There has been no shortage of discussion of late discussing the historically poor performance of the stock market in September. On the face of it September has been the has been the worst performing month for the stock market. This entire discussion has been ramped up by the strong performance of the market since the March low and the exceptionally poor performance of the market on September 1st. Indeed with stock market valuations getting stretched out once again it makes the possibility of at least a correction in September a decent bet.
We have no idea how this September will eventually play out, but it begs the question: what do you see when you read these results? Do you see a market anomaly that needs exploiting? Or do you see an exercise in data mining run amok? In a sense you can view your response to the September effect as a Rorschach test.
The September effect reared its ugly head again in a piece by Brett Arends at WSJ.com. Arends does a nice job citing the relevant academic finance research showing in a number of different ways how September is an outlier to the downside. However this research sheds little definitive light as to the causes. As Arends writes:
As for the causes of a possible September effect, most are stumped.
Arends also asks what investors should do with this information. Not much given the size of the effect and the potential costs. However investors should:
Perhaps the best you can do is brace for turmoil.
This sort of approach is echoed in a piece by Steve Smith at Barron’s. Smith notes that if the market were to hit rough waters in September we would likely see a jump in options volatility. A trader could get long the VIX to take advantage this relationship. However Adam Warner at the Daily Options Report notes that the VIX may already be reflecting this scenario.
Mark Hulbert at Marketwatch.com also weighs in the topic noting the fact that the September effect has been known about for quite some time and persists. Hulbert notes that here is a risk that this statistical result is simply a result of data mining, especially in light of the lack of concrete explanations why it should persist. Hulbert asks what is an investor to do?
It depends on how much importance you place on having theoretical explanations for statistical patterns. While most statisticians place a lot of importance of them, there are many on Wall Street who disagree.
There is good reason to be skeptical of the September effect. In a broader sense, do we really have enough good historical data to make meaningful statements about relative monthly performance. Carl Richards at behavior gap notes that our stock data really is insufficient to really say much about the equity risk premium.
I am not saying that 82 years of data is worthless, I am just suggesting that we move forward with a sense of caution instead of a false sense of precision.
Doug Short at dshort.com takes a longer term look at the historical seasonality data and notes that:
As we can readily see, there has been little in the way of long-term consistency in monthly seasonality based on monthly averages of daily closes.
Short also takes an alternate look at monthly seasonality based not on calendar closes but on average daily closes. In this example September looks not nearly as bad. The question is whether all of the noted research adds up to anything.
Jeff Miller at A Dash of Insight takes to task the purveyors of the September effect. Given the fact that historical stock market data has been thoroughly researched we should not be surprised to find a September effect arise.
There are twelve months. There will be a distribution. Some will be good and others will be bad. Always.
In short, we really can’t test the September effect in any sort of scientific manner. This data has already been thoroughly snooped. Miller believes there are four reasons why investors should discount the noise surrounding the September effect. Quoting (at length):
- This is a widely advertised theory. Even if you do not believe in completely efficient markets, one would expect some anticipation.
- We have already had a decline of nearly 2%, exceeding the expected monthly decline on the first day. Should we now expect normal trading for the rest of the month?
- The evidence shows that the September pattern is not a statistically significant deviation.
- Other seasonal methods (Sell in May, Presidential Cycle) have not worked well in this time of turmoil.
Miller’s critique aside, we couldn’t resist positing our own explanation for the September effect not often mentioned: Congress. There has been academic research showing that the stock market performs less well when Congress is in session.* Guess what. Congress’ summer recess typically ends after Labor Day (early September). So September may simply be a month weighed down by the Congressional effect.
On first glance the September effect is compelling. However absent some sort of fundamental explanation, is it really there? On second glance the statistical evidence is less compelling. As Jeff Miller notes we can easily spin a story about why the September effect exists and why it may persist in the future. Given this approach, an investor should give wide berth to the hype surrounding the September effect and focus on other more relevant fundamentals.
Another approach might be to acknowledge that markets and market participants are not always rational creatures. Given the superficial validity of the September effect it may very well be the case that the lore of September causes enough investors to change their portfolios to make it a self-fulfilling prophecy. In this case it would be foolhardy not to try and beat the crowd to the punch.
Two different takes on the September effect. Two different takes on how to invest. It begs the question, what kind of investor are you?
*Ferguson, Michael F. and Witte, Hugh Douglas,Congress and the Stock Market (March 13, 2006). Available at SSRN: http://ssrn.com/abstract=687211