The present thesis consists of three researches. The first research sought to ascertain the relationship between FXvol and future exchange rate and stock market index returns as the FXvol volatility index is viewed as a thermometer of investor uncertainty for a period ahead. The contemporary relationship between FXvol, Ptax and Ibovespa, as well as the ability of FXvol to capture the possible relationship between the level of uncertainty present in the market and the relative future return of the exchange rate and the stock index. The second research compared the traditional GARCH models and the GARCH model with regime changes regarding its power to predict the exchange rate volatility. We attempted to compare the performance of each of the models in a real situation of use, in this case, in the calculation of the Value at Risk of an exchange portfolio. The third research sought to identify the existence of the herd effect in the Brazilian market and to understand the influence of the exchange rate in this effect, due to the importance of the exchange market for the Brazilian market. The methodology comprised two steps, initially attempting to analyze the mean of the effect through regressions and in a second moment to study the variation of the effect over time through the Kalman Filter method.