A credit scoring model for development banks: an African case study

Loan portfolio quality could improve through a suitable creditworthiness analysis specifically designed to adapt to the particular LDC's context. In a broader perspective, an effective control of credit risk is a crucial aspect of bank management, especially in a development setting. This is the case of the Caisse Nationale de Crédit Agricole of Burkina Faso where the model proposed in this paper is tested. Two major problems concerning loan analysis are evident: 1. the difficulty and the high cost of collecting relevant information on borrower-customers businesses; 2. the absence of benchmark procedures by loan analysts consistent with bank lending policy. As regards the first point, loan evaluation procedures can be improved through a more efficient diffusion and use of the information necessary to estimate the probability of repayment. The paper identifies the relevant information and defines the criteria through which it can be weighed in the loan evaluation process. Two hypotheses are offered. The first assumes there are some factors affecting the borrower's behavior that are likely to increase the probability of default. Second, the problem of the quality of information is explored. It is assumed that some mistakes can be made in credit risk evaluation related to the difficulty of obtaining the commonly used information on the customers' ability and willingness to repay. To overcome the widespread absence of correct information, evaluation models are used to identify symptoms of the borrower's performance and behavior. For this purpose symptomatic variables are employed to create indicators of the borrower-customers' ability and willingness to repay. As concerns the second point – the credit evaluation procedures – there is often a need for shared decisional frameworks and coordination between the head office and bank branches in order to avoid arbitrary decisions. For this purpose an application of a quantitative model for creditworthiness based on multivariate discriminant analysis is proposed; input variables are factors obtained from a wider set of variables. The quantification of the judgement in a score is also important because it gives responsibility to loan officers for the decisions they take according to the score or for an alternative choice. This last aspect is crucial in development banks, where accountability of loan officers is often a problem. The proposed model is also relevant for the theoretical interpretation of credit risk determinants: objective circumstances – personal or firm specific –, the ability of the bank to correctly evaluate these circumstances and the bank capacity to establish loan conditions to reduce uncertainty. The model demonstrates that the theoretical framework presented in the paper is applicable in an LDC's context. This clear relationship between theory and the interpretation of results can be an important tool in improving portfolio management for development banks and the bank personnel decision-making and learning processes.