This paper studies why non-financial firms invest in risky financial assets. Within a dynamic corporate finance model with macroeconomic fluctuations, I show that firms can use risky financial assets to transfer liquidity from states with low aggregate investment opportunities to states with high aggregate investment opportunities. Specifically, when investment funding demand is pro-cyclical and external financing is costly, risky financial assets with pro-cyclical returns can increase rm value by improving the
match between internal cash flows and investment opportunities. Therefore, investing in risky nancial assets can be naturally optimal for the firm from a macro perspective. Based on U.S. firm data scraped from the SEC 10-K filings using a machine learning algorithm, I find empirical evidence consistent with this mechanism: (1) time-serially, the value of risky nancial assets is positively correlated with the corporate investment rate; (2) cross-sectionally, firms with pro-cyclical investment funding demand in excess of profits hold more risky financial assets. The empirical results are robust to adding variables to control for the risk-seeking, poor corporate governance and CEO overconfidence channels. 

Keywords: Business Cycles, Investment/Budgeting, Risk Management, Risky Financial Assets, Machine Learning, Financing Frictions, Cash Holdings, Portfolio Choice​ 

JEL Classifications: E32, G11, G32