Essays in Mechanism Design and Market Design.

摘要

'Full Implementation and Belief Restrictions' considers how information about agents' beliefs might be used to achieve full implementation, which aims to resolve the problem of multiplicity in mechanism design. We find that minimal knowledge about beliefs (described by moment conditions) can be used to reduce strategic externalities induced by the incentive compatible transfers by adding belief-based adjustments to the transfers. When strategic externalities are reduced to the extent that the best reply map becomes contractive, then uniqueness is achieved for a Delta-Rationalizability. We further show that (1) this result often obtains by very little information about agents' beliefs, therefore the uniqueness result holds for a large class of beliefs and (2) suitable moment conditions can be found in many economically interesting information structures, for example in quadratic smooth environments with independent or affiliated types.;'Shared Information Sources in Exchanges' explores implications of heterogeneous information sources available to market participants---due to regulation, choice or comes as a constraint. Traders in financial markets recognize that shared forecast services, differential access to information technology, targeted advertisement induce correlation in inference errors. We show that common information sources, seen as a departure from the private information acquisition assumption, qualitatively affect information aggregation and efficiency properties of markets. Even when traders' values are independent, inference from prices can be useful for learning about valuations. From a market design perspective, we show that imposing differential access to sources can improve informativeness, restricting participation to certain trading venues can be optimal.;'Privacy-Preserving Market Design' is motivated by the increasing concern about revealing information on past trades, income, liquidity needs. With improved data collection, preserving privacy has become a de facto participation constraint in exchanges. We suggest an incentive-based approach by formulating a mechanism design problem to study the joint design of the allocation rule, bidding language, observable outcomes (prices, quantities at various levels of aggregation, and other statistics). We show that privacy-preserving market design is feasible, in that the publicly observable outcome is minimally informative about private information. In contrast to the view in the literature, there need not be a trade-off between privacy preservation and efficiency.