Details Matter: Product Design for a Better Loan Experience

A number of studies show that around 20 percent of poor people accept a formal loan when offered one from a microfinance institution (MFI). This take-up rate is surprisingly low given ample evidence that poor people do borrow, often from multiple informal sources at a time (family members, friends, employers, moneylenders, etc.).

One explanation for why there is not more demand for formal microcredit is that the product is not flexible in its design. As highlighted in the recent CGAP paper, Latest Findings from Randomized Evaluations of Microfinance, one of the most common design features for microcredit has recently come under scrutiny, namely the group liability model pioneered by Muhammad Yunus. The group liability approach matches borrowers into lending groups and makes every group member responsible for full repayment of all loans. With this approach, the MFIs turn borrowing into a public act. Such scrutiny works to discourage default, yet development experts increasingly question whether group liability is necessary. The primary concern is that the model causes clients who can put credit to good use to avoid formal loans exactly because they don’t want the shoulder risk from their neighbors. Group liability also punishes good borrowers with higher borrowing costs, while not-so-good borrowers—whether they earn that designation through will or circumstance—make off with the cash.

Those questions have encouraged researchers and microfinance institutions to experiment with the traditional group liability design to see if it is really necessary. This work largely aims to understand how group liability works to keep default in check, and if that quality can be uncoupled from the group approach and applied in a way that is more flexible for the client. Latest Findings from Randomized Evaluations of Microfinance highlights one of the first studies conducted to assess whether group liability was really necessary to ensure repayment. Using a randomized controlled trial, economists Dean Karlan, from Yale, and Xavier Giné of the World Bank found that clients of Green Bank in thePhilippines did not default more often when offered individual liability loans. Not only that, the bank found it easier to recruit new clients for individual liability loans.

Since Karlan and Giné conducted their study there has been a rapid acceleration in the shift from group to individual liability lending, and with it an increase in evidence on alternative methods for encouraging repayment. Harvard’s Ximena Cadena and Antoinette Schoar of MIT find in Uganda, for example, that bonus payments, reduced interest rates on subsequent loans, and text message reminders all increase the likelihood that borrowers will pay on time by seven to nine percentage points above the average. In Peru, Karlan finds with Jonathan Zinman of Dartmouth and Harvard’s Sendhil Mullainathan that sending a written letter to individual liability borrowers that fall behind in their payments results in significantly higher repayment (especially when the letter is also sent to the borrower’s co-signer). And Karlan also finds in thePhilippines that text message reminders that include the loan officer’s name encourage repayment—these Peru andPhilippines findings point to the role played by personal relationships in repayment patterns.

These studies are only beginning to get below the surface of group liability to understand different ways to expand the pool of borrowers while maintaining high repayment rates. Ongoing work is testing different approaches to the group liability model, co-signatory requirements, and even whether the trustworthiness of an existing client can predict whether a borrower she refers will default.

Low default levels are necessary if banks and MFIs are going to offer products profitably and at scale. It is important, however, to keep the broader goals in mind. If credit is to have a positive development effect on the lives of poor people it needs to reach the borrowers who can put the funds to good use.