Research

A New Lens on Anomaly Premiums with Dynamic Exits

(with Nagpurnanand Prabhala and Nitin Kumar)

Anomaly premiums are traditionally based on long-short portfolios held in entirety for a fixed 1-month horizon. We estimate premiums when individual stocks in anomaly portfolios can exit early at stock-specific times. Using machine learning techniques based on random convolutional kernels that recognize and encode time series features of a small set of daily returns, we show that with early exit, profitability, value, and momentum have robust, significant anomaly premiums of about 100 basis points per month between 1975 and 2023 in large-cap stocks with value-weights. These anomaly premiums are not seen in shuffled non-anomaly portfolios, random non-machine learned exits, or in learning methods without the random kernel encoding step. The premiums reflect strategies not spanned by simpler, familiar rules, and are not static but evolve in complex ways over time.

When Diversity Rules

(with Joseph Kalmenovitz, Abhinav Gupta, and Kairong Xiao )

Can gender diversity influence regulatory outcomes? We explore this question in the context of U.S. federal rulemaking, leveraging a novel dataset that tracks 15,000 rulemakers responsible for drafting and advancing 38,000 regulatory projects. We find that teams including women deliberate longer and extensively revise their initial drafts, yet are ultimately less likely to see their projects approved by the agency leadership and issued as final rules. To explain these patterns, we develop a simple model of rulemaking in which output gaps arise from four sources: selection (women are assigned lower-quality projects), cost (differences in the cost of exerting effort), incentive (differences in the return to effort), and evaluation (differential treatment by agency leadership). Reduced-form estimates, aided by a machine-learning C-BERT model, reveal that women are assigned to lower-quality projects and face harsher evaluations, but receive stronger incentives tied to project completion. Structural estimates confirm that disadvantages in selection, cost, and evaluation each contribute meaningfully to the observed gender gap, partially offset by females receiving stronger incentives. Our findings shed light on how female representation shapes regulatory performance and contribute to the broader debate over whether diversity enhances or hinders the effectiveness of public institutions.

Working Papers

Disagreement and Fund Flows