Ride-sharing apps are bad, actually

Or why Matthew Yglesias has no clue what he's talking about

I love an incurious essay and Matthew Yglesias' recent one on why Uber is actually good as an attempt to snipe at the left is one of the more stunning examples in recent memory. I want to go through his argument then talk more generally about why this sort of analysis is particularly noxious.

Let’s start with a key thrust of Yglesias’ essay:

Uber's success should be celebrated.

The company's "clever arbitrage around municipal taxi regulations" broke Big Taxi, a series of taxi cartels which relied on business models that prioritized inefficiency and anti-competitiveness. Uber has improved prices, led to great social outcomes (e.g. reducing drunk driving and racial discrimination in pickups), and at minimal cost (namely traffic congestion).

Yglesias starts with an article that American University law professor Hilary Allen wrote for the Law and Political Economy blog titled “Why We Need to Stop Subsidizing Venture Capitalists” where she opens:

For many readers of this blog, Uber represents a cautionary tale. While the company attributed its initial success to cutting-edge technology—such as dynamic pricing, matching algorithms, real-time data—subsequent analysis has demonstrated that its growth was largely driven by ignoring, breaking, and then bending taxi regulations to suit its business model.

For companies in the fintech sector, however, Uber’s approach represented a blueprint to follow. From lending to payments to stock trading to crypto, prominent fintech businesses have found a competitive edge not in technology itself, but in using narratives about technology as a smokescreen for the profitable arbitrage of financial regulations. This modus operandi is encouraged by Silicon Valley’s venture capitalists, who decide which businesses to fund and often provide advice, gin up hype, and lobby for the businesses they’ve chosen. Our society continues to shower VCs with public subsidies, but as I argue in this brief post, if regulatory arbitrage is what we’re getting from Silicon Valley’s VCs in exchange, it’s well past time to reconsider this relationship.

Yglesias only quotes the first paragraph (more on that later) but responds to it with:

What I want to talk about here is not Allen’s forward-looking argument, which focuses largely on fintech issues. But the breezy way in which she asserts that Uber is a “cautionary tale” and assumes her audience will agree without argumentation. She thinks the fact that Uber busted up the old regulatory system and got a whole new category of companies legalized is a dangerous precedent. I think it’s good!

Yglesias sees Allen’s argument as "breezy" and in turn breezes by it, but it's worth spending some time actually engaging with Allen's argument to actually flesh out why we might have problems with Uber.

Allen's essay argues that VC-backed firms use an innovation "bait-and-switch" where they develop a business model built on a specific technology, craft a narrative about its ability to "democratize" and "disrupt" a legitimate issue, circumvent laws that get in the way of that business making a profit, and eventually convincing "lawmakers and regulators to change the law so that you never have to comply with it and those who are harmed have no recourse."

Allen hones in on fintech firms because they've been used this model to circumvent laws that protected consumers, investors, deposits, retirement funds, and the like from being gambled away on volatile assets or speculative activities that are incredibly risky for all but the businesses making these bets.

Financial regulations were adopted over time to protect consumers from harm, protect our economy from debilitating financial crises, and further law enforcement and national security objectives. All this regulatory arbitrage is profitable for the fintech industry and the VCs who fund it, but it leaves most of us worse off. In addition to exposing the public to harm, fintech has rarely delivered on its promises to improve financial inclusion, or to make the financial system more efficient, competitive, or secure. Not content to keep skirting regulations, many fintech businesses and their VC backers have successfully lobbied regulators and Congress to change rules and laws to permanently accommodate and legitimize their business models. These efforts erode faith in our democratic process, and in the law’s ability to protect people from harm.

To make matters worse, they are doing so with public funds. VC funds are limited partnerships that are only on the hook for what they invest, not what they lose. They successfully lobbied to gamble with pension funds, increasing appetite for risk by gambling with larger sums of other people's money. And they've secured tax loopholes that ensure fees and profits are taxed at substantially lower rates.

So what happens when a sector is constantly advancing business models that are subsidized by the public, aimed at rewriting laws to realize previously illegal levels of profit-making, and doing so with little regard for (to borrow a word from economics) “negative externalities”?

Allen writes:

Leading VC firms are well-aware of the important role played by this special legal treatment. Lobbying, media blitzes, and hiring government officials for access are all part of the VC industry’s standard operating procedure, but the playbook has been deployed with particular zeal when it comes to crypto. The crypto industry was responsible for 44% of all corporate expenditures during the 2024 election cycle—with most of it coming from Andreessen Horowitz, Coinbase, and another crypto company called Ripple. It’s worth bearing in mind that the public is not merely subsidizing a financial grift when it comes to crypto—it is subsidizing an ideological project with techno-libertarian goals of avoiding regulation, taxes, and eliminating central banks. Crypto is intended, for instance, to provide the infrastructure for the Network State movement backed by Andreessen Horowitz partner Marc Andreessen and Coinbase CEO Brian Armstrong, which aims to create tech-CEO led dictatorships outside the boundaries of democratic governance.

Yglesias dismisses Allen’s essay as being solely about fintech, but if he actually read it he would see it is about the consequences of letting predatory firms rewrite our laws to maximize their profits and to maximize their ability to convert those profits into political power that can further boost said profits. Should firms be able to subordinate the public interest to their business models or should it be the other way around? And is it any surprise that Yglesias is less interested in thinking through this question than in making a contrarian argument that affirms his priors?

Back to Uber. Yglesias' argument around Uber is relatively simple:

The taxi industry as a whole is not that large or important in the grand scheme of things. But the dynamics around ride-hailing really do, I think, illustrate the core differences between economic analysis and the LPE worldview in which Uber serves as a cautionary tale.

We will return to the idea that Yglesias’ misunderstanding of Uber is “economic analysis” while LPE’s is anti-economics, but for now we’ll focus on the narrative he advances. Yglesias explains that the taxi market was highly fragmented and localized, suggesting LPE opposes Uber because of fidelity to small businesses as opposed to a commitment to "high output and good prices." Taxis extract fares from riders and distribute them to owners of taxi firms or medallions, taxis enjoy massive barriers to entry that inflated prices, and they incur massive deadweight loss (a lot of people who would take taxis don't because the prices are too high).

Uber's arbitrage, then, broke the rickety old system because Uber’s sleek tech collected extensive data on customers and drivers that allowed it to calibrate prices and wages which would increase trips and better compensate drivers. It gave drivers flexibility to set their own hours, increased utilization rates (drivers spent more time with a passenger in the car), reduced drunk driving, reduced racial discrimination, and only cost a little traffic congestion.

The bad guys in this instance were not the richest people in America or huge corporations wielding concentrated power — they were a disaggregated network of largely anonymous small business owners. We have more competition now, but also a marketplace that’s dominated by a much smaller number of large global companies. And that’s good!

This all sounds good and well, but is it true?

Back in 2017, economist Hubert Horan's "Will the Growth of Uber Increase Economic Welfare" article made the case that even a cursory look at Uber's actual economics revealed it to be a deeply parasitic enterprise:

Uber has no ability—now or in the foreseeable future—to earn sustainable profits in a competitive marketplace. Uber's investors cannot earn returns on the $13 billion they have invested without achieving levels of market dominance that would allow them to exploit anti-competitive market power. The growth of Uber is entirely explained by massive predatory subsidies that have totally undermined the normal workings of both capital and labor markets. Capital has shifted from more productive to less productive uses, the price signals that allow drivers and customers to make welfare maximizing decisions have been deliberately distorted, and the laws and regulations that protect the public's interest in competition and efficient urban transport have been seriously undermined. Absolutely nothing in the "narrative" Uber has used to explain its growth is supported by objective, verifiable evidence of its actual competitive economics.

To show Uber's growth enhanced welfare, it would need to demonstrate three things: (1) the ability to earn sustainable profits in competitive markets or demonstrate scale/network effects that could realize these profits; (2) the capacity to provide services "at significantly lower cost than existing competitors" or "produce service that consumers value much more highly at similar costs"; (3) significant competitive advantages based on its product, tech, or innovations that cannot be replicated.

Then and now, Uber fails each welfare enhancement test.

Uber has a long history of publishing incomplete, opaque, and adjusted financial reports that do not adhere to generally accepted accounting principles (GAAP) until relatively recent. Data that Horan cobbled together on its operations from 2012 to 2016 showed Uber’s rapid growth did not yield rapid margin improvements observed in other Silicon Valley startups it compared itself to in PR (Amazon, Facebook, etc.). What Horan observed is that when Uber’s margins improved was not because of efficiency or network effects, but because of driver pay cuts.

How have things looked since? A 2018 JP Morgan study found that on-demand delivery and ride-hail workers made 53 percent less in 2017 than they did in 2013. A 2018 EPI study found that a third of passenger fares went to Uber via fees, driver compensation averaged $11.77 after fees and expenses (lower than average private-sector worker hourly pay of $32.06 and lower than service workers, the least-paid major occupation with an average hourly pay of $14.99), $10.87 an hour after deducting mandatory Social Security/Medicare taxes that self-employed drivers must pay, and $9.21 an hour after deducting the cost of a "modest" benefits package equivalent to what a W-2 employee might have.

The Uber driver W-2 equivalent hourly wage is roughly at the 10th percentile of all wage and salary workers’ wages, meaning Uber drivers earn less than what 90 percent of workers earn. The Uber driver W-2 equivalent hourly wage falls below the mandated minimum wage in the majority of major Uber urban markets (13 of 20 major markets, which include 18 cities, a county, and a state). The Uber driver “no benefits” hourly wage or discretionary compensation—the hourly compensation adjusted for an assumption that Uber drivers pay the extra payroll taxes that the self-employed must pay but do not provide a standard benefits package for themselves—falls below the mandated minimum wage in nine of 20 major markets, including the three largest (Chicago, Los Angeles, and New York).

Back then, it took an incredible amount of mental effort (or none at all in the case of some commentators) to ignore that that a key part of Uber’s arbitrage scheme was figuring out how to transfer more cash from labor to capital than was previously legally allowed—the cost of attempted margin improvements was externalized on drivers, whose working conditions degraded as they were paid out starvation wages.

In 2023, Columbia Business School professor Len Sherman affirmed much of what Horan argued when he observed Uber's 72 percent annual revenue growth and profit margin improvements that year were driven by accounting shenanigans (“Uber has never disclosed sufficient data to allow a fully substantiated assessment of its operational and financial performance”) alongside the fact that "Uber has been raising US ridehail prices four times faster than the rate of inflation while squeezing driver pay" in pursuit of revenue growth and profits. Since 2018—as part of a bid to juice investor appetite for its 2019 IPO—Uber raised prices by the equivalent of 18 percent per year while trips were down 29 percent from pre-pandemic Q319 to Q322 and fares were hiked 41 percent over the same period.

Sherman goes on to write that Uber has relatively few options for significantly growing its core revenue: improving productivity to generate more revenue per trip (batched delivery, pooled rides, etc.), new revenue sources (namely advertising), and increasing its take by hiking fares or delivery fees or cutting driver pay. Sherman makes it clear that Uber has prioritized the third plank of options:

These are the actions that undoubtedly are driving Uber’s recent financial performance improvements and the clearest manifestations of Dara’s “new math” to grow revenues (passenger fares) faster than expenses (driver pay). The sketchy data Uber does disclose confirms that the company has been increasing its take rates over the past few years. As already noted, passenger fares have been trending significantly higher for the past five years. While driver comp data is harder to come by, Uber’s recent shift to “upfront prices” has effectively decoupled consumer prices and driver pay from actual travel time and distance, giving the company cover to increase the spread between consumer prices and driver pay.

And in this regard, Uber enjoys a massive competitive advantage: more data on consumer and driver behavior on a global scale than any other mobility or delivery provider. Armed with such market insight, Uber is in an ideal position to practice what economists call first-order price discrimination — that is, charging each customer prices based on their known willingness to pay and setting each driver’s pay based on their known willingness to serve. The resulting upside potential of such price discrimination is enormous, and the opportunity (massive data + AI algorithms + upfront pricing policies) and need (growing investor pressure for near-term profitability) to exploit it is urgent.

Uber has historically denied (without evidence, of course) that they in any way set prices or pay based on individual passenger or driver characteristics. However, their recent tactics de facto promote a race-to-the-bottom in driver pay by operating a bidding process on their driving app where competing drivers have literally seconds to accept low pay offers (usually after Uber’s initial low pay offer to a specific driver was rejected). Anecdotal evidence also suggests that Uber may be using similar tactics to set and revise passenger fares, lowering them only if a passenger rejects its initial higher price offer.

Let’s skip ahead to the present. In 2025, Uber has now consistently reported quarterly profits since 2023. Uber’s second-quarter results on Wednesday: it beat revenue expectations, gross bookings were up 17 percent from last year, revenue up 18 percent since last year, monthly riders are up' posted a $1 billion quarterly profit, and so on and so on. It might be instructive to look at the past few years of improvements that led us here, so let’s rewind and look at Uber’s P&L improvement from 2019 to 2024. Here, again, I’ll refer to Horan who poured over the books to try and piece together a more complete picture of Uber’s finances since, as Horan and Sherman lay out, Uber regularly omits, obscures, or otherwise obfuscates what is offered.

In 2024, Horan looked at Uber's full year financial reports—it's a notable year because after losing $33 billion for its first 13 years, the company seemed to have finally turned a corner. Horan spends the first chunk of the piece explaining that a significant part of Uber's operating results trace back to dubious practices such as "its estimate of the changes of value in untraceable securities it received after shutting down operations that had been hopelessly unprofitable" such as shares in firms that beat Uber worldwide (Didi, Yandex, Grab) or in firms that acquired its failed moonshot projects (Aurora's acquisition of Uber's failed autonomous vehicle unit).

Horan estimates 2023 and 2024 "net profitability" were each inflated by $1.8 billion due to the aforementioned practices. He also finds that $6.4 billion of Uber's 4Q24 bottom line came from a tax valuation release, thanks to the massive losses it incurred for years (it has claimed over $41 billion in deffered tax assets). These add up and overstated Uber's P&L by 18 points in 2024, made "net profitability" a useless metric, but still allowed the firm to claim profit margins substantially higher than reality. To make matters worse, the firm publishes deceptively simplistic metrics that shed light any real changes in the way its services are consumed, their unit economics, differences across locales, or anything detailed enough to invite scrutiny. Operational costs are similarly blended and not broken out at the level of detail needed to clearly figure out Uber's take or driver revenue, driver incentives and bonuses, and so on.

These practices persist to this day and there’s little reason for Uber to change them as investors and reporters continue to report out Uber’s invented metrics, accept its garbled P&L data, and parrot back the talking points about its sustainability and viability and value that are built atop the distorted results.

After correcting Uber's numbers, however, Horan still found a massive improvement—$4 billion in 2022, then another $2 billion in 2023 and 2024. How?

Three major factors appear to have driven these improvements: Uber has been keeping a larger share of each passenger dollar (and giving drivers less), it eliminated major corporate costs during the pandemic, and developed more sophisticated price discrimination tools allowing it to charge higher fares to customers more likely to accept them and to reduce compensation offers to the minimum they thought specific drivers would accept.

We’ve talked extensively about Uber’s successful campaign to hike fares while cutting driver pay by taking a larger cut of each trip. The second boils down to shifting more costs onto the driver by pulling back on marketing and subsidies (incentives and bonuses), instituting lockouts to reduce the number of drivers (especially in cities like NYC which instituted a minimum wage law tied to driver utilization rates), and slashing its corporate workforce. The third factor—sophisticated price discrimination tools—has proven to be especially lucrative and is key to bolstering how much can be gained with the first two.

University of California, Irvine School Law Professor Veena Dubal’s law review article “On Algorithmic Wage Discrimination” extensively fleshes out how a firm like Uber can leverage their vast surveillance apparatuses, built out in the name of data-collection, and use them to squeeze their workers:

As a labor management practice, algorithmic wage discrimination allows firms to personalize and differentiate wages for workers in ways unknown to them, paying them to behave in ways that the firm desires, perhaps for as little as the system determines that the workers may be willing to accept. Given the information asymmetry between workers and firms, companies can calculate the exact wage rates necessary to incentivize desired behaviors, while workers can only guess how firms determine their wages.

…

[A]lgorithmic wage discrimination also creates a labor market in which people who are doing the same work, with the same skill, for the same company, at the same time may receive different hourly pay. Digitally personalized wages are often determined through obscure, complex systems that make it nearly impossible for workers to predict or understand their constantly changing, and frequently declining, compensation.

Dubal’s article also features eight years of embedded ethnographic research centered on ride-hail drivers in California from 2014 to 2022.

In defining algorithmic payment structures as unfair and unjust, workers frequently complained of their low hourly wages, even though they were not paid hourly. In describing the harms they suffered, they drew on the language of antidiscrimination law, condemning the variability of their income not just over time but more specifically compared to other drivers. The fact that different workers made different amounts for largely the same work was a source of grievance defined through inequities that often pitted workers against one another, leaving them to wonder what they were doing wrong or what others had figured out. This feature of algorithmic wage discrimination—because of its divisive effects—may also undermine workers’ ability to organize collectively to raise their wages and improve their working conditions.

In addition to complaints about the unfairness of low, variable, and unpredictable hourly pay, workers made two other moral judgements about the techniques through which they were paid. First, as the techniques of algorithmic wage discrimination deployed by on-demand firms both lowered pay and became increasingly obscure, drivers described the process of attempting to earn not through the lens of gaming but through the lens of gambling. Second, they portrayed the algorithmic changes or interventions that prevented them from earning as they had hoped or expected as trickery or manipulation enacted by the firm. Vacillating between feeling possibility and impossibility, freedom and control, workers experienced algorithmic wage discrimination as a practice in which the machine boss’s structures and functions were designed to take advantage of them by providing the illusion of agency. As Dietrich, a part-time driver in Los Angeles said, “[It’s] constant cognitive dissonance. You’re free, but the app controls you. You’ve got it figured out, and then it all changes.”

I want to emphasize this is only one prong of what is wrong with Uber but it is a large part of it. If I could sum up everything we’ve covered and a few other key points for you, it would go as follows:

• Uber is an enterprise that regularly uses accounting tricks to obscure its failure to realize profits through innovation. It has taken advantage of superficial coverage in the business press and crafted an aggressively deceptive PR campaign built on older taxi deregulation lobbying playbooks as well as company-sponsored academic research.

• Uber has realized stupendous growth by using capital as a weapon (investor subsidies), breaking the law then rewriting it (regulatory arbitrage), embracing anti-competitive business practices, and deploying incredibly successful political operations in the United States and worldwide that have codified its arbitrage efforts.

• Uber has realized profits largely through predatory behavior such as algorithmic wage discrimination and perpetual fare hikes. Expansions into other lines of business have benefited from its years of experimentation here (namely food delivery).

• Uber has offered a roadmap for other firms interested in immiserating their workers, growing via the misallocation of public subsidies. The metastasis of the so-called “gig economy” will continue unabated as Uber’s bastards proliferate.

• It is bad when a firm uses investor subsidies to distort markets such that it can realize profits through predatory behavior, even worse when it does so with public subsidies, and even worse when this allows the firm to rewrite laws to realize profits locked behind activities made illegal because they are against the public interest, and creates a model for other firms to do so. That Yglesias cannot understand this suggests he is an idiot or operating in a different moral universe.

It’s not clear why Yglesias looks at all of this and sees innovation. The previous system was not great, and we did not end up with a better system. Arguably it is worse, in part because it attracts commentators eager to defend it from criticism for no reason other than its use of technology. Is this what passes for “economic analysis?”

If you are so focused on innovation, how can you confuse it with plain old predation given a new digital skin? If you claim to be concerned with good outcomes, how can you overlook a business model that prizes its ability to sacrifice its workers, distort markets, break laws, and undermine democracy?