Competitive Rent Extraction
Competitive rent extraction is Nathan’s name (2014-10-17) for a market failure that standard antitrust has no category for: multiple firms, ostensibly in competition, simultaneously extracting rents from the same captive populations. The observation predates the current algorithmic-pricing literature by eight years, and the 2022 MacKay & Weinstein analysis supplies its mechanism.1
The journal observation (2014-10-17)
“Somehow it has become possible for multiple companies that are ostensibly in competition with each other to extract rents from the same populations. Google and Facebook compete with each other — but their users are captive audiences, ripe for rent extraction… The competition is real, but it’s competition for the captives, not competition that frees them.”
The insight: classical monopoly rents require one seller; platform-era rents require only that users be locked in somewhere — the firms can fight over who holds the key while the rent flows regardless. Competition for the captive, not competition that frees them. A 2023-03-24 entry completes the thought: intelligence itself becomes “something that can be just imbued into a tool as needed” — the platform imbues intelligence into its pricing and targeting, and the captive audience is monetized at machine speed.2
The mechanism (MacKay & Weinstein 2022)
The HBS working paper “Dynamic Pricing Algorithms, Consumer Harm, and Regulatory Response” demonstrates that the 2014 intuition is provable in equilibrium — and worse: algorithmic pricing raises prices even with zero collusion.3
The models (Brown & MacKay) show three paths, all ending above the competitive (Bertrand) price:
- Asymmetric frequency. One firm’s algorithm reprices many times a day; the rival’s only weekly. The fast firm can always undercut, so the slow firm stops trying to compete on price and charges above competitive — and the fast firm prices just below it, also above competitive. Both firms earn supracompetitive margins; consumers pay more; nobody colluded.
- Asymmetric commitment. A firm whose algorithm autonomously observes-and-reacts to rivals’ prices has a strategic advantage over one without — the same leader-follower logic, again without coordination.
- Symmetric commitment. Even when both firms have reactive algorithms, equilibrium prices never reach the competitive level — and can reach the fully collusive price, “even when algorithms are prohibited from employing collusive strategies.”
Their term for this is extractive innovation: an innovation (the pricing algorithm) whose private value to the firm is positive but whose social value is negative — it raises prices without improving the product. Antitrust, built on collusion and exclusion, can’t reach it: “winning market share through a superior pricing algorithm… has no antitrust remedy under current law.”
What is actually new: the commitment technology, not the game
A literature review commissioned 2026-08-02 (report in the research vault: ~/research/Algorithmic_Pricing_Collusion_Research_20260802/) sharpened the mechanism and the novelty question. The short answer: reacting to rivals’ prices is ancient; what is new is that software makes a firm’s reaction function a credible, observable, high-frequency, costly-to-revoke commitment. A pricing team with a “match competitors” playbook has none of those properties — it can reconsider in the meeting where it matters, rivals know it, and it cannot respond within the hour. Brown & MacKay’s three features of an algorithm versus a human agent are exactly the embodiment difference: lower cost of sophistication, lower cost of frequency, and a short-run commitment device (the algorithm is revised less often than it prices).
Three qualifications the review surfaced, all of which temper the novelty claim without overturning it:
- The game is old. Brown & MacKay explicitly nest their model in pre-computer theory — conjectural variations, Maskin–Tirole alternating-move games, price-matching guarantees (Salop 1986), supply-function competition. “Firms commit to rules rather than prices” is a conjectural-variations game with a new commitment device; Salcedo (2015) notes that hierarchical firms already embody algorithms as “protocols for lower level employees to implement.”
- The empirics are one clean study. Assad, Clark, Ershov & Xu (JPE 2024, German retail gasoline) find margins rise only in duopoly/triopoly markets and only after all local stations adopt — consistent with learned softening of competition, but also with a “failure to compete” reading the authors cannot rule out. The flagship experimental result (Calvano et al. 2020, Q-learning) has been substantially challenged as imperfect exploration rather than sophistication (Abada & Lambin 2023; den Boer et al. 2022).
- Type matters more than embodiment. Miklós-Thal & Tucker (2024): demand-forecasting algorithms can undermine collusion by sharpening the temptation to undercut in predicted-high-demand states; only rule-based competitor-pegging algorithms (the category Brown & MacKay model and M&W would regulate) raise concern. “What algorithm?” is the right first question.
The verdict the review defends: a phase transition in the commitment technology, not in the game — concentrated, as theory predicts, in concentrated markets.
The two-layer picture
Put the journal’s platform observation on top of the paper’s pricing mechanism and the structure is complete:
- Layer 1 (attention rents): platforms hold captive audiences and sell access to them (advertising, targeting). Competition for the captives doesn’t free them.
- Layer 2 (price rents): algorithmic pricing lets firms charge supracompetitive prices without coordinating, via frequency/commitment asymmetries. Competition between algorithms doesn’t discipline prices — it inflates them.
The shared structure is the point: in both layers, the normal discipline (competition) is present in form but absent in effect, because the thing being competed over is us and we are the captive input, not the choosing customer. The rent flows whether the captors fight or not.
The regulatory gap
MacKay & Weinstein’s two proposed remedies map the difficulty: restrict when firms price (eliminate frequency asymmetry — e.g., one price change per day, simultaneously) or restrict how (bar algorithms from incorporating rivals’ prices). Both require a new algorithm-reviewing bureaucracy and both risk dulling pricing innovation. The deeper problem: the harm is non-collusive, so the entire apparatus of competition law — agreements, conspiracies, coordination — is aimed at the wrong thing. The 2014 observation and the 2022 proof agree: this is a market failure the current regulatory vocabulary can’t name.
Cross-domain connections
- moral-economy-price-tag — the personalization layer: when the reference price disappears, the fairness check disappears with it (note: distinct from the market-level reactive pricing analyzed here — the two “algorithmic pricing” literatures should not be conflated)
- invisible-hand — the cleanest refutation in the wiki of the caricature hand: here self-interest plus competition reliably produces public bad, no coordination required
- intelligence-as-tool — the platform imbues intelligence into pricing/targeting as needed; the tool is pointed at the captive
- institutionally-constrained-technology-adoption — the same shape at the governance layer: optimizing internal metrics over external effectiveness
- prisoners-dilemma / cooperation-and-defection — the classical coordination problem the pricing algorithm dissolves by replacing it with unilateral commitment
- evolutionary-game-theory — adaptive learning dynamics as an alternative route to supracompetitive prices (Hansen–Misra–Pai misspecified learning; Calvano et al. Q-learning)
- farcaster-protocol — credible neutrality as a protocol-design goal, the mirror image of credible commitment as a pricing weapon
- bullshit-jobs — Graeber’s managerial feudalism is the employment-side analog: hierarchy extracting status-rent rather than productive output
- filter-bubble — the attention layer: the captive audience is sorted before it’s monetized
- life-liberty-happiness-trilemma — rent extraction as a robustness tax on the society that tolerates it
Further reading
- Literature review (research vault, 2026-08-02): Is Algorithmic Pricing New? A Literature Review Centered on MacKay & Weinstein (2022) —
~/research/Algorithmic_Pricing_Collusion_Research_20260802/research_report_is_algorithmic_pricing_new.md. Covers the novelty debate, the empirical record (German gasoline), the skeptical literature (Kühn & Tadelis, Schwalbe, Abada & Lambin), and the “what algorithm?” distinction (Miklós-Thal & Tucker). - Brown & MacKay (2023), Competition in Pricing Algorithms, AEJ: Micro — the theory foundation.
- Brown & MacKay (2025), Algorithmic Coercion with Faster Pricing, NBER w34070 — speed + commitment can be worse for consumers than collusion.
- Assad, Clark, Ershov & Xu (2024), JPE — the German gasoline field evidence.
Sources
- Selected journal entries (Day One export, 2012–2024)
- 2022
See Also
- ufc-mma-business-model — the labor-market mirror: captive suppliers (fighters) mined by a monopsonist buyer; same rent-extraction shape, input side