Provisão em crédito nos bancos de varejo : a aplicação de um modelo estatístico para análise de risco de pessoas jurídicas

Abstract

This project discusses credit risk rating with the purpose of determining provisions for doubtful credit. It begins with the following proposition: clients with the same characteristics get more or less the same ratings when evaluated by retail banks using quantitative methods. Within this context, credit provision distributions of these banks would be more or less homogeneous. The analysis utilizes robust statistical technique for testing the existence of homogeneity, and the analyzed data came from quarterly statements from March 2000 to December 2002 published by Banco Central do Brasil (Brazilian Central Bank). This work also suggests that the retail banks' credit portfolios of short-term loans destined to current use be granted through risk credit analysis based on quantitative methods. Those banks have a high demand for this kind of loans, and they should utilize scientific methods of analysis in order to rate all loans by the same standards and through a swift process. Unlike the scientific method, the judgmental method is a slow one, for it depends on the experience of a large (and expensive) group of analysts. The model suggested by this work was developed through statistical analyses that explain credit risk by financial statements and credit behavior data. These data comprise 500 firms, their financial performance in the years 1999, 2000 and 2001, and their credit behavior relative to loans granted in 2002. The default probabilities calculated by the model should set the basic pattern for the provisions for doubtful credit in accordance with Resolução CMN 2.682 (National Monetary Council Statement # 2.682)

This project discusses credit risk rating with the purpose of determining provisions for doubtful credit. It begins with the following proposition: clients with the same characteristics get more or less the same ratings when evaluated by retail banks using quantitative methods. Within this context, credit provision distributions of these banks would be more or less homogeneous. The analysis utilizes robust statistical technique for testing the existence of homogeneity, and the analyzed data came from quarterly statements from March 2000 to December 2002 published by Banco Central do Brasil (Brazilian Central Bank). This work also suggests that the retail banks' credit portfolios of short-term loans destined to current use be granted through risk credit analysis based on quantitative methods. Those banks have a high demand for this kind of loans, and they should utilize scientific methods of analysis in order to rate all loans by the same standards and through a swift process. Unlike the scientific method, the judgmental method is a slow one, for it depends on the experience of a large (and expensive) group of analysts. The model suggested by this work was developed through statistical analyses that explain credit risk by financial statements and credit behavior data. These data comprise 500 firms, their financial performance in the years 1999, 2000 and 2001, and their credit behavior relative to loans granted in 2002. The default probabilities calculated by the model should set the basic pattern for the provisions for doubtful credit in accordance with Resolução CMN 2.682 (National Monetary Council Statement # 2.682)