A Quantitative Theory of Credit Scoring

Starting in the early 1990s credit scoring became widespread and central in credit granting decisions. Credit scores are scalar representations of default risk. They are used, in turn, to price credit, and as a result alter household borrowing and default decisions. We build on recent work on defaultable consumer credit under asymmetric information to develop a quantitative theory of credit scores. We construct and solve a rich and quantitatively-disciplined lifecycle model of consumption in which households have access to defaultable debt, and lenders are asymmetrically informed about household characteristics relevant to predicting default. We then allow lenders to keep record of inferences on the hidden type of a borrower, as well as a binary 'flag' indicating a past default. These inferences arise endogenously from a signalling game induced by borrowers' need to obtain loans. We show how lenders’ inferences evolve over the lifecycle as a function of household behavior in a way that can be naturally interpreted as 'credit scores.' In particular, we first show that lenders' assessments that a household has relatively low default risk matter significantly for the interest rates households pay. We then show that such assessments rise most sharply an d interest rates paid by borrowers fall most sharply (on the order of 5-6 percentage points) when the bankruptcy flag is removed, consistent with work of Musto (2005). Lastly, we compare allocations across information regimes to provide a measure of the social value of credit scores, and the dependence of these measures on lenders' ability to observe borrower characteristics.