Given that the market is again near a median bear bottom it is timely to look at the reliability and performance of the Coppock Indicator as a tool for picking a safe re-entry point. I have used the monthly data generously made available by Colin Nicholson of Building Wealth Through Shares.
In summary, market history since 1960 suggests the Coppock indicator is worthy of consideration as a way of determining a safe re-entry point to the market, but consider selling if the market has a deterioration of more than say 7.5 %
or if there are two consecutive months each with lower performance
(compared to percentage difference to the index value at the signal
month end) compared to the prior month.
First a chart showing the results of all signals since 1960 shows that relying on the Coppock Indicator is generally potentially rewarding (partially depending on the timing of any sale) but not without risk.
Note the two false signals, one in August 1970, the other in April 1974. They would have been avoided by selling if the performance of the market went to -6% or worse at any month end after the signal was given. Two false signals from 15 is quite good performance. Thirteen of 15 signals provided the opportunity for profit.
We can also look at a table of the performance after the signal is given. I have chosen to look at the 24 months after the signal is given but note that 6 of the series considered ran on for over 4 years.
The number of months since the signal is shown in the left hand column, the average outcome (based on the number of continuing performances) is in the far right column. I have coloured some cells in the table. The two heavily pink series of results are the two very dangerous false signals from 1970 and 1974. The orange in the Average column shows that there are even months where the average outcome falls from the previous month. Yellow shows a month where the performance falls from the immediately preceeding month, but is better than for the month prior to that. It can be seen that a fall in performance measured at month end is not unusual, but that most of them pick up the following month. There were 65 such falls in performance. Of those, 27 were followed by a second fall in performance and I have coloured these pink and they could be considered as possible sell signals (although that is not part of the Coppock methodology). I have coloured the next rise in monthly performance green which could be considered as a possible buy signal, again not part of the Coppock methodology. None of this has been backtested other than by "eyeballing" the table.
(As an aside, after the transaction and whipsawing costs of timing systems the results seem likely to be broadly commensurate with the historical returns of a 50/50 portfolio of shares and long bonds and to more than halve the volatility and maximum drawdowns which frighten the life out of many self funded retirees.)
In considering when to sell if you are not buy and hold investor, always remember 1987 when the market was virtually at its bottom before most sell timing signals triggered, other than stop loss signals. From looking at various BEV charts I have done of recoveries from the bottom, there is great danger when the market falls more than about 7.5%. It rarely whipsaws immediately after a fall of 7.5%. After a bottom the market spends 72% of its time before the next bottom at less than 5% below it's most recent high during the recovery. The risk of frequent whipsawing is very high if you sell the market on a fall since last high of less than 5%. Selling after a fall of 7.5% would have given you a chance to avoid the worst of the falls in 1987 and 2008.
The best summary might be that if you had purchased immediately after the Coppock buy signal was given since 1960 you would have earned on average 24% in the following 24 months, even taking into account two false signals which could have been avoided with a simple rule: sell if the month end performance is more than 6% below the month end on which the signal was given. (but be aware the next time it might reverse after a 7% fall! Whipsawing is everywhere!).
For interest, I also looked at the lead in to the Coppock signal being given.
While there is quite a variation between individual series, on a median basis compared to the month end on which the Coppock signal is given, the month end bottom was two months prior at 5.5% below the closing index value on the month end of the signal. On that basis it seems that a variation of the Coppock indicator which gave its signals a month earlier would, based on median results, have picked up an extra 4% of the upturn over the series reviewed. I am looking at Coppock 11,8,7 as a possible way to generate an earlier signal (but that might result in more false signals and so might not be as potentially profitable because of whipsawing).