Assistant Professor of Economics, University of Pennsylvania
Department of Economics
University of Pennsylvania
3718 Locust Walk
Philadelphia, PA 19104
Curriculum Vitae pdf
Hidden Rust Models pdf, supplement pdf
Hidden Rust models answer the need for convenient models of unobserved dynamics in a dynamic discrete choice context. They can be thought of as classical Rust models with an additional unobserved state variable. As opposed to their classical counterparts, hidden Rust models are not Markovian. I study their econometrics in terms of: identification, time-series asymptotics, practical estimation. I illustrate their use in a model of dynamic financial incentives inspired by Duflo, Hanna and Ryan (2012). Several lessons carry over to other structural models with unobserved dynamics.
A Weakly Dependent Bernstein–von Mises Theorem pdf
The Bernstein–von Mises theorem is the central result of classical asymptotic theory applied to Bayesian methods. I isolate a convenient set of assumptions and prove a new weakly dependent Bernstein–von Mises theorem along the lines of Le Cam (1986). This new theorem is valid under few assumptions beyond local asymptotic normality. An application in a microeconomic model of dynamic choice can be found in my job market paper.
Research Papers in Progress
A Model of Strategic Retirement for Supreme Court Justices
Joint with Charles Cameron, Professor of Politics and Public Affairs, Princeton University
Supreme Court justices nearing the end of their mandate may face conflicting incentives before presidential elections. By retiring early, they can make sure that the current majority will nominate their successor, but have to renounce an often enjoyable and remunerative occupation. Building on a newly collected data set as well as a state-of-the-art dynamic structural choice model, this paper examines the question whether the strategic dimension is a major incentive force in the justices’ retirement decision.
Consistency of Maximum Likelihood in Dynamic Economic Models: The Entropy Theorem Beyond Markov Dominating Measures
This paper develops a generic argument for consistency of the maximum likelihood estimator in dynamic economic models. In the independent and identically distributed context, the familiar argument in terms of minimum Kullback–Leibler divergence is a consequence of the entropy theorem of information theory. It is known since Barron (1985) that the entropy theorem remains valid for data having joint marginal densities with respect to a Markov dominating measure. I extend the entropy theorem to more general dynamic settings and examine the consequences for several non-Markovian dynamic economic models, such as hidden Rust models (as in my job market paper) and more general dynamic game models.