Outliers and the Halloween Effect: Comment on Maberly and Pierce

Maberly and Pierce (2004) re-examine the work of Bouman and Jacobsen (2002) that documents significantly lower monthly stock market returns over the period May to October than over the period November to April. The finding has been called the Halloween effect and is present to varying degrees in most equity markets worldwide. Maberly and Pierce focus on the Halloween effect in the United States and contend it is driven by two negative-return outliers. We argue that controlling for two outliers is somewhat arbitrary. We apply robust regression methods—including all the data but limiting the influence of extreme returns—to the estimation of the Halloween effect in the United States. Contrary to the Maberly and Pierce findings, our results indicate statistical significance of a Halloween effect in the U.S. at levels similar to those originally reported in Bouman and Jacobsen.