An econometric estimation of a retail credit rating model: case of banking institutions in Zimbabwe

Abstract

Zimbabwe This study was done against the background of an increasing trend of non-performing loans in the Zimbabwean financial services sector between the periods 2009 to 2012. This was exacerbated by the absence of a national credit bureau in the country which resulted in many borrowers getting credit that was beyond their credit worthiness. This study sought to come up with an econometric estimation of retail credit rating model that could be used by the Zimbabwean financial services firms. It was envisaged that this could significantly contribute to the reduction of non-performing loans which was affecting the sector. Various literature sources guided the development of the model. The study adopted the cross sectional research design. The population comprised of the sixteen operational banks at that time. The study relied on secondary data collected from the loan officers of these banks. The sample comprised of 312 clients of these banks from both defaulters and non-defaulters. The study made use of stratified random sampling technique. The study used a multiple logistic regression model to come up with a proposed econometric estimation of a retail credit rating model. Information regarding the borrowers income, marital status, educational level, age, age squared, residential address, sex, vehicle finance and mortgage loan was captured and logistic regression analysis was applied. The results were analyzed in relation to the clients' default or non-default behavior. The results indicated that income and marital status are the strongest predictors of default behavior. The study concluded that there was a relationship between the probability of default and the individual characteristics of the borrowers. The study recommended the same research to be repeated when the Zimbabwean economy would have stabilized to establish if there would have been any changes on the relationships between the individual characteristics and their default pattern.