Resumen
The Halloween effect relates to the notion that stock market returns tend to be higher in the period from November to April than from May to October. In this study, we analyze the robustness of this trading strategy taking into account the individual returns of stocks traded in the Brazilian stock market during the period from August 1994 to June 2014. Using standard dummy regression approach introduced by Bouman and Jacobsen (2002), our results suggest the existence of the Halloween effect in the Brazilian market, which has shown to be economically and statistically significant, with a positive sign and a slight drop trend over the past few years. In addition, when reassessing these results using the "Superior Predictive Ability Test" of Hansen (2005), we have found that an investment strategy based on the Halloween effect generates a statistically significant returns superior to a buy-and-hold strategy when the effects of data-snooping when data-snooping effects are not neglected in the stock returns series, as in Bouman and Jacobsen (2002).