Behind every loan approval, there stands great information about a client

Financial inclusion has been of central interest for policy makers and researchers. However, we know less about the incentives of private sector participants for advancing financial inclusion. Think for instance of commercial banks. When deciding whether to lend to a new borrower, banks consider factors such as screening costs, the capacity of the borrower to repay a loan, or the time banks expect the borrower to remain a client. This final consideration is particularly important for new borrowers, as the first lender will incur the cost of establishing their reliability.

Economists have typically framed this issue based on whether a borrower will eventually switch to a competing lender (e.g., Petersen and Rajan 1995), but in principle the switch to a different lender could occur even before the first loan is issued. If competing lenders are more likely to approve borrowers who are already approved by other lenders, then the first lender that incurs the cost of screening new borrowers may not reap the resulting benefits. That is, if lenders free ride on the screening efforts of their competitors, the incentives to be the first lender to screen a new borrower (and to advance financial inclusion) are reduced. In such cases, policy intervention can be a way forward. 

In a recent paper, we find empirical evidence that free riding in loan approvals does indeed occur. We worked with a large Peruvian bank that was interested in expanding credit access to small and medium-size enterprises (SMEs). Our partner bank conducted a pilot to test a new screening technology to determine which SMEs to lend to based on a scoring rule with a strict threshold. Borrowers above the threshold were automatically granted a loan, whereas borrowers below the threshold were offered a loan only if a loan officer deemed it appropriate. During the pilot, 1,883 SMEs applied for a working capital loan with our partner bank. Of these, 366 were considered thin-file applicants (with little to no prior credit history) at the time of their application, who are particularly difficult to screen due to the lack of information on them. Exploiting the scoring rule threshold along with credit bureau data from Equifax Peru on SME loans from regulated financial institutions, we document several findings.

While thin-file applicants who scored above the threshold were more likely to receive a loan than those who scored below it, three-quarters of the additional loans were issued by competing financial institutions rather than our partner bank. Importantly, most of these borrowers never took even a single loan from our partner bank. Because the only differences between borrowers on either side of the threshold were whether they were approved for a loan from our partner bank and the resulting loan terms, this is evidence of free riding in loan approvals (figure 1). In the paper, we also show that free riding in loan approvals is higher in regions where our partner bank faces more competition. The pilot test led to higher profits for competing financial institutions but not our partner bank.

Note: Plots generated using the “rdplot” Stata command (Calonico, Cattaneo, and Titiunik 2014) for a bandwidth of 20 around the threshold, with a global polynomial of order one and 95 percent confidence intervals for each bin. EFL Score = Continuous score of the new screening technology; FI = financial institution.          

What mechanisms may be behind the free riding in loan approvals we observe?

• On the supply side, other lenders may use the loan approvals of our partner bank to update their own loan approvals. This may be the case if borrowers share their loan approval documents with competing lenders.
• On the demand side, borrowers who received a loan approval from our partner bank may have updated their beliefs about their own credit worthiness and redoubled their shopping around efforts.

And what mechanisms can we rule out?

• On the supply side, we can exclude any mechanism that operates through the credit registry. Our findings are based on loan approvals, which are not recorded in the Peruvian credit registry.
• On the demand side, we can rule out complementarities in borrowing, whereby an initial loan from our partner bank increases demand for credit from other lenders. In the data, very few borrowers in our sample who received loans from competing lenders first borrowed from our partner bank.

Taken together, our findings paint a stark picture. Although our partner bank incurred the costs of the novel screening technology, the benefits accrued largely to its competitors. The straightforward implication is that banks may underinvest in expanding credit to underserved borrowers, as doing so entails a private cost but produces a public good.  This underinvestment may justify subsidies to private sector efforts to expand financial inclusion.

References

Calonico, Sebastian, Matias D. Cattaneo, and Rocio Titiunik. 2014. “Robust Data-Driven Inference in the Regression-Discontinuity Design.” Stata Journal 14 (4): 909–46.

Petersen, Mitchel A., and Raghuram G. Rajan. 1995. “The Effect of Credit Market Competition on Lending Relationships.” Quarterly Journal of Economics 2 (110): 407–43.