# Determining the probability of default for Agrifinancial’s loan portfolio

## Date

## Authors

## Journal Title

## Journal ISSN

## Volume Title

## Publisher

## Abstract

Credit risk rating models are tools used by nearly every financial organization across the world. These models allow for organizations to measure risk more accurately within their credit portfolios, ensuring more confident credit approvals and capital requirements in the event default may occur. AgriFinancial is proposing the use of a credit risk rating system to make funding decisions using capital from its parent company in addition to participation funding that is currently the sole source of funds. This thesis evaluates AgriFinancial’s portfolio risk of originated and serviced loans, using ten years of historical data to determine the probability of default for the portfolio of originated loans and if the variables considered are statistically significant. The evaluation includes the estimation of a binomial logit regression model of default defined as loans that are ninety days or more past due. Credit factors measured at the time of loan origination are independent variables that predict default. Regression results were applied to the portfolio of loans to calculate their respective probability of default. Overall, the model has explanatory power in predicting the probability of default. Originally, eleven variables were considered; two were removed due to multicollinearity and a ratio used more predominately for loss given default, a default risk calculation not considered. It was concluded that four of the nine independent variables were statistically significant in predicting that a loan will default at some time during the life of that loan. One variable was statistically significant at the 1% level, two variables at the 5% level, and one at the 10% level. Credit score has the highest statistical significance at the 1% level for determining probability of default. Signs for the statistically significant independent variables were correctly predicted; pro forma debt/asset positive resulting in a higher probability of default as the variable increases, three year average total debt coverage negative resulting in a lower probability of default as the variable increases, credit score negative resulting in a lower probability of default as the variable increases, and the original principal positive resulting in a higher probability of default as the variable increases. Based on the results, recommendations were made regarding implementation of the model and the advantage of proceeding with the new business opportunity.