The Halloween effect: an alternative approach and new evidence from the US market

Abstract

The Halloween effect refers to a calendar anomaly that can be easily exploited and calls for buying the market index in the end of October each year and switching to treasury bills at the end of April the following year. The effect has only been studied on a 'calendar-month' basis and primarily on large capitalisation indices. By examining two popular indices for the US stock market for the period December 1983 to December 2009, on a calendar-month basis we find the effect to be statistically non-significant for large cap stocks (S%P 500) and statistically significant for small cap stocks (Russell 2000). Since we see no reason to constrain seasonality on a calendar-month basis, we go further to explore all possible entry and exit point combinations on a trading day basis and find that the interval between the first week of June (trading day n.107) and the last week of October (trading day n.208) captures the rationale of the 'Halloween' calendar effect at its peak. The resulting trading strategy remains robust even after controlling for outliers, short-term volatility and the January effect.