Home / Wiki / I / Incentive Compatibility

Incentive Compatibility

Economics Markets and Financial TheoryUpdated 16 Nov 2026

Incentive compatibility is a property of a mechanism, meaning a set of rules for making a collective decision, such as an auction, a voting system, a contract or a market. A mechanism is incentive compatible if every participant achieves the best result for themselves by behaving truthfully: bidding what an item is really worth to them, reporting what they really believe, or exerting the effort they have promised. Honesty is then the best policy by design, not by goodwill.

The idea comes from mechanism design, a branch of game theory that works backwards from a desired outcome to the rules that will produce it when participants pursue their own interests. The economist Leonid Hurwicz introduced the term in the early 1970s, and he shared the 2007 Nobel Memorial Prize in Economic Sciences with Eric Maskin and Roger Myerson for laying the foundations of the field. The concept addresses the problems caused by asymmetric information, where one party knows something the other cannot observe.

01

Formal definition

Suppose each participant ii has private information θi\theta_i, called a type, and is asked to report it. Writing uiu_i for the participant's payoff, a mechanism is incentive compatible when, for every participant and every possible false report θi′\theta_i':

ui(θi∣θi)≥ui(θi′∣θi)u_i(\theta_i \mid \theta_i) \ge u_i(\theta_i' \mid \theta_i)

The left side is the payoff from reporting the true type and the right side is the payoff from reporting anything else. Two strengths of the condition are distinguished. Under dominant-strategy incentive compatibility, also called strategy-proofness, truth-telling is best whatever others do. Under Bayesian incentive compatibility, it is best on average provided that others are also truthful, which makes truth-telling a Nash equilibrium of the game. A result known as the revelation principle shows that any outcome achievable by some mechanism can also be achieved by an incentive-compatible one in which participants simply report their types, which is why designers can restrict attention to truthful mechanisms.

02

Worked example: the second-price auction

In a sealed-bid second-price auction the highest bidder wins but pays the second-highest bid. Take a bidder who values an item at 100. If the highest rival bid is 80, a truthful bid of 100 wins and pays 80, for a surplus of 100−80=20100 - 80 = 20. Bidding 120 changes nothing, since the price is still 80. Bidding 70 loses the item and the surplus falls to zero.

If the highest rival bid is 110, a truthful bid loses, with a surplus of zero. Overbidding at 120 would win but cost 110, a loss of 100−110=−10100 - 110 = -10. In every case the truthful bid does at least as well as any other, so truth-telling is a dominant strategy. A first-price auction, where the winner pays their own bid, lacks the property: bidders gain by shading bids below their value. The field of auction design is largely concerned with such comparisons.

03

Applications in finance

  • Contracts. In the principal-agent problem, pay is structured so that the agent prefers the action the principal wants. Performance fees, deferred bonuses and deductibles in insurance are devices for reducing moral hazard, and menus of contracts are used to counter adverse selection by leading customers to sort themselves by type.
  • Security auctions. Governments choose between uniform-price and pay-as-bid formats when selling debt by competitive tender, partly on the basis of which encourages more aggressive and honest bidding.
  • Benchmarks. LIBOR was compiled from banks' own estimates of their borrowing costs, and submitters could profit by misreporting. Its replacements are calculated from actual transactions, which removes the opportunity.
  • Blockchains. A consensus mechanism is designed so that validators earn more by following the protocol than by attacking it.
04

Prediction markets and scoring rules

A prediction market is intended to be incentive compatible in a practical sense: a trader who believes an event is more likely than the price suggests profits, on average, by buying, and so reveals that belief. The price that results is read as an implied probability, and the wider process is called information aggregation.

Forecasting competitions achieve the same end with proper scoring rules. Under the Brier score a forecaster who states probability rr is penalised by (1−r)2(1 - r)^2 if the event occurs and r2r^2 if it does not. For someone who believes the true probability is 0.7, an honest report gives an expected penalty of 0.7×0.09+0.3×0.49=0.210.7 \times 0.09 + 0.3 \times 0.49 = 0.21, while exaggerating to 0.9 gives 0.7×0.01+0.3×0.81=0.250.7 \times 0.01 + 0.3 \times 0.81 = 0.25. The honest report scores better, and this holds for any belief.

05

Limits

Incentive compatibility is usually proved under assumptions that real settings violate. Participants may collude, hold stakes outside the mechanism, or be able to influence the outcome they are betting on. A trader with a large position elsewhere may accept a loss in a prediction market in order to move its price. Risk aversion, limited budgets and trading costs can also stop informed participants from acting on their beliefs.

There is in addition a trade-off with other goals. Theoretical results show that incentive compatibility, full efficiency, voluntary participation and a balanced budget cannot always be achieved together, so designers must decide which to relax. Mechanisms that are strategy-proof in theory, such as the second-price auction, are also less common in practice than their properties would suggest, in part because they are vulnerable to collusion and to false bids placed by the seller.

Last updated 16 Nov 2026