Best and Worst Days and Weeks of the Year
Posted in Calendar Effects
November 16, 2009
Are there particular calendar dates or week-length intervals that are especially bad or good for U.S. stocks, perhaps because of culture, regulation or law? Using S&P 500 Index daily closes for 1/3/50 through 11/12/09 (nearly 60 years), we find that:
The following chart presents histograms of the average daily returns for the S&P 500 index by calendar date during 1950-1989 (40 years) and since 1990 (almost 20 years). The average daily return for all dates (equally weighted) is 0.037% (0.030%) for 1950-1989 (since 1990), with standard deviation of daily returns for all calendar dates 0.18% (0.34%). In the much longer 1950-1989 subsample, extreme behaviors for particular dates are more likely to cancel and produce averages near the center of the distribution.
The outlier at the far left of the 1950-1989 subsample is October 19 (driven by 1987). The outlier at the far right for the 1990-present subsample is October 28 (driven by 2008).
What are the ten best and ten worst calendar dates, on average, for U.S. stocks?
The following set of tables list the ten best and ten worst calendar dates for the S&P 500 Index during 1950-1989 and since 1990. There is only one date that appears in more than one of the four lists: 6/6 is the best date for 1950-1989 and the worst date since 1990. Lack of consistency suggests that investors should not view any calendar date as reliably very good or very bad for stocks.
However, 11 of the 40 dates listed fall in October, suggesting a highly volatile month.
Might there be consistency among best and worst for intervals of five trading days?
The next set of tables list the 15 best and 15 worst intervals of five trading days by starting date for the S&P 500 Index during 1950-1989 and since 1990. The average daily return for all intervals of five trading days by starting date (equally weighted) is 0.18% (0.15%) for 1950-1989 (since 1990). Note that overlap of intervals causes clustering of starting dates. Three interval starting dates appear in more than one of the four lists: 10/28 and 11/23 are among the best starting dates for both subsamples, and 10/13 is among the worst starting dates for 1950-1989 but the best starting dates since 1990. Results and raw data suggest that:
- The interval around the Thanksgiving holiday may be especially good for stocks.
- There may be a tendency for investors to enter the market during the end of October.
- As noted above, October appears especially volatile.
- Extreme returns on a single dates (such as 10/19/87) can dominate the average returns around that date over long periods.
For additional insight on potential patterns, we scan returns and volatilities across the calendar year.
The following chart shows the average daily S&P 500 Index returns by calendar date for 1950-1989 and since 1990. Intervals on the horizontal axis are 30 calendar days. On the possibility that the calendar influences investor behavior in a complex way, the chart includes sixth-order polynomial fits for both series. In general, results for the 1950-1989 subsample show less variation because more data increases the likelihood that extremes will cancel. Principal areas of agreement between subsamples are: late winter through early spring is relatively strong; summer through early fall is relatively weak; and, November through December is relatively strong. These results generally confirm the cumulative return pattern shown in the Trading Calendar.
Next we scan volatilities by calendar date.
The final chart shows daily S&P 500 Index return volatilities (standard deviations) by calendar date for 1950-1989 and since 1990. This chart also includes sixth-order polynomial fits for both series. Daily volatilities are generally higher for the smaller, more recent subsample, which has fewer opportunities for cancellation of extremes. The best-fit polynomial curves both indicate that volatility peaks around the end of October.
In summary, examination of stock returns by calendar date offers no evidence of persistently good or bad specific dates, but there is some support for beliefs that: (1) October is especially volatile; (2) there is a persistent transition from weakness to strength at the end of October; and, (3) stocks tend to be strong around Thanksgiving.
