Since microfinance banks often provide credit without much collateral, these institutions must minimize the risk of client default. These banks face high operational costs employing loan officers to visit clients regularly, sometimes in remote areas. Finding a reliable and accurate mechanism for identifying clients with low default risk is critical for maintaining a sustainable bank. However, traditional loan application models that use quantitative data like client repayment history, salary, and assets, may overlook potentially creditworthy applicants at the bottom of the pyramid who may be taking formal loans for the first time. This evaluation assesses the effectiveness of a new credit risk model, incorporating qualitative data, in predicting repayment and studies its influence on loan committees’ application decisions.
For additional information on current SME Initiative projects, click here.
Bancamía is a Colombian microfinance bank that has developed a credit risk model to streamline its loan application process. This risk model incorporates the bank’s historic quantitative data with new qualitative data, like loan officer perceptions, to automatically approve creditworthy clients. Bancamía’s unique system of standardizing qualitative indicators to be used in the model’s algorithm was created to target clients who might otherwise be rejected by a more traditional model. The goal of this statistical modeling system is to improve identification of the best and worst clients, while decentralizing the loan committee process and reducing costs.
This pilot will assess how the score produced by a credit risk model influences loan committees’ decisions to accept or reject applicants, as well as how accurate the model predicts loan performance of approved clients. Bancamía’s credit risk model is designed to produce a score and application decision by calculating the probability of default based on quantitative and qualitative characteristics collected by loan officers.
Branch directors and loan officers at the eight participating bank branches were invited to a short training workshop before the implementation of the pilot. The training presentations covered topics such as: explanation and importance of a credit risk model, development of the Bancamía tool, and objectives of researching the accuracy of the credit risk model in predicting client performance.
After the training, credit committees, consisting of a branch director, loan officer who collected applicant information, and one or two additional loan officer witnesses, met to review and discuss loan applications. During the meetings, IPA research staff used a mechanism that randomly assigned each of the applications to an information treatment: score from the model presented before the discussion about the application, score from the model presented after a preliminary decision on the application had been made, and a comparison group (no information). Research staff was responsible for providing the credit risk score information at the assigned time and recording final and intermediate (second treatment group) application decision with details of approved and requested loan amounts. After each application decision, both branch directors and loan officers (some of whom collected the original client data) on the committee completed forms describing their perceptions of the loan decision. Over a period of six months, about 1700 applications were reviewed, with about 550 reviews in each information treatment group. The loan performance of approved clients will be collected over a 10-16 month period and will be used to analyze the impact of the different information treatments. Loan portfolio data will also be collected from eight other comparison branches of Bancamía. Results from this pilot will inform a full scale randomized evaluation.
Results forthcoming.