Partitional Information Revelation under Renegotiation [Job Market Paper]
Consider a long-term relationship between a seller and a buyer whose valuation (for a per-period service) is private. Contracts are long-term and subject to renegotiations. In such a dynamic game, a new dimension of mechanism design, intertemporal type separation, arises as its induced belief-updating affects the rent extraction-efficiency tradeoff. For a finite horizon, I derive intuitive and natural equilibrium properties in such gradual separation of types, namely a refining and a pure strategy property, of the seller-optimal equilibrium: at each history with non-degenerate belief, the seller refines the contract by partitioning the posterior support into countably many intervals and offering a pooling contract to each of these intervals. More generally, I show that any PBE is outcome equivalent to a seller-optimal equilibrium. The set of equilibrium outcomes of any PBE of a finite-horizon converges to the set of MPE outcomes of an infinite-horizon game as the horizon goes to infinity. Lastly, I provide simple implementations.
Quality Disclosure and Price Discrimination
This paper studies the design of information and pricing in a monopolistic market with product quality dispersion and consumer valuation heterogeneity. The resulting optimal mechanism involves a partial disclosure: it bunches qualities serving consumer types neighboring a region where the virtual valuation is decreasing. Somewhat surprisingly, this mechanism is independent of the quality distribution when the seller solely provides quality information. Moreover, allowing general production cost function, quality bunching arises when the cost function does not exhibit local economies of scale. The model and its extensions explain several common phenomena in product markets: nonuniform disclosure on the quality spectrum and existence of mid-range qualities that are region exclusive. Insights are provided about how social surplus and production level vary with the restriction on the disclosure rule.
Penalty, Voting, and Collusion: a Common Agency Approach to Industrial Regulation and Political Power
This paper studies moral hazard models with a limited penalty and with a voting committee as the principal. Voting aggregates welfare-relevant information but faces corruptive incentives. Such incentives push the optimal voting rule towards unanimity, while the optimal rule depends on bribery cost and maximal penalty. Results have implications on the welfare performance of instruments like whistleblower rewards, multi-verdict penalty, and on the vote-buying behavior.
The Horizon Effect in Learning Games
This paper studies a general learning game with a continuum (payoff-relevant) state space with a two-parameter conjugate prior to the signal processes of learning. The players Bayes update the two parameters through the signals they observed. One parameter gives the variance of the posterior belief, and is named as the horizon of the game: it represents how much the players value the current experimentation. Unlike two-type prior commonly used in the literature where the horizon effect is missing, such effect gives rise to non-stationarity of the optimal stopping rule. Implications are derived on the qualitative findings in a few popular games and agency models with learning aspects.
Experimentations under Private Learning
Many economic decisions under uncertainty involve a delegation of experimentation, and such agency relationship is often susceptible to information exclusivity: learning is private as the experiment outcomes are only observed by the agent that works on the field. This paper studies a scenario where the principal is to make an irreversible investment, and she can delegate to the agent the task of evaluating such real option based on experimentation. It explores how to optimally designs the experimentation rates and compensations (including wages and a stock option when the principal chooses) to incentivize the agent to report truthfully throughout time while being able to make timely and accurate decisions based on the revealed learning. Under formal contracting, full efficiency can be achieved; under relational contracting, the optimal incentive provision involves delay of learning and efficiency loss.
The IV Method for Estimating Local Average Treatment Regime Effects (with Thai Pham)
This paper proposes the instrumental variable regime (IVR) method, a sequence of instrumental variables, to estimate the causal effects of sequential treatments. This method serves to address the problem of endogenous sequential selections of treatments, generalizing Imbens and Angrist's LATE model. Compared to existing literature, our method is novel in that it does not require any functional-form assumption, hence is robust to model mis-specifications (a main concern in treatment regime settings). The estimator is motivated and illustrated through a contextual example -- the effect of ads on purchasing behaviors when they are displayed across time.