Around this time of year, pundits and commentators invariably trot out the recommendation to "sell in May and go away," on the basis that returns from May through September have historically been awful relative to those earned during the complementary months. There is no shortage of folk tales in investing, so in the spirit of a superb book I'm currently reading -- Expected Returns by Antti Ilmanen -- I thought I'd listen to the data instead. Before you liquidate your stock portfolio, I recommend you review my findings below.
We're off to the races!
First, a bit of market history: "Sell in May and go away" did not originate on Wall Street, but rather in the City, London's financial district. In fact, the full saying is "Sell in May and go away; come back on St. Leger's Day." The St. Leger Stakes is the oldest of England's five horseracing classics and is the last to be run.
As far as I know, the data that are most widely cited by investors and brokers regarding this phenomenon are that of Ned Davis Research (not so for academics -- the seminal paper in that realm is Bouman and Jacobsen's The Halloween Indicator (link opens PDF file).) The following table contains one of NDR's findings regarding the S&P 500
Current Value of $1,000 Invested in the S&P 500 Beginning on April 30, 1950*
|Sell in May, buy back in October**||$75,539|
|Buy in May, sell in October||$1,032|
Source: Ned Davis Research. *At March 31, 2012, does not include dividends. **Money is invested in stocks from Sept. 30 through April 30 annually and is in cash (no yield) during all other periods.
The numbers certainly look impressive: Stocks' price appreciation occurred almost exclusively (on average) during the period October through April. By comparison, holding stocks from May through September appears to have barely preserved the nominal value of the initial investment over a 62-year period!
I'm perfectly willing to believe that there is a seasonal component to stocks' price appreciation that is inconsistent with efficient markets, but these data aren't enough to judge the efficacy of a seasonal switching strategy. In that regard, NDR's methodology suffers from several shortcomings: For one, it assumes that when the money isn't invested in stocks, it earns no return whatsoever instead of being invested in Treasury bills. Furthermore, their data do not account for dividends, a critical component of stock returns. Finally, there is no benchmark data corresponding to a straightforward "buy and hold" strategy.
In order to address these issues, I performed my own calculations, using data series from Ibbotson Associates (a unit of Morningstar) that begin in 1926. The following table contains the results of my analysis:
S&P 500: Annualized Return (including dividends)
April 30, 1926 to March 31, 2012
|Sell in May, buy back in October||8.4%|
|Buy in May, sell in October||5.1%|
Source: Ibbotson Associates, Standard & Poor's, Federal Reserve Bank of St. Louis, author's calculations.
And your winner is...
There are two key observations here:
The "sell in May" strategy soundly beat the converse strategy, with a margin of outperformance that exceeds 3 percentage points on an annualized basis.
However, "Sell in May" underperformed buy-and-hold; in fact, the outperformance of buy-and-hold is understated because the returns in the table assume no transaction costs and no tax impact. Investors selling in May incur taxes on short-term capital gains and higher transaction costs than their buy-and-hold counterpart.
About the same
If you're clinging to the notion that "sell in May" could yet be superior to buy-and-hold on a risk-adjusted basis, you should know that both strategies have identical Sharpe ratios of 0.12 (the Sharpe ratio measures the incremental return that an asset or strategy generates per unit of volatility). That's not surprising, given that depending on where you are in the calendar, "sell in May" is either equivalent to buy-and-hold or simply earning the risk-free rate.
I think it's quite likely there is a seasonal effect to stock price appreciation that is more than simply a spurious historical observation. However, that's far from enough justification for reducing (much less eliminating!) one's exposure to stocks as May rolls in. The primary consideration when deciding one's allocation to stocks should be valuation, not the date on the calendar.
Right now, U.S. stocks look at least somewhat overvalued, based on the Shiller P/E, which uses a trailing 10-year average of real earnings. Furthermore, I think we can expect a change in volatility regime as the year unfolds, with the exceptional uncertainty linked to the European sovereign crisis simply biding its time to manifest itself. As such, an underweighting in U.S. stocks is prudent for those whose equity exposure is in index funds such as the SPDR S&P 500 ETF