Working Paper: NBER ID: w26110
Authors: Jeremy Burke; Julian Jamison; Dean Karlan; Kata Mihaly; Jonathan Zinman
Abstract: How does the large market for credit score improvement products affect consumers and market efficiency? For consumers, we use a randomized encouragement design on a standard credit builder loan (CBL) and find null average effects on scores. But a generalized random forest algorithm finds important heterogeneity, most starkly with respect to baseline installment credit activity. CBLs induce delinquency on pre-existing loan obligations, suggesting that even a seemingly modest additional claim on monthly cash flows is too much for many consumers to manage. For the market, CBL take-up reveals information: takers experience future score improvements relative to non-takers, which, given null average treatment effects, implies positive selection. However, we find suggestive evidence that the CBL weakens the score’s power for predicting default in some cases. We propose simple changes, to CBL provider strategy and credit bureau reporting categories, that could produce more uniformly positive effects for both individuals and the market.
Keywords: credit builder loans; consumer behavior; credit scores; market efficiency
JEL Codes: D12; G14; G21
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| credit builder loans (CBLs) (G51) | credit scores (G51) |
| credit builder loans (CBLs) (G51) | credit scores for bottom tercile of baseline installment credit activity (G51) |
| credit builder loans (CBLs) (G51) | credit scores for top tercile of baseline installment credit activity (G51) |
| credit builder loans (CBLs) (G51) | delinquency on preexisting loan obligations (G33) |
| CBL takeup (Y20) | credit scores for CBL takers (G51) |
| credit builder loans (CBLs) (G51) | predictive power of credit scores for default (G32) |