RANDOM WALK, ANTI-PERSISTENT OR PERSISTENT SERIES: AN INVESTIGATION IN THE BRAZILIAN STOCK MARKET

Abstract

The concept of efficient market still provokes intensedebates in academic community. Numerous academics have dedicated themselves tosearch for past data, which could explain and hint stock market future values.However, according to theoretical assumptions of the efficient markethypothesis, the task of predicting future prices, based upon past behavior of afinancial asset, is an inefficient procedure. On the other hand, the distributionof a number of stock time series denotes a brownian motion, which is designedfor random and independent characteristics. Thus, this study aimed toinvestigate whether the time series of returns of the Brazilian stocks presentlong-term memory existence in order to confirm or rule out the hypothesis ofrandom walk. Therefore, it was defined as the sampling plan of this study, theactions belonging to the Index of the Stock Exchange of São Paulo (BovespaIndex), provided that these actions had some trading in at least 98% of theStock Exchange of São Paulo in the period of January, 3rd 2000 to February, 9nd2012. After filtering, 29 assets were to be investigated. As a methodologicalchoice, we calculated the Hurst exponent of each stock. The results indicatedthat all series of assets surveyed had a long-term memory at a level of significance of 5%. First of all, it was concluded that it is possible toidentify behavior patterns in the price of assets, thus establishing desirableand basic conditions to use historical data as pricing model. As second pointof completion, it was possible to position itself in the list of studies that inquireabout the absolute and unquestioned use of the concept of efficient market,which has been weakened by empirical foundations.