On July 3, the CFPB published a report on its study of the use of remittance histories in credit scoring, which found that (i) remittance histories have little predictive value for credit scoring purposes, and (ii) remittance histories are unlikely to improve the credit scores of consumers who send remittance transfers. The report follows a 2011 CFPB report on remittance transfers, which was required by the Dodd-Frank Act and assessed, among other things, the feasibility of and impediments to using remittance data in credit scoring. At that time, the CFPB identified a number of potential impediments to incorporating remittance history into credit scoring, and noted the need for further research to better address the potential impact of remittance information on consumer credit scoring.
To conduct its supplemental analysis of the potential effect of including remittance histories in credit models, the CFPB collected a random set of 500,000 consumers from a single remittance transfer provider. The CFPB was able to match approximately 212,000 of those remitters to a credit record held by one of the three major consumer reporting agencies. The CFPB analyzed the remitter data set in comparison to a control set of 200,000 consumers with credit records.
First, the CFPB estimated a credit scoring model that used only credit history information to serve as a baseline estimate of the level of predictiveness that credit history information alone produces. The CFPB then evaluated how large of an increase in predictiveness results when remittance histories are added to that baseline model. The CFPB determined that including remittance histories in credit scoring models is unlikely to increase the predictiveness of the models enough to warrant generating scores for otherwise unscorable credit records. The report cautions that the CFPB’s analysis of the predictiveness of remittance histories is limited in numerous ways.
Second, the CFPB assessed the impact of remittance histories on the credit scores of three different populations: (i) remitters without credit files or who could not be matched to a credit file—the majority of the original 500,000 sample size; (ii) remitters with credit profiles insufficient to be scored (“unscoreable”); and (iii) remitters with credit scores.
The CFPB determined that for remitter sample members without credit records remittance histories likely would not allow those individuals to develop a credit profile. The CFPB hypothesizes that given the limited utility of remittance history for predicting future performance, credit model builders would not construct a credit scoring model for this population without any credit records. Similarly, although the CFPB found that although remittance transfers appear to be associated with better credit outcomes for the small segment of remitters with “unscorable” credit, model builders still would be unlikely to score these consumers because remittance histories offer little with regard to predictiveness,. Finally, the CFPB reported that for remitters who already have credit scores, remittance history information appears to drag down credit scores.
The report adds that remittance transfers also are likely to correlate with race or ethnicity, which could raise fair lending risk for credit model developments. Further, the CFPB states that the credit scoring value of remittance histories appears even less valuable compared to other potential alternative data—specifically rental and utility payment data. The bureau suggests that future efforts to enhance consumer credit scoring models should focus on activities that involve regularly scheduled payments, as opposed to voluntary payments like remittance transfers.