Simple Tests of Sy Harding’s Seasonal Timing Strategy
Several readers have inquired over the years about the performance of Sy Harding’s Street Smart Report Online (now unavailable due to Mr. Harding’s death), which included the Seasonal Timing Strategy. This strategy combines “the market’s best average calendar entry [October 16] and exit [April 20] days with a technical indicator, the Moving Average Convergence Divergence (MACD).” According to Street Smart Report Online, applying this strategy to a Dow Jones Industrial Average (DJIA) index fund generated a cumulative return of 213% during 1999 through 2012, compared to 93% for the DJIA itself. For robustness testing, we apply this strategy to SPDR S&P 500 (SPY) since its inception and consider several alternatives, as follows:
- SPY – buy and hold SPY.
- Seasonal-MACD – seasonal timing with MACD refinement.
- Seasonal Only – seasonal timing without MACD refinement.
- SMA200 – hold SPY (13-week U.S. Treasury bills (T-bills) when the S&P 500 Index is above (below) its 200-day simple moving average at the prior daily close.
Using daily closes for the S&P 500 Index, daily dividend-adjusted closes for SPY and daily T-bill yields during 1/29/93 (SPY inception) through 9/25/15, we find that:
Calculations/assumptions for the Seasonal Timing Strategy are:
- Calculate MACD for SPY using the Exponential Moving Average (EMA) template at StockCharts.com as the difference between the 26-day EMA price and the 12-day EMA price. A bullish (bearish) crossover occurs when MACD moves above (below) its 9-day EMA.
- For each calendar year, sell SPY at the close on April 20 if MACD is bearish or otherwise at the close on the first day with a bearish MACD after April 20. If April 20 is not a trading day, shift to the last trading day before April 20.
- For each calendar year, buy SPY at the close on October 16 if MACD is bullish or otherwise at the close on the first day with a bullish MACD after October 16. If October 16 is not a trading day, shift to the last trading day before October 16.
- The strategy is in the market at the beginning of the sample period (1/29/93).
- When out of the market, assume a return on cash equal to the contemporaneous T-bill yield.
- Ignore tax implications of trading semiannually.
Assumptions are the same for the other strategies as applicable.
The following table compares average gross (frictionless) daily returns, standard deviations of gross daily returns, gross annualized Sharpe ratios (from daily data), gross compound annual growth rates (CAGR) and maximum drawdowns over the entire sample period for the four strategies specified above. Buying and holding SPY wins based on average daily return and CAGR. SMA200 wins based on standard deviation of daily returns, Sharpe ratio and maximum drawdown. Seasonal Only modestly outperforms Seasonal-MACD.
For another perspective we look at cumulative performances.
The following chart compares on a logarithmic scale gross performances of $10,000 initial investments over the entire sample period for the same four strategies. Terminal values are $67,026, $58,921, $63,251 and $60,362 for SPY, Seasonal-MACD, Seasonal Only and SMA200, respectively. Buying and holding SPY is highest most of the time, but also the most volatile. SMA200 often is higher than seasonal strategies. Seasonal-MACD is sometime above and sometimes below Seasonal Only.
How do the four strategies compare based on annual returns?
The next chart shows gross annual returns over the entire sample period for the four strategies. Years 1993 and 2015 are partial only. Average gross annual returns (including 1993 and 2015) are 10.3% for buying and holding SPY, 8.4% for Seasonal-MACD, 8.9% for Seasonal Only and 8.9% for SMA200. Rrespective standard deviations of annual returns are 18.6%, 10.0%, 11.0% and 13.6%.
To explore the value of the MACD refinement, we calculate Seasonal-MACD returns relative to Seasonal Only by year.
The next chart shows the difference in gross annual returns between Seasonal-MACD and Seasonal Only, including partial years 1993 and 2015. Seasonal-MACD is usually the winner, but three bad years more than offset, such that the average of the series is -0.4% per year.
Are trading frictions material?
The final chart summarizes effects of trading frictions, ranging from 0.05% to 0.50% per one-way trade, on CAGRs for the four strategies. SMA200 is much more sensitive to trading frictions than the seasonal strategies. Seasonal-MACD and Seasonal Only switch between SPY and T-bills 45 times, while SMA200 switches 153 times. A monthly SMA would fare better in this regard.
The friction penalty of timing strategies is generally more severe for small investors, who usually bear higher frictions than large investors.
In summary, evidence from simple tests on available data for SPY offers some support for preferring (based on risk) Sy Harding’s Seasonal Timing Strategy over buy-and-hold, but not for preferring the MACD refinement of seasonal entry and exit.
Cautions regarding findings include:
- As noted, Sy Harding’s work is no longer available.
- Testing many strategies on the same data introduces snooping bias, such that the best performance tends to overstate expectations.
- Sample size is small in terms of number of bull and bear markets, and test results therefore vary considerably for different start and stop dates.
- As noted, tests ignore potential impacts of timing strategies on capital gains taxes.
Independently, Peter Brimelow and Mark Hulbert report in MarketWatch columns of…
…10/2/15 that “Harding’s modification of the Halloween Indicator [produced] an 8.8% annualized return [since mid-2002] — 1.6 percentage points per year more than a purely mechanical application of this seasonal pattern, and 1.4 percentage points ahead of a buy-and-hold. Even better, this market-beating return was produced with 39% less risk, which means it’s even further ahead of a buy-and-hold on a risk-adjusted basis.”
…4/1/15 that “Harding’s modification of the Halloween Indicator is in fourth place for risk-adjusted performance since mid-2002 out of the 91 stock-market-timing strategies the Hulbert Financial Digest has tracked over this period.”
…10/1/14 that “Harding’s tweaking of the Halloween Indicator has improved significantly on the mechanical version of the seasonal pattern. …good enough [8.9% annualized return from 5/31/2002 to 8/31/2014] to suggest that followers of the Halloween Indicator seriously consider getting a jump on those who wait until Halloween itself.”
…10/4/13 that “a hypothetical portfolio that, using Harding’s method, switched between the Wilshire 5000 index, which reflects the performance of the entire U.S. stock market, and 90-day Treasury bills…would have gained 8.5% annualized over the past 12 years, according to the Hulbert Financial Digest, assuming dividends were reinvested. That is better than the 6.9% annualized produced by the original, unmodified version of the Halloween Indicator, which in turn is better than the 6.2% return of simply buying and holding the stock market. Better yet, Harding’s portfolio was completely out of the market half the time, so it incurred only half the risk. That is a winning combination.”
…10/5/12 that from the end of May 2002 through September 2012 “the Sy Harding service has significantly improved on the Halloween Indicator, which itself is significantly ahead of a buy-and-hold.”
…4/2/12 that “since mid-2002, a buy-and-hold has produced a 5.4% annualized return. A purely mechanical application of the Halloween Indicator (automatically entering the market on Halloween and exiting on May Day) would have produced a 7.0% annualized return. Harding’s modification of the Halloween Indicator produced a 9.0% return (annualized) over the same period, or 2.0 percentage points per year more than a purely mechanical application of this seasonal pattern, and 3.6 percentage points ahead of a buy-and-hold. This is good enough to place Harding’s version of the Halloween Indicator in 4th place for performance since mid 2002, out of the 122 timing strategies the Hulbert Financial Digest has tracked over this period.”
…12/26/11 that Sy Harding’s Street Smart Report (6.5%) is one of the best-performing newsletters for 2011 [8th place] among those tracked by Hulbert Financial Digest. However: “Harding, a market timer who uses fundamental and technical methods…has a long record. The problem: It’s terrible. By HFD count, he’s down 6.03% annualized over the last three years vs. a gain of 15.24% for the dividend-reinvested Wilshire.”
…12/23/10 that Sy Harding’s Street Smart Report (-13.8%) is one of the worst-performing newsletters for 2010 among those tracked by Hulbert Financial Digest.
…12/24/09 that Sy Harding’s Street Smart Report (-12.5%) is one of the worst-performing newsletters for 2009 among those tracked by Hulbert Financial Digest.
…11/5/08 that “Over the year to date through October, Street Smart Report is up 3.6% by Hulbert Financial Digest count, vs. negative 32.9% for the dividend-reinvested Dow Jones Wilshire 5000. Over the past 12 months, the letter is up 5.26% vs. negative 36.31% for the total return DJ-Wilshire 5000. Over the past five years, the letter has achieved an annualized gain of 3.18%, vs. 0.78% annualized for the total return DJ-W 5000.”