A pricing war is a repeated prisoner’s dilemma
A pricing war is a prisoner’s dilemma when each competitor can gain by cutting price alone, yet all competitors earn less if everyone cuts.
The pattern appears in supermarkets, airlines, software subscriptions, freelance bids, petrol stations, and online marketplaces. By the end, you will be able to build a pricing payoff matrix, identify a dominant strategy, explain why individually sensible discounts can damage every seller, and judge which responses protect profit without breaking competition law.
The original prisoner’s dilemma is a game with two players and two choices. Each player receives a payoff determined by both choices. In a pricing version, each business chooses either to hold its price or cut it. A cut attracts customers if the rival holds. If both cut, neither keeps the advantage for long, and both collect less money per sale.
Two lunch shops each sell a meal for £10. Either shop can cut its price to £8 before Monday. A lone discounter attracts more customers. If both shops discount, customer numbers barely shift between them, but each earns £2 less per meal before considering any extra sales.
This is not a claim that every discount is harmful. A lower price can clear excess stock, introduce a new product, match lower costs, or serve buyers who would otherwise walk away. The dilemma exists only when the incentives fit the game: each business wants to undercut the other, while mutual undercutting leaves both worse off than mutual restraint.
What does the pricing payoff matrix reveal?
A payoff matrix shows each possible pair of decisions and the resulting profit for both firms. It turns vague fear about a competitor into a testable model: compare one firm’s outcomes down each column, then do the same for the rival across each row.
Suppose two meal shops, A and B, must choose between a £10 price and an £8 price for one trading period. The figures below are invented business profits for a worked example, not market statistics. They follow from assumed customer numbers and a £5 variable cost per meal.
| Shop A choice | Shop B choice | A profit | B profit |
|---|---|---|---|
| Hold at £10 | Hold at £10 | £500 | £500 |
| Cut to £8 | Hold at £10 | £540 | £200 |
| Hold at £10 | Cut to £8 | £200 | £540 |
| Cut to £8 | Cut to £8 | £300 | £300 |
Here is the arithmetic. At £10, a shop makes £5 contribution per meal. If both hold price and each sells 100 meals, each earns . If A alone cuts, assume it sells 180 meals at a £3 contribution, while B sells 40 meals at £5. Their profits are £540 and £200. If both cut and each sells 100 meals, each earns £300.
The matrix separates joint interest from individual temptation. Combined profit is highest when both hold: £1,000. Yet each shop sees a £40 gain from cutting when the other holds. The feared outcome also matters. If the rival cuts, cutting produces £300 rather than £200. The same move looks attractive under optimism and under fear.
Why does the rational choice produce a worse result?
Each shop chooses without controlling its rival, so it compares its own outcomes under both possible rival actions. Cutting pays more in either case. Both shops therefore cut, producing £300 each even though holding would have produced £500 each.
A dominant strategy gives a player a better payoff regardless of the other player’s choice. For Shop A, cutting beats holding if B holds, because £540 exceeds £500. Cutting also beats holding if B cuts, because £300 exceeds £200. The matrix is symmetric, so B reaches the same conclusion.
In the example: £540 > £500 and £300 > £200, so cutting dominates holding.
The symbol means Shop A’s profit. The first letter inside the brackets is A’s decision; the second is B’s. Once both choose their dominant strategy, neither can improve by changing alone. If A raises its price while B stays cheap, A falls from £300 to £200. This makes mutual cutting a Nash equilibrium.
An equilibrium is stable against one-sided changes. It is not necessarily fair, efficient, or profitable. That distinction matters because people often hear “equilibrium” and imagine balance in an approving sense. Game theory uses the word more narrowly: given what everyone else is doing, no player wants to move alone.
Each shop cuts because the matrix gives it a higher profit for either rival choice.
Both shops would earn more if both held, but neither can secure that outcome by acting alone.
The conflict is between incentives, not intelligence. Owners can understand that a price war will hurt and still start one. A promise to hold price carries little weight if breaking it pays immediately. The model explains why good intentions do not repair a reward system that favors defection.
When does a real discount become a prisoner’s dilemma?
A real market resembles the dilemma when sellers choose independently, a unilateral cut steals enough demand to pay, matching cuts erase that advantage, and no seller can safely commit to restraint. Remove one of those conditions and a different model may fit better.
Start with the customer response. Price elasticity of demand measures how strongly quantity demanded responds to a price change. If buyers barely react, a discount can reduce revenue without adding enough sales. If buyers switch readily, the first discount may be tempting. The methods taught in measuring demand and supply responses help distinguish those cases.
A lower price does not guarantee a larger profit. Extra units help only if their added contribution covers the margin lost on units that would have sold anyway.
A simple break-even calculation makes this visible. If a product costs £6 to supply and sells for £10, its contribution is £4. Cutting the price to £9 reduces contribution to £3. To preserve a £400 contribution, sales must rise from 100 units to , so at least 134 whole units are needed.
Costs shape the game too. A firm with spare capacity may serve extra customers cheaply. A firm near its limit may need overtime, new equipment, more support staff, or expensive delivery slots. The payoff belongs after these costs, not at the revenue line. Basic algebra for rearranging business formulas lets you solve for the volume needed at any proposed price.
Finally, ask whether the choices really are binary. Competitors can alter quality, delivery speed, bundles, warranties, loyalty rewards, or advertising. If customers value these differences, the market is not simply “hold or cut.” A payoff matrix still helps, but its rows and columns should represent the decisions managers can actually make.
Why do repeated pricing games behave differently?
Repeated competition adds memory and future consequences. A seller may reject a profitable cut today if it expects the rival to respond tomorrow. Cooperation can become self-enforcing, but only when future profit matters enough and each side can observe what the other actually did.
In a one-period game, there is no later punishment or reward. Take the best current payoff and the game ends. Real competitors often meet again. Petrol stations post prices every day, online sellers monitor listings, and contractors bid for recurring work. Today’s action changes beliefs about tomorrow’s response.
It wins a temporary advantage while the rival still charges more.
It matches or beats the lower price to recover customers.
Customers now expect the cheap price, so raising it alone risks lost sales.
Economists represent patience with a discount factor, usually , between zero and one. A profit received next period is valued at times an equal profit received now. If cooperation yields profit every period, its present value in an indefinitely repeated model is:
If and , the model values the stream at .
The calculation does not predict that firms will cooperate. It shows why the future can outweigh a short gain. Cooperation becomes harder when managers expect the relationship to end, cannot observe rival prices accurately, face sudden cash pressure, or believe a cut will be mistaken for a temporary promotion.
Repetition also creates a legal danger. Firms may recognize their shared interest without speaking, but explicit agreements between competitors about prices can violate competition law. Game theory explains the incentive to avoid a war. It does not provide permission to coordinate.
What information does your competitor hope you miss?
Your competitor benefits if you focus on its advertised price while ignoring its costs, capacity, customer mix, and time horizon. The visible price is one move; the hidden payoff determines how long that move can be sustained and what response it is meant to provoke.
A low posted price may signal several different facts. The rival may have lower unit costs. It may be clearing stock that will soon lose value. It may earn profit from add-ons after selling the main item cheaply. It may also be making a mistake. Copying the move before identifying its economics turns the rival’s assumptions into your decision.
Customer segments can overturn the simple matrix. A competitor might offer a student discount, annual contract, or basic tier while keeping its standard price. That move targets buyers with greater price sensitivity without surrendering margin on every sale. Good analysis asks who can claim the discount, what they would have bought otherwise, and how easily existing customers can switch.
Capacity is equally revealing. If the rival cannot serve many more customers, its cut may be brief or tightly limited. If it has unused warehouses, automated support, or perishable inventory, extra volume may be more valuable to it than to you. The same £2 discount can therefore represent aggression, housekeeping, or an attempt to fill idle capacity.
How should a business respond without starting a race to the bottom?
A sound response begins by rebuilding the payoff matrix with current evidence, then testing targeted moves before changing the headline price. The goal is not passive price holding. It is choosing the cheapest response that protects valuable customers and preserves contribution.
Check product quality, contract length, delivery fees, eligibility, stock limits, and promotion dates. Two prices are comparable only when the offers are comparable.
Subtract avoidable unit costs from each candidate price, then include extra service and acquisition costs. Record the sales increase needed to protect total contribution.
Identify which customers noticed, which can leave, and which value features beyond price. Do not subsidize buyers who were staying anyway.
Test a narrow segment, bundle, service promise, or limited promotion. Set a review date and a stopping rule before launch.
Reversibility matters because price cuts create reference points. A customer who sees £8 today may treat £10 next month as an increase, even if £10 was normal before. A bundle can be easier to change because the offer itself changes. Better delivery, clearer guarantees, or a simpler product can also compete on value rather than price alone.
For software businesses, experimentation needs the same discipline as product work. Clean tests, monitoring, and rollback plans connect strategic choice with the practical skills involved in shipping and monitoring software applications. A discount launched without measurement produces activity, not knowledge.
Set the stopping rule in advance. For example: end the test if contribution per visitor falls below the old offer for two complete sales cycles, or expand only if the gain persists after support and refund costs. These are decision rules, not universal thresholds. Each firm must choose measures that match its economics.
What does the model leave out?
The prisoner’s dilemma is a deliberately small model, so it can mislead when market structure, changing costs, brand differences, new entrants, or uncertain demand drive the result. Use it to expose incentives, then test whether its assumptions match the market in front of you.
Many pricing contests are not symmetric. A large chain may tolerate losses longer than an independent shop. A new entrant may value customer acquisition more than current profit. A platform may subsidize one side of a market because another side pays. Giving both firms identical numbers would hide the feature deciding the contest.
Two players choose between holding and cutting, with known payoffs and a clear result.
Many sellers, uncertain demand, unequal costs, product differences, capacity limits, and buyers who react over time.
Some price cuts reflect genuine efficiency rather than strategic defection. If a new process lowers the cost per unit, the efficient seller can profit at a price that harms an older rival. Mutual high prices are then not the best social outcome. Buyers benefit from the innovation, and the pressure may move resources toward the lower-cost method.
Consumer welfare also matters. The two firms’ combined profit is not the whole economy. In the worked matrix, customers pay less when firms cut, so some value moves from sellers to buyers. Regulators do not exist to protect competitors from healthy competition. They focus on protecting the competitive process under the relevant law.
A richer model might use several prices, uncertain probabilities, sequential moves, or entry and exit. Those extensions draw on optimization, probability, and functions across the wider mathematics curriculum. The best model is the smallest one that preserves the incentive causing the behavior.
A payoff matrix turns pricing fear into a decision
A competitor’s discount should trigger calculation, not imitation. Write the available moves, estimate profit under each pair, mark each player’s best response, and test how the answer changes when demand, costs, capacity, and future reactions change.
The prisoner’s dilemma gives a precise warning: a move can be rational for one firm and damaging when every firm copies it. The warning is strongest when cutting is a dominant strategy and mutual cutting is a Nash equilibrium. It weakens when products differ, future punishment is credible, or one firm has a real cost advantage.
The takeaway: Never answer a rival’s price before estimating the payoff behind it. Protect profit with evidence, narrow tests, and reversible moves, while keeping every competitive decision independent and lawful.
A useful pricing decision therefore ends with numbers and conditions. State what the rival must be assuming, how many extra sales your response requires, which customers it protects, and when you will stop. That record makes later learning possible. It also keeps a visible competitor move from becoming an automatic order to cut.