Author(s)

H M P Capobianco, A M Cister & B F Maceio

Abstract

The aim of this study is to test the Efficient Market Hypotheses (EMH) against the Sao Paulo Stock Exchange Index, Ibovespa. Given the long time span used, from January 1968 to December 2001, an attempt has been made at a classical forecast of a time series. However, as the results were unsatisfactory, neural networks (ANN) have been used to verify whether there was any pattern which might help the forecast. This article therefore covers forecasts using an ANN technique. The evidence points to acceptance of the HME. However, other, more robust approaches to forecasting this market have not been exhausted. 1 Introduction The efficiency of the capital market depends on the quantity and quality of the relevant information available in the market. Thus, it can be supposed that the Brazilian market is likely to be less efficient than the US and European markets, as generally commented by Bodie et al. [1], and there are therefore more opportunities for abnormal gains. In Brazil’s case, the Brazilian Securities Commission (CVM – Comissao de Valores Mobiliarios) has less rigorous disclosure requirements for companies listed on the stock exchange than those of the Securities and Exchange Commission, SEC, in the USA. In Brazil, a company with shares traded on the

Keywords