Blog - Investing Notes
December 9, 2005 – Reader Contribution to Trading Calendar Analysis
Reader Aongus in Narberth PA has done some serious number crunching to extend our Trading Calendar analysis, constructing year-to-date average and median performances for the S&P 500 index over 1976-2004, the NASDAQ 100 index over 1986-2004 and the Russell 2000 index over 1988-2004. Aongus has also agreed to share the results here. How do these indices tend to trend over the calendar year for these periods?
The following chart shows the average and median year-to-date behavior of the S&P 500 index over the period 1976-2004. The average shows weakness and volatility from July through October. The median emphasizes the July-October volatility and indicates that many April-May intervals are weak. Our Trading Calendar analysis, which covers January 1990 through November 2005, suggests that the April-May weakness is a relatively recent feature.
The next chart shows the average and median year-to-date behavior of the NASDAQ 100 index over the period 1986-2004. The average shows weakness during March-April and July-October. The median indicates that a few bad Junes and a few very good November-December periods drive the average. In most years the NASDAQ 100 index does not have a fall rally.
The next chart shows the average and median year-to-date behavior of the Russell 2000 index over the period 1988-2004. The average shows weakness during March-April and June-October, with greater year-long volatility than the S&P 500 index. The median shows high volatility during the extended summer doldrums, with frequent pronounced buying opportunities in mid-October. The median also indicates that a few bad years drag down the average November-December rally.
This last chart compares the average year-to-date behaviors of all three indices from above. The profiles are similar, with summers consistently weak. Note from above that only a few very strong years drive the November-December outperformance of the NASDAQ 100 index. Note also that the failure of the S&P 500 index to show weakness during March-April is at least partly due to its longer sample period.
In summary, Aongus has generally confirmed across indices the seasonality trends found in our Trading Calendar analysis.
The methodology used by Aongus differs slightly from ours. He uses daily closes starting with the first day of the year, thereby excluding price moves for that first day. Over an annual cycle, this difference in method has very little effect.
The sample sizes above are somewhat larger than that used for our Trading Calendar, but they are still small, so the statistical confidence in results is low. For our calendar (and many other analyses), we designate 1990 as the beginning of a "modern" era for equity investing/trading. We assume that the financial/technological/regulatory/cultural environment of earlier times diminishes the usefulness of old data for predicting the near future. One could argue that calendar effects are mostly cultural and therefore ought to change slowly over time. But one could also argue that the recent ascendancy of quants and hedge funds (the market's phagocytes?) might first amplify and then kill every anomaly. Without better theory as to why there should be calendar anomalies, timeframe selection seems mostly arbitrary -- a vague tradeoff between historical relevance and statistical significance.
For related research, see Blog Synthesis: Calendar Effects and the Trading Calendar.
