Recently, financial institutions have developed improved internal risk rating systems and emphasized the probability of default and loss given default. The default characteristics are studied for 756 loans from a French bank: CIC- Banque SNVB. A binomial logit regression is used to estimate several models of the probability of default of agribusiness loans based on information available at loan origination. The results show that leverage, profitability and liquidity at loan origination are statistically significant indicators of the probability of default. As leverage increases, profitability decreases, or liquidity decreases, the probability of default increases. As the length of loan increases, the probability of default also increases. Finally, it is more accurate to develop a model for each type of collateral (activity). By developing more quantitative credit scoring models,banks may benefit from lower capital requirements while borrowers may see better rates where the risk of loans is appropriately priced.