Halloween or January? Yet another puzzle

Recent works suggest a potentially exploitable effect in US markets, the 'Halloween Indicator'. This suggests that the greater part of changes in equity markets arises over the November-April period, with little change over the summer months, simultaneous with no evident changes in the risk profiles of the two six-month periods. We re-examine this and find contradictory evidence. Over the 1926-2002 period we find rather that the effect demonstrated may well be a reflection of the well-known January anomaly. Our conclusion therefore is that the jury remains out on the existence of a semi-annual seasonality.