Relative Strength of Indexes as a Future Return Indicator
Posted in Technical Trading
June 1, 2009
A reader pointed to the article “A Simple & Powerful Timing Indicator”, which discusses the strength of the (risky) NASDAQ Composite Index relative to the (conservative) S&P 500 Index as a market timing indicator, and requested confirmation. The article cites the book Technical Analysis – Power Tools For Active Investors as the source for the indicator, which is based on the ratio of weekly index levels with respect to its 10-week moving average. This book has a copyright year 2005 and a second printing date of May 2005. We consider the use of both the NASDAQ Composite and Russell 2000 as risky indexes compared to the conservative S&P 500 Index. Using weekly closes for the S&P 500 Index, the NASDAQ Composite Index, the Russell 2000 Index, the readily tradable S&P Depository Receipts (SPY) and the 13-week Treasury bill (T-bill) yield from February 1993 (as limited by availability of SPY data) through May 2009, we find that:
We make the following assumptions for implementing a trading strategy based on an index relative strength indicator:
- Based on weekly data, when the ratio of the NASDAQ Composite Index (or Russell 2000 Index) to the S&P 500 Index rises above (falls below) its 10-week simple moving average, we buy (sell) SPY at the close (assuming a trader can do the calculations just before the close). We adjusted closes for SPY to include dividends.
- When not in SPY, we accrue interest at the T-bill yield to approximate the broker cash rate.
- One-way trading friction when entering and exiting SPY is 0.2% of funds. Actual friction depends on specific broker fees, account size and market liquidity.
- For cumulative return calculations, we begin with $10,000 as an initial investment.
To exclude any data mining bias impounded during a pre-2005 search for indicators, we first look at post-2005 indicator performance. Since the beginning of January 2005, the NASDAQ-S&P 500 relative strength indicator:
- Signals 21 buys and 20 sells (in stocks 53% of the time).
- Produces an average weekly return of +0.11% with standard deviation of weekly returns 1.86%, compared to -0.03% with standard deviation 3.04% for SPY buy-and-hold.
- Beats buy-and-hold on an average weekly return basis as long as one-way trading friction is below about 1% of funds.
- With 0.2% one-way trading friction, generates a cumulative value of $12,242 from an initial $10,000 investment made at the beginning of January 2005, compared to $8,459 for SPY buy-and-hold.
Results confirm persistence of the value of the NASDAQ Composite relative strength signals with reasonable trading friction.
As a test of concept robustness, we substitute the Russell 2000 Index for the NASDAQ Composite Index. Since the beginning of January 2005, the Russell 2000-S&P 500 relative strength indicator:
- Signals 27 buys and 26 sells (in stocks 52% of the time).
- Produces an average weekly return of -0.09% with standard deviation of weekly returns 1.69%, compared to -0.03% with standard deviation 3.04% for SPY buy-and-hold.
- With 0.2% one-way trading friction, generates a cumulative value of $7,821 from an initial $10,000 investment made at the beginning of January 2005, compared to $8,459 for SPY buy-and-hold.
The Russell 2000 Index relative strength signals underperform buy-and-hold, undermining a hypothesis that index relative strength indicators exploit rising and falling aggregate investor risk appetite.
The following chart summarizes average weekly returns for both indicators since January 2005, with and without trading frictions. As noted, a trader using the NASDAQ Composite signals could have outperformed SPY buy-and-hold by keeping one-way trading friction below 1.0% of funds.
As a further test of the robustness of the index relative strength indicators, we extend measurements back to the initial availability of SPY as a practical trading vehicle in February 1993. Since then, the NASDAQ-S&P 500 relative strength indicator:
- Signals 75 buys and 75 sells (in stocks 55% of the time).
- Produces an average weekly return of +0.17% with standard deviation of weekly returns 1.61%, compared to +0.15% with standard deviation 2.47% for SPY buy-and-hold.
- Beats buy-and-hold on an average weekly return basis as long as one-way trading friction is below about 0.3% of funds.
- With 0.2% one-way trading friction, generates a cumulative value of $38,423 from an initial $10,000 investment made at the beginning of February 1993, compared to $27,279 for SPY buy-and-hold.
Since February 1993, the Russell 2000-S&P 500 relative strength indicator:
- Signals 81 buys and 81 sells (in stocks 50% of the time).
- Produces an average weekly return of +0.04% with standard deviation of weekly returns 1.57%, compared to +0.15% with standard deviation 2.47% for SPY buy-and-hold.
- With 0.2% one-way trading friction, generates a cumulative value of $12,997 from an initial $10,000 investment made at the beginning of February 1993, compared to $27,279 for SPY buy-and-hold.
The following chart summarizes average weekly returns for both indicators since February 1993, with and without trading frictions. As noted, a trader using the NASDAQ Composite signals could have outperformed SPY buy-and-hold by keeping one-way trading friction below 0.3% of funds.
Results for NASDAQ Composite-S&P 500 signals confirm some timing advantage over this longer period with low trading friction. Investors with small accounts may not have been able to realize the advantage. The underperformance of Russell 2000-S&P 500 signals again undermines a hypothesis that index relative strength indicators exploit rising and falling aggregate investor risk appetite.
As a final test, we construct a scatter plot to determine whether the size of the gap between the weekly value of the NASDAQ Composite-S&P 500 ratio and its 10-week simple moving average (SMA) relates systematically to the return for SPY the next week. We define the gap as the percentage difference with respect to the SMA. If the gap measures a persistent aggregate risk appetite, the larger the magnitude of the gap, the stronger the implication for future stock returns.
The Pearson correlation between the two series is 0.02, and the R-squared statistic is 0.00, indicating no useful relationship. In other words, the size of the gap between the ratio of indexes and its 10-week SMA appears not to be useful for weekly trading of SPY.
Note that the January 2005 through May 2009 sample period for the first set of analyses above is fairly short for testing the indicator and that all trading strategy returns ignore any tax consequences of trading.
In summary, evidence from several tests offers mixed support for a belief that the intermediate-term strength/weakness of “risky” stock indexes relative to a “conservative” index generates reliably profitable trading signals. Strength/weakness of the NASDAQ Composite Index relative to the S&P 500 Index may have merit for market timing.
