September by the Numbers

by Mark Hulbert

Wednesday, September 1, 2010

Commentary: Little statistical hope that stocks will beat month's bad reputation

I have good news and bad news when it comes to slicing and dicing the historical data as it pertains to September.

The good news is that it is possible, by carefully reading the statistical tea leaves, to get advance insight into whether any given month is likely to do better or worse than average.

The bad news: Those tea leaves provide no such hope that this September will be able to beat its historical reputation as being awful for stocks.

The Historical Record

Let me begin by reviewing the dreadful details of September's record. Since 1896, when the Dow Jones Industrial Average was created, the Dow has lost an average of 1.15% in September. The average gain for all other months is 0.71%. That spread of 1.86 percentage points is statistically significant at the 95% confidence level that statisticians often use to determine if a pattern is most likely genuine.

Furthermore, there has been a remarkable consistency to the stock market's dismal performance during September. During each of the past nine decades, for example, September's rank relative to other months in terms of performance was never higher than ninth. It was dead last in five of those nine decades -- including the most recent one.

To be sure, September's terrible reputation is widely known, and patterns often stop working once too many investors begin trying to exploit them. But the pattern has been widely known for many years already, and shows no signs of weakening.

For example, it was more than 20 years ago that (as far as I can tell) the first academic study appeared in which September's significantly-below-average return was noted. Since that study was completed, the spread between September's average return and that of all other months has been even wider than it was up until that point.

Beating the Odds

This terrible record notwithstanding, however, hope could still have been held out for September -- provided other statistical patterns had fallen in place. Unfortunately, those other patterns have failed to do so.

The strongest of those other patterns has to do with how stocks have performed in the months leading up to September. The market exhibits momentum, and September tends to at least somewhat beat the historical odds if the market is riding a strong wave of momentum going into the month.

Over the past 110 years, for example, September has produced an average loss of 2.7% whenever the stock market lost ground during August and was in the red for the first eight months of the year. It lost "just" 0.6%, in contrast, following stock-market gains during August and over the year-to-date period.

Unfortunately for this September, however, this momentum factor is working against the stock market. The Dow enters the month having not only lost ground during August, but as of Aug. 31 is also sporting a loss for the year-to-date.

Another pattern that has had some modest success in forecasting stock-market returns is the level of the CBOE's Volatility Index as of the end of the previous month. Despite the VIX's reputation as a contrarian indicator, my analysis shows that September tends to be a better month for stocks whenever the VIX is relatively low going into the month.

For example, since 1990, when the volatility index was created, September has lost an average of 2.8% whenever the VIX was above 20 at the end of August. In contrast, the stock market on average was flat during September whenever the VIX at the end of the previous month was below 20.

Once again, however, this pattern holds out little hope for this coming September in particular: The VIX as of the end of August stood well above this 20 threshold.

The bottom line? The stock market is not guaranteed to fall during September, of course. But, if you want to nevertheless bet on stocks rising over the next month, you will have to base it on something other than how the stock market has performed in prior Septembers.