Credit Cycles with Model Misspecification


Francesca Bastianello and Paul Fontanier. In Preparation. “Credit Cycles with Model Misspecification”.


We propose a behavioral theory of credit cycles that rests on model misspecification. Banks infer information about the underlying quality of the pool of borrowers by looking at credit volume, but use a misspecified model to do so. Their inferred beliefs then influence their current lending standards, which in turn lead to changes in aggregate credit volume and future beliefs, thus giving rise to a two-way feedback between outcomes and beliefs. We highlight three sets of results. First, following a positive shock, agents' beliefs become decoupled from fundamentals, and banks perceive the quality of the pool of borrowers to be increasing even when it is in fact decreasing. This helps rationalize the well-established fact that booms are associated with decreasing credit spreads and a deteriorating quality of funded borrowers. Second, we allow the quality of the pool of borrowers to be endogenous, and we show how the interaction of our behavioral bias with dynamic strategic substitutabilities in lending standards generates endogenous credit cycles with systematic reversals. Third, we turn to forecast errors to show that since the influence of beliefs on aggregate credit volume is state-dependent, the size of the behavioral bias is also state-dependent, and the response to positive and negative shocks is asymmetric.