Bankruptcy Prediction with Incomplete Accounting Information

Abstract

How does a creditor’s learning from a firm’s strategic actions affect bankruptcy prediction, debt values, and optimal capital structure? We investigate a Leland (1994) setting augmented by asymmetric information on the firm’s asset value. Observing the firm’s survival of apparently distressed periods, the creditor excludes asset value estimates that are too low to be consistent with the observed survival. We show that the expected bankruptcy threshold decreases as result of the learning. While expected asset and debt values decrease upon reaching new all-time-low asset values, they are persistently higher once the observed asset value recovers to a given level, but the creditor remembers the all-time low. In terms of selecting the capital structure, high quality firms can separate and signal their quality by over-leveraging if the information asymmetry is high enough. Moderate information asymmetry implies a pooling equilibrium.

Keywords: Learning Dynamics, Strategic Interaction, Quantitative Debt Models, Signaling Game