Will September Hurt Equities?

Historically, September has on average not been a good month for equities. A well-known fact, which has been mentioned many times before. But, that does not mean it’s useless to check up on this stock anomaly once in a while. And after doing so, it’s confirmed that the average monthly return on the S&P500 index is negative in September. However, my analysis also shows that in the last couple of decades September was no longer the worst month for equities. Furthermore, taking a look at all calendar month returns over different time periods reveals some interesting findings.

September month, losing month

The graph below shows the average monthly return on the S&P500 index per calendar month since 1928. The starting point is aligned with that of Bloomberg, as their S&P500 index data goes back to 1928. As can be derived from the graph, September is indeed the worst month for equities. Since 1928 equities have on average lost 1.1%. The average return is at least about 1.0% less than any other month. The difference between the best (July) and worst month (September) adds up to 2.63%. That is huge, especially when taking into account that this is more than half the historical equity risk premium (which I estimate to be 3.5%-4.0%). Also, a long-short strategy that goes long equities only in July and short equities only in September would yield a pretty decent return, considering the fact that you only have to bear the risk on equities in two out of the twelve months a year. The risk-return profile of this strategy would be pretty impressive.

Click to enlarge

1928-now is a long time horizon. The question rises if September is also bad for equities on shorter horizon? To find an answer to that question I have calculated the average monthly returns for two other time periods, 1970-now and 1991-now. 1970 is about halfway of the whole data sample (it also represents a good starting point for a comparison with other countries as the MSCI global and country indices start in 1970). The 1991-now period is used to take a closer look at calendar returns in recent history.

Cruelness of September reduced

The average monthly returns of these two time periods are plotted in the graph below. And in each of them September has been a lousy month for the S&P500 index, resulting in negative returns. In the first sub period (1970-now), the results are reasonably comparable to those of the whole period. September is the cruelest month. On average you would have lost 0.7%. The difference between the best and worst month is 2.50%, pretty close to that of the whole period. However, July is no longer the best month. That honour goes to December, which was already near the top spot when using the whole data sample. July has fallen back to only seventh place.

Click to enlarge

In the most recent period (1991-now), things look a bit different. September is still a month to keep your money in the bank. But the average loss since 1990 has ‘only’ been -0.3%. More importantly, September is by no means the ultimate losing month. August and June both realized bigger losses and February is not that far behind September. So, if the average monthly returns per calendar months were studied only during the last 20 years or so, September would not been given the honour of being the cruelest month for the S&P 500 index.

September overrepresented within worst months

That said, there are, of course, other ways of judging the cruelness of the September month. For example what is the percentage of the worst months, those with the biggest losses, for the S&P500 index that are represented by September? The graph below shows the percentage share of each calendar month of the 50 worst return months since 1928. It is clear that September has a disproportionately high chance of realizing a very bad return. Of the 50 worst months since 1928, 10 were September months. That represents 20% of all 50 months, almost 2.5 times more than expected when the worst months are evenly distributed among the calendar years. And, the very worst month for the S&P500 index since 1928 was also a September month. In September 1931 stocks went down a staggering 30%.

Click to enlarge

Now, how does September rank within the 50 worst return months in the most recent period starting in 1991? Well, although September was not the worst month based on average return, it is still the worst month when looking at the biggest losses. Out of the 50 worst return months since 1991, 7 of them were September months, or 14%. This is still almost twice as much as expected. August and June, which realized lower average returns, had fewer hits within the sample of 50 worst return months. So, from the perspective of maximum losses, September scores pretty badly in both the whole and most recent data sample.

So far, this analysis was mainly focused on September and its reputation of being a cruel month for equities. But the comparison of average returns of calendar months during different time periods also reveals interesting information about other months as well. The graph below sheds some light on this information. The bars show the difference in average monthly return between different time periods. The blue bars show the difference per calendar month between the most recent period (1990-now) and the whole period (1928-now) and the orange bars show the difference between the most recent period and the period starting in 1970 (1970-now).

Click to enlarge

Shift in January Effect?

There are a couple of observations that stand out. First, January, traditionally considered to be one of the best months for equities, has lost some of its charm. Since 1991 the average return has been 0.6%, ranking only seventh out of all months. Compared to the two other periods the average return has decreased by over 0.5%. This could indicate that, as a result of increasing acquaintance with the January effect, the performance of the first month of the year has deteriorated. At the same time, December has become even more profitable. In every sample December stands out as an exceptionally sweet month for the S&P500 index. Still it managed to increase its average return in the most recent period. Maybe investors are anticipating the January effect and therefore start investing in December.

Second, the average return of the summer months (June, July and August) has come down quite a bit. Especially since 1991 the summer has been the season to avoid stock markets. Two out of the three summer months realized a negative return. At the same time the spring months (March, April and May) have done much better since 1991. April is actually the best month in the sample (the late T.S. Eliot would not have been pleased)[i]. Third, and last, combining the monthly changes reveals that both in the period starting in 1970 and in 1991 the stock market seasonality effect has become stronger. The difference between the ‘winter’ (November – April) period and ‘summer’ (May -October) period has increased compared to the 1928-now sample. The difference between the winter season and summer season is over 5.5% in both periods, which is massive given the fact that the average return for equities has been around 8%. For more on the stock market seasonality, the reason for this split up in winter and summer periods and explanations for this anomaly please see Doeswijk (2005)[ii].

Click to enlarge


September has historically been a bad month for the S&P500 index. The cruelty factor is, however, by no means constant. Especially during the most recent period, starting in 1991, September does not really stand out because of its cruelty. Other months have realized bigger negative returns. This in contrast to the average monthly return in September measured from 1928 onwards. Over this period September underpinned its reputation as the worst month for equities. When looking at worst returns, September also pops up in all data periods. September represents a disproportionate number of large negative returns.

[i] en.wikipedia.org/wiki/The_Waste_Land

[ii] Doeswijk, Ronald Q., The Optimism Cycle: Sell in May (November 2005). Available at SSRN: ssrn.com/abstract=643583 or dx.doi.org/10.2139/ssrn.643583

(click to enlarge)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s