Financial Institutions

Measuring the welfare impact of new product introductions is a long-standing challenge for economists. In this study, we make progress on this problem by leveraging the informational efficiency of equity markets and a scalable consumer demand model. We construct a novel database of new product announcements covering 20 years ( 2002 - 2021 ) and use stock market reactions to estimate the profits that these new products generate for the inventor firms.

This paper analyzes the AI infrastructure boom as a physical capital buildout centered on data centers, power infrastructure, cooling systems, and specialized chips. It studies how this buildout is financed through hyperscalers, third-party developers, REITs, private credit, and structured finance, and discusses the implications for leverage, risk allocation, and financial stability.

Working from home resulted in a sharp contraction in office demand. We built a valuation model to find that the office stock lost about 45% in value. More for low-quality buildings and in cities with a larger IT sector and less for trophy buildings. We discuss the implications for mortgage lenders and the vitality of cities.

There has been growing interest in applying reinforcement learning (RL) to inventory management, either by optimizing over temporal transitions or by learning directly from full historical demand trajectories. This contrasts sharply with classical data-driven approaches, which first estimate demand distributions from past data and then compute well-structured optimal policies via dynamic programming.

Almost one-third of U.S. commercial mortgage dollars sits on regional bank balance sheets. Recent commercial property revaluations have sparked concerns that this substantial exposure may create fractures in the banking system and spill over to the wider economy. To assess commercial real estate (CRE) risks in regional banks, we construct a novel loan-level dataset from county records. While many regional banks have benefited from exposure to better-performing markets thus far, reported delinquencies understate risks from undercollateralized loans by a factor of four.

In the Newsvendor problem, the goal is to guess the number that will be drawn from some distribution, with asymmetric consequences for guessing too high vs. too low. In the data-driven version, the distribution is unknown, and one must work with samples from the distribution. Data-driven Newsvendor has been studied under many variants: additive vs. multiplicative regret, high probability vs. expectation bounds, and different distribution classes. This paper studies all combinations of these variants, filling in many gaps in the literature and simplifying many proofs.