I am an assistant professor at California State University, Fullerton and a Project PIPE-LINE data science faculty fellow. My research interests are macroeconomics, firm dynamics, and regulatory policy. Additionally, I am actively developing techinuqes to apply machine learning and big data to macroeconomic questions. When I am not wading through data, I enjoy cooking, gardening, and snowboarding.
Antitrust Policy and Economic Growth
Abstract: It has become increasingly apparent to policymakers that optimal antitrust policy re- quires looking beyond traditional static analyses and considering the dynamic effects of policy. Such analysis is challenging as limited studies exist concerning dynamic compe- tition policy. This paper attempts to bridge this knowledge gap by developing a novel structural growth model containing the major motivations of mergers and acquisitions (M&A) activity. To enable estimation of the model, frontier natural language processing (NLP) techniques are employed to classify whether parties to an M&A transaction are currently operating in similar markets or whether acquirers are using M&A as an entry mechanism into new markets. Examining the overall impact of M&A on growth reveals a double-edged sword: policies that either completely shut down M&A or allow unre- stricted M&A both result in significantly lower growth rates than the baseline estimate. This motivates an optimal antitrust policy that accounts for dynamic effects.
Using Machine Learning to Classify M&A Transaction
Abstract: This paper develops a novel methodology for classifying relationships between parties in a merger and acquisition transaction. Understanding the motives of an M&A transaction is essential in researching the economic impact of M&A. Given the high number of transactions per year, manually classifying every transaction is unfeasible. This paper proposes a novel methodology using a large-language model to determine if a transaction has possible horizontal or vertical linkages. With sufficient information, this model is highly accurate. Focusing on transactions involving U.S. firms, 38% of transactions were only horizontally linked, 38% were only vertically linked, 12% were both horizontally and vertically linked, and 10% had no linkages. This pattern is robust across time and presidential administrations. The resulting data can be used by any researcher studying M&A transactions.
Using Operator Learning To Solve Hetergenous Agent Macroeconomic Models with Aggregate Uncertainity