ELIZIÁRIO, M. A. A.; ELIZIÁRIO, Maria Aline Andrade.
Resumo:
Discriminant analysis is based on the use of multivariate statistical techniques through which
it is possible to predict insolvency of companies, based on data obtained from the financial
statements. With this focus, this study stands out for assisting in the process of credit
concession, playing a relevant role for development of several researches in the area. It is
inferred that in the first research studies conducted in Brazil, with the intention of to estimate
bankruptcy prediction models aimed at improving the credit analysis process, date from the
1970’s. To the insolvency or solvency prevision models construction were used the statistical
probability prediction from resources of discriminant analysis. The models analyzed in this
work were developed by Stephen Charles Kanitz (1978), Roberto Elizabetsky (1976) and
Matias Alberto Borges (1978). The models highlighted were estimated from a combination of
the financial indicators that measure and classify statistically a solvency/insolvency state of a
company. Given the relevance of authors' perceptions and the use of different indexes for
models prediction, we intend to evaluate the following question: Are there differences
between discriminant analysis models about the solvent and insolvent company’s
classification proposed by Kanitz, Elizabetsky and Matias? The objective of this work is to
propose an exploratory study into companies listed on the New Market of the Stock Exchange
of São Paulo (BOVESPA), with approach mainly at checking the differences between
insolvency models prediction proposed by Kanitz, Elizabetsky and Matthias. This study
discusses the issue concerning 75 non-financial companies listed on BOVESPA. The research
methodology can be classified as exploratory according to the purposes and bibliographic,
according to the means. The results obtained show a significant number of contradictions
when occurs the confrontation of the models. The study suggests that credit analysis from
discriminant models should be run based on more than one technique for a good level of
prediction on insolvency forecast.