Job Market Paper I:

Abstract: 

We study intertemporal decision making problems with endogenously preferences. A preference state is defined, which affects the decision maker's instantaneous utility. Each period this state is, in turn, determined by past actions chosen by the decision maker. We formulate the dynamic process as a simple closed-loop feedback system: current state as an input is fed into the decision maker's preference, which determines the action chosen in the current period, this then produces the new state and starts a new round of looping. In particular, we examine positive feedback systems in which instantaneous optimization always drives the state further and further away from the initial state. Outcomes are investigated according to the decision maker's different levels of self-knowledge of his own endogenous preference. For welfare comparisons, we use the initial preference, and show that under a mild sufficient condition a sophisticated person is always better off than a naive person. However the payoff is in general not monotonic in the level of self-knowledge.

==============================================================================================

Job Market Paper II:

Abstract:

We provide a general model of two-person finitely repeated game where players' preference states are cultivated over the history of play and the payoff functions are endogenously determined by these states which the players may or may not be self-aware of. The general model is then applied to internet auctions in which sniping (waiting to the last minute to bid) is frequently observed, while standard auction theory fails to provide an explanation. It has been argued in the literature that early bids tend to bring about a “bidding war” among bidders; hence sniping arises as an optimal behavior due to this concern. We show that a necessary assumption for this argument is that bidders' preference is endogenously changed by early bids; therefore, the framework we developed is called for. The equilibrium prediction confirms the optimality of sniping in online auctions.

==============================================================================================

Working Paper:

A Model of Job Market Signaling with Discounting

Abstract: The paper studies an extension of the traditional Spence job market signaling game. Workers differ in two dimensions -- discount factor as well as productivity and education cost -- both of which are privately known only to the workers themselves. When the second dimension of idiosyncrasy is introduced, single crossing property between high and low productivities can be violated. In general, multiple sequential equilibria exist even after Cho and Kreps' intuitive criterion is used for refinement. It is shown that if Grossman and Perry's perfect sequential equilibrium concept is applied, an "essentially" unique equilibrium arises when it exists. Roughly, when discount factors do not differ too much, the unique equilibrium is still the least costly separating equilibrium for the high productivity types. But an "essentially" unique pooling equilibrium is selected by the refinement when discount factors are relatively far apart. The result shows the equivalence of outcomes of signaling and screening games in two dimensional information case.