Alfonso Maselli

We study a decision maker who approaches an uncertain decision problem by formulating a set of plausible probabilistic models of the environment but is aware that these models are only stylized and incomplete approximations. The agent is effectively facing two layers of uncertainty. Not only is the decision maker uncertain regarding what model in this set has the best fit (model ambiguity), but she is also concerned that the best-fit model itself might be a poor description of the environment (model misspecification). We develop an axiomatic foundation for a general class of preferences that capture concern toward these two layers of uncertainty and allow us to compare individuals’ degrees of aversion to model misspecification and model ambiguity independently of each other.

We study a decision maker who employs a set of probabilistic models to solve an infinite-horizon problem under uncertainty and learns over time. The decision maker faces two layers of uncertainty: model ambiguity, as she does not know which model has the best fit, and model misspecification, as she is concerned that none of the hypothesized models accurately describes the environment. We provide an axiomatic foundation for forward-looking preferences that exhibit flexible attitudes toward model misspecification and model ambiguity. In a central specification of our model, we propose a new behavioral axiom that justifies updating beliefs over the set of models via Bayes' rule even when concerns about model misspecification and model ambiguity are present.

We study updating in decision-making under model misspecification. The lack of separation between tastes and beliefs leads us to study ``optimal beliefs.’' Since they incorporate misspecification concerns, they are taste-laden and no longer only reflect information. We show that they evolve according to an optimal updating rule and that under them the decision maker behaves in a traditional expected utility fashion. When concerns about misspecification are present but misplaced, the decision maker will eventually learn the correct model. When, instead, misspecification is present, optimal beliefs concentrate asymptotically on models that best approximate the correct one according to a long-run entropic measure of fit.

This paper studies the determinants of entry in centralized versus decentralized over-the-counter (OTC) secondary markets. We develop a model of asymmetric information in the lending market in which borrowers have access to two costly signals. Creditworthy borrowers signal their type by liquidating non-pledgeable assets in a centralized market or exchanging them for collateralizable assets in an OTC market. Equilibrium prices and haircuts determine signaling costs endogenously. In the optimal separating contract, the cheapest market in terms of signaling costs is accessed. We establish conditions for existence of equilibria in which different markets are accessed - CM-only, OTC-only, and dual-market - and rank them by the utility they provide to borrowers. We show that OTC-only equilibria offer the highest utility, followed by dual-market and CM-only equilibria. 

Information Acquisition under Model Uncertainty, with Simone Cerreia-Vioglio and Massimo Marinacci