A contribuição da informação contábil para a previsão de retornos de ações: avaliação a partir de modelos de séries temporais

• Ano de defesa: 2012
• Autor(a) principal: Octavio Valente Campos
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais; UFMG
• Palavras-chave em Português: Retornos das Ações; ARIMA; VAR; Indicadores Contábeis; Previsões; Contabilidade
• Link de acesso: http://hdl.handle.net/1843/BUOS-8ZTLA9

The objective of this research was to determine if the use of financial indicators improves the capacity of the stock return forecast over forecasts made only from the returns past history. Univariate models forecasting (ARIMA) were specified only with the stock returns past history, after these results were compared with the predictions made by multivariate models (VAR) using financial indicators as models explanatory variables. The sample was composed by 20 companies, and the results were analyzed in three different moments of time, to prevent the analyzed results from having problems related to data-snooping (results due to chance). The indicators that showed greater predictive power were sequentially: Net Margin, Return on Equity, Asset Turnover; Breakdown of Debt and Liquidity. It was found out that the univariate models contained higher precision in all enterprises; however, by incorporating new observations to series prediction, that is, when the analysis series have more observations, the multivariate models become more accurate. Comparing the portfolios returns formed from the best forecasting models, univariate and multivariate is observed, in general, that the forecasts and the decisions made based on multivariate models tend to provide investors with superior returns on long-term investments (1 year). On the other hand, the univariate models tend to provide investors with superior returns on short-term investments (1 quarter). Therefore, it is verified that the Brazilian capital market has some market inefficiency, either as in weak form as in semi-strong form, and we can conclude that the financial information is relevant, especially in investment strategies in the long term. Thus, we conclude that the inclusion of financial indicators in multivariate models forecasting extends the capability of prediction of stock returns, especially for forecasts in the medium and long term and when the series in analysis have greater number of observations.