It is human nature to learn and to cooperate.

When do learning and cooperation lead to socially desirable outcomes?

When they don't, what difference can public policy make?

I study firms' belief formation in a new market by examining private electric utilities' beliefs about the future sulfur dioxide allowance price under the Acid Rain Program, the first large-scale cap-and-trade program.  Beliefs have important implications for implementing cap-and-trade programs under utility regulation, and change their efficiency relative to taxes.

R&D in Low-Carbon Technologies and the Value of Carbon Capture, Utilization and Storage

I study optimal dynamic allocation of R&D efforts among low-carbon technologies, allowing for structural uncertainty, technology interdependency, and investment irreversibility. I then use this framework to discuss the value of carbon capture, utilization and storage and its policy implications.

Evaluating the Paris Agreement’s Mitigation Pledges with Statistical Forecasting Models

with Joseph Aldy and William Pizer

We develop forecasting tools, drawing from the economics and machine learning literatures, to produce country-specific emission forecasts to enable an assessment and comparison of expected mitigation effort by nearly every country participating in the Paris Agreement. 

Pledge and Agree: The Path to Cooperation

with Antonio Alonso Arechar and Richard Zeckhauser

We augment a generic public good provision game with a pledge-and-agree structure, and study how this structure facilitates cooperation using theory and experiments.

On Identification in Infinitely Repeated Games

with Jose Miguel Abito, Joao Granja, and Arkadiusz Szydlowski

We derive a sharp identified set for structural parameters of infinitely repeated games without assumptions on equilibrium selection nor additional equilibrium restrictions beyond subgame perfect Nash equilibrium. We then investigate consequences of imposing equilibrium selection assumptions using experimental data.

Game theoretic models in structural industrial organization have assumed that firms observe each other's behavior perfectly, and yet important problems such as collusion feature imperfect monitoring. I use incentive compatibility constraints, in which continuation payoffs come from the limit equilibrium payoff set, to deliver bounds on the structural parameters.