Essays on Financial Institutions

Abstract

My dissertation aims to understand the economic determinants of the forbearance behavior of financial institutions and their cross section of equity returns. It contains three chapters. Chapter One shows that higher capital requirements create a regulatory arbitrage incentive for banks to forbear on loans suboptimally. I develop a dynamic bank model with a capital requirement, where a bank can roll over bad loans without reducing their face value. When the capital constraint binds, banks hold excess non-performing loans (NPLs) and reduce the credit supply. I solve the model globally with occasionally binding capital constraints and calibrate the model to the pre-crisis banking sector in both the US and Italy. The model quantitatively explains about two-thirds of the difference in NPL ratios in the two countries following a simulated recession. I provide direct causal evidence of the effects of the capital constraint channel on banks’ NPL holdings using the Euro Area crises, supporting the predictions the model generates. Chapter Two studies the information externality of banks’ forbearance behavior in a sequential game with incomplete information. Follower banks observe less liquidation in the market due to leader’s forbearance and take it as a false positive signal of the aggregate state, leading to more forbearance and zombie firms. This chapter shows that the size of the externality decreases with the prior belief of the aggregate state of the economy being good. In other words, my model predicts a higher probability of bank herding in suboptimal forbearance during bad times. Chapter Three constructs a dynamic disaster model with implicit government guarantee to explain the hump shape relation between bank size and stock returns. The model shows two opposing effects on the bank expected returns. Lower cost of debt induces more risk shifting behavior of larger banks while the safety net effect provides insurance to equity investors during market downturns. A size threshold increasing with disaster probability determines which effect dominates, thus contributing to the hump shape relation.