Black cats or black swans? Outliers, seasonality in return distribution properties, and the Halloween effect

Abstract

Purpose – The purpose of this paper is to determine the extent to which outliers have persisted in augmenting the Halloween effect over time and to offer an econometric test of seasonality in return skewness that might provide a partial explanation for the Halloween effect. Design/methodology/approach – The authors split the Morgan Stanley Capital International data for 37 countries into two subperiods and, using median regression and influence vectors, examine these periods for a possible change in the interplay between outliers and the Halloween effect. The authors perform a statistical assessment of whether outliers are a significant contributor to the overall Halloween effect using a bootstrap test of seasonal differences in return skewness. Findings – Large returns (positive and negative) persist in being generally favorable to the Halloween effect in most countries. The authors find seasonality in return skewness to be statistically significant in many countries. Returns over the May through October timeframe are negatively skewed relative to returns over the November through April period. Originality/value – This paper offers the first statistical test of seasonality in return skewness in the context of the Halloween effect. The authors show the Halloween effect to be a more complex phenomenon than the simple seasonality in mean returns documented in prior research.