Sell in May: It's the silly season again

(MoneyWatch) As the calendar approaches May 1, you're sure to hear about the old investment legend that you'll get better returns if you sell stocks at the beginning of May and buy them back at the end of October. It's right up there with the "January Effect" (which never existed in any implementable way) and the "Santa Claus Rally." As you'll see, like most other investment legends, this is one has just enough truth to it to maintain the legend.

The thinking behind the "sell in May and go away" strategy is that your portfolio will have higher returns if invested in stocks from November through April (as opposed to the entire year) because they've had higher returns during those months than May through October.

• Research reputations can be misleading
• Short selling: Evil or necessary evil?
• How to avoid selling stocks at the wrong time

It is in fact true that stocks have higher returns from November through April. Going back to 1926, the average monthly return for May, June, July, August, September and October was 0.72 percent. The average monthly return for November, December, January, February, March and April was 1.16 percent.

However, this matters only if such a portfolio really outperforms the basic buy-and-hold strategy. To test the theory, we created two portfolios. One portfolio buys the S&P 500 Index on Jan. 1, 1926, and holds that position through April 30, 1926. On May 1, 1926, this stock position is swapped from the S&P 500 Index to 30-day U.S. Treasury bills. This portfolio remains in short-term Treasuries through Oct. 31, 1926. On Nov. 1, the Treasury bills are sold and the proceeds invested in the S&P 500 index. This is process is repeated every May and November, ultimately ending with data from October 2012.

The results tell the story: If an investor followed this strategy, they would have been worse off than had they just held stocks. The "sell in May and go away" portfolio produced an annualized return of 8.3 percent. However, compare this to the return of the S&P 500 Index, which produced an annualized return of 9.8 percent. And that's even before considering any transactions costs, let alone the impact of taxes (you'd be converting what would otherwise be long-term capital gains into short-term capital gains, which are taxed at the same rate as ordinary income).

The gap between the "sell in May" approach and investment in the S&P exists because there remains a risk premium for investing in stocks. As mentioned previously, the average monthly return to stocks for the months May through October was 0.72 percent. From January 1926 through October 2012, the average monthly return for 30-day U.S. Treasury bills was 0.29 percent, well below the market's return of 0.72 percent per month during that same period.

As with any data-mining exercise, you can find periods when the strategy "worked," at least before considering implementation costs (trading costs or taxes) and periods when it didn't. Consider the following:

From 1960 through 1979, the "sell in May" strategy provided an annualized return of 9.7 percent, outperforming a buy-and-hold strategy by 2.8 percent per year. The problem is that prior to that period the strategy had underperformed. From 1926 through 1959 it underperformed the buy-and-hold strategy by more than 5 percent (10.3 - 5.1) per year. Thus there's no way anyone would have known that the strategy would work beginning in 1960. By the time a sufficiently long period had passed to convince an investor that the strategy actually might succeed, it would have stopped working.

The above example of cherry picking a starting period and ignoring expenses is often the explanation for how myths arise. Randomly, we would expect almost any "system" (at least before expenses) to work, just as there were times when the "sell in May" strategy "worked."

Image courtesy of Flickr user Candie_N (Welcome Spring).