers isn’t something we know,” says Northeastern University researcher Christo Wilson.
Wilson, who specializes in “auditing” algorithms, and two Northeastern colleagues set out to reverse engineer how Uber’s surge pricing works in two cities–San Francisco and Manhattan–over the course of four weeks. In a paper being presented at a conference in Tokyo this week, they described how they mimicked 43 Uber customers using the app and also used Uber’s public interface for software developers to reconstruct what the company’s pricing system looks like behind the scenes.
Uber’s pricing algorithm is generally “fair,” they found, in the sense that it’s based on the laws and supply and demand and doesn’t seem to arbitrarily jack up the price. (An exception was a few, brief price “jitters” that the researchers uncovered. When they reported their findings to the company, Uber said they’d found a bug in the code and fixed it quickly.) Moreover, that system is responsive–surge prices seemed to be recalculated every five minutes. But their findings also indicated to them that surge pricing didn’t always work as intended.
“The system is truthful, but it’s not clear if it’s having the desired effect,” Wilson says. They estimated that ride demand dropped off when prices increased, but there was seemingly only a “weak” bump in car supply. In other words, drivers weren’t rushing to respond to higher prices. (One uncertainty in their analysis is that they couldn’t know if they were seeing every car in an area, though they tried to account for this by how they spaced out their fake customers.)
Even if Uber is playing by its rules–it’s still setting the rules. And Wilson thinks the public might want to know what those rules are. By virtually “placing” the GPS coordinates of the fake apps in a grid around each city, the researchers found that most surge periods last less than 10 minutes (and the majority less than five).
Also, it seemed that actual Uber humans are setting discrete surge “areas.” So especially in dense areas with patchy demand, like Manhattan’s Times Square, there are times that two people standing across the street from each other might see different prices.
For Wilson, the lesson is that sometimes by waiting awhile or by knowing the boundaries of surge areas (especially in a place like Times Square), it’s possible to avoid paying more. Drivers, too, the paper says, could theoretically collude to manipulate the system. Finally, for reasons that were not clear, surge pricing was three times more common in San Francisco than New York, even though there are more drivers in the Big Apple.
Uber said in a statement that Wilson’s study was based on “extremely limited, public data” and defended surge pricing as the tried-and-true system that allows Uber users to quickly call rides within minutes. As the researchers also noted, Uber said a big limitation of the study was it wasn’t able to track how each unique Uber car was being utilized, coming on and offline the system as it picked up and dropped off passengers.
“What we find in our data is that drivers really beeline for areas of surge,” Keith Chen, Uber’s head of economic research, told Co.Exist.
If the system works as intended, surge pricing is a “self-defeating prophecy” that quickly rebalances supply and demand, Chen says. That would explain why some veteran drivers may have a more “nuanced view of whether it’s worth their time” to travel to a surge area if they are, say, 15 minutes away. A bigger effect, he says Uber’s own data shows, is that a surge price signals to drivers that it’s worthwhile to stay out in their cars for a longer shift, adding to supply. And, as the Northeastern team found, higher prices do reduce demand as some users opt for alternative transport or to wait until prices drop again.
The broader question the researchers raise is whether people really want to get behind an opaque pricing system. Some certainly don’t: One New York lawmaker is trying to ban surge pricing altogether. Even the bug the study authors found (though minor) is to them a cautionary tale against making complex, widely-used algorithmic systems unaccountable to public scrutiny.
“We argue that Uber’s reliance on black-box algorithms makes their system more vulnerable to manipulation than other online marketplaces. The forces at play on markets like eBay and Airbnb are well understood: the supply of goods is transparent, and prices are set by competing individuals. In contrast, Uber does not provide data about supply and demand, and the pricing algorithm is opaque,” they say.
Chen, however, believes Uber is more clear about its prices than almost any other dynamic pricing system out there, and it has built a system that functions well. During times when surge pricing is in effect, its app now tells users exactly how much more they’ll be charged, and sometimes makes them type it, just to make sure they understand. He notes other widely accepted demand-based pricing systems aren’t transparent at all:
“I just bought a flight home for Thanksgiving, and there’s no explanation for why it’s three times more than the normal price.”