Previsão de insolvência baseado em dados contábeis de empresas brasileiras listadas na Bolsa de Valores nos anos de 2018 e 2019
The use of models to predict company insolvency and bankruptcy are common themes in the literature. It is in the interest of managers and other stakeholders, such as shareholders, to use indicators to try to identify whether a company is close to bankruptcy or not. The objective of this work is to u...
| Autor: | |
|---|---|
| Formato: | tesis de maestría |
| Estado: | Versión publicada |
| Fecha de publicación: | 2021 |
| País: | Brasil |
| Recursos: | Universidade Federal do Ceará (UFC) |
| Repositorio: | Repositório Institucional da Universidade Federal do Ceará (UFC) |
| Idioma: | portugués |
| OAI Identifier: | oai:repositorio.ufc.br:riufc/59231 |
| Acesso em linha: | http://www.repositorio.ufc.br/handle/riufc/59231 |
| Access Level: | acceso abierto |
| Palavra-chave: | Insolvência Logit Lasso Contabilidade |
| Resumo: | The use of models to predict company insolvency and bankruptcy are common themes in the literature. It is in the interest of managers and other stakeholders, such as shareholders, to use indicators to try to identify whether a company is close to bankruptcy or not. The objective of this work is to use an econometric model to predict insolvency. The post-lasso logit regression model was used to predict the insolvency of Brazilian companies listed on B3 for the years 2018 and 2019. The data used consist of the financial statements of these companies. As explanatory variables, accounting indices from the literature were used. The year 2019 was used to classify between solvent and insolvent companies, while the year 2018 was used to estimate the models, so that there is a time difference between the accounting indicators and the insolvency event. For comparative purposes, two other logit models with different variables were estimated. The results show good predictive performance of the models, as well as a good fit to the data. The post-lasso logit model had the best predictive performance when compared to the other two models, with the correctly predicted percentage, sensitivity and specificity all above 90%. |
|---|