GOOD BOOMS, BAD BOOMS

Credit booms are not rare, some end in a crisis (bad booms) whereas others do not (good booms). We document that credit booms start with an increase in productivity growth, which subsequently falls faster during bad booms. We develop a model in which a crisis happens when a credit boom transits toward an information regime with careful examination of collateral. As this examination is more valuable when collateral backs projects with low productivity, crises are more likely during booms that display larger productivity declines. We test the main predictions of the model and identify the default probability as the main component of measured productivity that lies behind crises.