Should You Sell in April and Go Away?

Should you try to get a head start on those who are planning "sell in May and go away?"

If so, then you will be looking for an opportunity to sell your stocks in April and go to cash, thereby beating the many investors who will instead wait until a month from now.

But before you rush to sell everything, bear in mind that the odds of success are quite low. Stock market timers in general have very poor success rates, rarely doing better over the long term than simply buying and holding. Why would we think that they can do any better timing their entries and exits in October and April than in any other month of the year?

Well, the proof of the pudding is in the eating.

And over the past eight years, one of the two market timing services that I monitor that regularly second-guess the "Sell in May and Go Away" system has significantly increased that seasonal pattern's performance. While the other one has not improved on the Halloween Indicator, it at least hasn't done appreciably worse -- and has still beaten a buy-and-hold strategy.

If you are not familiar with "Selling in May and Go Away," you may know by its other name: The Halloween Indicator.

According to a comprehensive review of its historical legitimacy that appeared in the December 2002 issue of the prestigious academic journal, American Economic Review, this pattern has existed historically in 36 of 37 countries studied. In each of those countries, average stock market returns from Halloween through May Day (the so-called "winter" months) were significantly higher than equity returns from May Day through Halloween (the "summer months").

In fact, the study found, the summer months' returns have averaged so much less than those of the winter months that almost all of the stock market's long-term returns have been produced during the winter months. That implies that simply going to cash between May Day and Halloween will have only minor impact on long-term returns while dramatically reducing risk -- a winning combination that would show up in a much improved risk-adjusted performance.

To be sure, the results of this study are based on long-term averages, and there have been many individual years in which the overall pattern did not hold up. Several such exceptions came during the recent bear market, for example, when the winter months from November 2007 through April 2008, as well as from November 2008 through April 2009, saw the stock market buck the favorable seasonality and fall sharply.

Nevertheless, the "Sell in May and Go Away" strategy has more than made up for these and other missteps and therefore is in the rarefied ranks of those select few market timing systems that truly have worked over the long term.

But, not willing to leave well enough alone, two market timing services I monitor have for a number of years tried to second guess the exact days on which a follower of this seasonal pattern would enter and exit the market. The first is the Almanac Investor Newsletter, edited by Jeffrey Hirsch, and the other is Sy Harding's Street Smart Report, edited by Sy Harding.

Both pursue surprisingly similar modifications to this basic seasonal pattern: Each relies on a technical indicator known as MACD to pinpoint the precise day on which they should get back into the market at the beginning of the six-month seasonal period, as well as when to get out at the end of that period. (MACD, of course, stands for moving average convergence divergence. It is a short-term momentum indicator, created several decades ago by Gerald Appel, editor of the Systems & Forecasts advisory service. It compares the relative movements of several moving averages of different lengths.)

Sometimes the modifications recommended by Harding and Hirsch are quite minor. In 2004, for example, Hirsch's version of this seasonal pattern for the S&P 500 index re-entered the market on Oct. 28, just three days before it otherwise would. But in other years the differences have been quite significant.

The Hulbert Financial Digest has data for both market timers' modifications of the Halloween Indicator back to mid-2002, eight years ago. The HFD calculates their returns on the assumption that, when they are invested in stocks, they earn the return of the Wilshire 5000 Index; otherwise they are assumed to be invested in 90-day Treasury bills.

To put these newsletters' success into context, consider that since mid-2002, a buy-and-hold has produced a 3.4% annualized return. A purely mechanical application of the Halloween Indicator (automatically entering the market on Halloween and exiting on May Day) would have produced a 4.3% annualized return. Note that this 4.3% return beats the market by even more on a risk-adjusted basis, since it was produced with half the risk.

Now consider the performance of Harding's modification of the Halloween Indicator: It produced a 6.1% return (annualized) over the same period, or 1.8 percentage points per year more than a purely mechanical application of this seasonal pattern, and 2.7 percentage points ahead of a buy-and-hold.

In fact, this performance is good enough to place Harding's version of the Halloween Indicator in 13th place for performance since mid 2002, out of the 132 timing strategies the Hulbert Financial Digest has tracked over this period.

To be sure, Hirsch's modification of the Halloween Indicator performed less well, producing a 3.9% annualized return. Though that is less than the 4.3% return of the pure version of the Halloween Indicator, note that it is still better than buying and holding.

What about this April? To be informed when these two newsletters actually trigger their buy signals, of course, you will need to subscribe to their services.

But one way in which their MACD-based systems are likely to trigger an early sell signal (but not the only way) is if the market is strong for a week or two and then quickly drops back.

If that happens, traders interested in taking the rest of the spring and summer off from stressing about the market may want to consider going to cash.