Share

Policy Research, Working Paper 5438, Finance and Private Sector Development Team, Development Research Group, The World Bank, Washington, DC. Giné, Xavier, Jessica Goldberg, and Dean Yang (2010)

A randomized control evaluation in Malawi finds that using a fingerprint technology to identify customers will increase repayment rates and cause low-quality lenders to decrease loan size, suggesting that reducing information asymmetry will increase incentive to repay loans.

Obtaining and extending credit in developing countries is known to be difficult, particularly in the case of smallholder agriculture where infrequent cash flows and shocks to production are the norm. The process is rendered more difficult if punishment for defaulting cannot be enforced. Different information about the riskiness of the borrower can be exploited in this case, either by higher-risk-of default borrowers taking out larger loans (adverse selection) or by diverting resources away from the agricultural production they are intended for (moral hazard). In a context where information about borrowers is not tracked, a method of identification is introduced and evaluated for its impact on loan repayment.  A fingerprint is a unique identification, cannot be stolen or lost, and serves as a direct tie between the borrower and their loan repayment obligation.

A randomized control evaluation in Malawi studies the repayment rates of a sample of smallholder farmers seeking credit to finance production of a cash crop, in this case paprika. Half of the sample (the treatment group) is randomly chosen for fingerprinting if their loan is approved and, to create an empirical counterfactual, the remainder of the sample has no knowledge of the fingerprinting process (the control group). The incentive of being denied access to credit is dynamic in that punishment will affect future loan requests, and fingerprinting ensures that the threat is credible. Baseline and follow-up surveys provide data for a sample of 1,147 loan-approved farmers belonging to either treatment or control groups.

Characteristics of the treatment and control group prove to be statistically equivalent. Additionally, the rate of loan approval and take-up are similar across groups. Where the treatment group differs, however, is where the study’s main implications are found.

First, the loan repayment rate is found to be higher for the treatment group than for the control. Further examination of this results shows that the effect is entirely concentrated in the quintile of borrowers identified by the baseline study as having the highest default risk. In other words, the incentive improves the repayment rate, but only for borrowers expected ex-ante to be least likely to repay. The effect is demonstrated by lower outstanding balances, a higher fraction paid, and an increased likelihood of paying on time.

Secondly, smaller loan sizes are found for the quintile of high ex-ante default risk borrowers. Likewise for repayment rates, no impact is found on the loan size of low-default-risk borrowers. This result is consistent with the model’s prediction that ‘bad borrowers’ will react most when a dynamic incentive is imposed. It is interpreted as evidence that fingerprinting reduces ex ante adverse selection, as is the strong positive correlation found between loan size and default in the treatment group.

Finally, since the loan agreements specify that only paprika may be seized by the lender upon default, moral hazard would manifest itself as lower land allocation to paprika and lower use of agricultural inputs. In other words, actions that would cause loan default. The highest default-risk subgroup is found to increase their land allocated to paprika and their use of agricultural inputs, thus providing evidence of a decrease in moral hazard.

By reducing information asymmetries, fingerprinting technology can improve incentives to repay loans in a rural credit market.  A conservative cost-benefit analysis shows that a finger-print ID system is beneficial for micro credit institutions. The results suggest that, first, a more efficient credit-extending process will expand the supply of credit, especially benefiting ‘good borrowers’ who would see constraints lifted and loan sizes rise. Secondly, a unique identifier could lead to a system for building credit histories, perhaps through a credit bureau, which would further streamline the borrowing-screening process and reduce information asymmetries.