Can Machines Learn Capital Structure Dynamics?
59 Pages Posted: 22 Oct 2019 Last revised: 30 Oct 2020
Date Written: October 29, 2020
Yes, they can! Machine learning models that exploit big data identify leverage determinants and predict leverage better than classical methods. By allowing for nonlinearities and complex interactions, machine learning boosts the out-of-sample R-squared from 36% to 56% over linear methods such as LASSO. The best performing model (random forests) selects market-to-book, industry median leverage, cash & equivalents, Z-Score, profitability, stock returns, and firm size as reliable predictors of market leverage. Improved target measurement through machine learning yields 10%-34% faster adjustment relative to LASSO. Machine learning identifies uncertainty, cash flow, and macroeconomic considerations among primary drivers of leverage adjustments.
Keywords: Machine Learning, Target Leverage, Speed of Leverage Adjustment
JEL Classification: G0, G17, G30, G32, C10, C50
Suggested Citation: Suggested Citation