Blog - Investing Notes

February 15, 2007 - Long-Term Outperformance from Trends Defined by Moving Averages

Does trading based on simple moving average crossovers outperform? In his November 2006 paper entitled "A Quantitative Approach to Tactical Asset Allocation", Mebane Faber presents a simple moving-average timing model that improves the risk-adjusted returns across various asset classes (represented, for example, by the S&P 500 index, the Morgan Stanley Capital International Developed Markets Index, the Goldman Sachs Commodity Index, the National Association of Real Estate Investment Trusts index and 10-Year Treasury notes). Using a model that mechanically buys (sells) an index when it crosses above (below) its 10-month simple moving average, he shows that:

• Applied to the S&P 500 index over 1900-2005, the model produces a 10.66% compound annual growth rate, compared to 9.75% for buy-and-hold. The timing model is less volatile than buy-and-hold. The model underperforms the index in about 40% of all years but avoids the worst bear markets. (See the first chart below.)
• Simple moving average periods of eight, 12 and 14 months produce similar results over 1900-2005.
• Over 1990-2005, the timing model slightly underperforms buy-and-hold based on raw returns, but suffers far less volatility. (See the second chart below.)
• Over the period 1972-2005, the timing model improves raw (risk-adjusted) returns for about 70% (90%) of 20 other indexes across asset classes.
• A straightforward application of the timing model across asset classes generates equity-like returns with bond-like volatility and drawdown, and over 30 consecutive years of positive performance. It modestly outperforms a simple equal-weighted asset allocation approach (and both beat buy-and-hold). A leveraged timing model substantially increases returns, at the cost of higher volatility.
• The timing model generates on average less than one round-trip trade per asset class per year, but taxes would still cut into the model's advantage over buy-and-hold unless trading in a tax-deferred account. [See the related review of Jim Rohrbach's timing system.]

The following chart, taken from the paper, compares on a log scale the results of the simple moving average timing model and a buy-and-hold approach for the S&P 500 index (including dividends) over the period 1900-2005. The timing model avoids much of the drawdowns of the significant bear markets of the 1930s and 2000s.

The next chart, also from the paper, compares on a linear scale the results of the simple moving average timing model and a buy-and-hold approach for the S&P 500 index (including dividends) over the period 1990-2005. The model slightly underperforms based on raw returns, but has a much lower volatility. Note that the trend following model, while avoiding most of the 2000-2002 bear market, underperforms during the preceding strong bull market. Results suggest that trend-following may have been stronger prior to 1990 than after.

In summary, this simple moving average trend-following model is a risk-reduction technique that signals when to be long a risky asset class with potential upside, and when to be sitting in cash. This is a very long horizon trading strategy, better suited to a tax-deferred account.

For related research, see Blog Synthesis: Momentum Investing/Trading. See also the much more modest analysis of moving average crossovers as trading signals in our blog entry of 7/19/07.