What’s the best length for a moving average? Commentary: Trend-following indicators won’t pick the top
Those are the questions that are being asked, in one form or another, by market timers the world over as they figure out which indicator(s) they will use to tell them when to exit the incredible party Wall Street is throwing.
Three weeks ago, you may recall, I focused on the 200-day moving average, one of the more widely followed indicators for determining shifts in the market’s major trend. I found that it left much to be desired: For example, its performance has markedly diminished in recent decades — so much so that some researchers have begun to wonder if it has lost its market-timing ability.
Another reason some market timers are dissatisfied with the 200-day moving average is not a criticism per se, but an inherent feature to any trend-following indicator: It by definition will not pick the top. That’s because a sell signal will not be triggered until the market has dropped below its average level of the previous 200 trading days. By that time, of course, the market may have already suffered a sizeable loss.
For both reasons, a number of you who read my three-weeks-ago column urged me to measure the performance of much shorter-term moving averages. So that’s what I did for this column.
Unfortunately, I did not reach appreciably different results with any of the shorter moving averages I studied. To be sure, the shortest-term of the moving averages do a better job than the 200-day of getting out sooner when the market turns down. But they also get whipsawed for a loss more frequently as well. On balance their track records over the long term are not significantly different than that of the 200-day moving average.
Furthermore, each of the moving averages I tested suffered from the same marked diminution in returns in recent decades as I found with the 200-day average.
Surprised by these results? Norm Fosback, the former head of the Institute for Econometric Research and currently editor of Fosback’s Fund Forecaster, argues that we should not be. In the textbook he wrote three decades ago, entitled “Stock Market Logic,” he wrote:
“There are no magic numbers in trend following... Some moving average lengths may have worked best in the past, but, after all, something had to work best in the past and by testing everything possible, how could one help but not find it? It should be a basic requirement of any moving average trend following system that practically all moving average lengths predict successfully to a greater or less degree. If only one or two lengths work, the odds are high that successful results were obtained by chance.”
What about the death cross?
Before I leave the subject of moving averages of different lengths, I want also to say a few words about attempts to combine two moving averages of different lengths into a single trend-following system. Many consider it to be bearish when the shorter moving average crosses below the longer one, and bullish when the shorter rises above the longer one.
By the way, in the case of the 50-day and the 200-day averages, these two crossovers are called the “death cross” and the “golden cross.”
I investigated all death and golden crosses over the last century for the Dow Jones Industrial Average. As before, I found that their predictive prowess has declined significantly in recent decades.
Notice from the accompanying table that, over the entire period the Dow has existed since 1896, these two crossover events did a respectable job. However, note also that since 1970 they have done a much poorer job, with the market over the one, three and six months following death crosses actually doing better on average than following golden crosses.
||Average Dow gain over next month||Average Dow gain over next 3 months||Average Dow gain over next 6 months||Average Dow gain over next year|
|All Golden Crosses||0.23%||1.54%||3.26%||5.90%|
|All Death Crosses||-0.68%||0.37%||1.69%||2.19%|
|All Golden Crosses||-0.14%||1.58%||3.04%||5.78%|
|All Death Crosses||0.14%||1.69%||3.44%||4.66%|
Once again, the picture that emerges is one in which moving-average trend following systems used to work quite well, but which have steadily declined in recent decades.
The search continues for the optimal market timing indicator or indicators.
Mark Hulbert is the founder of Hulbert Financial Digest in Annandale, Va. He has been tracking the advice of more than 160 financial newsletters since 1980. Follow him on Twitter @MktwHulbert.