← all systems · see it in The System
Oligopoly
In economics an oligopoly is a market in which a few sellers hold pricing control, so that each must reckon with how the others will respond. Industrial organization models the interdependence as a game: quantities chosen at once (Cournot, 1838), prices chosen at once (Bertrand, 1883), or a leader moving first and followers responding (Stackelberg, 1934). Repeated play lets the few sustain prices above the one-shot level by a tacit understanding backed by the threat of retaliation (tacit collusion), or by an explicit agreement (a cartel), which competition law prohibits; how easily this holds depends on the number of firms, entry barriers, how often they meet, how well they see each other's prices, demand swings, cost and capacity asymmetries. Incumbents can deter entry by irreversible commitments. Walked here as that market form and the conduct it carries; the firms are sys:firm, the explicit agreement sys:cartel.
Properties published scienceour proposal
Sourced fields, and our own readings.
| born at | per market: any time concentration rises to a few sellers; described in theory from Cournot's spring-water duopoly (1838) |
| born precision | n/a |
| boundary | Inside: the few sellers of one relevant market, their interdependent choices of price, quantity, capacity and investment, and whatever conduct regime holds among them (rivalry, tacit coordination, an explicit cartel). Outside: the buyers, potential entrants kept out by barriers, input suppliers, competition authorities and courts, and other markets. The line is drawn by market definition and entry barriers: an open oligopoly has none, a closed one has prominent barriers. In practice the relevant market is drawn by the hypothetical monopolist test: whether a hypothetical monopolist over a candidate market could profitably raise price. |
| nested level | Built from a few firms and the rules of play among them; a kind of market structure (between monopoly and many-firm competition); an industry within the economy; runs on its firms' decisions. |
| scale | social |
Measures (22) published science
| critical discount factor for collusion, n identical price-setting firms | delta* = 1 - 1/n: 0.50 for two firms, 0.67 for three dimensionless |
| HHI threshold for a highly concentrated market (US 2023) | above 1,800; a change of more than 100 points is significant; 10,000 for one firm index points |
| four-firm concentration ratio for a tight oligopoly | above 60 (loose: 40-60) percent of market |
| cartel overcharge, raw case estimates | mean 31-49% (Connor and Lande 2008); mean 50.4% for successful cartels (Connor 2010) percent above competitive price |
| cartel overcharge, bias-corrected | mean 15.47%, median 16.01% percent above competitive price |
| US base cartel fine, share of affected commerce | 20% (10% base plus 10% for buyers priced out) percent of affected commerce |
| cartel duration | average about five years across studies; median five to six years; bimodal, many under one year, some decades |
| annual probability a functioning cartel is discovered | roughly 10% (as cited by Wachs and Kertesz); at most 13-17% for US conspiracies (Bryant and Eckard 1991); 12.9-13.3% for EU cartels (Combe, Monnier and Legal 2008) per year |
| Perth retail gasoline price cycle length after the focal point | 7 days for all firms from April 2009 (7 to 35 days before) |
| Perth average daily retail gasoline margin, before and after March 2010 | difference-in-difference against other Australian cities +3.49 cpl (SE 0.39), between 1.00 and 3.49 cpl (19-64%) depending on sampling; raw: 4.85 cpl (SD 4.19) before, 10.01 cpl (SD 4.68) after, which partly reflects a national trend Australian cents per litre |
| settling time of a coordinated equilibrium (Perth gasoline) | about three years |
| Perth gasoline station shares 2005-2015 | BP 22%, Caltex 16%, Woolworths 14%, Coles 16%, remaining 32% largely independents percent of stations |
| experimental Cournot markets by number of firms | 2 firms: some collusion; 3: near Nash; 4-5: never collusive, at or above Cournot output qualitative |
| Theocharis stability bound, linear Cournot best-response dynamics | stable with 2 firms, oscillatory with 3, unstable with 4 or more (discrete best-reply adjustment, linear demand) |
| satellite duopoly constellations, leader and follower | leader 29,750 satellites at 500 km; follower 1,945 satellites at 603 km satellites; km altitude |
| welfare lost to duopoly in orbit use | up to 12% (about USD 1.1 billion per year) below a welfare-maximising public utility system; per-hand-off efficiency of orbit use under duopoly about 0.88 or more of the regulated benchmark percent of welfare |
| share of natural and medical science papers from the top five publishers | a little over 20% (1973), 30% (1996), 50% (2006), 53% (2013) percent of papers |
| public contracts screened for cartel groups (Georgia 2011-2016) | nearly 150,000 contracts |
| Lerner index and the Cournot margin | L = (P - MC)/P, from 0 (competitive) to 1; Lerner rule L = -1/E_d for a profit-maximiser with market power; Cournot firm i: (P - c_i)/P = s_i / eta; the HHI tracks the share-weighted average margin under Cournot with linear costs dimensionless |
| average US markup over marginal cost (firm-level data) | 18% in 1980 rising to 67% (2017 working paper; the later published figures not read) percent above marginal cost |
| common-ownership-adjusted concentration, US airlines | increases in concentration 10 times larger than what antitrust authorities presume likely to enhance market power |
| experimental Cournot output against prediction | the ratio of actual to predicted total output rises significantly with the number of firms qualitative |
Open contentions (11) open contention
Where sources or readings disagree. Both stay on the map.
- C-3462 Can pricing algorithms learn to collude on their own?
booked: The map keeps the competing readings; none is ruled.
the other reading: the Q-learning results depend on long training runs and particular settings (the specialist's lead, no citation in hand) - C-3466 How much do cartels overcharge?
booked: The map keeps the competing readings; none is ruled.
the other reading: Connor and Lande: the 10% presumption in US sentencing is far too low; Boyer and Kotchoni: raw estimates carry model, estimation, endogeneity and publication bias - C-3471 Does shared ownership of rival firms raise prices?
booked: The map keeps the competing readings; none is ruled.
the other reading: strongly contested (the specialist; the critical papers not read) - C-3478 Do concentrated markets make firms profitable, or do efficient firms become both?
booked: The map keeps the competing readings; none is ruled.
the other reading: game-theoretic IO: industry characteristics dictate outcomes case by case - C-3479 Can the mere threat of entry keep a few firms honest?
booked: The map keeps the competing readings; none is ruled.
the other reading: - C-3480 Does a market of quantity-setting firms settle down as firms are added?
booked: The map keeps the competing readings; none is ruled.
the other reading: capacity constraints can prevent the destabilisation (Puu) - C-3518 Does the kinked demand curve explain sticky prices?
booked: The map keeps the competing readings; none is ruled.
the other reading: - C-3531 Does market power help or hurt innovation?
booked: The map keeps the competing readings; none is ruled.
the other reading: market power funds innovation (Schumpeter) / competition sharpens the incentive (Arrow): the specialist's framing, not read here - C-3533 Have US markups really risen since 1980?
booked: The map keeps the competing readings; none is ruled.
the other reading: NBER w34513 comments on the paper, with a reply and a rejoinder (listed on the NBER page, not read); critics: the estimates depend on how overhead costs are treated (Traina 2018; Basu 2019; the specialist's lead, not read) - C-3538 Do firms in a small market compete on price or on quantity?
booked: The map keeps the competing readings; none is ruled.
the other reading: Bertrand's critique that firms choose prices, not quantities (1883); Pareto unimpressed by the critique; Fisher siding with Bertrand; draft framing (before the specialist pass): quantity, price and leader models as rival descriptions - C-3545 Do price wars come in booms or in slumps?
booked: The map keeps the competing readings; none is ruled.
the other reading: evidence from the Joint Executive Committee (Porter 1983) and Ellison's re-analysis said by the specialist to fit Green-Porter broadly (not read)
Cross-scale patterns seen here (24) our proposal
Shapes that recur when the same kind of problem is solved at different scales, and how each shows in Oligopoly. Each links to every other system that shows it.
- conservation law plus closure (structural)
Profit maximisation (marginal revenue against marginal cost) leaves the rivals' reaction open; it predicts nothing until a conjecture closes it: rivals' output fixed (Cournot), rivals' price fixed (Bertrand), the follower's best response (Stackelberg), or a collusive path. Different closures, differ - runaway from an unstable balance (structural)
One-shot Bertrand: identical prices above cost are a balance point where a small undercut feeds itself ('a destabilizing incentive for each firm to undercut the other') down to cost; the linear discrete Cournot adjustment is oscillatory at three firms and unstable above (Theocharis), though stable f - same role, different builds (structural)
One job, coordinating sellers' prices, done by different builds: explicit cartels with meetings and sales agencies, industry associations, tacit focal points and price leadership, Q-learning and LLM pricing agents, state-run cartels. - symmetry breaking (name the cut) (structural)
Two identical firms, symmetric in Cournot; the cut is who moves first with commitment (Stackelberg); after it the leader is larger. In orbit the leader takes 29,750 satellites at 500 km, the follower 1,945 at 603 km. - trade one property for another (structural)
Price-war length trades discipline for profit: longer wars deter cheating better but cost more when bad luck triggers them, so the best scheme sets T just long enough. In orbit, lower shells buy latency and pay with more satellites. - who opens the door (structural)
Rivals' price signals arrive into doors the receivers opened ahead of time: monitoring arrangements and sales reporting built by cartel organisations, surveillance tools bought by firms; the receiver installs the port. - barrier-gates-complexity (partial)
Behind entry barriers the few can build and keep coordination (cartel organisations, monitoring, focal points); without the barrier the case is shown: entry is the most frequent cause of cartel breakdown and an open oligopoly with free entry behaves more competitively. Escalating sunk spending (adve - causal ceiling (partial)
Where each firm sees only its own price and sales, no firm can tell a rival's cheating from bad luck, so perfect collusion is impossible whatever the firms do: a limit set by the market's information structure, not by effort. - coordination-before-mapping (partial)
LLM pricing agents not given the demand function, and unable to communicate except through prices, still coordinate on supracompetitive prices; tacit collusion more widely coordinates without an agreed plan. The agents do reason about rivals, so a map of rivals exists; the market's map does not. - holding / entrainment (partial)
In Perth, once Thursday price jumps became the focal point, every firm's station-level cycle length converged to seven days and stayed there: the market's coordinated regime set its members' clock. Dark-twin check: the firms do not share an outside driver; the leader's experiments set the day. - outside-to-inside (partial)
A merger takes rivalry that ran between two firms across the market and puts it inside one firm ('the elimination of competition between the merging firms'); the outside phase (two rivals) comes first. - red-queen (partial)
Edgeworth cycles: each firm undercuts the other in turn to gain share, and the cycle returns to where it started (a sharp rise, then undercutting again); in a close patent race both spend heavily and neither pulls ahead. Stalemate shown in the cycle; the race ends when one leads. - same shape, opposite sign (partial)
One process, an incumbent's irreversible investment before entry, flips sign with the market: under price competition (strategic complements) a tough incumbent under-invests to accommodate and over-invests to deter; the conclusions reverse if the investment makes it softer or competition is in quant - same toolkit, independent builds (partial)
Q-learning agents, LLM agents and human cartels separately arrive at reward-punishment schemes whose punishments decay over several periods, from the same parts (repeated play, observed prices). Dark-twin caution: LLMs are trained on human text, so the build may be partly inherited. Calvano et al. ( - templating (partial)
One equilibrium module reused across industries: the trigger-strategy condition (stick to the collusive price if the discounted share outweighs the one-period gain) is applied unchanged to rail, shipping, gasoline and lab markets; Cournot equilibria are a subset of Nash equilibria. - The carry (partial)
The wait face: coordination in Perth took a gradual three-year transition to settle; overflow face in Bertrand-Edgeworth: when one seller cannot supply the whole market the price-equals-cost result may not hold, the unmet demand falling to rivals. - The Payment to Entropy (partial)
Thermal-free form, 'welfare' form: a duopoly in orbit loses up to 12% of the welfare a regulated system would get from the same orbits (the toll); the cartel overcharge itself is a hand-off from buyers to sellers (dark twin, booked as the buyer exchange), while the buyers priced out are the toll (th - the whole sets the local reference (partial)
The focal price (kinked demand) or the leader's price is the reference each seller prices against; each reads its own demand locally, yet the reference is a market-level object. - Umwelt: each system carves its own slice of one shared medium (partial)
Each seller takes its own slice of the shared market: its own price and sales (Green-Porter), or all prices plus its own demand (the LLM experiments); the market price is common, the views are not. - lightning (surface)
A price war releases a stored margin all at once and changes what follows for T periods; sources read gate it (a trigger strategy), which is the lightning dark twin, so surface. - nested ladder (surface)
Concentration bands ordered as a ladder: unconcentrated, moderately concentrated (HHI 1,000-1,800 in 1982), highly concentrated (above 1,800), monopoly at 10,000; loose (CR4 40-60) and tight (above 60) oligopoly. Ordered by degree; the implication between rungs is not shown. - partner-becomes-part (surface)
Rivals merge into one firm and publishers acquire journals into their own houses; but the rivals were not in a gain/gain loop before (the dark twin), so surface. - seed (surface)
Game theory entering industrial organization in the 1980s unpacked a large family of industry models (strategic investment, Markov perfect equilibria, patent races); that seed belongs to the field, not to the market, so surface here. - smooth to sharp in finite time (surface)
In the base duopoly the sustainability of collusion jumps from none to full at a finite discount factor (1/2), a 'knife-edge' the source says is specific to the simple example. The source's own caveat: robustly there are two thresholds, with more collusion achievable between them as patience rises;
Valence: what it trades, and with whom (12) published science
Each exchange with a neighbour, read from both sides: who gains, who pays, and what carries it.
- Monopoly via sys:market · Oligopoly even
- Human (Homo sapiens) via sys:human-society · Oligopoly gain
- State (polity); the nation state its predominant modern form via sys:human-society · Oligopoly mixed
- Law (rules and law) via sys:human-society · Oligopoly loss
- LLM-based agent via sys:written-record · Oligopoly gain
- Firm (the economic unit) via sys:market · Oligopoly gain
- Consumer via sys:market · Oligopoly gain
- Trade association (business association) via sys:human-society · Oligopoly gain
- Potential entrants via sys:market · Oligopoly mixed
- Low Earth orbit (as a shared resource) via sys:gravitational-field · Oligopoly gain
- Market competition via sys:market · Oligopoly mixed
- Tacit collusion via sys:market · Oligopoly mixed
Partners and parts published science
Built from / builds (4)
Sits inside / holds (4)
expressed by (18)
- Templating / Standardization
- Niche Construction (alter environment)
- Internal Coordination/Regulation
- Sensing/Scanning
- Mapping/Modeling
- Uncertainty Reduction/Learning
- Active Experimentation / Probing (Optimal Experimental Design)
- Surprise / Change-Point Detection
- Social Learning / Public Information Use
- Evade Detection
- Resource Denial/Attrition
- Prepositioning & Latency
- Battlespace Shaping
- Reactive Response
- Self-Monitoring
- Resilience & Repair
- Deterrence
- Quality Control & Error Correction
floor filled by (13)
- Price: a count of the unit bound to a good (role: quanta)
- LLM-based agent
- Market and trade
- Firm (the economic unit)
- Economy
- Consumer
- Game theory
- Trade association (business association)
- Tacit collusion
- Cartel
- Entry barriers
- Pricing algorithm
- Herfindahl-Hirschman index
represents (1)
Kinds (10)
- Market structure
- Firm (the economic unit)
- Market competition
- Duopoly
- Perfect (pure) oligopoly
- Imperfect (differentiated) oligopoly
- Closed oligopoly
- Tight oligopoly
- Collusive oligopoly
- Dominant firm with competitive fringe (partial oligopoly)
Notes published science
- In experiments with simulated markets (not real firms), two AI pricing agents built on large language models, told only to maximise long-run profit and never to collude, quickly settled on prices well above the competitive level, and their written reasoning showed they were avoiding a price war. Changing a few innocent-looking words in their instructions pushed prices higher still, close to what a single monopolist would charge (src:fish-2024-llm-collusion). In real German petrol markets, margins rose where pricing software arrived, but in two- and three-station markets only when every station adopted it (src:assad-2024).
- In Perth, Australia, petrol stations learned to coordinate without meeting: from April 2009 price jumps clustered on Thursdays, every chain's price cycle settled at exactly seven days, and the average retail margin went from 4.85 to 10.01 cents a litre (src:byrne-deroos-2019).
- Oligopoly reaches into space: in a model of two satellite internet companies, the first mover takes the low, fast orbit with 29,750 satellites at 500 km and pushes its rival up to 603 km with 1,945, and the duopoly leaves up to 12% (about 1.1 billion dollars a year) of orbit's value on the table (src:guyot-2023-satellite-oligopoly).
Layers
The ways it takes in its world, where they meet, and how it acts back, each from L1 (where it meets the world) up.
oligopoly as industrial organization describes it: a few sellers' market to their moves (reader's view)
- L1: Buyers' demand and the few sellers' offers on one relevant market; prices posted where buyers can compare them
- L2: The few: number of sellers and their shares (concentration: HHI, four-firm ratio), held few by entry barriers (investment, loyalty, regulation, scale)
- L3: Watching the rivals: observability of prices, terms and sales; detection of a deviation
- L4: Interdependence: each seller's best response to the others' quantities (Cournot), prices (Bertrand) or a leader's committed move (Stackelberg)
- L5: The conduct regime: rivalry, tacit coordination (focal points, price leadership) or an explicit cartel, held by the threat of retaliation and by patience
- L6: The settled outcome: market price, outputs, shares and margins between the competitive and the monopoly level
- L7: The moves acted on the world: prices and output set, irreversible commitments (capacity, R&D, advertising), entry deterred or accommodated, price wars launched
the boardroom (shared floors)
- L4: Interdependence: best responses to the rivals' quantities, prices or committed moves (Cournot, Bertrand, Stackelberg); conjectures about rivals' reactions
- L5: The conduct regime of repeated play: compete, coordinate tacitly, or agree explicitly; retaliation threats and the patience that sustain them
- L6: The settled outcome: price, outputs, shares, margins (between Bertrand cost pricing and the monopoly level)
- L7: The sellers' moves: set prices and quantities, commit capacity and R&D, deter or accommodate entry, launch or end price wars
way in: Competition law: prohibition, merger review, detection and fines
- L1: Statutes and agencies that prohibit price-fixing and review mergers
- L2: Merger review by concentration: structures conducive to coordination are blocked
- L3: Detection and penalties: roughly a 10% yearly chance of discovery; fines set on affected commerce; leniency for early confessors, the cornerstone of enforcement, turning members' trust problem against the cartel
way in: Contact in other markets and shared owners
- L1: The same rivals met in other geographic or product markets; rivals held by the same institutional investors
- L2: More frequent interaction across markets; cross-ownership lowering the gain from undercutting
- L3: Asymmetries in single markets softened by trading advantages across markets
way in: Costs and capacity: inputs, scale and technology
- L1: Inputs and technology; economies of scale
- L2: Marginal cost and capacity: whether one seller can supply the whole market
- L3: Cost asymmetry: a low-cost firm fears a price war less
way in: Demand: buyers' purchases and the sales each seller makes
- L1: Buyers choosing among the sellers' offers, observing their prices
- L2: Each seller's own sales and market share; large buyers can concentrate orders to make interaction less frequent (buyer power)
- L3: Demand shocks read through own sales: a drop may be bad luck or a rival's cheating; booms raise the gain from deviating
way in: Potential entrants: the threat of entry
- L1: Outside firms able to enter if profits are high and barriers low
- L2: Entry barriers that filter them: investment, loyalty, regulation, scale, sunk costs
- L3: Contestability: the credible threat of short-term entry disciplining the few
way in: Pricing algorithms: software that reads rivals' prices and sets one's own
- L1: Price data streams and the tools that track or predict competitor prices
- L2: The pricing agent (Q-learning or a large language model) taking in all prices and its own demand, without the demand function
- L3: Learned strategy: reward and punishment answering a rival's low (high) price with low (high) prices, decaying over periods; prompts shift the level
way in: Rivals' prices and quantities
- L1: The rivals' prices, outputs and terms offered to customers
- L2: Observability: how promptly and easily rivals' terms can be seen (published prices, trade publications, monitoring)
- L3: Detecting a deviation: an undercut must be identified before it can be punished
way in: Talk among sellers: meetings, associations, announcements
- L1: Meetings, industry and trade associations
- L2: Information exchange, public or private, raising observability
- L3: The terms of an agreement: when and how high, who leads, how fast to follow, what counts as violation
way in: The history of play: past price wars, past coordination, focal points
- L1: The whole history of play that strategies may condition on
- L2: Focal points that emerge from play (Thursday price jumps, 2-cent cuts)
- L3: Reputation of the market: prior coordination or failed attempts read as susceptibility
way out: Irreversible commitments: capacity, R&D, limit prices, aimed at entrants and rivals
- L1: Outside firms face a market they cannot profitably enter
- L2: The barrier raised: excess capacity or a limit price that makes entry unprofitable
- L7: Investment chosen to deter or accommodate entry (over- or under-invest)
way out: Prices and output to buyers
- L1: Buyers pay the price; overcharges under cartels
- L6: The margin above cost (above the competitive level when coordination holds)
- L7: Price or quantity chosen
way out: R&D races and new technology
- L1: New technology and patents in the world
- L7: R&D spending set against rivals' in a race to a patent
way out: Retaliation: price wars and targeted punishment
- L1: The deviant's sales and profits cut (fighting ships)
- L5: The trigger: revert to competition, or a price war for T periods
- L7: Retaliation launched after a deviation
way out: Signals to rivals: price leadership and price experiments
- L1: Rivals follow, raising or lowering by the same amount
- L3: The move made observable to rivals
- L7: A dominant firm leads a price rise or experiments with prices
Who fills each layer
- L5: Tacit collusion