The Halloween effect suggests that investors should be in stocks during November through April and in cash during May through October. Is there a connection between the January effect and the Halloween effect, or are they distinct market anomalies? In their March 2006 paper entitled “Halloween or January? Yet Another Puzzle”, Brian Lucey and Shelly Zhao examine seasonal returns to determine whether the Halloween effect is just an imprecise reflection of the January effect. Using monthly return data for U.S. stocks allocated to capitalization-based size deciles over the period 1926-2002, they conclude that:
- The average return for the winter months consistently exceeds that for the summer months.
- The magnitude of this excess return declines as firm size increases.
- Volatility tends to be lower in winter months than in summer months.
- January returns drive the outperformance of the winter months. In fact, the existence of a January-less Halloween effect is dubious.
- Over the entire 76-year period a Halloween strategy (sell in May…) outperforms a buy-and-hold strategy in eight of 10 deciles. However, during 1986-2002 buy-and-hold beats a Halloween strategy in eight of 10 deciles, and during 1966-2002 buy-and- hold outperforms in seven of 10 deciles.
The following table, Table 1 from the paper, shows the average returns for U.S. stocks by month during 1926-2002 for a value-weighted index, an equal-weighted index and for each of ten portfolios based on firm capitalization. The smallest firms are in Decile 1, and the largest are in Decile 10. The table shows the strong outperformance of small stocks in January, as well as the generally poor performance of stocks in September and October. (See page 15 of the paper for a larger version.)
In summary, there is probably no Halloween effect distinct from the (diminishing) January effect in the U.S. stock market.
