For Stocks, September May Be the Cruelest Month

September is fewer than three weeks away. Feeling nervous? Maybe you should be. For investors, the period between Labor Day and Halloween is proving an annual fright show. And no one knows why.

It was, of course, in September last year that Lehman collapsed and everything fell apart. But then it was also September-October 2002 that the last bear market plunged to its lows.

The 1998 financial crisis? It began late August, and rolled on for two months.

The famous crash of 1987 came in October. But most people have forgotten that the market actually started sliding downhill in late August.

That's almost exactly what happened in 1929 too. The big crash came in October, but the market peaked just after Labor Day. Prices began falling through September, then tumbled further still.

The worst month of the Depression? September, 1931, when the Dow fell about 30 percent.

It was also in September, 2000, that the bear market really got going.

The 9/11 crisis, of course, came in September. That was hardly caused by investors. But what is forgotten is that the stock market was already looking wobbly. In the two weeks before the terrorist attacks, the Standard & Poor's 500-stock index fell 7 percent.

The great panic of 1907? October. The great crash of 1873? September.

Yikes.

So is there really a September, or a Halloween, effect?

Since 1926, investors have lost nearly one percent on average during September, according to market data tracked by finance professor Kenneth French at Dartmouth's Tuck School of Business. It's the only month with a negative average return.. For each of the other 11 months, investors gained nearly one percent on average.

Other research takes the idea of an autumn dip even further. Georgia Tech doctoral student Hyung-Suk Choi studied the so-called "September effect" as part of his recent Ph.D. thesis. He looked at data for 18 developed stock markets around the world spanning up to 200 years, and found that in 15 of those markets, September brought red ink for investors.

Fund manager Sven Bouman and finance professor Ben Jacobsen concluded that investors in most world markets have historically fared poorly from May through October each year. They made their money between November and April.

Hence the old British investors' saying, "sell in May and go away, don't come back till St Leger Day." (But since St. Leger Day is in the middle of September, even that date may be premature.)

Some of the September or Halloween effect is caused by a few really bad years. But that's not the whole story. To reduce the influence of outliers I looked instead at the median result since 1926 instead of the statistical mean. The performance gap between September and the other months shrank from 2 percent to 1.4 percent. That's smaller, but it's still a difference. The median September saw losses of just 0.07 percent. But the median month for the rest of the year gained 1.37 percent.

As for the causes of a possible September effect, most are stumped.

"There haven't been any good academic stories to explain it," admits Michael Cooper, finance professor at the University of Utah's David Eccles School of Business. "One credible explanation is just luck."

It's been suggested that mutual funds drive down the market by selling their losing stocks before their October 31 year end. Or that third quarter profit warnings come in early September, raising fears about full-year results. Or that these autumn crashes used to be related to the harvest, as Midwestern banks withdrew capital from New York.

(Still another theory cites seasonal affective disorder. Investors simply get more risk-averse, and more prone to sell, as the days get shorter. That's the case argued by York University finance professor Mark Kamstra and others.

So what, if anything, should you do?

In practical terms, maybe not that much. For most people, even a performance difference of one or two percentage points isn't going to cover the transaction costs of selling before the end of August and re-entering the market a month later. And stock market patterns aren't ironclad. The market may even jump in September, as it did in 2006 and 2007.

Perhaps the best you can do is brace for turmoil.