Credit Card Redlining Revisited

Using a proprietary dataset of credit bureau records, Cohen-Cole (2008) finds that banks set credit limits on revolving accounts based in part on the racial composition of the neighborhood in which each borrower resides. This paper evaluates the evidence presented in that working paper using the same proprietary database of credit bureau records. The replication effort presented in this paper suggests that decisions about how to calculate the variables used in that study may have resulted in the unnecessary exclusion of one-fifth of available observations from the estimation samples and may have increased the size of the reported effect by over 25 percent. Furthermore, this analysis suggests that when a control for neighborhood income is added to the estimations, the results presented as evidence of redlining activities disappear.