Expectations and Credit Slumps

Abstract
Since the 2008-09 financial crisis, U.S. bank lending has been slow to recover, despite the period of very low interest rates. We show that banks do not process information efficiently, and this is a quantitatively important explanation for credit slumps after 2008. Using a new dataset of bank expectations, we find that banks over-extrapolate the past, and their lending decisions are sensitive to beliefs. The behavioral bias matters more for large banks, whose loan portfolios are more sensitive to beliefs. To quantify the economic significance of imperfect expectations, we build a dynamic model with heterogeneous banks that are over-extrapolative and face a small risk of economic disaster. We show that a realistic degree of over-extrapolation estimated from the data generates the pace of credit and real recovery observed after the crisis. Banks' distorted beliefs hamper the effectiveness of policy tools, such as quantitative easing (QE) programs. Joint paper with Antonio Falato (Fed Board).