Do a search under "seasonal investing" on Google and you'll get a Charles Schwab article called "Does seasonal investing really work?".  Some research guys at Schwab tested the "sell in May" thesis with data going back a thousand years, and concluded that the strategy doesn't work.  Therefore, SEASONAL INVESTING DOESN'T WORK.  There you have it.

The article begins with "At Schwab, we believe in..."

Oh well, give Schwab credit for discouraging churning, at least in this particular article.  There's also a nice bar chart on the page showing monthly market performance since 1926.

I'm not big on the "sell in May" philosophy myself, but I do believe that if the data were more recent (going back to 1950 or thereabouts, as opposed to 1926), then the case could be made that this sort of "seasonal investing" actually does work with respect to the major indices.  It's an easy statistic to manipulate, depending on how far back you go in history, which market you choose, where you put the money between May and November, etc.

I'm reminded of football commentary, where someone proclaims something like "the Detroit Lions haven't won in Dallas since 1973".  Are we supposed to believe that this fact has any bearing on the Lion's current chances of winning in Dallas?  I do believe the Lions have replaced some of the players that were on the roster in 1973.

Likewise, does data that goes back to Lindbergh's time really have any relevance for an investor in 2003?  Seasonality has underlying causes, however fuzzy they might be, and when these causes...tax laws, corporate laws, accounting trends, politics, saving rates, speed of information transfer, whatever...change, so should seasonality.  Schwab treats "seasonality" as something that is set in stone, and then conveniently proceeds to "debunk" it.

Now, having argued against Schwab's dismissal of "sell in May", we offer our own dismissal.  Click here for our reasons..

**********************

March 31, 2004...Here's a link on the "sell in May" strategy.  The gyst of the article, in case the link dies, is that the philosophy may not work particularly well in election years, where the Dow has averaged better than 5% gains from May through October.  Congratulations to Mark Hulbert, the writer, who actually noted that this election year trend is not significant at a 95% confidence level...rarity of rarities in the popular financial press!

**********************

September 2, 2004...We came across an interesting article on the subject of the historical tendency for September to be a lousy month for the general market.  Unfortunately, the URL is impossibly long and we can't offer a link for technical reasons.  Just go to www.cbsmarketwatch.com  and look for Mark Hulbert's columns.  In the article,  Hulbert points out that September has been THE worst performing month in 6 out of 10 decades since 1900. What's more, in only one decade (1911-1920) did the market perform better than average in September. 

He then goes on to say that he wouldn't advise taking any action based on this historic trend. Why? Because no one has identified a reasonable-sounding cause for this phenomenon. Here, we find ourselves in the rare position of chastising a market analyst for being overly skeptical. 

Skepticism takes many forms. One that we and Hulbert endorse is that directed at false or questionable causes. For example, one might find that the stock market tends to peak when sunspot activity is highest. It's awfully hard to imagine how sunspot activity could be relevant to that of the stock market. What's more, it's difficult or impossible to know how many different phenomenon were examined before noting that sunspots and the market have some sort of correlation. This, as Hulbert knows, falls into the category of "data mining". 

But that's not what we're seeing with respect to the "September effect". Here, we've identified a phenomenon but haven't posited a cause. Should our lack of imagination, or perhaps our inability to grasp complex, chaotic, multi-variable causes, be cause enough to dismiss a phenomenon that certainly does fall inside the usual definitions of statistical significance? 

Given his writings on the subject, it does appear that Hulbert endorses the "sell in May" philosophy. The difference is that there are ready-made explanations for the tendency for the market to be weak from May to late October, whereas it's more difficult to explain the "September effect". 

If a drug works well within the usual measures of statistical significance, but scientists don't understand the chemical basis for its activity, should the FDA refrain from approving it? Certainly not! Consumers take the medicine and scientists do further studies. 

Exactly how significant is the September effect? Well, it all depends on precisely how you ask your question. If we take Hulbert's own decade-to-decade data, we calculate there's about a 0.15% chance that this effect is a fluke. You simply generate a list of 10 random numbers from 1 through 12 (1 being the best month in a decade of data), sum the 10 numbers, go through the above exercise repeatedly, and see with what frequency the sum equals or exceeds Hulbert's sum (101). You could do the same exercise with averages, instead of sums, but the result would be the same. 

One could argue that recent decades should be given extra weight. After all, we're living in an era where ordinary people have cars and telephones and rapid access to information. In that case, the numbers should become even more significant, since September has been the worst month in each of the last three decades. 

You could simplify things to make it a bit more intuitive. If a decade of Septembers outperforms, you call that a "white" September. A decade of underperforming Septembers is "black". Now, what are the odds that, given an unlimited supply of black and white marbles (in a 1:1 ratio), you could randomly pick 10 marbles and get one or zero white marbles? The answer is 11/1024...about 1%. 

To complicate the above result a bit, it goes without saying that six months HAD TO underperform in the 1901-1910 time frame. So it might be better to ask "what are the odds of a repeat performance" in 8 out of 9 of the next decades? Then the answer becomes 10/512...about 2%. 

You can argue with my methodology. But you'd be hard-pressed to make a case that the September numbers are simply a fluke, falling outside the usual standards of statistical significance (p>5%). 

Suppose that September had been a negative investment month in the 1911-1920 time frame. Or, suppose that September was the worst month in EVERY decade. Or, suppose that September was the worst investment month in every year of every decade. At what point does one think in terms of investing with the "September effect" in mind? It appears that Hulbert would say "never". 

Having said all that, we've got a lot of respect for Mark Hulbert. He's one of the few prime-time financial writers who pay any attention to statistics and probabilities and reason when dissecting the market. 

But what should an investor do? Well, we've identified several areas of the market that tend to fare well in September. So we'd argue that one needn't move into cash...one should simply place additional emphasis on certain sectors. However, just as we can't promise that any particular September will be negative for the most general market, we can't promise that any particular sector will be positive...diversification remains important. Goes without saying, of course.

May 5, 2005:  Over at CBSMarketWatch.com , Mark Hulbert has some thoughts on the "sell in May and stay away [for six months]" strategy.  Historically, this period of the year has been a bad time to own stocks.  The question he addressed was whether a weak six months prior to the "sell in May" period alters the dynamics of the strategy (might we expect a relatively strong May-October this year, given the lousy performance of the market over the last six months?). 

Looking at Dow data, he observes that the Dow has lost money in 43 of the Nov-May periods going back to 1896.  In these instances, the Dow lost 13 times (13/43...30% of the time) in the following May-October period.  Percentage-wise, he found that the Dow lost 35% of the time following a strong Nov-May showing.  He then concluded that the market's Nov-May performance has no bearing on the May-Oct performance.

We'll accept that.  However, one should note that if the Dow loses 35% of the time after a strong Oct-May period, and 30% of the time after a weak Oct-May period (with the Dow only being sliced into those two categories...it's either strong or weak), it's clear that the Dow has lost less than 35% of the time during the time of the year when you're supposed to be putting your money into bonds (or whatever the accepted wisdom is).

Of course, if a handful of the historical May-Oct periods lost in a very big way, that could mean that the period has been unprofitable for investors.  But then, of course, one has to question the significance of a "trend" that has been skewed by just a handful of exceptionally rotten May-Oct performances. 

No need to speculate about the above, actually.  Just take a look at our Dow seasonal charts going back to 1929.  Note that if you'd divested your stocks precisely at the early May peak, and re-entered the market precisely at that late October trough, the market would have been flat and you could have done better by putting your money in bonds.  But you've really got to "cherry pick" that period with precision.

We've commented about the "sell in May" strategy on numerous other occasions.  It certainly has been an uninspiring time of the year to buy a big index like the Dow, but we've identified any number of sub-groupings of stocks that have fared well during the period.  Looking at the seasonal chart, it's also apparent that there is a June to late August "sub-period" that has fared well.

At the risk of blathering, we'd also question, for the umpteenth time, whether it really makes any sense at all to be dredging up market data from the late 1800's when you're trying to aid investors in their buy-sell decisions in the year 2005.

Copyright © 2010 MarketSynopsis.com. All rights reserved.