Seeing the Forest Through the Trees: Do Investors Underreact to Systemic Events?

Abstract

We examine whether investors efficiently incorporate the effect of financial sector shocks into the equity prices of non-financial firms. Shocks to the financial sector are complex macroeconomic events affecting many firms to varying degrees. Prices may adjust slowly in response to financial sector shocks if information processing capabilities are constrained. We test this hypothesis by examining the ability of systemic risk shocks to predict future returns of financially constrained firms. We find that in the month following the month of large systemic events, large debt constrained firms exhibit significantly lower returns than unconstrained firms with an average relative risk-adjusted return of -3.5% (-42% annualized). More generally, innovations in systemic risk exhibit significant predictability for returns of constrained firms in excess of unconstrained firms. To identify financially constrained firms we introduce a new classification method using random decision forests, a machine learning technique. This method allows us to extend the measures based on textual analysis of firm filings both in the cross-section and time series using only financial information. Overall, the results provide evidence that the market is slow to incorporate the effects of financial sector shocks into equity prices.

Keywords: financial constraints, systemic risk, machine learning, complexity

JEL Classification: G00, G01, G02, G10, G20, G30