Seeking Alpha
Research analyst, newsletter provider, portfolio strategy
Profile| Send Message|
( followers)  

Most human activities have seasonal cycles. The stock market's seasonal cycle is not reliable every year, but it is powerful over the long term.

You probably have already heard of its four best-known effects:

  1. "Sell in May and Go Away." This old saying tells that the odds are against you from May until October.
  2. "Halloween Effect." The odds are favorable again in November.
  3. "Santa Claus Rally." December is usually a good month.
  4. "January Effect." It is a small cap rally in January.

Globally, it is said that all the positive return of the stock market is done during the 6-month period from the 1st of November to the 30th of April, and that it's better to be out of the market from the 1st of May to the 31st of October. But is this just a Wall Street legend or does it really work?

Many research articles have been published on the subject, generally with partial data and unconvincing explanations. Until recently there was no exhaustive study to prove that one of the best quantitative indicators might be the calendar.

Late in 2012, Ben Jacobsen and Cherry Y. Zhang from Massey University (New Zealand) wrote two unbelievable articles: Are Monthly Seasonal Real? A Three Century Perspective (74 pages) and The Halloween Indicator: Everywhere and all the time (55 pages). This is the lethal weapon against skepticism. Compiling available data since 1693 in 108 countries, they claim that not only the discrepancy between Summer and Winter is as old as stock data, but that it can be observed worldwide and has increased in the last decades. I think that these academic papers are a must read for anyone seriously involved at any level in the stock market. At the time these lines were written, both of them are available on the Social Science Research Network website (SSRN eLibrary). I will borrow their global results here, in-depth statistics can be found in the original articles.

The following table shows monthly average returns in percentages from different sources. The first column is a summary of the most important numbers in Jacobsen & Zhang's papers, the second and the third ones are drawn from online data sources, the fourth one gives results obtained by my own simulations. Like Jacobsen, I call "Winter" the 6 months from November to April and "Summer" the 6 months from May to October.

Average Return % 316 years(1) 82 years(2) 50 years(2) 14 years(3)
1693-2009 1929-2011 1961-2011 1999-2012
January 0.69 1.0 1.2 -1.6
February 0.09 0.0 0 -1.3
March -0.03 0.4 1.1 2.7
April 0.49 1.4 2 2.7
May 0.02 -0.2 -0.1 -0.8
June -0.12 0.5 -0.6 -1.22
July -0.31 1.5 0.9 0.34
August 0.44 0.8 0.2 0.39
September -0.49 -1.3 -0.8 -1.28
October -0.5 0.0 0.5 1.32
November 0.35 0.8 1.2 1.2
December 0.81 1.5 1.5 2.6
Winter 2.42 5.20 7.20 6.36
Summer -0.96 1.28 0.09 -1.27
Whole Year 1.44 6.55 7.29 5.01
Difference Winter-Summer 3.38 3.92 7.11 7.63

(1) Jacobsen's Global Financial Data Index

(2) Dow Jones Industrial Average

(3) SPDR Dow Jones Industrial Average (NYSEARCA:DIA)

The best months of the year are April and December (joined by March in the recent years), the worst ones are May and September. September is the only month to be negative on the four timescales. Please note that January has been one of the best months in the past, and the worst one for at least 14 years. The least we can say is that the January effect didn't work so well from 1999 to 2012: U.S. big cap and small cap indexes have both a negative average return in January.

Jacobsen and Zhang have also noted that the U.S. stock market had historically a good average winter gain among developed countries.

Here is the Dow Jones Industrial Average return by season from 1950 to 2011. The Total Return and Total Drawdown columns give the return and the drawdown since 1950 to the end of the period (source of seasonal returns: Market Seasonality: Capitalizing Upon Summer Decline by Mateo Blumer on Seeking Alpha; source of total returns and drawdowns: my own calculation).

DJIA Return% May-Oct Total Return% May-Oct

Total Drawdown% May-Oct

Return% Nov-Apr Total Return% Nov-Apr Total Drawdown% Nov-Apr
2011 -6.7 -10.21 36.36 13.30 6919.65 0.00
2010 1.00 -3.76 31.79 15.20 6095.63 0.00
2009 18.9 -4.72 32.47 13.30 5278.15 8.69
2008 -27.3 -19.86 43.20 -12.40 4646.82 19.41
2007 6.6 10.23 21.87 -8.00 5318.75 8.00
2006 6.3 3.41 26.71 8.10 5789.94 0.00
2005 2.4 -2.72 31.06 8.90 5348.60 0.00
2004 -1.9 -5.00 32.67 1.70 4903.31 0.00
2003 15.6 -3.16 31.37 4.30 4819.68 0.00
2002 -15.6 -16.23 40.63 1.00 4616.85 0.00
2001 -15.5 -0.75 29.66 9.60 4570.15 0.00
2000 2.2 17.46 16.75 -2.20 4161.09 2.20
1999 -0.5 14.93 18.54 0.04 4256.94 0.00
1998 -5.2 15.51 18.13 25.60 4255.20 0.00
1997 6.2 21.84 13.64 21.80 3367.51 0.00
1996 8.3 14.73 18.69 16.20 2746.89 0.00
1995 10 5.94 24.92 17.10 2349.99 0.00
1994 6.2 -3.69 31.74 10.60 1992.22 0.00
1993 7.4 -9.31 35.73 0.03 1791.70 0.00
1992 -4 -15.56 40.16 6.20 1791.13 0.00
1991 6.3 -12.04 37.66 9.40 1680.73 0.00
1990 -8.1 -17.26 41.36 18.20 1527.72 0.00
1989 9.4 -9.96 36.19 0.40 1277.09 0.00
1988 5.7 -17.70 41.67 12.60 1271.61 0.00
1987 -12.8 -22.14 44.82 1.90 1118.12 0.00
1986 5.3 -10.71 36.72 21.80 1095.41 0.00
1985 9.2 -15.20 39.90 29.80 881.45 0.00
1984 3.1 -22.35 44.97 4.20 656.13 0.38
1983 -0.1 -24.68 46.62 -4.40 625.65 4.40
1982 16.9 -24.61 46.57 23.60 659.05 0.00
1981 -14.6 -35.51 54.29 -0.50 514.12 0.50
1980 13.1 -24.48 46.48 7.90 517.20 0.00
1979 -4.6 -33.23 52.68 0.20 472.01 0.00
1978 -5.4 -30.01 50.39 7.90 470.87 0.00
1977 -11.7 -26.01 47.56 2.30 429.07 1.69
1976 -3.2 -16.21 40.61 -3.90 417.18 3.90
1975 1.8 -13.44 38.65 19.20 438.17 0.00
1974 -20.5 -14.97 39.74 23.40 351.48 0.00
1973 3.8 6.96 24.20 -12.50 265.87 15.65
1972 0.1 3.04 26.97 -3.60 318.14 3.60
1971 -10.9 2.94 27.04 13.70 333.75 0.00
1970 2.7 15.53 18.12 24.60 281.49 0.00
1969 -9.9 12.49 20.27 -14.00 206.17 14.17
1968 4.4 24.85 11.51 -0.20 256.01 0.20
1967 -1.9 19.59 15.24 3.70 256.73 0.00
1966 -13.6 21.91 13.60 11.10 244.00 0.00
1965 4.2 41.10 0.00 -2.80 209.63 2.80
1964 7.7 35.41 0.00 5.60 218.55 0.00
1963 5.2 25.73 6.79 7.40 201.66 0.00
1962 -11.4 19.51 11.40 21.70 180.87 0.00
1961 3.7 34.89 0.00 -5.50 130.79 5.50
1960 -3.5 30.08 3.50 16.90 144.22 0.00
1959 3.7 34.80 0.00 -6.90 108.92 6.90
1958 19.2 29.99 1.12 14.80 124.40 0.00
1957 -10.8 9.05 17.04 3.40 95.47 0.00
1956 -7 22.25 7.00 3.00 89.04 0.00
1955 6.9 31.45 0.00 13.50 83.54 0.00
1954 10.3 22.97 0.00 20.90 61.71 0.00
1953 0.4 11.49 0.00 15.80 33.75 0.00
1952 4.5 11.04 0.00 2.10 15.50 0.00
1951 1.2 6.26 0.00 -1.80 13.13 1.80
1950 5 5 0.00 15.20 15.20 0.00

These statistics are from May to May. It means that the "N" line shows returns from May of year N to May of year N+1. It also means that if you compound the returns of summer and winter, you won't obtain the historical return of year N.

Since 1950, being out of the market 6 months from May to October would have avoided a 10% loss, and above all would have avoided major drawdowns, both in depth and in length.

Summary of drawdowns:

DJIA 1950-2011 Maximum depth Maximum duration
Buy & Hold -35.50% 6 years
Winters only -19.41% 3 years
Summers only -54.29% 47 years

Conclusion:

This is a (very) long-term vision of the stock market. Knowing that seasonal patterns don't work every single year, can an investor draw applicable ideas on the short term? I think so.

For example this article was written in mid-February. This is a dangerous month in the middle of the good season. The indexes are close to all-time highs, clearly due for a consolidation after a powerful rally. Even a bullish investor should be wary of taking new long positions in the next weeks. The risk is not worth the possible missed opportunity for just a couple of statistically dangerous weeks. March will be a better time to make a decision.

In an article to come, I will expose interesting conclusions from Jacobsen's publications about seasonal patterns in specific countries. Follow me if you don't want to miss it.

Source: Seasonal Patterns In Stock Markets: 319 Years Of Evidence