Working Paper: NBER ID: w13873
Authors: Li Gan; Roberto Mosquera
Abstract: This paper proposes an econometric model to identify unobserved consumer types in the credit market. Consumers choose different amounts of loan because of differences in their time or risk preferences (types). Thus, the unconditional probability of default is modeled using a mixture density combining a type-conditioning default variable with a type-determining random variable. The model is estimated using individual-level consumer credit card information. The parameter estimates and statistical tests support this kind of specification. Furthermore, the model produces better out-of-sample predictions on the probability of default than traditional models; hence, it provides evidence of the existence of types in the consumer credit market.
Keywords: credit market; unobserved consumer types; probability of default; econometric model
JEL Codes: C81; D12
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| unobserved consumer types (D11) | probability of default (G33) |
| time discount rates and risk preferences (D15) | unobserved consumer types (D11) |
| unobserved consumer types (D11) | borrowing decisions (G51) |
| borrowing decisions (G51) | probability of default (G33) |
| unconditional probability of default (G33) | mixture of conditional probabilities (C11) |
| inclusion of types (C46) | out-of-sample predictions of default probabilities (C52) |
| older individuals (J14) | responsible type (Y90) |
| gender (J16) | type classification (P50) |