Estatística Multivariada Aplicada: Construção do Modelo de Previsão de Insolvência Aranha & Gondrige

Abstract

The insolvency forecast, although it is a widely discussed subject, still presents a need to improve the existing models, due to the emergence of new predictor variables, such as several currency substitutions, economic scenarios, adaptations of accounting standards with the international standard, among others factors that affect the economy and performance of companies. The measurement of insolvency is conjectured as one of the countless difficulties to which organizations are susceptible, in which the analysis of financial statements helps to obtain information about the economic and financial performance of companies. Thus, one asks how to build an insolvency prediction model with the application of the discriminant function? The aim of this work was to empirically develop an insolvency prediction model using discriminant analysis. The work is justified by the need to seek to understand the financial situation of companies and the transition that occurs between solvent and insolvent companies so that it can serve as guidance in financial forecasts. For the development of the model, two samples with 30 companies were used, each consisting of insolvent companies because they are in judicial recovery, which have recurring losses or with Liabilities Uncovered to another sample composed of solvent companies. For the purpose of homogeneity between sample groups, companies from the same segment as well as similar asset volumes were taken into account. Economic and financial indicators were collected from the Economática® database for the year 2019. For modeling the discriminant function, the IBM SPSS Estatistics Software was used, as well as a Microsoft Excel® spreadsheet. The research, in terms of its nature, characterizes up as applied; as for the approach, it is quantitative; with regard to the objectives, it is classified as descriptive and uses bibliographic research and data collection methods. Statistically, the developed model has a discriminatory power of 90% and, when submitted to the validation test, using a sample of companies different from those used initially, it presented a 95% accuracy rate. In the comparative test with other existing models, the result was 93.33%, therefore, the Aranha & Gondrge Insolvency Prediction Model obtained an excellent representation, being considered a robust model.