This looks like another year where the “sell in May” strategy appears to be working. But is this nothing more than an anomaly? Some new research says the seasonal effect is not just an anomaly:
“A multitude of calendar or seasonal anomalies to the efficient market hypothesis (EMH) have been identi ed: the January e¤ect, day-of-the-week e¤ects such as the Monday effect, the Holiday effect, the Turn of the Month e¤ect, etc. However, as a predominantly non-experimental eld, nancial economics is vulnerable to spurious inferences from data mining.In fact, many if not most calendar anomalies dissipate after they have been identi ed. On the disappearance of January effect, Eugene Fama is quoted as saying (Smallhout, 2000):
“I think it was all chance to begin with. There are strange things in any body of data.”
An objective test of a calendar e¤ect must then include a test of the e¤ects persistence, out of sample in future data. Sullivan, Timmermann, and White (2001), who address the dangers of data mining for calendar e¤ects, note:
“New data provides an e¤ective remedy against data mining. Use of new data ensures that the sample on which a hypothesis was originally based e¤ectively is separated from the sample used to test the hypothesis.”
In this paper we perform the rst out-of-sample analysis of an anomaly identi ed by the old adage Sell in May and Go Away, also known as the Halloween e¤ect. In the fi rst academic analysis of this effect, Bouman and Jacobsen (2002) study 37 markets and find higher returns in 35 of these markets during the November to April semester as compared to the May to October semester. November-April returns are statistically higher in 20 of the 37 markets. Bouman and Jacobsens (2002) sample end in 1998. Here, we study the out-of-sample 1998-2012 period for those 37 equity markets.”
Read the full paper here.