Blog - Investing Notes

May 11, 2009 - "Sell in May" During Bull and Bear Conditions

A reader inquired:

"John Mauldin's 'Thoughts from the Frontline' of May 1, 2009 analyzes the 'Sell in May' concept separating out secular bull versus secular bear market cycles. There appears to be a stronger negative effect during the summer months in secular bear periods. I would be interested in your thoughts about this, given that you have already looked at the 'Sell in May' concept and found it to be fairly sound, at least since 1950."

One thought is that investors cannot reliably identify ex ante the beginnings and ends of secular bull and bear markets, so the retrospective observation is not confidently exploitable. Is there any exploitable (systematic) way to combine "Sell in May" with long-term stock market trend? To investigate, we return to Robert Shiller's dataset, which provides monthly levels of the S&P Composite Stock Index since 1871, and apply the following simple definition of stock market state: the stock market is in a bull (bear) state when its monthly close is above (below) its long-term simple moving average (SMA) of monthly closes. We consider both 1-year and 5-year SMAs to identify cyclical and secular trends. We split the investing year into two half-years (seasons): November (October close) through April (April close), and May (April close) through October (October close). We define the state of the market for a given season as bull or bear based on conditions at the end of the immediately preceding season. Using monthly closes for the S&P Composite Stock Index from 1951 through April 2009 (117 seasons), we find that...

The following chart summarizes average seasonal returns over the sample period for five cases, as follows:

• All seasons in the sample.
• Seasons that are cyclically bullish (index above its 1-year SMA at the end of the preceding season).
• Seasons that are cyclically bearish (index below its 1-year SMA at the end of the preceding season).
• Seasons that are secularly bullish (index above its 5-year SMA at the end of the preceding season).
• Seasons that are secularly bearish (index below its 5-year SMA at the end of the preceding season).

For all cases, the average return during November-April beats the average return during May-October by a substantial margin. The margin appears to be largest for "secular bear" conditions (for which the average November-April return is about zero).

For a closer look, we compare differences between November-April and May-October average returns for these market states.

The next chart compares differences between November-April and May-October average returns for different market states over the sample period. The chart shows that the "Sell in May" effect tends to be strongest for secular bear market states. However, as seen above, the May-October average return is no worse for secular bear conditions than for cyclical bear conditions. The seasonal gap is larger for secular than cyclical bear market states because of the difference in November-April average returns.

In summary, the "Sell in May" effect may be stronger during secular (but not cyclical) bear markets. However, there probably is no reward on average for being long stocks during either the good or bad seasons in bear markets. Conversely, there probably are rewards for holding stocks during both the good and bad seasons in bull markets.

As noted in our blog entry of 12/26/08, "Sell in May" does not work during the prior 1871-1950 period. Applying the above bull and bear definitions generally confirms that "Sell in May" is unreliable during both bull and bear conditions during this earlier period. Given this long-run inconsistency and comparisons by some commentators of the current environment to the 1930s, one might be skeptical of its reliability now.

For related research, see Blog Synthesis: Calendar Effects. See especially the blog entries of 12/26/08, 12/20/07, 3/14/06 and 8/31/05.