After reading this week's Barron's I saw an interesting little chart that showed the growth of $100 if the "sell in May" strategy were followed, also known as the Halloween indicator (original research paper). Surprisingly, if you had followed a strategy of investing $100 in the S&P 500 since 04/30/45 during the period from October — April versus that of May — September, your $100 would have turned into $9,329 in the former period versus $99 in the latter.
The chart in Barron's:
One would think, after seeing this chart, that selling in May, or maybe even April to get ahead of the curve, would be a prudent investment strategy. However, if you extend the timeline out a bit more you get a picture that looks like this (this chart uses May to October and Nov to April; however, it is more representative due to the fact that the months are split 50/50, as opposed to 58/42 in the above chart):
Now, you may have noticed that November to April still outperformed. However, clearly, if you had ignored the period from May to September you would have missed out on some nice gains during this time frame. So what's the difference due to? If you look at the average returns by month since 1925, it becomes quite evident:
It seems one of the major culprits is September. So why not sell in September?
As Ken Fisher points out in his book "Debunkery," where this chart was first found, the two worst Septembers (which are notably in the May to October period) were down 29.6% in 1931 and 13.8% in 1937, due to the Great Depression. All other months offered gains over the long haul.
What else accounts for the big downtrend in the May-to-October period?:
- Panic of 1901: May 17, 1901
- Panic of 1907: October 1907
- Black Monday: October 19, 1987
- Friday the 13th mini-crash: October 13, 1989
- Economic crisis in Asian: 1997 mini-crash: October 27, 1997
- September 11th attacks: September 11, 2001
- Internet bubble bursts: October 9, 2002
- Global Financial Crisis starts: September 16, 2008
- 2010 Flash Crash: May 6, 2010
Who is to say the next big stock market crashes won't happen to coincide with April or July, the best performing months on average?
Bottom line: Don't rely too much on statistical anomalies and try to "time the market" or you might be left holding the bag.