Any Recent Day-of-the-Week Anomalies?
Posted in Calendar Effects
June 23, 2009
Does recent (post-1980s) data suggest any day-of-the-week stock market return anomalies? To investigate, we examine close-to-close returns for the five trading days of the week during normal trading weeks (those having five trading days). In other words, we exclude weeks which have at least one day during which U.S. stock exchanges are closed all day. We do not exclude normal weeks adjacent to abnormal weeks, so a normal week occasionally follows or precedes a three-day weekend. Using daily closing prices for the S&P 500 index since the beginning of 1990 (847 normal weeks), we find that:
The following chart shows the raw average daily S&P 500 index returns for all trading days and for each of the five trading days of normal weeks since the beginning of 1990 and since the beginning of 2000. Results suggest that Tuesdays offer the best returns and Mondays or Fridays offer the worst. However, the standard deviations of daily returns are in the range 1.1% to 1.6% across the five days of the week for both sample periods, so variability tends to swamp average returns.
To isolate potential day-of-the-week effects from overall stock market drift, we detrend the results by subtracting the average return for the samples from the raw daily averages.
The next chart summarizes the detrended average daily S&P 500 index returns for the five trading days of normal weeks since the beginning of 1990 and since the beginning of 2000. Results suggest that the outperformance of Tuesdays and the underperformance of Mondays and Fridays derives mostly from post-1990s data. Again, the standard deviations of daily returns indicate that variability tends to swamp average returns.
As a robustness test, we calculate day-of-the-week returns for four non-overlapping subperiods since 1990.
The next chart summarizes the detrended average daily S&P 500 index returns for the five trading days of normal weeks during four consecutive five-year subperiods since the beginning of 1990 (with the last subperiod only about four and half years). Inconsistencies in results for specific days of the week across the four subperiods undermine belief in day-of-the-week anomalies. Tuesday’s outperformance and Monday’s underperformance since 1990 comes mostly from the past five years of data. Friday’s underperformance comes from two earlier subperiods.
In a separate test, we address a reader question about whether Tuesdays tend to reverse Mondays. The scatter plot below relates the return on Monday to the return on the immediately following Tuesday since the beginning of 1990 (847 Monday-Tuesday pairs). Since the slope of the best-fit line is negative, the result indicates a small tendency for Tuesday to reverse Monday. The R-squared statistic of 0.05 indicates that variation in returns for Mondays explains about 5% of the variation in returns for Tuesdays.
However, when we exclude the four Mondays with the highest returns and the four Mondays with the lowest returns (eight Mondays out of 847 in the sample), the R-squared statistic decreases to 0.01. This significant decrease in R-squared means that just a few very extreme Monday-Tuesday pairs (black swans?) drive the small amount of predictability implied by the total sample.
In summary, evidence from simple tests on recent data offers little support for belief in day-of-the-week anomalies in broad stock market returns.
