Incentive Compatibility

Incentive compatibility is a property of an economic mechanism or contract under which each participant finds it optimal to reveal private information truthfully or to take the action the mechanism prescribes. It is central to mechanism design, game theory, and contract theory. Rather than assuming that participants obey instructions, incentive-compatible arrangements make the intended behavior consistent with their own interests. The precise guarantee depends on the equilibrium concept and information assumptions used. (nobelprize.org)

Private information and strategic behavior

Incentive problems arise when an institution needs information that participants possess but cannot readily be compelled to disclose accurately. A bidder knows its willingness to pay; a supplier knows its production costs; a worker may choose effort that an employer cannot observe. Such information asymmetry creates opportunities to manipulate reports or actions. Mechanism design therefore treats the rules governing allocations and payments as determinants of behavior, not merely as procedures for processing information. (nobelprize.org)

A participant’s type represents relevant private information, such as a valuation or cost. A direct mechanism asks participants to report their types and selects an outcome from those reports. Asking for truthful information does not itself make a mechanism incentive compatible: its allocation and payment rules must ensure that misreporting is not profitable. (ocw.mit.edu)

Formal definition and principal forms

Let participant $i$ have true type $\theta_i$ , report $r_i$ , and utility $u_i(o;\theta_i)$ from outcome $o$ . Write $g(r_i,r_{-i})$ for the mechanism’s outcome, where $r_{-i}$ denotes everyone else’s reports. Utilities may incorporate an expectation over random outcomes. Incentive compatibility compares the utility from truthful reporting with the utility from every alternative report. (nobelprize.org)

Dominant-strategy incentive compatibility

A mechanism is dominant-strategy incentive compatible (DSIC) if, for every participant, true type, alternative report, and profile of others’ reports,

u_i(g(\theta_i,r_{-i});\theta_i) \geq u_i(g(r_i,r_{-i});\theta_i).

Truthfulness is therefore a weakly dominant strategy: it remains optimal regardless of what others report. This property is often called strategy-proofness. It does not require beliefs about other participants’ types or strategies. “Weakly” matters: some false reports may yield exactly the same utility as the truthful report. (nobelprize.org)

Bayesian incentive compatibility

A mechanism is Bayesian incentive compatible (BIC) if truthfulness is optimal in expectation, conditional on a participant’s own type and assuming that others report truthfully:

\mathbb E_{\theta_{-i}\mid\theta_i} \!\left[u_i(g(\theta_i,\theta_{-i});\theta_i)\right] \geq \mathbb E_{\theta_{-i}\mid\theta_i} \!\left[u_i(g(r_i,\theta_{-i});\theta_i)\right].

Truthful reporting then constitutes a Bayesian Nash equilibrium. Unlike DSIC, the guarantee concerns expected utility under the model’s beliefs rather than every possible profile of others’ reports. DSIC implies BIC under the same utility model, but the converse need not hold. (nobelprize.org)

Example: a second-price auction

A standard illustration in auction theory is the Vickrey auction, or sealed-bid second-price auction. The highest bidder wins and pays the second-highest bid; losing bidders pay nothing. Suppose a bidder privately values the object at $v$ , and the highest competing bid is $b$ . Its utility from winning is $v-b$ , while losing yields zero. (theory.stanford.edu)

If $v>b$ , bidding $v$ wins at a beneficial price. If $v<b$ , bidding $v$ avoids an unfavorable purchase. Changing the bid without changing the outcome provides no benefit; changing the outcome moves it in an unfavorable direction. Truthful bidding is consequently weakly dominant under this private-value, quasilinear utility model. By contrast, a first-price auction charges the winner its own bid, making strategic bid shading potentially profitable. (theory.stanford.edu)

The revelation principle

The revelation principle explains why direct, truthful mechanisms occupy such a prominent place in economic theory. Under the relevant assumptions, an equilibrium outcome of an indirect mechanism can be reproduced by an incentive-compatible direct mechanism. The direct mechanism asks for types and simulates the strategies participants would have used in the original equilibrium. (ocw.mit.edu)

The principle is an analytical reduction, not a claim that every desirable outcome is achievable. A Bayesian equilibrium produces a Bayesian incentive-compatible representation; it does not automatically produce a dominant-strategy guarantee. Nor does reproducing one truthful equilibrium establish that every equilibrium of the new mechanism yields the desired outcome. (nobelprize.org)

Incentive constraints in contracts

In adverse selection problems, participants possess hidden characteristics before contracting. An incentive-compatible menu must make each type prefer its intended contract to contracts intended for other types. These restrictions can require leaving participants information rents—utility above their reservation level—to discourage imitation. They may also produce a trade-off between extracting surplus and productive efficiency. (ocw.mit.edu)

In moral hazard problems, the hidden variable is an action, such as effort. For a wage rule $w(y)$ , an intended action $a^*$ must satisfy

a^*\in\arg\max_a \left\{\mathbb E[U(w(y))\mid a]-c(a)\right\},

where $y$ is observable output and $c(a)$ is the action’s cost. This is an incentive constraint even though no truthful report is involved. When agents are risk-averse, linking compensation to uncertain output can create a trade-off between incentives and insurance. (live.ocw.mit.edu)

Mathematical characterization

In a single-parameter quasilinear model, let a participant’s value be $v$ , allocation be $x(v)$ , and payment be $p(v)$ . Truthful utility is

U(v)=vx(v)-p(v).

For types on an interval beginning at zero, incentive compatibility is characterized by a nondecreasing allocation rule and the payment identity

p(v)=vx(v)-\int_0^v x(z)\,dz-U(0).

For a dominant-strategy interpretation, these conditions apply while holding others’ reports fixed. In an independent-type Bayesian model, analogous conditions apply to interim expected allocations and payments. Thus an allocation rule and its payment rule cannot be chosen independently. (ocw.mit.edu)

More general one-dimensional screening models use the single-crossing condition and an envelope identity to simplify incentive constraints. Under suitable regularity and increasing-differences assumptions, monotonicity plus the envelope identity characterizes global incentive compatibility. A local first-order condition alone is not generally enough. (ocw.mit.edu)

Distinctions and limitations

Incentive compatibility differs from individual rationality, which requires participation to offer at least as much utility as the outside option. A mechanism may make truthfulness optimal among reports while still making nonparticipation preferable. Likewise, incentive compatibility does not by itself establish Pareto efficiency, high revenue, or an appropriate allocation. For example, allocating an object randomly without charge can be strategy-proof while ignoring who values it most. (live.ocw.mit.edu)

The guarantee is also relative to the modeled preferences, information, available actions, and rules. Bayesian constraints depend on specified beliefs. Hidden-action contracts depend on the relationship between actions and observable outcomes. A proof under one model does not establish compatibility after those assumptions change. (ocw.mit.edu)

Applications include auction design, procurement, monopoly regulation, income taxation, and the provision of public goods. In each case, incentive constraints restrict the set of feasible institutional arrangements alongside resource and participation constraints. (nobelprize.org)

Historical development

Leonid Hurwicz’s work in 1960 and 1972 helped establish mechanism design and the explicit analysis of incentive compatibility. James Mirrlees’s research on optimal income taxation demonstrated how private information and self-selection constrain economic policy. Later developments of the revelation principle made truthful direct mechanisms a standard framework for studying allocation under asymmetric information. (nobelprize.org)

Roger Myerson extended revelation-based analysis to Bayesian environments and, in 1982, to coordination systems incorporating both adverse selection and moral hazard. In 2007, Hurwicz, Eric Maskin, and Myerson received the Nobel Memorial Prize in Economic Sciences for their foundational contributions to mechanism design theory. (nobelprize.org)

References

Mechanism Design Theory: Scientific Background to the 2007 Prize in Economic Sciencesnobelprize.org
Mechanism Design Iocw.mit.edu
124 Spring 2017 Review Notes: Contracts Iocw.mit.edu
CS364A: Algorithmic Game Theory, Lecture #2: Mechanism Design Basicstheory.stanford.edu
Principal Agent Theory and Incomplete Contractslive.ocw.mit.edu
Labor Economics, Lectures 6 and 7: Moral Hazard and Applicationseconomics.mit.edu
Advanced Information: Economic Sciences Prize 1996nobelprize.org
Roger B. Myerson – Biographicalnobelprize.org