Monday, January 21, 2008

The 17 and 43 week EMA crossover

I wrote this post a couple of days ago and it was actually posted to 1stmillionat33 this morning. However, in view of the world wide stock market carnage seen today, I'm not sure in what light I should present it. In the end, I decided not to change a word, even though the last paragraph may sound more than a little silly right now.

I learn about this long term indicator from an article on ContrarianInvestor (I’m a subscriber). I doubt they are the originator as moving average crossovers in general has been in use for a long time. The system looks at the relationship between the 17 and 43 week exponential moving averages (EMA) of the S&P 500 index. When the 17 week EMA is above the 43 week EMA, one should long the S&P, otherwise, one should short the S&P. The weekly EMAs are equivalent to the 85 and 215 day EMAs which is plotted below using the new Yahoo charts. I encourage you to play around with the time intervals for the averages. The signals are fairly “robust”, in the sense that buy/sell signals don’t change much given small variations in the intervals. For example, 85/200, 60/200, or even 60/170 give roughly the same thing.

The track record of this system is impressive: it correctly gave a buy signal in 1995, a sell signal in 2001 and a buy signal in 2003. One has to go back to 1991 to see a meaningful whiplash. If you go to a shorter interval, you’ll see that the two moving averages have been converging. Indeed, they crossed on Friday to give a fresh sell signal (using 85/215, shorter intervals would have generated the signal sooner)! Again, I encourage you to play around with the time periods to see how well this system worked (or not worked) before. Although I haven’t done the exact calculation, it seems that this system would handily beat a buy-and-hold approach while having shallower drawdowns since 1950 which is how far the Yahoo data goes back to.

So what’s so magical about 17 and 43 weeks? I can hear you ask. Of course there’s some leeway in those two numbers, but I guess what you’re really asking is the philosophical basis for this system. There is none, or anything a priori that I can tell. This system is based on a long history of observed facts, which by the way, is identical to the reasoning behind statements like “in the long run, stocks go up by x% a year”.

Personally, I’m taking this sell signal very seriously even though I recognize that the market is very oversold and ripe for a bounce. On the other hand, my portfolio is set up to benefit from the on-going global growth story thus is susceptible to a global recession. While I still regard that as unlikely, I will most likely treat any significant bounces in the general market as opportunities to build up a hedge.