This work aims to develop a statistical model for estimating the probability of Brazilian companies defaulting on credit agreements with banks. In addition, a validation methodology called Accuracy Ratio will be implemented to the developed model, so that it can be compared with other models already recognized in the literature of quantitative risk models. The database was constructed based on the accounting information of Brazilian companies collected on Bloomberg and information on the debt structure of the companies in the Central Bank Credit Information System (SCR). A panel logistic regression model is developed in which the explanatory variables of the default event are chosen to estimate significant parameters, with a high degree of economic-financial interpretability, as well as a reliable adherence of the model to the data. The data have a panel structure and, for this reason, will also be approached the strategy adopted to eliminate the fixed temporal effect. Finally, the probability of default will be introduced in a practical context that justifies its importance in the decision-making process in banks. The probability of default will be presented as a variable of the expected repayment size model of credit operations, known as RAROC (Risk Adjusted Return on Capital).