Working Papers

1. Entity Neutering [SSRN] [Joseph Engelberg, Asaf Manela, William Mullins, and Luka Vulicevic]
   • Abstract: Cutting-edge LLMs are trained on recent data, creating a concern about look-ahead bias. We propose a simple solution called entity neutering: using the LLM to find and remove all identifying information from text. In a sample of one million financial news articles, we verify that, after neutering, ChatGPT and other LLMs cannot recognize the firm or the time period for about 90% of the articles.  Among these articles, the sentiment extracted from the raw text and the neutered text agree 90% of the time and have similar return predictability, with the difference providing an upper bound on look-ahead bias.  The evidence here suggests that LLMs are able to effectively neuter text while maintaining semantic content.  For look-ahead bias, LLMs can be both the problem and the solution.
2. Bayesian Pockets of Cross-Sectional Predictability [Upon Request] [Luka Vulicevic]
   • Abstract: I show the predictive ability of factors is concentrated in finite-lived periods ("pockets''). To identify these pockets, I develop a Bayesian panel regression framework that estimates the state-dependent risk premia. Empirically, I find that pockets exhibit a 1.2% monthly risk premium on average, while factors outside of these pockets ("out-of-pocket'') earn a negative risk premia of -1.65% per month. Pockets typically last four years, while out-of-pocket periods are on average three to four years. I find that state transitions occur during episodes of high volatility and I document which factors work best during recessions which work best during expansions. These findings motivate a novel factor classification that distinguishes consistent from transient factors.

‍

Inactive Papers

1. Arbitrage Between Retail and Secondary Markets, Time Will Tell [SSRN] [Luka Vuliceivc and Mitchell Riddell]
   • Abstract: Arbitrage between retail and secondary markets is a recent economic phenomenon that emerged prior to the COVID-19 pandemic. Despite the lack of explanatory theories, we contribute to the literature by examining the relationship between product characteristics and arbitrage returns. Using cross-sectional data on wristwatches, we develop a predictive model to estimate the arbitrage return at a watch's release. Our findings demonstrate significant predictors from watch-specific attributes, even after considering non-contemporaneous economic effects. Our model offers valuable insights into the predictability of arbitrage returns, benefiting speculators, watch retailers, and researchers.

‍

‍

‍