Multiple selves and endogenous beliefs

This dissertation consists of three essays. In the first essay I propose a new theoretical model of decision making under risk and uncertainty, titled "Affective Decision Making" (ADM). That is, I model the decision maker as having two inner processes: the "rational account" and the "mental account". The rational account chooses an action (insurance), given perceived risk (a perceived probability distribution). The mental account chooses perceived risk, given an action, to maximize expected utility net of mental costs. This captures the desire to hold the most favorable beliefs that one can justify. The agent's decision is a consequence of the interaction of the two accounts. This interaction is modeled as a simultaneous move intrapersonal game. The intrapersonal game is a potential game where the potential can be interpreted as the "composite" agent's utility function. The agent's choice is a pure strategy Nash equilibria of this game, reflecting consistency between the two accounts and the resolution of cognitive dissonance. The model gives rise to, generally, multiplicity of equilibria, which can be interpreted as framing (attentional) effects or uncertainty. For the insurance markets, the model allows for negative correlation between loss size and insurance level, and shows that the absolute risk aversion property can not be concluded from the data.; The second essay proves that given bounds on risk belief, the ADM model is testable. That is, one can find a finite data set that refutes the model as well as one that is consistent with it. In addition, I show the existence of a mutual insurance equilibrium with affective agent and prove optimal sharing of perceived risk.; In the third essay, I consider a traditional screening model with a monopolistic firm facing two types of agents. The agents are Bayesian and observe a noisy private signal about their type. Consequently, the agents choose according to their perceived, rather than true, type. I show that the optimal endogenous information structure for the monopolist is always one of the corner solutions: either completely informed or completely uninformed agents.