This week, to much fanfare, Google introduced its new Universal Commerce Protocol (UCP). UCP was developed in collaboration with multiple retail partners, including Target and Walmart.. UCP involves artificial intelligence (AI) agents (i.e., programs that can perform some tasks autonomously) in the shopping process (aka, agentic commerce). Simply put, instead of searching for a specific product on the web, finding your preferred location, then going there to perhaps purchase it, you can perform the entire transaction within Google’s AI Mode through UCP.
Google’s announcement was met with immediate (and frankly, warranted) skepticism and concern, particularly given recent court decisions finding Google has engaged in anticompetitive conduct in both search and display advertising. Lindsay Owens, Executive Director of The Groundwork Collaborative (whose recent research on price discrimination by Instacart’s shopping algorithms prompted congressional calls for an investigation) raised concerns regarding the potential for UCP to result in supracompetitive consumer prices. Her two-part viral tweet appears below.
The concern here is less with upselling, which occurs ubiquitously, but rather more with the potential for data collected from consumers’ AI prompts to motivate price discrimination by creating or amplifying existing market power.
Google’s Defense Made Matters Worse
Google immediately denied such claims in an attempt to mollify any concerns that UCP would lead to consumer harm. But, in doing so, Google did the exact opposite. Look carefully at the highlighted text.
If you’re familiar with the ongoing Ad Tech litigation against Google, as well as the pricing parity cases against Amazon, you might have reacted with more than just a little surprise. This looks A LOT like the very same conduct for which both Google and Amazon have been accused of violating the antitrust laws. Simply put, the highlighted text reflects a most-favored nations agreement: merchants cannot sell the same product at a lower price elsewhere than the price at which they sell on Google. This reflects a reduction in choice.
Notably, in NCAA v. Board of Regents, the Supreme Court noted that widening choice reflects a procompetitive outcome. Consistent with this view, courts adjudicating the ad tech antitrust matters have recently found a similar policy that Google implemented with respect to display advertising to constitute anticompetitive conduct. The specific Google practice there is called Unified Pricing Rules (UPR). In the district court’s August 5, 2024 Memorandum Opinion, Judge Brinkema described UPR as “a policy that prohibited publishers using DFP [DoubleClick for Publishers] from setting higher price floors for AdX [Google’s ad exchange] than for other exchanges…Unified Pricing Rules also prohibited DFP publishers from setting higher price floors for Google AdWords demand than for demand from other ad networks or demand-side platforms.”
Google’s own description indicates that it implements a similar policy with respect to Google Shopping—namely that merchants cannot advertise lower prices on other platforms (including their own sites) than on Google. But in the ad tech case, publishers indicated that they had good reason to reject such a policy. In finding that UPC constituted anticompetitive conduct, the court explained,
But in implementing Unified Pricing Rules, Google simultaneously took away publishers’ ability to set higher price floors on AdX than on third-party exchanges, which was a primary tool that publishers had used to maintain revenue diversity and to mitigate Google’s dominance of the ad exchange market. Publishers viewed Unified Pricing Rules as not in their best interests, but felt stuck using DFP given its tie to AdX. Unified Pricing Rules is another example of Google exploiting its monopoly power and tying arrangement to restrict its customers’ ability to deal with its rivals, thereby reducing its rivals’ scale, limiting their ability to compete, and further compounding the harm to customers. Under these circumstances, Unified Pricing Rules constituted anticompetitive conduct because it involved Google using its coercive monopoly power to deprive its publisher customers of a choice that they had previously exercised to promote competition.
In the ad tech case, Google could have chosen to compete on the merits rather than imposing UPR. It could have reduced its AdX take rate to motivate publishers to choose its exchange. Instead, it chose an anticompetitive course of action (UPR).
The same concept applies here. A seller’s ability to set different prices across sales channels can benefit consumers. For example, suppose one shopping platform offers lower fees to sellers, just as some ad exchanges offered lower take rates than AdX. Sellers can take advantage of those lower fees and pass on the benefits to consumers in the form of lower prices. In turn, this places pressure on rival platforms to lower their own fees and offer consumers the same benefits. Similarly, if sellers can avoid such costs, they can offer lower prices on their own sites. This is how competition works. Imposing price parity requirements as Google indicates it that it does, avoids such competition, to the detriment of consumers.
This article focuses on Google, but Amazon has also previously implemented a price parity policy. Amazon dropped this policy in Europe in 2013, after facing multiple investigations from European competition authorities. (See FTC 2nd Amended Complaint ¶275.) Though it also abandoned the policy in the United States in 2019 after facing legislative pressure, particularly from Sen. Richard Blumenthal (D-CT), Amazon continues to face antitrust suits based on implicit enforcement of such policies. The FTC has alleged that Amazon implicitly enforced this policy “through an internal mechanism called Select Competitor – Featured Offer Disqualification” (See FTC 2nd Amended Complaint ¶277.). In other words, sellers who did not abide by the price parity policy could lose their Buy Box eligibility. For more information on how the Buy Box works and its role in algorithmic pricing, you can see my recent piece on The Sling on this topic.
Google’s public acknowledgment that it imposes a price parity policy seems at best an unforced error, thought it may also signal its confidence that its conduct can escape the “anticompetitive” label in this case. Google might argue that it does not have monopoly power in shopping, particularly given Amazon’s presence. But Google does have monopoly power in search advertising. UCP will integrate with AI Mode in Google Search, where Google continues to test ads. Google has also added Direct Offers, a new Google Ads pilot that “allows advertisers to present exclusive offers for shoppers who are ready to buy — like a special 20% off discount — directly in AI Mode.”
Leveraging Its Search Monopoly into Online Shopping
Google already serves ads in its other AI-powered search product, AI Overviews (the AI-generated summaries that appear above search results) as well as AI Max for Search. The competitive concern here are twofold: (1) that Google will leverage its market power in search to the online shopping industry and (2) the remedies contemplated in the Google search case, specifically the expectation that AI would begin to dilute Google’s market power in search, may prove less effective than anticipated, if at all.
Specifically, the competitive concern would arise because, as mentioned, Google is the dominant search engine, and it now offers AI Mode in Search, which uses Google’s family of Gemini LLMs. AI Mode allows a more in-depth, “conversational” interaction between an individual and the AI-powered Gemini-3 model. It is worth noting that Google already knows a lot about individual users from products other than search: Chrome, Gmail, Google Fiber, and so on. As Google itself has acknowledged, it “draws insights from across your Google apps to provide customized responses from Gemini.”
The information people feed into Gemini, ChatGPT, and other LLMs provide more information about that user, valuable data that allows platforms to monetize their user base. Take, for example, the OpenAI commercials of the sort (“let ChatGPT plan your vacation”, or “use ChatGPT to schedule your day”). The intent here is to integrate AI into every facet of life, maximizing a platform’s opportunities for commercial extraction. OpenAI’s Sam Altman provided perhaps the emblematic example of this goal when he told Jimmy Fallon “I cannot imagine figuring out how to raise a newborn without ChatGPT,” as though humans have not been doing this very same thing for millennia.
Suppose you type in your agentic commerce-empowered AI chatbot that “I’m looking for lightweight running shoes with a carbon plate and support for pronation” instead of “I’m looking for running shoes.” An LLM can glean more important information from the former than the latter, which in turn informs that back-end machine learning algorithms. (For more about how such algorithms can result in tacit algorithmic collusion, you can check out my new paper on this topic here.) The former description suggests you’re a serious runner, likely a racer, who is familiar with various purpose-designed shoe features. It can then “upsell” you on other products that similar individuals have purchased. You can see Google acknowledging this below.
Learning individual preferences “on a deeper level” from user interactions allows Google to build out your consumer profile more accurately. This practice also preys on information asymmetries. While some may regard LLM outputs as ground truth (note the tendency of Twitter posters to ask “hey grok is this true), platforms such as Google can exploit that misunderstanding. After all, consumers acting more financially responsible and making better decisions doesn’t keep the lights on in the data center, nor does it pay for those new NVIDIA Blackwell chips. It’s worth remembering Google’s own warning: “Generative AI is a type of machine learning model. Generative AI is not a human being. It can’t think for itself or feel emotions. It’s just great at finding patterns.”
Amazon’s AI Shopping Tool Also Inflicts Harms
Of course, agentic commerce has some ostensible appeal. Cross-platform checkout can potentially save time or otherwise improve comparison shopping (absent the price parity restraint that Google imposes). But such attempts by Amazon have met with mixed results. Businesses reprimanded Amazon for using its AI shopping tool through its Shop Direct program to list products on its site without their permission. Shop Direct allows potential customers to browse offerings from brands sites directly on Amazon. The individual can then complete the purchase by clicking the “Buy for Me” button, which prompts the AI agent to purchase product on the shopper’s behalf.
The problem arose when the AI agent attempted to purchase items that the shop does not even sell or when the seller does not even participate in the Amazon program. Hitchcock Paper, a stationary company in Virginia explained in an Instagram post,
I fiercely believe this is why we shouldn’t let AI control things with no human backup or accounting. Amazon should not be beta testing faulty programs on small businesses without ANY way for us to seek help when it inevitably goes wrong. @sellonamazon, unknowingly involving my business in this program – then requiring me to pay to get help – is deceptive and wrong.
Other sellers noted that Amazon’s AI agent attempted to buy discontinued products from third party sellers though Shop Direct, indicating that this was not an isolated incident but a program that affected sellers more broadly. Such practices point to another source of consumer and seller harm: platforms’ misuse of agentic commerce can impose transaction costs on sellers, which eventually translate into higher prices for consumers.
Amazon itself is not immune to the vagaries of agentic commerce. In November 2025, Amazon sued Perplexity, an AI-powered answer engine that operates the Comet web browser application. Comet AI incorporates agentic AI functionality, enabling it to take actions on users’ behalf, including placing orders on Amazon’s store. Amazon alleges that Perplexity did not identify its AI agents as such, and that Perplexity set up Amazon Prime accounts, enabling users to make purchases on Amazon and take advantage of Prime features without paying for them.
Agentic commerce makes many promises. Whether these actually manifest themselves remains to be seen. The outcome will depend, at least in part, on whether platforms engage in good-faith efforts to improve consumer experiences, or instead turn to exploitative practices that mirror those already challenged under antitrust statutes. If anything, consumers and sellers have cause for concern.
Housing prices are up, and would-be homeowners are shifting to rental units. The inventory of homes for sale is shrinking because investors are buying up properties with cash offers. Investors then rent the homes to households who cannot afford mortgages. Many feel that the American dream of homeownership is slipping away. Building more homes won’t alleviate the problem if those additions are not in the right location or if investors buy them first.
To address this pinch, right after the New Year, President Trump called for Congress to cap the holdings of institutional investors in the housing market. A very reasonable idea, so long as it is implemented correctly.
Within days of the president’s announcement, the New York Times Opinion section featured an essay titled “The Landlords Are Not The Problem,” noting that institutional investors collectively own “less than 1 percent of the nation’s single-family homes—and less than 5 percent of single-family rentals.” That figure presumes that the relevant geographic area for studying investor pricing power is the nation. But housing investors are not randomly acquiring properties across the nation. Instead, they are selectively acquiring properties in neighborhoods to maximize their pricing power. This purchasing strategy, sometimes called “rentlining,” entails buying homes that are most likely to permit rent extraction from tenants who lack options. Measuring investor ownership using the nationwide housing stock as the denominator artificially deflates the true investor share of the markets in which they operate.
The Wealth Defense Industry Strikes Back
President Trump’s proposal sent members of what Matt Stoller aptly calls the “wealth defense” industry into overdrive. Jay Parsons, former chief economist of RealPage, was quick to tout a study by the American Enterprise Institute (AEI), a libertarian think tank that has been one of the loudest opponents of recent federal and state efforts to restrict investor homebuying. The AEI report, from August 2025, estimated that institutional investors owned a small share of single-family homes when looking at the national, county, and even zip code levels. That report offers the following analysis of single-family home ownership:
Institutional ownership of single-family homes is highly concentrated and varies significantly at the county level. Just 162 counties (or 5% of U.S. counties for which Parcl Labs data are available) account for 80% of all institutionally-owned homes, according to Parcl Labs data. Yet not a single county has a share greater than 10%…[E]ven in metros that have received significant media attention for their more pronounced investor presence, such as Atlanta (4.2%), Dallas (2.6%), and Houston (2.2%), these investors do not dominate any single neighborhood[…] In Atlanta, for instance, the highest institutional investor share in any ZIP Code is 12.4%, while half of its ZIP Codes have a share below 1.5%. No ZIP Code in Houston has an institutional investor share of over 10%[.]
Although this analysis may inspire faint hope as far as it shows that institutional investors haven’t yet taken over most U.S. housing markets, this analysis misses two valid concerns of affordable housing advocates: (1) concentration can be caused by large local players, regardless of their national or statewide holdings, and (2) neighborhoods, not zip codes, are the relevant geographic market.
Institutional Investors Aren’t the Only Source of Pricing Power
The current policy discourse has poorly defined the valid concerns about institutional housing investors. In recent bills proposed to regulate institutional investors, these investors are generally defined as an owner of one hundred or more residences. For example, Florida’s House Bill 1593 regulates the home-buying of a business that “has an interest in more than 100 single-family residential properties in this state.” The AEI report quoted above uses the same one-hundred-home designation. This definition misses the crux of the pricing power issue.
No housing advocate or economist has ever drawn a bright red line at one hundred units as the threshold for corporate pricing power. When scrutinizing the ability of businesses to coordinate and set artificially high prices, economists measure the overall concentration of the relevant market, regardless of the asset portfolio of the participants. Moreover, sometimes different firms that purchase homes are subsidiaries of a larger firm, so a firm may appear to be a small investor but actually be just a tentacle of a larger corporate entity.
In this instance, a more accurate analysis of pricing power examines which local housing markets are controlled by a few big players, rather than analyzing arbitrary ownership thresholds. Owning just five homes in the same neighborhood reasonably conveys investor status, regardless of whether the owner is an institutional investor.
Taking this more expansive view of market concentration allows us to measure pricing power in individual housing markets regardless of national (or statewide) asset portfolios.
Measure Neighborhoods, Not Zip Codes
A second issue with AEI’s claims is that market power must be evaluated in a relevant geographic market. The Merger Guidelines compel us to ask, how much contiguous real estate would a hypothetical landlord have to acquire in order to raise rents over competitive levels? Although zip codes can be a useful and relevant unit of observation, zip code boundaries do not necessarily reflect the areas within which a renter or homebuyer considers their options. A medical student at University of Miami who lives in the trendy Brickell neighborhood for easy access to the metro station (and a short ride to the medical campus) would not consider any apartment in the 33130 zip code as a substitute.
To study this issue, we focused on Atlanta, as the AEI study provided institutional market shares by zip code there. Fulton County provides a map of all official neighborhoods in Atlanta, which we combined with a dataset containing all tax parcels in the county to identify all single-family homes in Atlanta by neighborhood.
Because there are approximately four times as many official neighborhoods as zip codes in Atlanta, neighborhoods provide a more narrowly defined set of smaller geographic housing markets relative to AEI’s zip code analysis. Furthermore, renters and buyers both seem more likely to deliberately target neighborhoods rather than zip codes when searching for their next home.
Neighborhood Market Power Analysis
We assembled a dataset of approximately 75,000 single family homes in Atlanta. We identified homes owned by an LLC or other business entity, and we also identified “investor owners,” which we define as homeowners who own at least five homes within a given neighborhood. The vast majority of investor owners are business entities. In line with AEI’s research, we find that investor owners hold a small share of homes in most Atlanta neighborhoods.
While AEI asserts that investor owners do not hold over 12 percent of homes anywhere in Atlanta, we identified Atlanta neighborhoods where investor owners held larger shares of single-family homes. Investor shares of the housing stock are especially large when we exclude owner-occupied homes from the analysis, considering only those homes that are currently empty or occupied by renters.
A prominent example is Historic Westin Heights/Bankhead, a neighborhood just outside of Atlanta’s downtown. Bankhead’s single-family housing stock is dominated by Canopy Development Group, which holds 11 percent of all homes in the neighborhood and 17 percent of homes that are not owner-occupied. (According to its website, Canopy “is leading the largest land and property acquisition effort in Atlanta’s Westside Beltline area.” Per the Atlantic Journal-Constitution, Canopy “has used anonymous limited liability corporations to buy up large portions of west Atlanta’s impoverished English Avenue neighborhood.” While Canopy has developed properties as well, it obtained its status via acquisition.). Mechanicsville, named for the mechanics who worked on the rail line, is another neighborhood community with substantial investor ownership, with investors holding 14 percent of all homes and 24 percent of homes that are not owner-occupied.
We identified three other Atlanta neighborhoods with at least 350 single-family homes where at least 10 percent of homes that are not owner-occupied are owned by investors.
It is worth noting that three of the five neighborhoods highlighted by this analysis (Bankhead, Collier Heights, and Capitol View) are among the “Beltline” neighborhoods, communities adjacent to a major urban renewal project. The Beltline project has displaced many long-tenured homeowners in Atlanta, making their former homes available to large-scale landlord investors.
High Shares in a Neighborhood Are Consistent with Direct Evidence
Given the low investor shares at the zip code level (implying lack of pricing power), and given the high investor shares in certain neighborhoods, one can look to direct evidence of investors’ pricing power to resolve the dispute. If investors can be shown to inflate rents, then the narrower geographic market is consistent with the direct evidence of pricing power.
There is a large and growing literature demonstrating the inflationary effect of these types of institutional holdings on rental prices in local housing markets. A July 2020 working paper from St. Louis Fed economists investigated the effect of institutional investors —defined as “entities who purchase multiple housing units under the name of an LLC, LP, Trust, REIT, etc.” — on rental prices. The economists found that institutional investors increase the price-to-income ratio of rental properties, especially in the bottom price-tier. In an antitrust court, such evidence would be considered “direct” evidence of the pricing power of institutional investors, which obviates the need to define a market, estimate share, and infer market power through high market shares.
In another study of rental pricing, Watson and Ziv (2021) analyzed the relationship between ownership concentration and rents in New York City, finding that a ten percent increase in concentration is correlated with a one percent increase in rents. This finding suggests that policymakers should be concerned about concentration of ownership, regardless of whether concentration is comprised of institutional investors or smaller investors.
These findings importantly hold even when looking within individual neighborhoods over time. Using mergers of private-equity backed firms to isolate quasi-exogenous variation in concentration of ownership at the neighborhood level, Austin (2022) found that shocks to institutional ownership cause higher prices and rents. This finding suggests that the association between institutional ownership and higher prices isn’t merely selection bias (institutional investors happening to invest in hot housing markets).
Neighborhood Ownership Caps Make Sense
Although AEI and other housing concentration skeptics are correct that regulating corporate landlords is not a silver bullet to address the ongoing affordability crisis, their analysis and rhetoric understate the reality of housing investor ownership.
A landlord does not need to own one hundred properties in a state to contribute to the concentration of economic power. Corporate landlords target neighborhoods where homes can be purchased cheaply and rented out profitably because renters in that area have limited choices. These renters’ options are limited in part because this targeted approach creates market power in the neighborhood-level housing market.
Landlords, especially those owning many homes in a single community, are part of the housing crisis, especially in rentlined communities like Bankhead in Atlanta. Regulating the accumulation and exercise of market power will always be part of a holistic solution to market failure. Our analysis suggests that a modest cap on the share of rental properties in a neighborhood that a single investor could own—say, of five or ten percent—could weaken the grip of investors and give renters some much-needed relief.
In recent months, public attention has returned to a business practice that many consumers intuitively recognize as unfair and has provoked varying attempts to regulate: surveillance pricing. Investigations into Instacart’s grocery pricing, along with renewed scrutiny of algorithmic pricing by the Federal Trade Commission (FTC), have revived concerns that firms increasingly tailor prices to individual consumers based not on cost or competition, but instead on perceived willingness, or necessity, to pay.
The debate is often framed as a question of technology and fairness: Can algorithms responsibly personalize prices? Do data-driven pricing systems benefit price-sensitive consumers? Are these tools simply the next step in efficient market segmentation?
These questions miss a more fundamental issue of industrial organization. Surveillance pricing is not primarily a technological innovation; it reflects a lack of competition. The profitability, persistence, and coercive nature of surveillance pricing depend on the presence of market failures—especially high concentration, entrenched frictions, and severe information asymmetries. In genuinely competitive markets, surveillance pricing would be unstable and self-defeating. In oligopolistic ones, it becomes a powerful mechanism for extracting consumer surplus.
The rise of surveillance pricing therefore offers a diagnostic insight into modern capitalism: it reveals how concentrated markets transform data and algorithms into tools of consumer exploitation rather than competition.
A Familiar Pattern of Abuse
Surveillance pricing is not new. For more than a decade, journalists and regulators have documented variations of the practice. Ticketmaster’s “dynamic pricing” has transformed concert tickets into auctions that capture nearly all consumer surplus from the most devoted fans. Orbitz infamously offered more expensive hotel options to Mac users, assuming they were less price-sensitive. Staples and Target experimented with GPS-based pricing that charged more to customers situated close to their stores and far from competitors. A ProPublica investigation revealed that Princeton Review charged higher prices to users from ZIP codes with more Asian people.
What unites these examples is not merely personalization, but exploitation. These pricing strategies succeed by identifying moments when consumers are least able to walk away: emergencies, deadlines, emotional commitments, and logistical constraints. The algorithmic sophistication matters less than the underlying logic—using asymmetric information to extract maximum payment from captive demand.
Surveillance pricing relies on exploiting market frictions that are endemic to many modern consumer markets.
Stark information asymmetries mean that sellers now know vastly more about buyers than buyers know about pricing strategies. Firms collect data on browsing history, purchase patterns, location, device type, demographics, mouse movements, and even phone-battery life. Algorithms can infer “need points,” or moments of heightened urgency, such as last-minute flights for the funeral of a family member or late-night ride searches with a dying phone. Consumers, by contrast, have little-to-no visibility into whether a price is individualized, how much it differs from others’ prices, or which data triggered the increase.
Search costs further weaken consumer discipline. Comparing prices once meant driving between stores or walking through malls. Online commerce promised to eliminate these costs, but instead replaced them with digital equivalents: loading multiple sites, navigating opaque interfaces, deciphering bundled products, ascertaining the value of unique features of slightly differentiated offerings, and deciphering subtly differing seller policies through terms and conditions. Each additional click imposes friction that pricing algorithms exploit.
Switching costs compound the problem. Loyalty programs, termination fees, stored payment information, and personalized settings make leaving costly. Consumers must weigh the potential savings from switching against the time and effort already invested in a transaction: accounts created, finance and shipping details entered, digital cart configurations, product configurations, or accrued rewards. Even modest price discrimination can succeed when a consumer feels exhausted by the time, effort, and other sunk costs of conducting market research to make an informed decision to switch.
Learning costs also matter. Choosing a new grocery store requires learning a different aisle layout and where your items are located. A new clothing retailer might have different return policies that burden a customer’s usual practice of trying and returning apparel. In e-commerce, new sellers that a consumer has never heard of may require complex research to distinguish legitimate options from increasingly sophisticated scams that pose a constant threat of financial ruin and require only a single lapse in vigilance and judgment. These cognitive burdens all discourage switching even when alternatives exist.
Product differentiation and bundling further obscure value comparisons. Airlines package seats, bags, boarding priority, and insurance; travel sites bundle flights and hotels; retailers vary sizes, features, and subscription terms. Surveillance pricing thrives in environments where consumers struggle to identify a clear benchmark price.
Finally, in markets like airline tickets, the time gap between purchase and consumption, combined with the difficulty of resale, allows firms to identify inelastic demand. When a product is essential or temporarily critical, like food, last minute transportation, or attendance at an important family event, a customer’s willingness to pay increases dramatically. Algorithms need only detect necessity to capture more consumer surplus.
Why Concentration Is the Key Enabler
Even in a market with the frictions described above, surveillance pricing would be unstable without concentration. In a competitive market, a firm that raises prices for less price-sensitive customers would invite rivals to advertise uniform pricing and capture those consumers. Transparency and rivalry would discipline discriminatory strategies. Even the threat of such competition would deter firms from deploying surveillance pricing at scale.
Oligopolistic markets change the calculus. When a small number of dominant firms control most sales, each can reasonably expect rivals to follow suit rather than defect. The profits from mutual adoption of surveillance pricing outweigh the risk of lost customers when alternatives are limited. Any firm that refuses to participate risks retaliation: price wars with temporary below-cost predatory pricing, increased output, aggressive advertising, complex bundling, or loyalty programs designed to lock in consumers. In concentrated markets, the cost of defection is high, and conscious parallelism becomes the rational equilibrium.
Even in markets with some competitive fringe, dominant firms can deploy partial surveillance pricing. They may offer competitive prices to the most price-sensitive customers while charging inflated prices on niche or low-volume items to less elastic buyers. Or they may use competitive pricing on headline items to shape price perception while extracting surplus elsewhere. The result is higher overall margins without provoking meaningful competition.
At the extreme, a perfectly concentrated oligopoly using surveillance pricing could capture nearly all consumer surplus. In a frictionless, competitive market, the same strategy would drive customers away. Surveillance pricing therefore scales with concentration.
The Illusion of Pro-Consumer Benefits
Defenders of surveillance pricing sometimes argue that it benefits lower-income or more price-sensitive consumers by offering them discounts. This argument collapses under scrutiny.
For surveillance pricing to be profitable, total surplus extraction must increase, otherwise the scheme would be irrational. Discounts for some consumers are outweighed by higher prices for others. In practice, the “discounted” prices approximate what consumers would have paid in a genuinely competitive market, while the inflated prices represent pure extraction from those deemed able—or forced—to pay more.
This is not redistribution; it is maximal surplus extraction. The poorest consumers don’t gain new surplus. Everyone else loses it.
Surveillance pricing has expanded during an era of wage stagnation, declining labor share, rising markups, and elevated corporate profits. These dynamics have steadily reduced household purchasing power and slowed economic growth by suppressing aggregate demand, contributing to a host of social ills related to economic anxiety.
Retail competition once expanded consumer surplus through price matching, coupon honoring, and aggressive rivalry. These strategies once reflected a modicum of buyer power and informational symmetry. Sellers could not individually tailor prices based on personal data; competition disciplined margins.
Surveillance pricing represents a reversal of that equilibrium. It shifts power decisively toward sellers by weaponizing data in markets already tilted by concentration.
Antitrust Law and the Limits of Current Enforcement
While several state and federal legal regimes such as consumer protection, privacy, and anti-discrimination laws offer potential avenues for addressing surveillance pricing (as well as surveillance wage setting) with varying degrees of potential limitations, competition law faces some unique obstacles.
Surveillance pricing sits uneasily within existing antitrust doctrine. It often resembles collusive price-setting, but without explicit agreement. An algorithm or third-party data mining firm that pools competitively sensitive information from multiple competitors to help determine prices could, in theory, constitute an illegal hub in a hub-and-spoke conspiracy under Section 1 of the Sherman Act. But firms could theoretically avoid liability by independently customizing their algorithms and preventing them from using aggregated data from third-party brokers or directly incorporating competitor prices and data. Even independent pricing algorithms could lead to anticompetitive outcomes.
Efforts to prevent algorithmic collusion—such as the California bill banning the use of competitor pricing data—are directionally helpful, but incomplete. Even strictly siloed algorithms could infer market conditions, albeit imperfectly, through demand elasticity tests, purchase rates, cart abandonment, page views, and customer churn rates. Conscious parallelism, or competitors engaging in mutually beneficial common conduct without explicit agreement, remains lawful, and courts have long refused to punish firms for independently adopting strategies that are mutually profitable even when they produce consumer harms identical to cartel price fixing.
As a result, antitrust enforcement alone struggles to address surveillance pricing when it arises from ostensibly lawful parallel conduct rather than explicit price coordination in markets where lax merger enforcement failed to prevent the concentration levels that enable its success.
Mandating Disclosure as Critical First Step
A blanket ban on surveillance pricing would be the most direct solution, but it faces political and legal obstacles. Mandatory disclosure could be a promising first step and is already incorporated in some algorithmic pricing bills and being considered in others. Firms could be required to clearly inform consumers, at the point of sale, that prices may be individualized based on surveillance data. This mirrors the recent FTC rule on junk fees, which forced upfront price disclosure in travel and event markets. Once the entire market was covered, no single firm faced competitive disadvantage for transparency—and many began advertising “what you see is what you pay,” as a consumer-friendly feature.
A disclosure regime would not regulate prices or restrict business autonomy. It would restore informed consent and reduce buyer-side frictions. Ideally, by making discrimination visible, it could reactivate competitive market pressures and be a significant step in curbing its abuse.
A private right of action for violating disclosure could further enhance enforcement, reducing reliance on resource-constrained consumer protection agencies and regulators and allowing harmed consumers to police abuses.
Surveillance Pricing as a Structural Warning Sign
Surveillance pricing should be understood not as an isolated abuse, but as a structural warning sign. It thrives where competition has failed, where consumers lack meaningful alternatives, and where firms can safely exploit moments of vulnerability.
The core question is whether a business strategy that succeeds only by exploiting market failure should be allowed to persist at all. If surveillance pricing is profitable only in oligopolistic markets, then its spread is evidence—not of efficiency—but of the urgent need for stronger competition policy and market transparency.
In that sense, surveillance pricing does more than raise prices. It exposes the costs of having allowed so many markets to reach oligopoly levels of concentration in the first place.
Randy Kim is an assistant city solicitor for Philadelphia and a recent graduate of the University of Pennsylvania’s Carey Law School. The opinions expressed here represent those of the author and not those of his employer.
Price inflation has been the dominant economic concern for Americans in the post-Covid era. The rising prices of cars, groceries, and healthcare (especially given recent Congressional inaction) have all imposed increasing burdens on the average American. Despite the consistent price hikes for those items, all of them pale in comparison to the skyrocketing rental costs that Americans have endured since the pandemic. Per Harvard’s Joint Center for Housing Studies, a record 12.1 million renter households were spending at least half of their incomes on housing in 2022, putting them at increased risk of eviction and homelessness.
According to Zillow, rental prices have increased by more than a third since the pandemic, while the median household income has risen by only 22 percent. As the gap between earnings and rent prices widens, families are forced to stretch their budgets to afford shelter. In cities, these higher rental prices have forced families out of their neighborhoods in search of more affordable housing. The press and politicians have both acknowledged the affordability problem that has resulted from the recent housing crisis wave. Indeed, in September 2025, the U.S. Treasury Secretary Scott Bessent noted that the Trump administration was considering declaring a “national housing emergency” to address affordability.
So it was odd that The Economist last week sought to deny the reality that is in front of our faces. In a briefing titled “America’s affordability crisis is (mostly) a mirage,” the magazine asserts that “on the economics, Messrs Trump, Bessent and Duffy have a point” when they claim that the affordability crisis amounts to a “hoax” and a “con job.”
The Economist brings this same skepticism to the rental affordability crisis. A widely used industry rule-of-thumb is that rents are affordable so long as they account for less than 30 percent of a renter’s income. While acknowledging that “the squeeze for [home] buyers is real,” The Economist dismisses the concerns about rental affordability:
Until rates began rising in 2022, the average home in most counties was affordable by the 30% rule-of-thumb, even for buyers with only a 10% downpayment. Now, most are not (see chart 4). Homeowners who fixed their mortgages before rates went up have dodged this. The average rate on all outstanding mortgages is still only 4.3%, nearly two percentage points less than the average rate on new mortgages. Still, the squeeze for buyers is real. Rents, which are less directly affected by mortgage rates, are more affordable: the average in most counties is still below that 30% threshold. (emphasis added)
Whether The Economist intentionally aims to gaslight its readers or simply has not given the issue the requisite thought it deserves is unclear (we’ll generously assume the latter), but either way, this curious statistic does not support the claim that rents are generally affordable.
Lies, Damned Lies, and Statistics
Let’s start by ascertaining where The Economist likely came up with this figure. The exact methodology is uncertain, but Figure 4 in its briefing relies upon Zillow, The Census Bureau, FRED, and the Insurance Information Institute. Because Zillow’s Observed Rent Index only has information for a subset of counties in the United States, the magazine’s county-level analysis likely came from the Census American Community Survey (ACS). The ACS collects annual county-level data on rent prices, income, population, and other demographic variables. That gives us the exact information we need to kick the tires. Given this uncertainty, we use the 2023 American Community Survey 5-Year Estimates.
By comparing the median rental price to the median household income in each county in the 2023 ACS, we can replicate The Economist’s conclusion—namely, that “most” of the counties enjoy rents of less than 30 percent of income. There are at least two fatal flaws, however, with that analysis.
First, the number of counties is not the relevant unit of analysis. We want to analyze the cost of living for people. This mistake is akin to the misleading map of the United States showing which candidate won county, which ignores the fact that people vote, not land. Second, looking at the median household income understates the problem because it combines homeowners and renters into the same category. The median homeowner household has nearly double the income of the median renter household. Using the median renter’s income is a better estimate of how the usual renter is doing.
Adjusting the rental analysis to measure affordability using renter’s income and taking into account population gives a much different result: the median rent is affordable—that is, it is below 30 percent of income—for slightly more than a third of the population. The rental affordability issue shouldn’t be seen as a coastal problem affecting a handful of cities; it’s endemic across the United States. The figure below shows that just these two adjustments to The Economist’s statistic make a significant difference when measuring rental affordability. The first adjustment (measuring renter income rather than household income) brings the proportion of affordable counties down from 99.8 percent to 64.9 percent. A second adjustment, measuring the population in affordable counties (rather than naively treating all counties equally) shows that only 35.2 percent of the population lives in counties meeting the rental affordability threshold.
Source: 2023 American Community Survey 5-Year Estimates.
We cannot be sure whether The Economist’s “average-in-most-counties” statistic was based on the first or second bar; in any event, both overstate rental affordability. The extent of the affordability crisis can be verified using more recent 2024 ACS data. Among the 46.1 million renters recorded in 2024, 22.3 million renters (48.4 percent) paid 30 percent or more of their income on rent. Of these, 11.2 million renters paid half or more of their income on rent. Despite the assurances of the neoliberal magazine, the rental affordability crisis is real for millions of people.
Being Honest About The Crisis
As the saying goes, the first step to recovery is admitting you have a problem. We have a serious rental affordability problem in this country that demands real solutions. There has been much progress in recent years with rent control, zoning reforms, tenant protections, permitting reform, and antitrust action (such as against RealPage). Yet these are small victories that need to be built upon and extended. Ignoring the rental problem will only cause the overburdened renter to fall further behind.
Given the flimsiness of The Economist’s rental affordability figure and the stridency of its advocacy—the briefing reviewed here serves as supplement to last week’s cover story—one wonders why the editors of such an esteemed publication want to deny there’s an affordability problem. Presumably the corporatist class whose concerns The Economist seeks to assuage fears any interventions in the market to the solve the problem. One such intervention that’s rightfully gaining traction among economists is rent controls. Neale Mahoney and Bharat Ramamurti recently endorsed price controls, including for rents, in the opinion section of the New York Times, explaining that “Rent caps focused on existing units, combined with government investment in new housing and reforms to zoning, permitting and other land-use regulations, can protect tenants from rent spikes, while encouraging new construction to build the three to four million homes that economists believe we need to make up the shortfall in the housing supply.” Similar to climate-change deniers, if neoliberal economists deny the affordability problem, they don’t have to address it.
Warner Bros. Discovery (“Warner Brothers”) announced on Wednesday that it is poised to reject a takeover bid by Paramount, clearing the way for Netflix to acquire Warner Brother’s studio and subscription streaming platform, HBO Max. As shown in the figure below, lifted from The Economist, Netflix and Warner Brothers comprise the first- and fourth-largest streaming platforms based on third quarter 2025 global subscribers.
Hence, a merger between the two streaming platforms will further consolidate the industry, similar to Disney’s majority acquisition of Hulu in 2019. But that clear concentration of economic power didn’t stop The Economist from endorsing the merger.
Seemingly at the behest of Netflix, The Economist devoted one of its lead stories to bolstering the claim that Netflix and HBO Max compete for viewers’ attention with YouTube, which mostly offers long-form (around ten minutes), amateur, ad-supported videos; and with TikTok, which offers short-form (around 35 seconds), amateur, ad-supported videos, in a purportedly immense streaming market. Never mind that Netflix and HBO Max offer studio-produced, paid-subscription streaming services of series and movies that often last over an hour. The technical term for what Netflix and HBO Max offer is subscription video on demand (SVOD). And the technical term for what YouTube and TikTok offer is ad-supported video on demand (AVOD). After reviewing purported evidence of substitution between SVOD and AVOD, the neoliberal magazine concludes that “This new competitive landscape means that trustbusters should not rule Netflix out of the Warner race, as many in Hollywood argue. It may be dominant in streaming, but under the broader market definition it is a smaller actor.”
It’s the smallest set of services, stupid!
The question for the relevant antitrust market seems to elude many in the business press and on Twitter. To bring them quickly up to speed, we offer this brief tutorial: When defining a market, the inquiry turns on the smallest collection of services such a hypothetical monopoly provider of said service could profitably raise prices above competitive levels. This test is referred to as the hypothetical monopolist test (“HMT”). Start with a hypothetical monopoly provider of SVOD services. That’s right—one firm that controls all the streaming services listed in the above figure. Could a single SVOD provider with just those assets raise prices over competitive levels of say $10 per month? If the price increase is deemed unprofitable, then the market must be expanded to include nearby substitutes, with the test repeated. (For those who want a deeper dive, see the 2023 Merger Guidelines Section 4.3.A.)
It strains credulity that a hypothetical monopolist of all SVOD services could not raise prices above competitive levels, without also controlling YouTube and TikTok. To where might (say) a Netflix subscriber turn when the price of all of Netflix’s streaming substitutes (HBO, Prime, Apple TV, etc.) also increase? Presumably those paying subscribers would stay put. What The Economist would have you believe is that, when the price of all subscription streaming services increases, a substantial share of those customers would terminate their paid subscriptions and instead spend their time watching free, amateur, short-form and long-form videos. That those outside options are not even priced speaks volumes about the lack of price-discipline imposed by YouTube and TikTok on the SVOD services—that is, the amateur producers of these videos have to give away these services for free (ignoring the ads) as an inducement to watch their content. Indeed, if AVOD services constrained the prices of SVOD services, then why has Netflix been able to raise its subscription for standard and premium service by 29 percent and 39 percent, respectively, since 2020?
Of course, there are other ways to prove a market, such as by invoking the Brown Shoe factors. These are also known as “practical indicia” of a market, such as industry recognition as a separate economic entity, unique production facilities, or distinct prices. See the Merger Guidelines Part 4.3. By any of these standards, SVOD services are a relevant market. CNET, Esquire, Yahoo!Tech, and Consumer Reports maintain rankings of the best subscription streaming services, none of which include YouTube or TikTok. In March of this year, Netflix co-CEO Ted Sarandos reportedly described YouTube as being for consumers interested in “killing time” rather than “spending time” with professionally produced movies and shows. Compared to YouTube, co-CEO Greg Peters said Netflix is “playing a specific and differentiated role in the ecosystem.” The Intelligencer quotes a top agent saying that traditional streamers like YouTube and HBO Max “do things that YouTube can’t. There’s a good 50 to 60 percent of the audience that literally has never been on YouTube. When you make A Quiet Place, it goes into the Zeitgeist forever, whereas YouTube shows don’t seem to have long-tail resonance.”
With respect to the second factor, production of movies and series for SVOD services take place at professional studios, as opposed to the basement in some amateur’s home. And SVOD services are similarly priced—Netflix’s ad-free service starts at $17.99 per month, while HBO’s ad-free service starts at $22.99 per month—whereas YouTube, TikTok and other AVOD service are generally free.
We note that there is also YouTube Premium, which is essentially identical to YouTube, but with fewer to no ads. While such a service is closer to SVOD in its monetization strategy, it is still not a substitute for Netflix, HBO Max, or other SVOD services considering that YouTube Premium has the same content as ad-supported YouTube. Additionally, there is YouTube TV, which allows for live TV streaming and costs $82.99/month, and which also does not compete with SVOD services (as evinced by content differences between the two service types and by YouTube TV’s significantly higher price point relative to Netflix and HBO Max).
To play up the degree of substitution between SVOD and AVOD services, The Economist notes that “Americans spend longer watching YouTube on tv than on their phones. At the same time Hollywood is relying less on cinemas in favour of tv, and moving to even smaller screens.” That viewers increasingly watch both services on a tv doesn’t imply that the (free) AVOD service disciplines the price of the (paid) SVOD service. The Economist next argues that SVOD platforms are introducing ad-supported offerings, while YouTube is offering no-ad plans, further blurring the lines. So? While the ads can reduce the price of Netflix or HBO Max, the ad-supported price premium is still substantially above the free services of the shorter streamers. And that premium reflects a substantial difference in quality. Finally, The Economist notes some overlap in content across the two products, such as Amazon Prime offering a series starring YouTube’s biggest star (MrBeast), while social platforms are showing television-like content such as YouTube’s “Chicken Shop Date.” This modest overlap hardly constitutes evidence of how consumers of SVOD service would respond to a small increase in the price of their services.
The relevant output market is highly concentrated
Having defined the relevant output market as SVOD, the next task is to assess the degree of market concentration, both before and after the merger. To compute market shares, we used global streaming subscription data from The Economist (pictured above) for our initial analysis.
Table 1: Concentration Index of the Subscription Streaming Services Market
Source: Subscriber counts are from The Economist.
The pre-merger HHI is 2,055, which the 2023 Merger Guidelines consider “highly concentrated.” The change in HHI owing to the merger is 829 (equal to 2,884 less 2,055). The Guidelines explain that a merger in a highly concentrated market (pre-merger HHI over 1,800) that involves an increase in the HHI of more than 100 points is “presumed to substantially lessen competition or tend to create a monopoly.”
The Economist cites Ampere Analysis as the source of its subscription data, which we could not access. As a sensitivity check, we also used 2024 U.S.-based subscription data from Statista. The pre-merger HHI falls to 1,778, barely below the 1,800 standard for highly concentrated markets by the Guidelines. The change in HHI is 546 (equal to 2,324 less 1,778). Notably, the combined share of the merging parties using the Statista data is 34 percent. The Guidelines note that “a merger that creates a firm with a share over thirty percent is also presumed to substantially lessen competition or tend to create a monopoly if it also involves an increase in HHI of more than 100 points.”
Price effects in the market for subscription streaming services
One common method used by economists to estimate the anticipated price effects of a merger is referred to as a “GUPPI” analysis (which stands for “Gross Upward Pricing Pressure Index”). GUPPI analysis follows a simple economic logic—when a firm unilaterally increases price, some of its customers substitute away to its competitors. For instance, if HBO Max were to raise its price, then the law of demand would imply that it would lose some subscribers, with some proportion of these lost subscribers diverted to Netflix. If Netflix were to acquire HBO Max, then these diverted sales from HBO Max to Netflix would remain under the same corporate umbrella, thereby dampening the competitive price discipline that Netflix would have otherwise imposed on HBO Max (and vice-versa).
There are three inputs needed to estimate a GUPPI. First, one needs an estimate of the “diversion ratio” between the merging entities, which measures the proportion of customers a firm would lose to the merging entity if it were to unilaterally increase price. Economists routinely use market shares as a proxy for diversion absent having more detailed, firm-level data. Based on the figure from The Economist above, I estimate that Netflix has a 33 percent market share in streaming, compared to HBO Max’s 13 percent. The share-based diversion ratio from HBO Max to Netflix is therefore 37.6 percent (equal to 0.33 / [1 – 0.13]), and from Netflix to HBO Max is 18.8 percent (equal to 0.13 / [1 – 0.33]).
Second, one needs prices for the merging parties’ products. For our analysis, we use monthly base tier prices for each service. As of December 2025, the price of the Netflix base tier package with ads is $7.99/month, whereas the price of HBO Max’s base tier package with ads is $10.99/month.
Third, one needs an estimate of the merging partner’s economic margin. Economic margin is equivalent to the Lerner Index—it measures the difference between price and marginal cost as a percentage of price. As far as we are aware, the marginal cost for either of streaming service to produce an extra stream is likely close to zero. While these services incur fixed and quasi-fixed costs—most prominently, content costs for either acquiring or producing shows and movies—these costs do not change on a per stream or per user basis. We do not use gross accounting margins, which are sometimes used in place of economic margins, as gross margins are contaminated by the amortization of content costs and other fixed costs over time (which do not reflect true marginal cost in an economic sense). For illustrative purposes, we assume that both platforms’ marginal costs equate to 10 percent of their revenues, which we think is a conservative estimate, thereby implying an economic margin of 90 percent (equal to [1 – 0.1] / 1). More precise information on the merging parties’ economic margins can be obtained via discovery.
The GUPPI for Netflix under a Netflix-HBO Max merger can be calculated as:
where DR_NF->HBO is the diversion ratio from Netflix to HBO Max, M_HBO is the economic margin for HBO Max, and P_HBO / P_NF is the ratio of HBO Max’s base tier monthly price to Netflix’s base tier monthly price.
The table below provides estimates of GUPPIs for both streaming services under the hypothetical merger. It bears noting that GUPPI is a pricing pressure index, but it does not necessarily equate to actual price changes—although the index can often approximate price changes under certain conditions as specified by Miller et al. (2016) and Koh (2025). Generally, antitrust authorities are concerned with GUPPIs greater than 10 percent. For both Netflix and HBO Max, we estimate GUPPIs in excess of 23 percent. These high GUPPIs raise significant alarm as to the potential for this merger to harm consumers, by allowing both platforms to charge higher prices.
Table 2: Gross Upward Pricing Pressure Indices (GUPPIs) Under Netflix-Warner Bros. Discovery Merger
To put into context, in the proposed Penguin Random House-Simon & Schuster merger that was blocked by a judge in 2022, the DOJ’s economist estimated GUPPIs of 3.7 percent to 7.4 percent (note that these GUPPIs correspond to percentage reductions in author compensation as opposed to here, where they represent increases in output market prices). In a 2020 FTC merger case involving consolidation in the hydrogen peroxide industry (FTC v RAG-Stiftung, which involved an output market GUPPI analysis similar to here), the FTC’s economist estimated GUPPIs of 5.5 to 13.2. Any efficiency justifications that Netflix and Warner Bros. Discovery may proffer here are almost certainly outweighed by the magnitude of the upward pricing pressure implied by our estimates.
Not to mention the harms in the labor market
It bears noting that the above results speak only to merger-induced price effects in the output market for streaming subscriptions. They do not address wage effects in the labor (input) market, which tend to be harder to estimate absent detailed, firm-specific data on substitution patterns of talent.
Going back to the book publisher merger, the DOJ’s economist used data on market shares and profit margins to estimate the effect of the merger on author advances using what is referred to as a second-score auction model (for books that acquired an advance of at least $250,000). In this model, the highest-bidder must only bid slightly above the second-highest bidder to win the auction. Such a model finds harm when the two merging parties would otherwise be the first- and second-highest bidders—in such a scenario, the merging party must only pay slightly above what would otherwise have been the third-highest bidder absent the merger. The DOJ’s economist also applied variants of GUPPI tailored towards assessing price effects in an input market (rather than an output market), and towards the market involving auctions. Similar methods could be tailored towards assessing the Netflix-Warner Brothers merger’s effects on streaming market input providers.
There are notable reasons to be as concerned with this merger with respect to its effects on actors, directors, producers, and other input providers in the production of professional long-form streaming movies or series. Netflix and Warner Bros. Discovery both represent major buyers of high-budget films and series in a market with only a handful of meaningful competitors. For directors and producers (and the actors and other input providers they otherwise employ), a greater number of studios implies a greater number of buyers to which input providers can sell their content. Fewer studios available with which to negotiate would mean less competition, driving down compensation for input providers (as was the case in the book publisher merger).
In summary, the proposed Netflix-Warner Brothers mashup is an audacious deal that would be challenged under most administrations. That Netflix is even attempting a horizontal merger in such a concentrated market suggests that its management believes that the Trump administration will not be faithful to merger law or the 2023 Guidelines. As consumers of these services, we can only hope that Netflix has miscalculated.
Amazon Prime Day and Black Friday have become de facto national holidays of impulsive shopping, ostensibly offering a bevy of “great deals” to tempt consumers. A flood of media articles accompany this celebration of capitalism that masquerades as an event worthy of news coverage. Legions of outlets receive compensation from Amazon in exchange for driving traffic to its platform, often by offering “advice” on the best bargains to snag (e.g. CNN Underscored).
But, as some have already noticed, those advertised promotional discounts can be a mirage and often offer higher price than those regularly found throughout the year. Rather shockingly, even an article on the Washington Post, which Amazon founder Jeff Bezos owns, acknowledged this, with the author explaining that “I would have saved, on average, almost nothing during Amazon’s recent fall “Prime Big Deal Days”—and for some big-ticket purchases, I would have actually paid more.” Even so, Prime Day 2025 was Amazon’s biggest ever, with record sales and volume.
In the past few months, I took a deep dive into algorithmic pricing, the machine learning methods used, as well as various case studies, including how repricers work in conjunction with Amazon’s Buy Box to raise prices. Repricing algorithms on Amazon and their interaction with Buy Box rotation indicates that focusing exclusively on common pricing algorithms misses the risk of industry-wide algorithmic standardization. Even though sellers on Amazon can use various repricing providers and even though they may not share competitively sensitive information, the outcomes on Amazon still reflect supra-competitive prices, similar to those achieved from explicit coordination.
Price discrimination is just the tip of the iceberg
Recently, an investigation by the Institute for Local Self-Reliance (ILSR) found that school districts paid widely varying prices for the same products often on the same day, a practice known as price discrimination. While the report focused on schools specifically, businesses that purchase on Amazon should pay attention as well. Segmenting business purchases from those made by consumers is an example of third-order price discrimination.
Employees making purchases for their employer may be less price-sensitive than when making purchases for themselves, allowing Amazon to charge higher prices to the former. Those higher prices eventually get passed on to those businesses’ own consumers, adding to inflationary pressure. Those advocating for lower interest rates would do well to concern themselves with such pricing practices.
The ILSR report attributed the pricing variance it found to Amazon’s opaque dynamic pricing algorithm, the details of which I want to address here. Millions of sellers market their products on Amazon so, of course, one might think that the pricing discipline they (should) exert on each other would result in competitive prices for consumers. After all, as the Lending Tree commercial reminds us, “when banks compete, you win”. So, why aren’t consumers really winning? Why are prices going up?
In my new paper, I discuss the role of algorithms in pricing as well as the various machine learning tools used to implement them in various industries, including E-commerce, hospitality, airlines, real estate, and others. I also talk specifically about pricing on Amazon, which involves the relationship between rules-based and machine learning algorithms.
How repricers’ algorithms interact with Amazon’s Buy Box
On this topic, there’s another factor at play that has flown comparatively under the proverbial radar: the role of Amazon repricers’ respective algorithms and how their interaction with Amazon’s “Buy Box” algorithm enables successful price hikes. Repricing refers to the process of dynamically changing the offer price according to various guidelines. Many companies such as Repricer.com, BQool, Seller Snap, Flashpricer, and Amazon’s own Automate Pricing repricer offer this service.
The Buy Box, now called the “Featured Offer” on Amazon, refers to the box that appears on the right of an Amazon page prominently, displaying the price and shipping details for the seller who currently holds the Buy Box for this product. To see other competing options, a consumer needs to click on “Other Sellers on Amazon,” which appear on a pop-up page.
Not every seller is eligible for the Buy Box—Amazon imposes various eligibility criteria, including prioritizing its own delivery service, Fulfillment by Amazon (FBA) over Fulfillment by Merchant (FBM), conduct that prompted an investigation by the European Commission.
Needless to say, the vast majority of purchases, around 80 percent, occur though the Buy Box, which is why sellers want to secure it. Eligibility criteria also matter, because eligible sellers can choose not to compete at all with sellers of the same product, such as those that do not qualify for the Buy Box. So, that seemingly vast landscape of competitors just got smaller. And note that we’re talking about two different types of algorithms that work in concert: Amazon’s algorithm to determine the Buy Box winner and the repricing algorithms that sellers use to set prices.
Repricing on Amazon offers a particularly interesting case study of algorithmic pricing because (1) sellers can use different repricers, (2) repricers can use different algorithms, AI-based (i.e., machine learning), simple rules-based algorithms (e.g., if-then-else statements), or a combination of both, and (3) though sellers can obtain their own data and limited competitor information using the Amazon SP-API, (e.g., though the getcompetitivesummary call), no obvious sharing of competitively sensitive information occurs. In other words, the conditions that have garnered the most focus in algorithmic pricing cases such as the RealPage litigation and Gibson v. Cendyn, do not seem to apply here (at least with repricing).
And yet, these repricers openly advertise that their products seek to “avoid price wars” and “look to raise prices 24/7” after their client seller acquires the Buy Box, which one would expect they could only obtain if they were the lowest price offer and would immediately lose upon raising the price.
Not quite. As BQool advertises, its algorithm (in the third bin pictured below) “matches the Buy Box price then increases the price to capture greater profits.”
Hold on, you say. If you raise the price after you get the Buy Box, would you not lose it immediately to a lower priced seller? After all, that’s how competitive markets work—a given seller is a price taker not a price setter, and any attempt to raise the price would result in losing sales to rivals.
Not in this case. Repricers can adopt similar strategies (e.g., “avoid price wars” and “raise prices at every opportunity,” including setting similar floor prices, ignoring each other’s pricing, slowing reaction time (a time delay), and “match but do not undercut.” In other words, repricing algorithms can tacitly collude without any explicit coordination by mutually recognizing each other’s strategy, a process known as “cross-platform recognition.” For example, observing that a seller alters its price every 15 minutes would suggest the use of a repricer.
Simply put, the issue isn’t so much that sellers on Amazon use a common pricing algorithm (though many sellers use one or more repricers), but rather that they use repricers that adopt common strategies based on a common knowledge structure, without directly coordinating. This reflects industry-wide algorithmic standardization. If algorithms settle on a common standard, collusive outcomes can occur even if no obvious rim to an alleged “hub-and-spoke” conspiracy exists. A bevy of research, which I review and describe in my paper, has already observed the same collusive outcomes with Q-learning algorithms (a type of algorithm that falls under reinforcement learning).
Focusing solely on common algorithms can create a tunnel vision that misses other conditions in which algorithmic pricing can harm competition and raise prices. Building a modern-day Maginot Line against the use of common algorithms may accomplish little to defend competition if sellers can outflank it through industry-wide algorithmic standardization.
How tacit coordination occurs
But wait, you say, if Seller A holds the Buy Box, wouldn’t Seller B still have some incentive to undercut Seller A’s price to secure the Buy Box for itself? Otherwise, how does the Buy Box change hands?
This is where the coordination that Amazon’s Buy Box rotation algorithm effectuates comes into play. With rotation, Amazon gives the Buy Box to one seller for a period of time, say two hours, then rotates to another similar seller for the next two hours and so on. Rotation allows sellers (even at slightly different prices) to adopt a “wait my turn” strategy and share the Buy Box rather than aggressively competing for it. The seller holding the Buy Box knows when it can profitably raise the price incrementally without being undercut, and the other sellers that use repricing algorithms have little incentive to undercut the price because they will get their turn to sell at the same higher price.
Basic game theory can illustrate the payoff scenarios here. Suppose Seller A has the Buy Box and prices at $12.50. Without rotation, Seller B cannot simply match, it must undercut to get the Buy Box. This will prompt A to respond in turn by undercutting B, resulting in a price war (the bottom right box in the “Without Rotation” scenario). This is the sort of aggressive price competition that would benefit consumers.
With rotation added, however, the incentives change. Seller B knows that by matching A at $12.50, it will eventually get its share of time with the Buy Box. So B doesn’t undercut A, and the price stabilizes at $12.50 (upper left box in the “With Rotation” scenario”). So, both Sellers A and B have the incentive to explore upward, not downward pricing scenarios.
Repricers themselves say as much, advertising that they look to raise prices 24/7. Here’s Flashpricer saying exactly this.
And here’s Sellersnap echoing it.
And here’s a case study from marketing agency BellaVix, titled “Successfully raising price while retaining the Buy Box,” that discusses how “A premium skincare brand selling on Amazon faced challenges when attempting to raise the price of their best-selling Crepe Repair Cream from $59.99 to $79.99” (a 33 percent price increase!).
And here’s how BellaVix describe its strategy to overcome those challenges and successfully raise the price to $79.99.
Note that BellaVix first changed the list price, which provides an anchoring effect, not only for consumers but also for Amazon’s algorithm. This move also exploits information asymmetries between the seller and buyers, well discussed in the literature. Such asymmetries result in a market failure, where the price no longer reflects the true market value of the product. Then, BellaVix gradually raised the price, exactly the process that rotation enables and that repricers openly advertise.
In my paper, I discuss the concepts of information asymmetries and anchoring and the latter’s effects on accepting or rejecting price recommendations from algorithms. Perhaps rather shockingly, BellaVix successfully raised the price by 33 percent without losing the Buy Box entirely (though the Buy Box likely rotated), offering a practical example of the strategy I described above.
But, you say, a shopper can just switch to Walmart to avoid the repricers. Sorry, repricers such as Flashpricer, which advertises “AI-powered algorithms for every competition scenario and business model that look for opportunities to raise prices 24/7,” as well as Streetpricer, which “Checks if we still hold the BuyBox after each price increase – backtrack if necessary” are there as well. In fact, repricers are ubiquitous across various platforms, such as Airbnb, Vrbo (e.g., PriceLabs) and others, not just e-commerce.
Bad AI risks driving out the good
Nothing discussed above means that prices always rise. Nor do they need to do so for anticompetitive harm to occur. Remember, the benchmark isn’t whether prices go up or down in the absolute sense, but rather relative to the competitive benchmark. Prices may fall in the absolute sense if, for example, a new entrant unfamiliar with the tacit agreement qualifies for the Buy Box and begins exerting some pricing discipline. Moreover, just because competition occurs in some cases on Amazon does not offset the anticompetitive conduct or render it trivial. After all, not everyone gets lung cancer after smoking, but we still warn against its dangers.
Of course, firms employ algorithms for various beneficial uses, such as identifying fraudulent sellers or counterfeit products. As such, anticompetitive uses of pricing algorithms has another harmful effect—nefarious uses of artificial intelligence can drive out the beneficial ones, an outcome known as Gresham’s Law that occurs through adverse selection.
Much of the problem here results from information asymmetries, both between sellers and buyers and between regulators and algorithmic pricing providers and the platforms on which they occur. Many machine learning algorithms that power AI are “black boxes,” such as neural networks, ensemble models like boosting, random forests, and so on. Having a rudimentary understanding of such algorithms goes a long way toward protecting against anticompetitive consequences they might cause.
Moreover, as this article shows and my paper discusses in some length, harm to competition can occur even using simple rules-based pricing algorithms from independent providers and even in the absence of sharing competitively sensitive information. The old Latin saying caveat emptor (buyer beware) has perhaps never been more poignant than in this dawning age of algorithmic pricing.
Common pricing algorithms can be used to coordinate prices among sellers, to the detriment of buyers. RealPage is the seminal case, but there are (alas) plenty of others. The problem is particularly acute in a two-sided transactional platform setting, where the platform influences—and sometimes coerces—the pricing decisions of its sellers.
Take the case of Airbnb. Even before considering its “Smart Pricing” tool aka common pricing algorithm (discussed below), Airbnb inflicts tremendous costs on society. The short-term rental platform keeps rents artificially high by converting capacity for long-term (e.g., monthly or annual) rentals into short-term (e.g., daily) rentals. When the supply of long-term rentals artificially contracts, holding demand for apartments fixed, rents zoom upwards.
And high rents keep residents from spending money on other things—a drag on economic activity—and even contribute to homelessness for those who are priced out of the rental market entirely. According to a study by Harvard’s Joint Center for Housing Studies, a record 12.1 million Americans in 2024 were spending at least half of their incomes on rent and utilities, putting them at increased risk of eviction and homelessness.
Economists have studied the inflationary impact of Airbnb on rents. Calder-Wang (2021) found that the presence of Airbnb in New York leads to a transfer from renters to property owners of $200 million per year or $2.7 billion in net present value. Barron, Kung and Prospero (2017) found that a one percent increase in Airbnb listings leads to a 0.018 percent increase in rents; in aggregate, the growth in home-sharing through Airbnb contributes to about one-fifth of the average annual increase in U.S. rents. Seiler, Siebert, and Yang (2022) found that Irvine’s short-term rental ban reduced contract rental prices in the long-term rental market by 2.7 percent between 2018 and 2021. (Airbnb consultants point to evidence that Airbnb puts downward pressure on hotel prices for travelers, but absent some redistribution mechanism, that purported benefit to out-of-towners cannot offset the harms to local residents from higher rents.)
To bring down rents, Barcelona recently moved to end licenses for Airbnb homes, requiring owners by 2028 to offer them as long-term lodging at capped rents or put them up for sale. Closer to home, since September 2023, New York imposed a requirement that hosts must be present for stays under 30 days, and limited guests to two, reducing available listings on Airbnb. Similarly, in Santa Monica, the host must be present during the guest’s stay and unhosted rentals are banned, and Las Vegas bans non-owner-occupied short-term rentals. New Orleans banned Airbnb rentals in the French Quarter.
Airbnb’s “Smart Pricing” tool
Converting housing into short-term rentals is not, on its own, a cognizable violation of the antitrust laws, notwithstanding the clear price and output effects. What is cognizable, however, is price fixing, or the coordination of pricing strategies and output between horizontal rivals. Here the horizontal rivals are homesharers in the same geographic market. And while evidence of an illegal agreement to fix prices is typically challenging to detect—after all, minimally savvy competitors are unlikely to leave breadcrumbs that trace back to illegal behavior—Airbnb has at least invited homesharers on its platform to participate in one such conspiracy.
Here’s how: Airbnb offers homesharers on its platform a tool called “Smart Pricing,” which is an internal pricing algorithm that automatically updates homesharers’ listing prices. Hosts can opt in to Smart Pricing and set certain parameters, including minimum and maximum prices, then Smart Pricing does the rest by pinning prices to the “competitive” price. Of course, “competitive” is a misnomer to the extent prices are no longer a function of independent price setting among homesharers, but instead an automated prediction by the algorithm of what the market will bear.
The primary harm from Airbnb’s Smart Pricing tool is inflated rents that flow from a price-fixing conspiracy. Airbnb’s Smart Pricing tool also raises concerns regarding price discrimination, as it not only considers the features of the property and economic conditions, but also the characteristics of the guests themselves—for example, Airbnb acknowledges that Smart Pricing considers guest behavior in its algorithm. While maximizing the quality of the guest experience is a worthy goal, exploiting information asymmetries to extract supra-competitive prices can evince a market failure. Alas, the antitrust laws generally condone price discrimination. (For a great example of price discrimination achieved via algorithm aka “surveillance pricing,” check out Groundwork’s recent study of Instacart, another online platform that sets prices for sellers such as Target and Safeway.)
While Smart Pricing offers homesharers a convenient tool for managing their prices, Airbnb’s interests may not be aligned with the interests of homesharers. Some hosts have expressed frustration at the Smart Pricing algorithm for automatically pinning prices to the high end of their price range. Other hosts have complained that applying additional “rule sets” to their properties kicks them out of the Smart Pricing tool, creating friction for hosts who otherwise prefer to automate their prices. In other words, Airbnb’s Smart Pricing tool betrays a conflict of interest, and what’s good for the platform may not be, in the end, what’s good for individual homesharers. Insight into where precisely a listing will be ranked is limited because Airbnb’s recommendation algorithm is private. For example, the algorithm might punish non-adopters of the Smart Pricing tool by lowering their placing on the results page.
Airbnb incentivizes use of its Smart Pricing tool by telling homesharers that setting a “competitive” price helps improve a property’s ranking in search results. And the easiest way to set a “competitive” price without “constantly monitoring it”? Well, by using Airbnb’s Smart Pricing tool, of course. Want to automatically change your price in response to “travel trends” in your area? Turn on Smart Pricing.
Even when a homesharer declines Smart Pricing and elects to use its own pricing algorithm, doing so does not extinguish the concern of inflated prices. The economics literature recognizes how independent algorithms can learn to collude with each other by avoiding price wars. Moreover, the mere existence of a default option establishes a price floor around which all other prices are established.
If not dispositive of an illegal price fixing scheme, these facts provide at least circumstantial evidence that Airbnb is coercing homesharers into adoption of its price coordination tool, including by withholding access to consumers through search page rankings. If so, both Airbnb and participating homesharers may be on the hook.
An unwelcoming legal environment
Companies across a large swath of industries, from meat processing to hotels to real estate, are increasingly using common algorithms to set prices—and facing federal enforcement actions for doing so. Despite a defendant-friendly legal terrain, many of these arrangements have been challenged by either private or public enforcers. In large part, these cases focus on the exchange of competitively sensitive, often non-public, information between competitors, from which courts have begun to infer the existence of an illegal agreement. Self-styled “revenue management” or price- and rate-setting services like RealPage or Yardi in the rental housing industry, Cendyn in the hospitality industry, or Agri Stats in the poultry processing industry, have in recent years defended themselves against protracted litigation alleging their facilitation of these information exchanges. (Disclosure: I served as the economic expert for plaintiffs in two Agri Stats cases.)
That the DOJ recently settled its litigation with RealPage on decisively unfavorable terms suggests an unfriendly legal terrain. (Alternatively, it could reveal the subversion of law enforcement by a politicized agency.) And if there was any doubt about the steep evidentiary hurdles faced by plaintiffs, one needs only look at a strange and economics-free decision in the Ninth Circuit.
The Ninth Circuit’s decision in Gibson v. Cendyn reveals a basic misunderstanding of the economics of pricing. To dismiss any vertical relationship between Cendyn and its hotel clients, the Court claimed that “While hotels may use Cendyn’s revenue-management software to maximize profits, the software is not an input that goes into the production of hotel rooms for rentals.” (emphasis added) Yet the revenue-management software is precisely an input in the selling of hotel rooms, the output that forms the relevant product market (not producing or constructing hotel rooms from scratch). While hotels could technically function without it, the common pricing tool improves a hotel’s ability to extract additional surplus from their guests in the sale of the relevant product (again, not constructing hotel rooms).
The Cendyn decision also asks, “Why don’t the independent choices of Hotel Defendants to obtain pricing advice from the same company harm competition, as alleged here? Because here, obtaining information from the same source does not reduce the incentive to compete.” Yet the entire purpose of a common pricing algorithm is to reduce the incentive to compete unilaterally. If firm A knows that its rival, Firm B, is going to default to the joint profit-maximizing price as determined by the common pricing algorithm, it is in Firm A’s interest to mimic that price and not undercut it. In this sense, the algorithm facilitates a coordinated monopoly outcome that would not be as easily achievable in its absence. And for many common pricing algorithms, the clients are further incentivized to accept the recommended pricing for fear of being disappeared in search results.
The Cendyn decision also identified the sharing of competitively sensitive information as a key ingredient that enables collusion. This is also wrong as a matter of economics. Turning over one’s pricing authority to a common agent—whether a dude named Bob or an AI-based algorithm—increases the chances of reaching the monopoly price relative to a world in which companies make independent pricing decisions. This is true even when information about a rival’s costs or capacity is commonly known. Can firms in an oligopoly setting with complete information feel their way to the monopoly price in a repeated setting? Perhaps. But at least with independent pricing, there’s a chance that your rival will undercut your inflated price to gain share. And that threat tempers one’s enthusiasm to raise prices. Once rivals agree to turn over pricing to a common agent, however, that threat is extinguished. (This is not to say that sharing of confidential information isn’t a viable pathway to a finding of liability. It just shouldn’t be a necessary condition. In any event, homesharers are likely sharing confidential information with Airbnb, including the number of days for which the seller plans to occupy her home.)
That’s enough of the economics. For a nice explainer on the legal flaws in the panel’s decision, check out this brief by the American Antitrust Institute (AAI), urging the Ninth Circuit Court of Appeals to grant rehearing en banc. By insisting on a causal link between the licensing agreements and a restraint in the relevant market, AAI’s brief explains, the panel confused proof of an agreement with proof of the agreement’s anticompetitive effects. The brief also explains how the decision conflicts with Board of Trade of the City of Chicago v. United States, 246 U.S. 231 (1918), by creating a new category of agreements not subject to rule-of-reason analysis.
The case against Airbnb
Common pricing algorithms, like Airbnb’s Smart Pricing tool, can erode the fair functioning of markets when they deprive competitors of their independent decision-making authority. In a well-functioning competitive market, a series of (ideally, atomistic) suppliers would set their price independently. But when sellers can coordinate their prices, it is easier to move from a competitive output to something that approximates the monopoly outcome. Despite the propensity for market distortion, enforcing the antitrust laws against common pricing algorithms may prove challenging absent additional circumstantial evidence of an acceptance of that invitation to collude.
Airbnb’s facilitation of pricing decision among horizontal competitors should be assessed under the per se standard, which eliminates any consideration of efficiencies and obviates the need to establish market power. If assessed under the more burdensome rule-of-reason standard, plaintiffs would have to establish that Airbnb has market power, either directly, via evidence that it has the power to raise prices over competitive levels, or indirectly, via evidence that it commands a high share of a relevant product market. Empirical evidence that Airbnb’s smart pricing algorithm has led to higher short-term rents on the platform would suffice for direct proof. Regarding indirect proof of Airbnb’s power, per one estimate, Airbnb commands 43 percent of the U.S. market for online travel agents (aka short-term rentals), with Vrbo and Booking.com occupying significantly smaller shares. This estimate includes “direct bookings” in the relevant market, however, which arguably do not provide the same services as Airbnb and thus could plausibly be removed from the market, resulting in an even higher Airbnb share.
Airbnb isn’t just facilitating a data exchange; it is incentivizing or coercing homesharers on its platform to participate in a common pricing scheme. A coercion-based approach to enforcement should obviate the need to provide heightened evidence of acceptance, because participation in the scheme is a condition of a participation on the platform. Platforms wielding access to non-price business services, like advertising or market research services, on the condition that sellers accept price recommendations deprives sellers of their independent pricing authority. Airbnb’s “Smart Pricing” tool coerces their participation in a price-fixing cartel. With luck, the authorities are watching.
On November 18, the Federal Trade Commission (FTC) lost its landmark case against Meta over its acquisition of Instagram. The opinion was issued by Judge James Boasberg. The FTC spokesperson commented to the press decrying the loss as is usual agency practice. But the whole statement was far from usual. FTC spokesperson Joe Simonson told the press: “We are deeply disappointed in this decision. The deck was always stacked against us with Judge Boasberg, who is currently facing articles of impeachment. We are reviewing all our options.” This attack on Judge Boasberg aligns the FTC’s leadership with the partisan attacks on Judge Boasberg emanating from the Trump administration and its supporters.
I will not focus on the substance of the Meta case—Tim Wu has treated that subject well in the New York Times and on The Sling’s podcast. Instead, I would like to focus on how the FTC has chosen to describe Judge Boasberg and the destructive impacts that choice could have on the FTC.
To my knowledge, the FTC has never issued a statement like this impugning a federal judge’s neutrality. Doing so when there is no evidence of such a bias would always be reprehensible. But to make such a claim against Judge Boasberg is particularly counter-productive.
The FTC does not even attempt to suggest why Judge Boasberg would be biased against the agency in its statement. That is likely because the FTC’s leadership believes no such thing and is merely joining the right-wing criticism of Judge Boasberg for ruling against the administration. Because there is no substance to these allegations, I will concentrate on why attacking Judge Boasberg is counter-productive.
Judge Boasberg has served as a judge on the District Court for the District of Columbia, known as DDC for short, since 2011 and has served as Chief Judge of the district since 2023. He is one of the most highly respected district judges in the United States. Moreover, as chief judge, he is responsible for the administering DDC’s operations. He also represents his colleagues on DDC at the Judicial Conference, the policymaking body of the federal courts. In the past five years alone, he has been the assigned judge in four FTC antitrust cases according to Westlaw.
But more important than anything about Judge Boasberg specifically, DDC is the most important district court to the FTC. As the FTC’s home district court, the agency litigates in DDC more often than any other district. This year, the FTC has six cases in DDC. In the last five years, the agency has had 37 cases before the court. The other judges on this court are almost certainly paying attention to the insults the FTC chose to bestow on their colleague and chief judge.
This childish statement by the FTC therefore jeopardizes not only the agency’s credibility in front of a prominent district judge but also its credibility with the most important district court to the agency.
But it is not FTC leadership that will pay the price for this choice. The career attorneys who must litigate these cases will have their hard work put on the line because Chair Ferguson wanted to score political points with President Trump.
The FTC has a deep store of credibility with the bench. FTC attorneys are careful, professional, and deliberate in how they litigate cases. But credibility is easy to burn and hard to earn. And with each unprofessional, craven political stunt Chair Ferguson pulls with the FTC—from investigating Elon Musk’s political opponents to threatening Google over allegedly “partisan” email filtering to this attack on Judge Boasberg—Chair Ferguson burns the FTC’s credibility. Regardless of one’s views of Judge Boasberg’s Meta ruling, antimonopoly advocates should denounce this statement by the FTC.
Bryce Tuttle is a student at Stanford Law School. He previously worked in the office of FTC Commissioner Bedoya and in the Bureau of Competition.
Home affordability is a pressing issue. Young people often enter the workforce saddled with student debt, limited work options, and faced with exorbitant housing costs. For the millennial generation, the prospect of buying even a “starter home,” an option available to previous generations, has all but disappeared in most urban and suburban areas. The affordability problem touches more than just youth—blue collar workers, the foundation of this country without whose labor the economy would grind to a halt, fact made obvious during the early days of the COVID pandemic, no longer have the same opportunities either.
Efforts to address such an affordability crisis by exploring the viability of various options are a worthy endeavor, to be sure. But, as the intrepid detective of Hollywood fame, Jack Reacher, reminds us, “in an investigation, details matter.” The latest potential solution, which President Donald Trump proposed via social media on November 7, contemplates offering 50-year mortgages, up from 30 years—that is, increasing the time horizon for mortgages by two-thirds. Presumably, the logic motivating this argument posits that spreading out the mortgage over a longer horizon will lower monthly payments, increasing affordability.
Offering 50-year mortgages will do no such thing.
To repeat, the goal of making home ownership more affordable is laudable. But offering 50-year mortgages does not make ownership more affordable. At best, it makes mortgages more affordable. This policy preys on an unfortunate consumer tendency to buy a monthly payment rather than buying the property itself (a home or a car).
Most people who have shopped for a car will have likely received the question, “What sort of monthly payment would you like to have?” If you’re shopping for anything that requires a loan, and you get this question, stop right there.
You are not buying a monthly payment. This question conflates two products: the property (e.g., home or car) and the loan. The monthly payment just combines the two, masking the price you end up paying for the product you wanted in the first place. Sure, you might be able to get a lower recurring payment, but at what cost? Knowing the answer can inform the difference between a good and a potentially catastrophic financial decision.
The End of the Dream?
The American Dream, whatever vestiges of it remain, has always been about owning one’s home, not owning a monthly payment on a mortgage. Extending the payment over a longer period not only does not advance that goal, it actively undermines it; the longer the loan period, the smaller the proportion of each payment that goes to principal. This just means that you end up paying a higher price for the home. The longer payment schedule merely allows you to spread out the payment over a longer period in exchange. But that costs money, so let’s look at some numbers.
Suppose you want to buy a $500,000 home. Assuming you have the 20 percent to put down (congrats!), you will have a $400,000 mortgage. Your monthly mortgage payments will just depend on the interest rate you obtain and the term of the loan (usually 15 or more commonly, 30 years). Calculating the monthly payment, ignoring for the moment closing costs, home insurance, and other associated costs that one pays through escrow, is a simple calculation that one can do in Excel, either by hand or using the built-in payment (“PMT”) formula. The table below shows the calculations if you choose a 15- or 30-year mortgage. Note that a 15-year mortgage also commands a substantially lower rate (currently about 5.6%) compared to a 30-year mortgage (currently about 6.4%), because the 15-year mortgage carries lower risk.
Table 1: 15 vs. 30-year Mortgages
| Home Price | $500,000 | $500,000 |
| Loan Amount | $400,000 | $400,000 |
| Interest Rate | 5.6% | 6.4% |
| Loan Term (Years) | 15 | 30 |
| Monthly Payment | $3,290 | $2,502 |
| Total Amount Paid | $592,128 | $900,729 |
| Total Interest Paid | $192,128 | $500,729 |
With a 15-year mortgage, you face a higher monthly payment, but you end up paying less than half of the amount in interest. Assuming you pay off the 30-year mortgage at the end of the term, you will end up paying more in interest (about $500k) than the $400k you originally borrowed!
Now, let’s see what happens with a 50-year mortgage. The table below has the same two scenarios as above (Options 1 and 2, respectively), and two additional 50-year mortgage scenarios: (1) with the same interest as a 30-year mortgage, and (2) with the more likely increased mortgage of 7.3%. The increase in the rate associated with a 50-year mortgage would occur for the same reason that 30-year mortgages command higher rates than 15-year ones (higher risk).
Table 2: 15, 30 and 50-year Mortgages
Let’s look at Option 3, assuming the same 6.4% rate for 30- and 50-year mortgages. Yes, the monthly payment on the latter drops by $277 per month, from $2,502 to $2,225. Ok, you say, this is nice. But is it more affordable? Look at the cost of that lower monthly payment: over the 50-year length of the mortgage, you end up paying over $934k in interest to borrow $400k—that is, nearly a million dollars in interest alone!
But wait, it gets worse.
The mortgage rate will likely increase by raising the term from 30 to 50 years. Let’s say the 50-year rate is 7.3%. Then, not only are you paying the same monthly payment ($2,502, Option 2 vs. Option 4), but you’re also paying $1.1 million in interest alone with Option 4. You’re now paying more than double in interest while not even getting a lower mortgage payment. Guess who benefits here? (Hint, it’s not you.)
Adding Insult to Injury
Of course, you’re unlikely to carry the mortgage to term. This is where familiarity with the amortization chart helps. The average amount of time people hold a particular loan before moving or refinancing is 12 years. We can re-calculate the total amounts paid in interest and principal using that term. Keep in mind that earlier in the mortgage term, the majority of your payment usually goes to interest not principal. So, you’re not really gaining much equity.
Here’s where things go from worse to terrible.
Table 3: 15, 30 and 50-year Mortgages with 12-year Amortization
Look closely at how much of your principal you’ve paid off under these scenarios. With a 15-year mortgage, you’ve paid off just under 75% of your total mortgage of $400k (equal to $291k divided by $400k). With a 30-year term, you’ve paid off less than a quarter (equal to $79k divided by $400k). Now the terrible part: with a 50-year mortgage at the same rate as a 30-year mortgage, you’ve paid off less than $20k in principal in 12 years. In other words, even though you will have made a total of $320,371 in mortgage payments over that period, $300,631 of that will have gone to pay off interest and only $19,740 went to principal. You’ve gained less than $20k in equity from your payments. With the more realistic higher rate on a 50-year mortgage, you end up paying just $15k in principal. In other words, maybe, just maybe, you end up paying off the equivalent of your closing costs.
Ostensibly, this idea was intended to at least superficially appeal to the younger generation aiming to salvage some scraps of the dream of homeownership. Younger people tend to hold mortgages for shorter periods: as they build families they seek out more the requisite spaces. On that note, let’s look at the amortization schedule after holding the mortgage for five years.
Table 4: Five-Year Amortization Schedule
By taking a 50-year mortgage, you end up paying between $133k (Option 3) and $150k (Option 4), but only between $6,447 and $4,722 goes to principal! Over 95% of your payment has gone to interest. Again, you’ll walk away with next to nothing after the sale.
Let me emphasize. As Option 4 shows above, after paying your 50-year mortgage for a full five years, you will have managed to pay down the principal on our loan by less than five thousand dollars, even though you paid a total of over $150k towards your mortgage. You’re not really becoming a homeowner, you’re financing someone else’s investment vehicle.
When you buy a home, you generally hope to gain equity. This comes (1) from the portion of your mortgage payment that goes to principal vs. interest and (2) market factors. As the amortization schedules above show, the longer the term of your mortgage, the lower the percentage of your payment that applies to principal. That means you’re left open to the vagaries of the property market. You can actively improve the home in the hope to achieve a positive return on investment, but money that you could have used to do so now goes to paying down interest.
This might be the time you expect to hear that there’s a silver lining somewhere. No. Just more clouds.
Remember, because of the transaction costs involved, once you buy a property, you’re very likely immediately underwater. You’ve likely paid some closing costs to get into the mortgage. And if you changed your mind and wanted to sell the property immediately and hire an agent, you will end up paying around 3-6% of the total sales proceeds. Assuming you could sell the home for exactly the price you paid, $500k, that means you’re $15k-$30k underwater, plus the title costs.
Let’s look again at Option 4 in Table 3. After 12 years, you’ve paid $345k in interest but only paid off $15k in principal, just barely enough to cover a 3 percent agent fee. Not only that, the $100k you’ve put as down payment has done nothing over this time, unless the home prices have increased. The 50-year mortgage option just creates more mortgage-backed investment vehicles, not more homeowners.
Better Options Than 50-Year Mortgages Exist
I’ve skipped over some additional details, all of which tend to make this calculation even worse for potential homeowners. Of course, there are downstream effects as well. Locking people down in debt restricts their mobility, with second-order effects on labor markets. Suffice it to say, even if this idea were well-intentioned, it should be dismissed immediately. I would expect even the Abundance crowd to concur, as 50-year mortgages would counteract attempts to increase home supply. Yes, maybe we’d see more mortgages, but not more homeowners. And isn’t greater home ownership what the American Dream was supposedly about?
So what’s the solution? Lowering rates would help. Most homeowners with mortgages have rates below 5% and are unwilling to forego them without compensation. I hear the common argument that “well interest rates were much higher in the 1980s.” This misses the point. The issue isn’t as much with the size of the interest rate as with the rapid change in rates from less than 3% to north of 6%. People have locked in at lower rates, and now they’re wearing “golden handcuffs”. For example, financing a $400k loan at 3% means a $1,686 monthly payment. Increasing the rate to 6.25% raises the monthly payment by nearly $800 to $2,463. But not only have interest rates risen, so have home prices. So, instead of a $400k loan, maybe you now have to take out a $500k loan to buy the same or similar house. That means a $3,079 monthly payment. So, you can see that the same house now requires precariously close to double the monthly payment in this case. Even if people want to move, they stay put. This means a lot of shadow inventory is sitting out there, and lowering interest rates would go a long way to inducing it to come on the market.
When it is funded, the Supplemental Nutrition Assistance Program or SNAP (colloquially known as food stamps) helps feed over 40 million people every month by dispensing $187 per average recipient. The program is especially critical for families—children represent about 40 percent of all SNAP recipients. Despite 78 percent of Americans (and 69 percent of Republicans) holding a positive view of SNAP, one of the many detrimental consequences of the longest government shutdown in U.S. history is that the program’s funding is about to dry up.
In anticipation of extraordinary conditions, such as a long-lasting shutdown, Congress had set aside a small “contingency reserve” of funds to be used “at such times as may become necessary to carry out [SNAP] operations[.]” Breaking from the USDA’s September 30 interpretation of this language, on October 24, the Trump Administration claimed it could not legally tap these funds during a shutdown—threatening to disburse no benefits in November. This novel understanding of the law (a continuation of the Administration’s hostility to a program from which it has already cut $187 billion over the next decade) was quickly rejected by the courts in a pair of Halloween rulings mandating that the Administration use these funds. In the midst of this court battle, the president waffled between the position that “If we are given the appropriate legal direction by the Court, it will BE MY HONOR to provide the funding” (October 31 post), and that SNAP “will be given only when those Radical Left Democrats open up the government, which they can easily do, and not before!” (November 4 post).
After the president’s hemming and hawing, the Administration set forth a plan to offer partial payments to SNAP recipients for November. Originally, payments were anticipated to be half of full benefits, but because math is hard for merit-based hires, this number has since inexplicably increased to 65 percent of full benefits. U.S. District Judge John McConnell has deemed these partial payment plans insufficient. SNAP advocates argue that the partial payments would only create bureaucratic bottlenecks that would slow access to these critical funds. Of course, the Administration appealed this “absurd” ruling, which would ensure the wealthiest country in the world allows its poorest denizens to afford food. For now, the Supreme Court has paused Judge McConnell’s full payment order as it awaits a decision from the First Circuit Court of Appeals. This pause occurred only after at least some states disbursed full SNAP payments to recipients, and, amid a deluge of conflicting orders, the USDA seems to want states to magically claw back these full benefit down to its previous 65 percent partial payment plan.
Treating SNAP and Non-SNAP Recipients “Equally”
Amid plans to provide only partial SNAP benefits, many grocers were intending to offer discounts to SNAP recipients to ensure their quantity of food consumed was minimally interrupted (and to avoid missing out on loss sales). For example, the McMinnville Grocery Outlets of Oregon tried to offer a 10 percent discount to all SNAP recipients whose food-assistance money had been frozen. Despite its court loss over halting payments completely, the Administration continues to selectively enforce policies to ensure some discomfort for those in need, including its own supporters. Hence, the USDA has warned grocers that these discounts, absent a waiver, run afoul of a rule requiring SNAP recipients to be treated “equally” compared to non-SNAP recipients. In response to that warning, many grocers ended their proposed discounts, though others retailers, like Instacart, continued to offer their planned discounts to SNAP recipients thanks to already having the needed USDA waiver.
Unfortunately, SNAP’s Equal Protection Rule (7 CFR 278.2(b)) does forbid “special treatment in any way” for SNAP recipients. The Biden Administration-era USDA also interpreted this to “prohibit[] both negative treatment (such as discriminatory practices) as well as preferential treatment (such as incentive programs).” The waivers that allow a small number of retailers (mostly farmers market) to offer discounts was codified in the 2018 Farm Bill (7 USC 2018(j)) to promote the consumption of healthier food. Ignoring the fact that such non-discrimination rules were premised on full-funding and delivery of SNAP benefits, the Administration can plausibly argue that its hands are tied.
As we have seen so frequently with this Administration, however, the Executive Branch has significant latitude in whether and how it will actually enforce statues and regulations. For instance, this decision stands in direct contrast to this same Administration’s abandonment of the antitrust case against PepsiCo under the Robinson-Patman Act; apparently, a supplier’s selective use of discounts to large stores does not warrant similar condemnation to grocers offering discounts to those in need. We would be remiss not to mention that PepsiCo donated half a million dollars to Trump’s inauguration ceremony, after a brief interlude of pausing political donations in the wake of the January 6, 2021 storming of the Capitol.
If anything is justifiably worth similarly lax enforcement, it would be the allowance of private discounts after SNAP recipients have faced an unexpected decline in governmental assistance. Hence, in our estimation, forbidding price discrimination in the name of equality—when such discrimination would assist impoverished people in a time of need—suggests that serving the poor does not rank too high in the Administration’s policy preferences. That the Administration has ignored price discrimination in other contexts, such as by airlines (when your trip is for business) or Uber (when your battery is low), reveals that Trump’s cruelty to the poor is likely the point.
Democrats have responded to the Administration’s enforcement of SNAP’s Equal Protection Rule with a bill—which does not appear to have Republican support—that would allow private discounts by grocers for SNAP recipients during government shutdowns. We think that Congress should push the boundary further, by amending the Equal Protection Rule to always allow discounts for SNAP recipients. Equality, while often laudable, is overrated here, as disparate conditions necessitate differential treatment.
To Limit Hunger, Allow Price Discrimination
So, what is price discrimination? And why should it be used to support SNAP recipients? In its simplest form, price discrimination is the ability for sellers to charge different prices to different buyers. Think of a discounted museum ticket to students or movie theaters offering cheaper matinees. Such common and innocuous pricing strategies might conflict with the layperson’s negative associations with the term discrimination; for instance, Merriam-Webster includes the word “unfairly” in its definition. Yet economists often use a definition that strips away this negative baggage. Under this conception of the word, SNAP discriminates on the basis of income—those below a certain income thresholds receive benefits and those above are deprived benefits.
Economists tend to support price discrimination because it can increase consumption and social welfare under certain conditions. Price discrimination necessarily benefits companies (otherwise they would never adopt it as a pricing strategy!) but also may benefit certain consumers, like SNAP recipients, with lower willingness to pay for (or lower ability to afford) goods. Under certain forms of price discrimination, these consumers can increased their quantity consumed thereby improving their welfare. Whether these benefits outweigh the potential increase in price to consumers with higher willingness to pay depends on the exact nature of the market and the form of price discrimination. For instance, one of us has a prior piece in The Sling explaining how perfect price discrimination (by airlines) in the face of quantity restrictions is undesirable.
Price discrimination is in theory infeasible in competitive markets because sellers in such markets are price takers (facing a flat demand curve) with no impact on market prices. In the real world, however, markets are rarely perfectly competitive, and firms often have a degree of market power (facing a downward-sloping demand curve) and can influence market price. For instance, the USDA has found that the market concentration (a measure related to the potential for market power) of grocers has increased over the last 30 years and is particularly problematic in rural communities. Indeed, the mere proposition of these discounts by certain grocers is suggestive that these sellers are already operating above marginal cost and possess some degree of market power. Regardless, if price discrimination is infeasible in the long term due to competitive forces, then a policy that allows for price discrimination would have no long term effect—SNAP and non-SNAP prices would inevitably converge.
If price discrimination is possible, allowing grocers to concurrently drop prices to SNAP recipients and raise prices to non-SNAP recipients, then there are difficult tradeoffs to consider: the welfare harm of higher prices for non-SNAP recipients may be greater than the welfare gained by SNAP recipients and grocers. (In the real world, however, there is a distinct possibility that grocers simply drop the price to SNAP recipients and leave the price to non-SNAP recipients the same, which greatly simplifies the welfare calculus. There is no lump-of-profit law in economics deeming that losses incurred on one group must be recouped on some other.)
The overall societal welfare effect of price discrimination by groups (or what economists call “third-degree” price discrimination) is not straightforward, but social welfare tends to increase when output increases. This can happen when price discrimination creates new markets that were otherwise unprofitable (e.g., in the long run, price discrimination may mean that certain food deserts will disappear because a grocer can enter the market). Output can also increase if the price cuts to SNAP recipients allow some SNAP recipients to obtain needed food when they otherwise could not.
Beyond the potential welfare effects from greater output noted above, there are several other reasons to believe that allowing discounts for SNAP recipients will be beneficial. First, because the cost of SNAP is likely borne mostly by non-SNAP recipients (through taxes on higher income individuals), the decrease prices paid by SNAP recipients may allow for SNAP itself to need less money to cover the dietary needs of beneficiaries; in effect, allowing non-SNAP recipients to have a slightly lower tax bill. Second, and more significantly, there are several externalities at play; to name one, hungry people are less productive workers, and hungry children perform worse in school and jeopardize their long-term productivity. One study found that for every $1 spent on food stamps for a child, $62 of value were generated over that child’s lifetime. Third, there are also non-economic (equity) reasons to support such discounts: Non-SNAP recipients may be willing to face higher costs so their neighbors can afford to eat. In other words, society cares more about lifting up the downtrodden than the relatively minor costs borne by the better off.
Turn the Spigot and Let the Discounts Flow?
Because the Trump Administration has already indicated its opposition to the American welfare system, its open hostility to SNAP does not come as a shock. Nevertheless, while its true motives are hard to decipher, the Administration’s stance against allowing private discounts for SNAP could be defended by certain economists (not us) by appealing to concerns about the risk of inflation for non-SNAP recipients. Such concern would be unwarranted, however, if grocers are offering private discounts merely to mitigate short run losses in the face of a government-induced demand shock. In such situation, there is a low risk of the concomitant increase in the price paid by non-SNAP recipients ever materializing.
Regardless of these arguments, it is our contention that the economics suggests that allowing private discounts for SNAP recipients is simply good policy. The enormous lift to each SNAP recipient (getting to eat) swamps the likely trivial costs to each non-SNAP recipient, who are in a better position to absorb such costs. As such, we would urge Congress to explore allowing grocers to offer SNAP recipients private discounts not just during shutdowns but also in times of continuing resolutions and the oh-so-rare actual fiscal year budgets.
Advocates of all stripes will pounce on a Nobel prize in economics to promote their particular policy agenda. They find a strand of the work by the winning economist, or a snippet from the Nobel committee, spin it into their narrative, and voila, their pet theory is proven right. Even a Nobel prize winner says so!
We should take such claims with a grain of salt.
I confess that the initial coverage of this year’s economics prize, awarded on Monday to Joel Mokyr, Philippe Aghion and Peter Howitt for their work on the drivers of innovation, left me a little empty, as I couldn’t see any crisp policy implications. At least not at first.
But then Aghion was quoted in the New York Times, noting that he was optimistic about the prospects of artificial intelligence (AI), yet still “favored policies that promoted competition and did not consolidate power to just a few winners.” In a longer version of the story, he said “there needed to be competition policy so that AI innovators would not stifle rivals.”
Sometimes we don’t need to decipher the hidden meaning of a complex economic article written decades ago, or even the Nobel committee’s summary of an economist’s contributions. Sometimes we can just defer to what the economist actually says. Aghion is literally telling us that we’d be better off as a society if AI were competitively supplied, compared to a scenario in which AI is provided by a small handful of firms with near-monopoly power.
To take one possible scenario, imagine a “hypothetical” monopolist in the general search market with its own proprietary AI tool. It’s true that independent AI tools might shake up the general search market, assuming each AI tool competed on a level playing field. But if said search monopolist can preference its own AI tool to answer its users’ general search queries, then its search monopoly could be artificially extended.
Indeed, Aghion’s and Howitt’s own research suggests that monopolists can be stifling for innovation. The figure below is reproduced from their 2005 Quarterly Journal of Economics article titled “Competition and Innovation: An Inverted-U Relationship,”in which they studied the relationship between market power and innovation across UK industries.
The Y-axis plots the citation-weighted patents in an industry, a measure of innovation. The X-axis plots one minus the Lerner index or price-cost margin, a measure of competition in an industry. A value of one (far right along the X-axis) indicates perfect competition, as price equals marginal cost. Values below one indicate some degree of market power. As the figure shows, when industries become monopolized—one minus the Lerner index approaches zero—the amount of innovation decreases from its peak. In the words of the Nobel winners, “competition may increase the incremental profits from innovating, and thereby encourage R&D investments aimed at ‘escaping competition.’” The figure also indicates too much competition might be harmful for innovation as well.
The Pro-Monopolists Claim Victory
Despite this clear signal from one of the winners, the anti-anti-monopoly crowd, or more affectionately, the “pro-monopolists” as I like to call them, issued a press release trumpeting the exact opposite message—that this year’s Nobel prize stands for the proposition that antitrust enforcement is based on failed predicates and has swung too radically towards the interventionist side.
In a post titled “What Competition Scholars Should Know About the 2025 Economics Nobel,” ICLE’s chief economist Brian Albrecht argues that the work of Aghion and Howitt should cause a fundamental rethink on how we perform antitrust analysis:
The standard antitrust framework inherited from 1960s industrial organization focuses on market structure. Count the firms. Measure concentration. Assume that structure determines conduct, which determines performance. More firms mean more competition, which means lower prices and better outcomes. The U.S. Merger Guidelines embody this view with their use of Herfindahl–Hirschman index (HHI) thresholds.
The Aghion-Howitt framework tells a different story. Competition is a process of innovation and displacement. Firms compete by trying to make better products, not just by cutting prices on existing ones. What matters is not the number of firms at any point in time, but whether new innovators can challenge incumbents. Market structure is an outcome of this competitive process, not just a cause of competitive behavior.
Notwithstanding Aghion’s own words calling for vigorous competition policy, Albrecht’s spin is a mischaracterization of how antitrust works in practice. Antitrust decisions don’t turn on simplistic concentration metrics in a market. In a single-firm monopolization case, concentration metrics are rarely used; when proving a defendant’s market power indirectly, what matters is the defendant’s share of the relevant market, along with evidence of of significant entry barriers. Alternatively, market power can be shown directly with evidence that the defendant raised prices over competitive levels or excluded rivals. But market power by itself doesn’t constitute a violation of antitrust law, as Albrecht intimates; plaintiffs must criticallyshow anticompetitive effects.
Even in merger cases, where concentration metrics play a role in establishing a presumption of anticompetitive effects, plaintiffs still must show anticompetitive effects owing to the merger,using entirely different models (such as upward-pricing pressure models). And merging parties can overturn the presumption with evidence of procompetitive effects. If Albrecht is saying that antitrust courts merely look at market structure to decide cases, he is attacking a straw-man.
Other Policy Implications
Let’s get back to reality. Here’s how the New York Times described the contribution of Aghion and Howitt, as summarized by the Nobel committee:
Mr. Aghion and Mr. Howitt shared the other half of the award for what the committee described as “the theory of sustained growth through creative destruction.” They built a mathematical model for growth, with creative destruction as a core element.
The committee described creative destruction as “an endless process in which new and better products replace the old.” They used the example of the telephone, in which each new version made the previous one obsolete, from the rotary dial phone in the early 1900s to today’s smartphones.
Mr. Aghion and Mr. Howitt’s work shows how economic growth can continue despite companies being sidelined by the innovation of other firms. Their work can support policymakers in creating research and development policies, the committee said.
The laureates’ work shows “we should not take progress for granted,” Kerstin Enflo, a member of the Nobel committee, said during a news conference in Stockholm. “Instead, society must keep an eye on the factors that generate and sustain economic growth,” she added. “These are science-based innovation, creative destruction and a society open for change.”
A narrow reading of this passage suggests that the policy implications of Aghion’s and Howitt’s work should be limited to “research and development policies.” Even under this lens, proponents of policies like subsidizing medical R&D, green energy technologies, or broadband infrastructure could claim a new quiver in their bow. To take broadband investment as an example, high speed internet connectivity generates spillovers across the economy and promotes growth by enablingapplications in myriad connected industries, as demonstrated here and here.
When announcing their award, the Nobel Prize committee noted that “Policy should support the innovation process,” and that “science-based innovation” was among the factors that “generate and sustain economic growth.” Upon receiving the award, Howitt reportedly said that the biggest advances in technological progress “have involved cooperation between governments, universities and businesses.” This is confirmed by countless examples of public investment pushing technology forward. One notable recent example was Operation Warp Speed, a public investment into Covid-19 vaccine development and production. Another public investment, the BEAD Program, allocated billions to states to expand high-speed Internet access.
Would this year’s Nobel prize winners embrace other interventions in the economy? Say, to permit publishers to collectively bargain against Internet behemoths to ensure that content does not vanish? Perhaps, given their apparent fondness for competition and government programs. But rather than reading the tea leaves, we could just ask them. Or we can recognize that there are limits to what a Nobel prize can say about our pet policy ideas.
The proposed merger of the Union Pacific (UP) and Norfolk Southern (NS) railroads would consolidate ownership and control of a significant part of the central arteries or “trunk lines” of the rail network in this country. Currently, four railroads control most of these key components of the rail network, especially for the east-west service, and each has a clear regional dominance. The other two major trunk line railroads are the Burlington Northern Santa Fe (BNSF) and CSX. A number of “branch lines” extend out from the trunk lines, and they are often operated by short line railroads. Although there are 600 short lines, a number of them have common ownership.
Proponents of the UP-NS merger argue that it would allow the combined railroad to operate trains from the west coast to the east coast without having to switch the cars between rail lines at some transfer point. It is speculated that if the UP-NS merger is allowed, the BNSF and the CSX would then seek to combine, creating even more concentrated control over trunk lines. The current position of the BNSF and CSX is that, even without a merger, they will jointly offer through service from west coast ports to various locations in the eastern half of the country. This service would use the same train for the entire trip with crews changing as the train moves from one system to the other. Notably, the routes that they propose for this combined service will directly overlap and compete with the UP-NS lines. All the major trunk lines already share track usage in some places with each other or with short lines. Furthermore, Amtrak, which only owns a few lines in the Northeast, has operated trains on the trunk lines of all major railroads for its nationwide services since its creation in 1971. In addition, many of the freight railcars in use belong to third parties that use them for their own goods or lease them to others for use. All of this shows that ownership of trunk lines is not essential to the operation of freight or passenger service on those lines.
Increased consolidation of ownership of trunk lines might induce their rationalization and enhancement to facilitate more efficient movement of trains. But this would come at the cost of increased dominance of these essential transportation facilities. Moreover, much of the rationalization and improvement is rational conduct by each major trunk line without any merger. Hence, the proposed UP-NS merger should be forbidden as long as the resulting company both owns the tracks and controls their use.
Beware the Bottlenecks
Trunk lines can be understood as “bottlenecks” through which the bulk of rail freight must pass. Ownership of trunk lines confers the ability to control the use of the capacity—namely, whose trains at what price will operate on those lines. Because the owner also controls the dispatch of trains on the line, it has substantial capacity to affect the quality of service. For example, Amtrak, which enjoys a right to use trunk lines for its services, has a long history of problems of having its trains delayed so that the trunk line owner can send the latter’s freight trains ahead. Where freight lines share track use rights, similar disputes are not uncommon between the track owner and the other railroad sharing that track.
What is concentrated is the control over the track itself. Major interstate highways are a single system with multiple users. The same is true of the inland waterways. Should ownership of railroad trunk lines be separated from operation of freight service by multiple users? Amtrak’s and freight lines’ shared use of some trunk lines suggest that operating freight and passenger trains do not require owning the trunk lines on which those services operate. Moreover, while the capacity of any rail line is ultimately limited, with proper scheduling and dispatch these lines can accommodate substantial increased use. It follows that if ownership of the trunk lines were separated from the operation of freight service, it is probable that the number of competitors in providing such service would increase. Freight rates would likely decline, and assuming proper coordination of use, service itself would increase in efficiency. In particular, services such as the movement of container shipments would likely increase significantly, as trucking companies as well as new entrants would be able to develop a range of through services. The same might be true for passenger service, but given low total demand, this seems unlikely except in a few potentially high-volume routes
The collaboration of BNSF and CSX to provide a through freight service from coast to coast absent any merger demonstrates the feasibility of such service. Not surprisingly, that proposed service will focus on competing directly with the potential (merged) UP-NS services. This highlights the importance of competition in stimulating innovation in transportation service. Notably, the other ports on the west coast that BNSF serves, but which do not have UP-NS alternatives, are not in the through services package.
As noted, there are significant problems with the coordination in the use of trunk lines. There are ways to coordinate use to facilitate more efficient shared use of track. The railroad controlling the line, however, has limited incentive to resolve those problems because it has a conflict of interest. Its rational goal is to maximize its own use rather than facilitate the use by others whether they provide a distinct service like Amtrak or are competitors in freight service as in the case of shared track rights.
Experience with Separating Ownership from Operation
Some countries have experimented with separating ownership of the tracks and licenses for operators to use those tracks to stimulate competition in rail services. A leading example is the United Kingdom. In the 1990s, the publicly owned national rail system was terminated. Regional passenger services devolved into private ownership but retained a localized monopoly. In addition to the regional services, long-distance passenger services emerged that provided some competition. Freight service was similarly privatized but in a manner that encouraged competition. A distinct but private authority maintained the tracks and provided coordination of services.
By 2024, parts of that experiment failed. The government had retaken the responsibility for ownership and coordination of the rail system itself because of concerns about maintenance of the rail lines. It has similarly retaken ownership and operation of some regional passenger services that had persistent problems in both finances and operations. In 2024, the new Labor government determined that as the remaining contracts for the regional service expire, their passenger service will be absorbed into a state-owned entity that will provide service in the England and the parts of Wales and Scotland where those regions had not already recaptured public ownership of local passenger service.
On the other hand, the independent freight services and long-distance passenger services will continue to operate. The distinction is that the regional services were local monopolies of rail passenger service, while both the freight services and the long-distance passenger lines operated in a competitive environment. There are five major freight services competing in the UK market. The American short line holding company, Genesee & Wyoming, owns one of the leaders, Freightliner, which is particularly strong in handling containers, including offering its own trucking service to deliver the container.
Remedy Design for the STB
If the UP-NS merger is to be allowed, there should be separation between ownership of the tracks and operation of the freight services on those tracks. Because the law exempts these mergers from standard antitrust review, the decision will come exclusively from the Surface Transportation Board (STB), whose authority to regulate the operation of the rail system extends beyond standard antitrust criteria to encompass a broader concern for the overall public interest in the rail system. If it adopts this strategy, the STB should also require the BNSF and CSX to make a similar separation of rail lines from their operation of freight services. Indeed, this separation should also apply to the other two Class I railroads—the Canadian Pacific Kansas City and Canadian National. Doing so would immediately create six freight service operators even if the STB did not require subdivision of the operating companies. Of course, new entry would have to be authorized as well. The Genesee and Wyoming Railroad is among the most obvious entrants given its UK experience and expertise. The STB would have to develop standards for entry and define the rights of such firms. It might be necessary to have criteria for use of trunk lines such as those into the Los Angeles area, which might have capacity limits.
A central question would be the form of ownership and management of the trunk system. Private ownership would risk exploitation, as such an owner would have monopoly power over the use of its track. The STB would have to impose rate regulation to avoid overcharges. Perversely, the track owners will then have the incentive to make overly costly investments if rate regulation is based on the assets devoted to the business. This is the history of regulated public utilities where rates are based on the capitalized value of the investment in the facilities. If rates are not based on capital investment, however, there would be an incentive to underinvest in the maintenance and improvement of the rail lines. A related risk is the control over dispatch along lines. An owner might have incentives to favor some users over others or use this control to demand additional compensation for priority.
It is very hard to imagine that the American government would take ownership of the rail system even though it “owns” and maintains both the highway and inland waterway systems. Not only the UK, but many other countries have public ownership of their rail systems. Early in the development of America’s railroad system, states were often the owners of the lines that operating companies used and North Carolina still owns a major trunk line in that state. Indeed, Amtrak is a government entity reflecting the failure of private companies to maintain a successful national rail passenger system. But Amtrak’s problems with getting sufficient support to develop and operate its system cautions against government ownership. Indeed, the deferred maintenance for both the national highway system, especially bridges, and the failure to make timely investments in the inland waterways counsel against public ownership. All are dependent on Congressional appropriations.
The source of these problems, whether privately or publicly owned, is the conflict of economic interest between rail users and the track owner. Darren Bush and I have suggested that one solution to dealing with the operation of bottlenecks is to have the ownership rest in either a cooperative owned by the users or by creating a condominium type of ownership in which owners of specific use rights share ownership but not management of the overall entity. Both strategies assume that there are a significant number of users whose primary interest is in using the bottleneck as a means to provide some valuable good or service. Hence, the goal of these owners is to maximize the utility of the bottleneck for all its users. This means providing as much access as is technologically feasible, ensuring proper investment in maintenance, and providing the kinds of coordination that would again maximize the use of the resources involved. Our article showed that such strategies have in fact been used successfully in a variety of contexts—from cooperative grain elevators at a time when farmers faced very local, monopsonistic markets for their grain to electric transmission that facilitated the opening of competitive markets for the sale of electricity to local utilities.
In the case of railroads, given the goal is to have actual and potential competition for freight service between all major markets, the logical option for the trunk lines is that of a cooperative in which the users would collectively own and manage that essential element of the business. Given a large number of participants, a necessary element to ensure that the incentives to exploit or exclude are minimized, the day-to-day operations would have to be run by a board of directors and managers. The key here is that the incentives of management would be the efficient and open operation of the trunk system.
Remaining Challenges Are Big But Solvable
There are many other hurdles that would have to be overcome to make this transformation possible. During a substantial transition period, the STB would have to play a strong role in policing conduct and guiding the development of the new system. The lessons of path dependency tell us that there are almost certain to be major challenges from unexpected problems with such a transition, but there can also be unexpected positives. Among the issues that would need resolution are the ownership and relationship of the short lines and branch lines owned by the major truck lines to the system. While having all rail lines owned by one or more cooperatives is a possibility, given the limited and often focused use made of branch lines, creating a distinct ownership model for such lines might be a better strategy. A similar issue would arise with respect to the switching yards in which trains are made up, and cars are moved around either for unloading or to be dispatched to a new location. Who would own and operate those entities and how would they relate to the operating companies? The experience of the UK system should be instructive.
Yet another challenging issue is the development and operation of a dispatch system that would coordinate not only the freight services but also Amtrack and any other passenger service that might immerge. Interchange among carriers would require development of protocols. This is likely to be especially challenging for low volume freight service that smaller users require. With the technology from AI and other computer systems such as those currently coordinating air traffic, these problems are solvable but nonetheless challenging.
Finally, there is the major question of pricing. The cooperative presumably would have purchased the rail trackage from its prior owner. The price has to be set in some equitable way and the resulting access charges appropriately allocated among uses. Here there would be real questions about differentiation of prices for through service for trains carrying containers, passenger service, as well as unit trains carrying bulk commodities such as soybeans, coal, corn, or oil.
As the foregoing review suggests a great leap into a cooperative trunk system seems very risky. Perhaps in addition to blocking the UP-NS merger, the STB could consider requiring access to existing trunk lines for other users given the CSX-BNSF and Amtrak examples. As Neil Averitt has suggested, such mandatory access could be limited to specific routes or kinds of freight such as containers where the Genesee & Wyoming as well as national trucking companies are potential entrants. For the reasons discussed earlier, the conflicts of interest between the railroads and such users would be a problem. Still, experimenting with opening some types of rail service to third party participation would provide a way to test the viability of this concept in the American context, as well as a useful experience about the challenges that would emerge.
Peter Carstensen is Emeritus Professor at the University of Wisconsin Law School.