by Ghamal de la Guardia, Global Economic Intersection Associate
Recently a research note by perma-bear Albert Edwards, of Societe Generale, made reference to work done by Dominic Picarda, of Investor’s Chronicle and the Financial Times UK, who uses the Coppock indicator as a long-term buy or sell signal for the S&P 500 index. According to Edwards:
[Dominic Picarda] has identified eight killer waves in the S&P 500 over the last 83 years. All have been followed by substantial losses. The average fall following a killer wave has been 40 per cent over 20 months.
The above note was reported on by John Lounsbury in this blog.Whenever we hear of an indicator that is 100% effective over 83 years worth of data alarms start to go off, so we decided to look at the data.
This is a plot of the Coppock indicator and a monthly average of daily prices for the S&P 500 since July 2009.
The first top occurred in July 2010 right before the indicator dropped 3% from the previous month. Then, without crossing the zero-line, it began moving up after April of this year, but at the end of August it confirmed a new top was in place by dropping 3% once again. This pattern is what’s been defined as the “killer wave” and is basically a double top in the long-term momentum of prices .  Since the last drop in the indicator was of roughly 3% and this seems to have been the trigger that has Dominic Picarda and Albert Edwards worried, we will use a drop of roughly 3% for determining when a sell signal has taken place.
To illustrate what’s supposed to happen after a sell signal let’s take a look at the Coppock Curve of the Dow circa 2008.
The first sell signal of the “killer wave” pair (the set-up signal) triggered in July of 2004 and has been marked by an orange vertical bar in the above chart. The second sell signal, the “killer wave” signal, triggered in November 2007 and has been marked by a red vertical bar. A similar signal was triggered in early 2000, although this one took a lot longer to play out and the drop was not as steep.
As you can see not all “killer wave” signals are followed by an immediate plunge. In fact some are not followed by a plunge at all. Let’s look at how this signal fared during the early 1990’s, again in the Dow.
It seems that the indicator is not fool-proof and can give false signals despite its long time frame (and reputation.) The long time frame can also work against you when it comes to recognizing a false signal, unless it becomes immediately obvious that it’s a false signal because the indicator turns up, as happened in 1997.
Judging from the current momentum of prices in the Dow, and barring a miraculous rally that materializes out of nowhere, the newest signal is not likely to be proved false in the coming months, so the indicator will probably continue its downward trek. It will be interesting to follow along with the indicator as the drama unfolds.
On the other hand, if the stock market stabilizes with the S&P 500 holding a bottom around 1,100, the signal will have come at a time when half of a drop of approximately 20% had already occurred. This would make the signal’s timing less than auspicious.
Jason Goepfert of SentimenTrader and Sundial Capital Research wrote a piece about the Coppock Killer Wave in the Wall Street Journal, where he observed :
“The last six “killer wave” signals, in fact, led to quite bullish
outcomes, excepting the one from November 2007.
Overall, I don’t see much about the indicator that would have me
overly worried about stocks’ long-term prospects.”
My observations do not entirely agree with Goepfert because the 1997 signal actually never showed a loss (although a sustained rally did not start for 11 months) and the 2000 and 2007 signals were followed by major bear markets. See my graphs above for details. One difference between Geopfert’s work and mine is that some of my graphs are for the DJIA and all of his analysis was with the S&P 500. Some of my DJIA signals are not the same months as his S&P 500 signals, for example.
The above observations aside, Goepfert provided a table that shows when all the data is tabulated from 1928 to 2011 that (1) market performance is widely variable and (2) the averages for Killer Wave signals show no pervasive market direction compared to all identical length stock market performance periods.
Note: the median gain in Goepfert’s table (18.3%) is the smallest gain for the five periods when there was a gain; the median gain for the five periods with advances is 25.1%. This makes the median gain when one occurs much larger than the median gain shown for all 20 month periods). Combined with the median loss also being much larger than the median shown for all 20 month periods, maybe the indicator has use as a signal of larger than usual market swings (in both directions). Is it a Volatility Wave signal?
The Killer Wave by John Lounsbury