Blog - Investing Notes

June 9, 2008 – Update: A Slinky (Short-term Reversion) Effect?

Do investors/traders tend to overdo it during buying and selling frenzies, coming to their senses shortly thereafter? In other words, does the broad U.S. stock market tend to revert after extreme short-term moves up or down? To check, we test for short-term reversion of the S&P 500 index after its sharpest declines and advances since the beginning of 1990, hoping to identify reasonably frequent trading opportunities. Using daily closes of the S&P 500 index over the period 1/2/90 through 6/6/08 (4647 trading days), we find that...

The following chart provides a rough indication of the overall degree of short-term reversion for the S&P 500 index for various short-term cycles over the entire sample period, depicting the Pearson correlations between returns for:

• The last trading day and the next trading day (1:1);
• The last three trading days and the next three trading days (3:3);
• The last five trading days and the next five trading days (5:5);
• The last 10 trading days and the next 10 trading days (10:10); and,
• The last 15 trading days and the next 15 trading days (15:15).

Results suggest that there is a slight degree of short-term reversion built into the market, with weekly (5:5) reversion relatively pronounced. Due to use of daily data, sampling intervals overlap for for the 3:3, 5:5, 10:10 and 15:15 cycles, thereby potentially distorting results in the context of actionable trading.

The next chart summarizes returns for three reversion cycles (1:1, 3:3 and 5:5) and four categories of extreme events (25 biggest percentage advances/declines and 100 biggest percentage advances/declines) over the entire sample period. Results indicate that going long at the close after sharp declines may generate abnormal returns for all three cycles, with the 5:5 cycle most attractive. Going short at the close for one day after the most extreme one-day advances may also generate an abnormal return.

However...

Future returns for the S&P 500 index after extreme declines tend to be exceptionally volatile. To illustrate, the next chart displays the dispersion (standard deviation) of future returns around the means for the 5:5 cycle and the same four categories of extreme events over the entire sample period. Five-day future returns after the 25 biggest five-day declines are extraordinarily volatile, so trading on this signal is very risky.

And a further however...

The 25 biggest five-day declines occur (at the close) on the following dates:

8/22/90
8/23/90
10/27/97
8/31/98
9/1/98
9/2/98
4/14/00
4/17/00
10/12/00
12/20/00
3/14/01
3/15/01
9/17/01
9/18/01
9/19/01
9/20/01
9/21/01
7/12/02
7/19/02
7/22/02
7/23/02
7/24/02
8/5/02
9/3/02
1/22/08

Some of these dates are clustered such that a trader committing all funds immediately at the close upon a signal could not exploit them all for a 5:5 cycle. Realistically, there are only 14 exploitable signals from this list since 1/2/90, and the average five-day future return for this reduced set is 2.3% (compared to an average of 3.1% for all 25 dates). Moreover, there were no signals during 1991-1996 or 2003-2007 (periods of low volatility).

In summary, extreme market declines (much more than extreme market advances) present reversal trading opportunities with relatively high risk, but these opportunities are by definition infrequent and are very unevenly spaced over time.

The less extreme the declines, the less attractive but more frequent the trading opportunities.

For related research, see Blog Synthesis: Volatility Effects.