redback report, week ending 14 december, 2012, page-3

Hi Red,

As they say in Parliament, I thank the Honourable Member for his question.

In the past, I've been critical of the use of some cycles. I'll first deal with a couple that I've mentioned in the past.

The one that I've personally done the most work on is the Decennial Cycle of the Down Jones Index.

First, the sample is too small to have statistical validity. The Dow Jones Index was created in 1896. Since creation, there have only been 11 full decades. 11 iterations. Too small a sample.

Secondly, the performance of the Index in any one year of the decade is wide and randomly distributed. Using an average of a wide spread of random numbers as some guide to future performance is, quite simply, naive.

Thirdly, the composition of the index is continually changing. Initially it consisted of only 12 stocks. Currently, the only stock which has remained in the Index since inception is General Electric. When trying to measure the performance of the Dow Jones over time, we're not measuring the performance of the same thing. It would be like trying to predict the average speed of Ford cars in 2013, on the basis of the average speed of all Ford cars in the third year in a decade since Ford first began producing cars.

I could go on about the Decennial Cycle. But that's enough.

I've also been critical of the Armstrong Economic Confidence Model. I said not too long ago, that I wouldn't mention Armstrong again. But since you ask about cycles, I'll once again briefly deal with it. Armstrong has an Economic Confidence Model based on an 8.6 Year Cycle. A magic number related to Pi. The first problem with Armstrong is: What is he actually talking about? It's virtually impossible to pin down. Confidence in what? Which economy? Which stock market? Which commodity?

Here's a predictive example taken from his website:

On this forum, Armstrong is often used to predict what will happen in the stock market. A review of the model and how it performed against the American Stock Market in recent years provides insufficient evidence to believe that the Armstrong model provides much utility in timing major highs and lows in the American stock market. Sometimes it seems OK - other times it does not. Do your own research, just compare the performance with the actual performance of the Dow Jones - as I did in the not too distant past on this forum. Perhaps it works for other stock markets? Gold? Corn? Well - it's never really specified. I doubt you'll find that it performs with any outstanding reliability. Armstrong uses snake oil methods - he uses examples that fit, but ignores the rest that don't. That's unscientific.

Armstrong selectively picks, retrospectively, examples that fit his model. That's cherry picking. As a forward predictor of What?, it seems to have little practical use.

A guestion: Have I ever said that I don't consider evidence for a "Santa Rally" to be statistically significant? I don't remember having said that.

There is a widespread belief that a reliable Santa Rally exists.

One of the problems with that is, once again, defining what is meant. Some people define it as the period from Thanksgiving to New Year. That has a good record. Since World War II, those five weeks have been up an average of 2.1%. The Dow Jones has been up 73% of the time. The yearly average rise for the Dow Jones during that time has been 8.4%. So the five weeks from Thanksgiving to the end of the year (10% of the year) accounts for 25% of the yearly rise. That's looking good.

Other people define the Santa Rally as occurring in December. Occording to Bespoke, the Dow Jones has been up 73% of the time in the past 100 years with an average rise of 1.39%, 68% of the time in the past 50 years with an average rise of 1.51%, and 70% of the time in the past 20 years with an average rise of 1.8%. Over time, the performance has been fairly reliable. These figures may, of course, be subject to criticism like the decade figures. Averages may hide wide variability. I haven't done a detailed statistical analysis of these figures, so I can't say. But intuitively, they look OK.

Others define the Santa Rally much more narrowly than either of the above two definitions. One narrow definition is the last five trading days of the year plus the first two trading days of the New Year. Here's what Stock Traders Almanac says about that period:

",,, according to Stock Trader's Almanac, the appearance of "Santa" is a rather dependable event at the end of the year. The last 5 trading days of the year and the first two trading days of the New Year (12/27/2010 - 01/04/2011) combine for an average return of 1.6% since 1969. Gains of some kind are generated in that trading window 76% of the time, with just 3 declines of 3% (or more) in 41 years of history (and 11 rallies of greater than 3%). At a 76% success rate the "Santa Claus Rally" is among the highest batting averages for the market ... "

This provides some insight into average returns and variability.

When I referred in my post to "seasonals", that's the period I was referring to.

A guestion: Have I ever said that I don't consider evidence for a "Santa Rally" to be statistically significant? I don't remember having said that.

I hope that long winded response goes some way to answering your question.

Redbacka