When Information and Incentives Shape Each Other

Designing rules around behavioral responses

Mechanism design studies how economic rules can induce desired outcomes when participants have their own interests and private knowledge. Tian’s preprint, A General Theory of Dynamic Mechanism Design with Endogenous Information and Hidden Actions, brings several connected problems into one framework: eliciting truthful information, motivating actions that cannot be directly observed, and deciding what evidence to disclose.

The central argument is that these problems generally cannot be optimized separately. An auction format changes the value of learning about a rival. A contract’s disclosure policy changes the rewards for producing evidence. Those choices then affect future information, payments and behavior.

Tian models a principal who commits in advance to rules governing allocations, transfers and disclosure. Participants repeatedly observe private information, send messages and choose actions. Actual actions—not merely recommended ones—help determine subsequent economic conditions and information.

Truthfulness is not enough

A key contribution is an equilibrium concept called joint perfect correlated Bayesian equilibrium. It requires participants to choose optimal reports and actions while imposing consistency on beliefs across histories, including feasible histories that equilibrium play never reaches. It does not make hidden actions observable.

Under explicit regularity and information-preservation conditions, Tian establishes a dynamic revelation principle: outcomes of an indirect mechanism can be represented through truthful reports of private histories and contingent action plans. This is not an unconditional claim that reporting only a current valuation or productivity measure is sufficient.

The distinction matters because someone who lies can also change subsequent behavior. A contract must therefore withstand coordinated deviations in reporting and action, not just false reports followed by obedient implementation.

In the private-values analysis, Tian characterizes incentives using four requirements: optimal private choices, a marginal reporting condition, a global condition ruling out profitable current-type misreports, and an additional condition covering other private-history misreports. Local first-order calculations alone do not establish full incentive compatibility.

Three different economic costs

The paper derives an accounting identity separating the principal’s expected payoff into total economic surplus minus participants’ information-related rents and lowest-type continuation rents. The identity permits correlated initial private types and signals under its stated distributional and regularity assumptions.

Tian distinguishes these rents from two other costs. Hidden-action implementation costs arise because some desirable choices cannot be induced. Direct information-production expenditures consume real resources and already enter economic surplus.

Comparing four contracting environments—neither friction, private information alone, hidden action alone, and both together—produces a branching hierarchy. Removing either friction cannot reduce the principal’s best attainable payoff, but neither single-friction environment generally dominates the other. The combined losses need not equal the sum of separately measured losses; their interaction can have either sign.

An auction ranking can reverse

One constructed example uses two bidders whose common object value is the sum of two independent private types, each uniformly distributed between zero and one. Acquiring information about the other bidder’s type is free, and public disclosure is uninformative.

In Tian’s specified second-price equilibrium, neither bidder acquires the additional information. Expected seller revenue is two-thirds. In the specified first-price equilibrium, both acquire it and bid the realized common value, producing expected revenue of one.

The difference is one-third despite identical expected allocated value. Tian explicitly limits the conclusion to these equilibria: it neither overturns the classical Milgrom–Weber theorem under its assumptions nor establishes that first-price auctions always earn more. Costless acquisition and the selected equilibrium strategies are important features of the construction.

Transparency depends on the environment

Disclosure is constrained by available evidence. A mechanism can selectively process or withhold information that its technology produces; it cannot freely manufacture any desired informative signal.

In a quadratic contracting benchmark, Tian finds full disclosure of available evidence uniquely optimal under the specified assumptions, even when hidden information-producing effort cannot be induced. Revealing the available signal is not equivalent to revealing the underlying state.

A separate financial-contracting example yields no disclosure, partial disclosure or full disclosure depending on project value and disclosure costs. Fees and evidence-contingent bonuses jointly support truthful reporting and hidden production. That optimum applies within the specified bonus–fee family, with collectible fees and unrestricted transfers—not to every possible financial contract.

These are mathematical results and constructed examples, not empirical estimates or tested policy interventions. The framework maintains full commitment, monetary transfers and one-dimensional screening types; particular results add differentiability, independence, concavity or other restrictions. The preprint’s broader lesson is conditional but clear: evaluating institutional rules requires accounting for the information and behavior those rules induce.