Lyft disputed CR’s findings, arguing that there was an “observer effect,” or that CR had dozens of people check prices for the same route at the same time, artificially inflating demand for its rides and potentially affecting the final fares seen by volunteers. Uber said it is “impossible” to guarantee ride requests occur at exactly the same time because ride prices change “almost every second.”
In other words, Uber and Lyft argue that no two rides on their platforms can be exactly the same, no matter how close their time or location may seem.
“In an open and dynamic market like ours, with approximately 1.7 million mobility and delivery trips per hour, trips are defined not just by where they go, but also by when they are requested and what is happening nearby,” Uber said in a statement to CR.
However, several experts with whom we shared our findings dispute this argument. They noted that on almost every route we tested, we found that at least some volunteers flocked to the same ride price at about the same time. Experts also said it would be difficult to artificially spike demand with CR testing alone, given the relatively small number of volunteers used and the test drive locations mostly in large, densely populated locations.
“Are you saying a few dozen people would have caused such a dramatic impact? It might have been rush hour heat, from the airport to downtown, right in the surge area, but that’s not the case here,” says Christo Wilson, a computer science professor and associate dean at Northeastern University in Boston who previously audited Uber and Lyft pricing models in San Francisco.
So what explains why the prices volunteers see vary from company to company? Uber and Lyft said a variety of factors, including rider demand, supply of available drivers, location, time, estimated ride time and distance, weather, promotional offers, and traffic patterns all influence both the original and final price.
“The price difference reflects actual market trends,” Sid Patil, Lyft’s executive vice president of marketplaces, said in a statement. “More drivers may be available at any given time in a given region, at different demand levels, or at different promotional activities. Overall, our markets ebb and flow depending on location, time of day, events, weather, and other factors.”
Uber and Lyft said the only truly personalized pricing on their platforms is through promotional offers like new ride discounts and “re-engagement offers,” which they use to bring back customers who haven’t used their apps in a while. Neither company provided a complete list of all the elements used to personalize the promotion.
But both Uber and Lyft spell out what data they collect and how it’s used in U.S. patent applications and corporate privacy policies.
Lyft said in a statement that it does not group or “segment” customers or use behavioral data to set base prices. But the company admitted it uses “a wide range of signals” for promotions and discounts. Lyft’s Privacy Policy details some of the customer data that Lyft collects. If you consent, your address book and calendar. and making inferences about who you are. Lyft provides two examples in its privacy policy. If you frequently travel to and from the airport, you may be perceived as a frequent traveler. Lyft says it may also infer your gender based on your name.
Lyft’s patent goes further, outlining a “sensitivity” score and model that can be used to predict the “importance” or “priority” of a particular trip, arrival, or drop-off location. “Intent” model. User demographic information can be used to predict rides before they are requested. The willingness to pay score is defined as the “mobile requesting device’s willingness to pay a higher amount for transportation services.”
Uber said in a statement that the company also does not use “protected characteristics” such as race, gender, ethnicity or disability status in its base fares or promotions. It also does not use “rider-specific behavioral characteristics.” However, it did not address the use of behavioral data for large groups of customers. Uber has acknowledged that it uses personal data for promotions and discounts. Uber’s patent shows that the model can use cell phone sensor data and a user’s past behavior. That data may include how quickly and accurately you entered your address. Your gait and walking speed. It can be used to estimate height, weight, and body shape. The exact angle at which you hold your phone also identifies deviations from the norm. Your riding history is also a powerful predictor of who you are and where you’re likely to go. Uber outlines one such example in one of its advertising patents. If someone regularly requests an Uber to a daycare center or school before going to work or college, they may be identified as a single working parent. From there, Uber says it can determine the passenger’s age, gender, and approximate age of the child just by looking at the ride history.
“Earlier generations of these pricing systems focused on time, supply and demand, price elasticity and efficiency. But now many companies are actively using behavioral and contextual data to inform their models. They don’t necessarily need personal data,” said M. Keith Chen, a behavioral economist and professor at the University of California, Los Angeles, who previously worked as head of economic research at Uber and helped create the surge pricing algorithm.
Uber and Lyft also denied offering fictitious discounts to customers. Lyft attributes these discount findings to the fact that “prices constantly change based on real-time market conditions.” Uber said its platform cannot establish a true reference price and said our test was “fundamentally flawed.”
“When one user’s undiscounted price matches another user’s discounted price, it’s simply because those prices were different to begin with due to changes in real-time market conditions,” Uber said in a statement.
Experts also disputed these arguments, saying the fact that so many volunteers saw exactly the same final price for many of the routes we took suggested that, at least in some cases, there was a true starting price determined by an algorithm.
“I disagree that there is no standard price even for ride-sharing,” says UCLA’s Chen. “What we see in the data is certainly a baseline.”
Uber also took issue with bogus discount analysis. We counted the fare as discounted if what appeared to be the original price was crossed out and a lower price appeared. Uber said in a statement that when these prices are accompanied by labels such as “lower-than-usual fares,” they do not imply a discount, but are intended to be “historical” or “informational” comparisons.
Finally, Uber and Lyft said the percentage of each fare they receive is much lower than what CR calculated. They put the U.S. “take rate” at “approximately 20%” and “significantly lower than 30%,” respectively.
The disagreement is largely due to differences in accounting practices, with Uber and Lyft arguing that our calculations don’t acknowledge the large and growing amount of money drivers spend on auto insurance to cover them on and off trips.
However, the experts we spoke to argue that a company’s insurance costs are simply a cost of doing business and should not be excluded under standard accounting practices. (Columbia’s Sherman said excluding them is “very misleading.”) And he points out that both companies have established their own in-house insurance subsidiaries, with billions of dollars in reserves available for claims.
Uber also claimed that because drivers and passengers were in the same location, the experiment minimized ride distances and “created an artificial scenario that was not representative of reality.” Although our test was designed to have riders and volunteers in close proximity to each other, our analysis of how much money Uber and Lyft receive from each fare is similar to other studies. Experts say Uber and Lyft actually take in nearly half of their customers’ fares. When Columbia’s Sherman analyzed more than 50,000 rides provided by three Uber drivers, he found that Uber’s market share had risen to more than 50 percent in many cities. In another analysis conducted on CR using ride-hailing trip data from Oregon, Princeton’s Workforce Algorithm Observatory calculated that on average Uber pays 44 percent and Lyft pays 52 percent of what riders pay for a trip.
Technical accounting issues aside, the Uber and Lyft drivers we spoke to said that the portion of the take-home pay their riders are actually paying continues to decline and is far less than what they have been led to believe they will earn. For example, Lyft announced that in 2024 it will guarantee drivers more than 70 percent of their weekly passenger payments “excluding outside fees.” (Later that year, the company settled with the Federal Trade Commission, paying a $2.1 million fine for what the agency called “false earnings claims” about the amount drivers could expect to earn, and announced this year that it would cap fees at 30%.)
A lawsuit filed against Uber says that before Uber changed the way it pays drivers, drivers expected to keep 80% of their fares.
Mohamed Drissi, a 43-year-old driver from Morocco in Portland, Oregon, said his take-home pay has gradually declined over the six years he’s been driving for Uber. “There’s insurance premiums, city fees, Uber fees, whatever. And at the end of the day, it’s not $70.” [out of $100]. It’s much less,” he says.
In fact, on Mohammed’s six test trips, passengers paid about $126 in fares, not including tips. Of this amount, $66.73 went to Mr. Mohamed, $58.41 went to Uber, and $16.66 went to city and airport fees. Excluding government fees, about 53 percent went to Mohammed and 46 percent to Uber. (Again, according to Uber, much of that 46% goes to insurance.)
