Job Market Paper

""Analyzing Idiosyncratic Volatility through Machine Learning"

Ever since Ang et al (2006) found that idiosyncratic volatility has a negative impact on returns in data, the relationship between these two factors has been an ongoing controversy. The linear regression models they favor to analyze this puzzle explain about 5% of the variation in returns. This paper uses a neural network model to capture more of the variation in returns, and analyze the role that idiosyncratic volatility plays in predicting asset returns. Fascinatingly, my findings show that while idiosyncratic volatility does improve regression based forecasting, it does not add value to neural network forecasting. This suggests that volatility impacts forecasting by reflecting an important feature of common risk factors that traditional models have yet to detect. Any unique information contained in volatility is not the reason it's important to forecasting. This methodology can be applied to evaluate the information composition of other controversial variables.

Work in progress

"How will firms react, when machines come to town? The emergence of robo-analysts and their impact on firm communication strategy"

Artificial intelligence methodologies have become increasingly commonplace in firm analysis. This study investigates how firms change their communication strategies when they are being scrutinized by machines, which will not respond to the same factors as human analysts. We expect to see greater similarity among firms in their communication style as they put less focus on embellishment and other forms of rhetorical techniques, owing to these elements not impacting a robo-analyst's evaluation of a firm. We also expect to see information releases to become formatted in more formulaic, easily coded ways to more clearly communicate what they want the machine to understand.