The Friction Dividend: Why Ride-Hailing Apps Don't Need to Be Cheaper

Based on the research ofJeffrey Fossett, Michael Luca, and Yejia Xu, "Leaving Money on the Dashboard: Price Dispersion and Search Frictions on Uber and Lyft," NBER Working Paper 34441, 2026

In February 2025, researchers from Harvard Business School, Johns Hopkins' Carey Business School, and Theia Insights hailed the same New York City ride on both Uber and Lyft, 2,238 times. The two apps quoted different prices for the identical trip at the identical moment, an average gap of $3.50, or roughly 14% of the fare. Riders had no way to know it without opening both apps, and almost none did: device-level data from Comscore, covering 4,016 mobile devices over three months, show only 16% of people who opened one app on a given day also opened the other. Two firms, fierce headline rivalry, apps one tap apart on the same phone, and most riders still pay a price they never had the chance to compare.

The conventional read on ride-hailing competition is that two apps on every phone should crush price differences to noise. Economists have known since George Stigler's 1961 paper "The Economics of Information" that even modest search costs can sustain price dispersion in an otherwise competitive market. But Stigler was writing about phone calls to five hardware stores, costs that should shrink to nothing on a smartphone with both apps installed and logged in. Ride-hailing, of all markets, was supposed to be the one where search friction finally died: no store to drive to, no salesperson to avoid, just two icons on the same home screen. Instead, the new working paper by Jeffrey Fossett, Michael Luca, and Yejia Xu shows that the very algorithm making ride-hailing efficient is the one keeping the friction alive.

Call it the Friction Dividend: the extra margin a platform collects not because its price is better, but because its own real-time pricing engine makes checking the competitor's price too costly to bother with. Uber and Lyft need no agreement to avoid competing away that $3.50 gap: their pricing systems manufacture the friction for them, in three distinct ways.

The Quote That Vanishes

A grocery shelf price sits still long enough to compare. A ride-hailing quote does not. Both platforms compute fares continuously from live, location-specific supply-and-demand signals rather than a fixed rate card, so a price fetched at 8:14 p.m. describes conditions at 8:14 p.m., not 8:15. There is no receipt from last week's ride to check it against, no menu to memorize, only a moving number that has often already shifted by the time you would switch apps to verify it. Where Stigler's shopper faced a fixed price and a costly search, the ride-hailing rider faces a cheap search and a price that will not hold still for it. If you have ever glanced at one app, flicked to the other, and just tapped "confirm" on whichever quote loaded first, you have paid the Quote That Vanishes firsthand.

Uber and Lyft need no agreement to avoid competing away that gap; their pricing engines manufacture the friction for them.

The Silent Toll

That instability turns a small annoyance into real money at scale. Fossett, Luca, and Xu estimate the friction costs New York City riders roughly $300 million a year in forgone savings, about 6% of the two platforms' combined gross bookings in the city (see the exhibit). For scale, Uber's global Mobility segment booked $83 billion in 2024 and Lyft booked $16.1 billion across 828 million rides, so a nine-figure friction dividend in a single city is not a rounding error. None of that $300 million is money either company prices explicitly; it is money riders leave on the table because checking costs more, in time and hassle, than the $3.50 they would typically save. Multiply one overlooked comparison by millions of rides, and an individually trivial friction becomes a line item that never appears on anyone's invoice.

The Gatekept Comparison

The dividend is not purely an accident of app design, either. Uber's API terms of service restrict how third parties can build tools on top of its pricing data, a policy that, whatever its stated rationale, keeps the one product that could neutralize the Friction Dividend off the market. Airfare has Kayak; gas prices have GasBuddy. Ride-hailing has no equivalent clearinghouse, a full decade into the category, and that absence is not an accident of platform youth. It persists because both incumbents have more to gain from riders not comparing than from riders comparing, and the market structure protects them: Uber holds roughly 76% of US rides to Lyft's 24%, so a comfortable duopoly has little reason to hand riders a referee.

The Rule for Real-Time Marketplaces

You might object that $3.50 is not worth the seconds it takes to switch apps, and for any single ride you would be right. That is precisely the mechanism. The Friction Dividend is not extracted in amounts large enough to fight over; it survives because ignoring it is the rational choice every single time, even as it compounds into nine figures a year across one city. It is also a dividend neither platform has to defend in public. Nobody at Uber or Lyft has to justify a $3.50 gap on an earnings call, because no dashboard anywhere aggregates it into a number a reporter could ask about.

Any marketplace that prices in real time and per user is very likely collecting some version of the same dividend (see the exhibit). Food delivery is the clearest parallel: DoorDash alone books more than $80 billion in annual marketplace order value at roughly a 67% US share, all of it priced dynamically and rarely comparison-shopped. Event resale, freelance bidding on the same job, and surge-priced grocery slots run on the identical logic. Most simply have not had 2,238 matched receipts pulled on them yet.

The fix that would eliminate the Friction Dividend is simple in principle: a trustworthy, real-time comparison layer, the ride-hailing equivalent of Kayak. It does not exist for one reason, that neither Uber nor Lyft has an incentive to build it, and both have reasons, contractual and competitive, to make sure nobody else does either. So the rule for any manager pricing in real time is this: if your product refreshes faster than a customer can compare it, you are already earning a friction dividend, and the only choice left to you is whether you name it and price it deliberately, or wait for someone with 2,238 receipts to name it for you. Until that reckoning comes, the dividend keeps paying out, $3.50 at a time, to whichever app you happened to open first.