Volatilidade das bolsas de valores : um estudo através de modelos lineares e não lineares estruturais

Abstract

This study aims to understanding and analyzing the volatility of returns in the prices of the New York Stock Exchange (NYSE), the electronic stock exchange of New York (NASDAQ) and the Brazilian stock exchange (BM&FBOVESPA). The time series were composed by data collected from January 2000 until March 2015, obtained from the Dow Jones, NASDAQ composite and Ibovespa. To explain the moments of turbulence in stock markets were used structural statistical models nonlinear, ARCH and GARCH, which enabled the measurement of the absolute mean errors. According to the results, the conditional heteroscedastic model that best fit the three temporal series was the ARCH model, which captured the sharpest volatility of the BM&FBOVESPA and adjusted satisfactorily to the more stable markets of New York. To make predictions from the behavioral history of the index of stock exchanges were used statistical models Arima and Arfima. The forecasts were made for three days, and the model that is closer to the actual prices of stock exchanges was model Arfima.