First, we'll look at the fourth quarter and then October itself...

The quarter has averaged about 4.4% gains since 1993.  Volatile stocks have had some very good years (50%+ gains in Q4 2002), and so are high on our list of non-risk-adjusted long strategies.  They've also suffered (1993-1997, 2000, 2008) in individual years.  In fact, 1993, 1995 and 1996 were positive quarters, yet volatile stocks underperformed.  

After risk adjustment, volatile stocks drop out of our lists.  Good risk-adjusted groups include late September losers, stocks with high market-caps, and non-volatiles.  Stocks that finish the third quarter with strong gains are short candidates..

One impromptu calculation we did was the following:  buying the 4% of stocks with weakest "momentum" coming into October and selling at the end of the quarter resulted in compounded gains of 568% over the 12 quarter period of 1993-2004.  Bear in mind that we selected out a basket of about 50 stocks in each of these 12 years...a couple fluky gains are not going to bias the result to any great extent.

Regarding industry groups, computer stocks do well, particularly software.  REIT's and oils tend to suffer.

Investors with a quarter-long investment horizon should be aware that December has a way of reversing a number of more general quarterly trends.  Take a random selection of our yearly seasonal charts...long term or short term data, hi-tech/low-tech... and you'll probably be surprised with the number of situations that have inflection points at the beginning of October and the beginning of December.  Thus, for example, we wouldn't argue with the strategy of holding software stocks until the beginning of December.

There does not seem to be any special tendency for stocks that prospered in this period over the previous three years to perform any better or worse than other stocks.  In fact, there's a slight tendency for stocks that have been weak in this period over history to bounce back with a better-than-average performance.  For fans of such strategies, however, we offer the following table of stocks with particularly strong or weak historic fourth quarter performances:

"Industry momentum" does not seem to play much of a role as a predictor of gains or losses in the quarter.  As usual, this sort of strategy works best on shorter time scales, and not at the end of quarters, when losing groups have a habit of turning around.

Concerning the month of October...

Despite the reputation October has for being a losing month (via several famous monster crashes in the 80's), it has only lost around .7% when you glom all the data together since 1993.  This "flatness" doesn't mean there isn't potential for nice long or short gains, though.  Going back further in time, to 1984, the month has lost the same .7%..

Cutting the month into halves, the first half is negative (1.5% losses, on average), and the second half is positive, gaining almost 2%.  Yes, this results includes the ugliness of 2008.

Probably the most interesting trend is the tendency for large caps to gain and small caps to lose.  There are a few strategies that have proven more potent, but this one comes to the fore after risk-adjustment.   The effect is strongest in the first half of the month, and loses significance if you get too picky...don't try to improve on it by choosing the absolute highest market-caps (or lowest caps if you intend to go short).

On the long side, a significant strategy in both positive and negative Octobers has been to purchase late September losers...depending on exactly how you played it, you could have realized upwards of 7% average gains in the month since 1993 following this sort of approach.  The strategy has worked well in any number of Octobers...we were a bit taken aback by the pronouncements of several well-known gurus to the effect that 2002's gains in beaten-up stocks were anomalies.  These pronouncements are especially astonishing when you look at the monster gains that were available in Oct 2001 via purchase of stocks that lost in a big way over the third quarter.  

The "buying September losers" strategy works best in the first half of the month, but isn't entirely consistent...in 1999 and 2003 a better first half strategy would have been to stick with yearlong winners for the first couple weeks of October, and then go with recent losers.  Easy to say in retrospect...difficult to recognize in practice, of course.

For shorts, simply reverse the above...sell late September winners.

Note that the strategy of buying losers, particularly yearlong losers, is also significant in November, so there's not necessarily any compelling reason to dump this group at the end of October.

Again, beaten-up stocks also fare well during the second half...the effect was monstrous in 2002 (gains of 30%+ for big 3 month losers in a span of just 10 days!), but also was strong in 1998, 2000, 2001, and 2004, with 1995 as the only year where the trend did a minor flip-flop.

We've heard some analysts marvel at the gains available in October 2002 via the purchase of "junky stocks"...money-losing or debt-ridden operations teetering on the edge of survival.  At the same time, we've noted that some folks who are aware of seasonal effects choose to exclude such junky stocks from their buy lists.  It's possible, however (possible, we emphasize...we can't make a judgment either way given the lack of a clear-cut junkiness indicator), that it's precisely these high-risk, fundamentally-unsound companies that reap the biggest gains at certain times of the year, and by filtering such stocks from your lists you are reducing profits.

Software stocks have done well over the years, particularly those that have suffered in September.  

The concept of "industry momentum" does not seem to play much of a role in deciding which groups will prosper and which will gain in the full month of October.  If you're married to such an approach, look for industries that made nice gains on the last session in September to continue to gain in October.  In the second half of the month, you might be able to profit by going with stocks whose industries have taken a beating over the last 3 months to 1 year.

Last October's strongest performers have performed better than the market in general, though the level of significance is weak...this result could be a fluke.  Stocks with historically weak performances, however, also have a habit of performing well in October.  The table below lists stocks that have performed particularly well in October on a historic basis.

REIT's have not performed well in either half of the month, particularly if you adjust for risk.  The effect is quite significant.   Searching the internet (try "REIT" and "seasonality" on Yahoo or Google), you'll find numerous articles on REIT seasonality, examining these equities from the point of view of large-cap vs. small-cap, mortgage REIT's vs. equity REIT's (better than 80% of publicly traded REIT's fall in this category), etc.  The bottom line:  these equities exhibit strong seasonal behavior in terms of gains and losses, as well as volatility.  Gains and losses in one month have also been shown to predict gains and losses in the subsequent month.

In 2003, REIT's suffered as expected.  We decided to test the notion that it's not REIT's per se, but stocks with high dividends that suffer in October.  Comparing 55 high dividend REIT's against 55 low dividend REIT's, though, we found that both groups performed similarly in October of 2003.  Looking more deeply at the data (an extracurricular experiment not on the website), there did seem to be a tendency for REIT's that performed well in October 2002 to do so again, and for last October 2002's losers to lose yet again.  Interesting, though the statistical significance was not mind-blowing.

In another series of tests, we looked at October S&P 500 data going back to 1983.  Most of the above-mentioned trends manifested yet again, though at lower significance levels in this smaller, larger-cap, dowdier group of stocks.  Going with stocks with late September losses was again indicated.  Tech stocks have performed well.  Stocks that performed well in the previous October or 3 Octobers continued to perform well.  The results here did differ where volatility was concerned...non-volatile stocks outperformed...but bear in mind that the period examined included a couple of October market crashes...something we haven't seen for quite a while.  

Another difference between the periods 83-93 and 93-06 is the growing tendency for mutual funds to end their fiscal years in September or October...one of many examples of why it may not be best to look at the oldest data available when trying to extract seasonal trends.