Some stories are just great for recycling. And the ‘September is bad for equities’ story is one of them. You don’t have to dwell too long in the financial markets to understand that September is generally believed to be the ‘danger’ month for equities. But it can’t hurt to test this hypothesis from time to time. Especially since recent history suggests September is not that abysmal anymore.
An appropriate way to get a grip on the negative effect September has on stock markets, is to look at history. The graph below shows the average return per calendar month on the S&P 500 index since 1928. During this period, September comfortably ranks as the worst month for equities. An investor who entered the market only during September would have lost more than 1% each time on average. That is roughly a full percent more than in the other months that have generated negative returns since 1928, February and May.
Click to Enlarge
But a lot can happen in more than eighty years. Therefore I calculated the average return realized in September for a number of other, shorter periods as well. The results are shown in the graph below. The graph reveals an interesting pattern. Over time September has become less damaging for stock markets. In fact, over the most recent period, starting in 2002, the average return on the S&P 500 index is positive. Moreover, in the last two sub periods, those starting in 1991 and 2002, September was no longer the worst month for equities. It ranked ninth and sixth, respectively. September is moving up in the ranking of calendar month returns over time.
Click to Enlarge
Let’s take a closer look at the average calendar month returns for the most recent period starting in 2002. As the graph below shows, a couple of interesting shifts have occurred. First, instead of September, August now ranks amongst the worst months for the S&P 500 index. August has pretty much traveled in the opposite direction as September, falling in ranking consistently over the different time periods (not shown). Second, January, traditionally perceived as one of the best months to invest in equities, has also completely reversed. And, as for August and September this happened gradually as well. At the same time, December affirmed its status of ‘investor-friendly’, hitting top spot in the sub periods starting in 1946, 1970 and 1991.
Click to Enlarge
The changing pattern of average calendar month returns, with August getting worse, while September improved and the reversal of January returns in combination with firming December returns, could imply that investors are anticipating historical return patterns. Perhaps investors no longer wait for September to hit them, by selling (some of) their exposure in August, shifting the average calendar return in the process. Similarly, in anticipation of solid returns in January, investors already double up in December, reducing the positive bias in January.
The calendar month data offer an additional opportunity to check whether investors are anticipating historical return patterns. The so-called seasonality effect is also closely linked to calendar months. Historically, almost all of the (excess) return on stocks is made during the ‘winter’ period (defined as the six month period from November until April), while equity returns are very modest during the summer period (spanning from May until October). Hence, the expression ‘Sell in May and come back in November.’
If investors behave in the same way, as the shifts in returns for August/September and December/January suggest, the seasonality effect should also move forward in time. Investors will start selling equities already in April and re-enter the market one month earlier, in October. The graph below shows the results of both the new and traditional seasonality effect for different time periods. The seasonality effect is calculated by subtracting the return during the summer period (short equities in summer) from the return during the winter period (long equities in winter).
Click to Enlarge
The results are a bit inconclusive. This has mostly to do with the most recent period starting in 2002. During this period the pattern that is revealed by the four other sub periods is broken. Based on the period starting in 2002 the conclusion would have to be that there has been no shift in the seasonality effect. The return difference between winter and summer, with winter now starting in October, instead of November, is practically zero. For the traditional seasonality effect the return difference is approximately 4%, which is substantial.
However, the other sub periods do seem to suggest some change in investor behavior. The return difference between the ‘new’ winter and summer steadily rises from 1.3% to almost 5% per year. This could imply that investors do anticipate the Sell in May effect by selling already in April and buying again in October. The difference between the new and the traditional seasonality effect declines from 1.7% for the period starting in 1928 to only 0.6% in the period starting in 1991. That said; it is also true that the new seasonality return never tops the traditional seasonality effect.
Looking at average calendar month returns through time yields some interesting results. First, September, perceived by many as the ultimate horror month for equities, has become much more investor-friendly in recent years. At the same time August has steadily fallen in the calendar month ranking over time. Although less pronounced, a similar return patterns seems to appear for the months December and January. This could suggest that investor behavior changes over time as investors try to anticipate ‘good’ and ‘bad’ months for equities. When excluding the most recent sub period possible investor anticipation also seems to appear in the seasonality effect. However, the most recent sub period starting in 2002 does not confirm this pattern. Hence, while there are indications that investors anticipate historical return patterns, the evidence is not conclusive at this point in time.