I’ve gotten a few emails, mostly from new readers, inquiring why the Moose missed most of the recent stock rally. The answer, of course is that it is an intermediate term framework. It is weighted toward slower-reacting nine to fifteen month technical trends, which are traditionally considered the most efficient way to time markets over the long haul. In “normal times” a longer hold and fewer transactions eventually make for a higher return. In abnormal times, however– when stocks fall 60% in six months and then rise 50% in five months, for example– an intermediate term construct seems out to lunch.

It’s always difficult to determine whether what was previously considered “abnormal” is now actually the “new normal”. What is “normal” over the last hundred years or so may not be normal over the last year, or even the last decade. The Moose was conceived in the late 80’s and perfected in the early 90’s, pre-dating both the 2000-03 and the 2008 market crashes, both of which were “abnormally” severe compared to the “average” 20th century bear market. In both cases, the model exited stocks in time, but got back in late.

The “Encyclopedia of Technical Market Indicators” (first edition, 1988) was my primary reference in building the model. It back-tested all the popular indicators to the early 1900’s and sorted out the wheat from the chaff. Years later, I bought the second edition (2003) which aspired to more of the same, but because the back-test now included the 1987 crash, and the 2000-02 crash, the wheat and chaff had changed. The optimal trend was no longer intermediate term, but a shorter, sub-intermediate trend.

So if you’re building and testing your own models, the lesson is that any back-test period will skew your observations. My guess is that a lifetime (or one Kondratieff cycle) of 65-75 years is probably optimal. Longer than that and you run the risk of including two great crashes, as the second edition of the encyclopedia did. Shorter than that and you risk not including at least one “worst case” scenario. Cyclically, I equate 2000-03 crash to 1929-32, and the 2008 crash to 1937, which followed the ’29 crash by eight years too. Ultimately, ’29-’37 proved to be an abnormal time compared to the rest of the century.

So I doubt that the “new normal” insures wildly volatile and seriously manic-depressive U.S. financial markets. Although our government, for the last half-century, has been doing everything it can to destroy the Dollar by replacing private initiative with public largess, we are still a potent economic force. We may be descending toward banana republicdom with all the vigor of a boatload of chimpanzees, but we aren’t there quite yet. At this rate, we’ll get there soon enough, but the first stop would appear to be Japanization, arguably a more respectable (and less volatile) form of failure. Consider it a preliminary (and doomed) effort to save face.

I may be wrong about volatility eventually reverting to the mean. Perhaps the Greenspan Fed did indeed mark the beginning of a new era of monetary and economic instability leading to market chaos. Of course, cultural changes could also be a factor in shortening the optimal technical time frame, increasing volatility, and leading to a “new normal”. Technology has made it much easier to trade and do research, for example. While this democratization of the markets is a good thing, it is worth remembering that democracy is generally messy.

On the off-chance that the intermediate term is indeed a dinosaur, I’ve spent the greater part of this quarter trying to figure out how the model could be made more responsive to near term events. In that regard, I’ve been working on a more temporally balanced technical approach, on incorporating volatility into the mix, and on changing the way the model treats cash. (You might not think it, but cash is by far the most volatile asset in the model during a crisis. Price, the inverse of yield, can rise or fall by huge percentages in a short time. It can take months to mathematically eliminate the distortion, delaying the switch to other assets.)

So far, so good. The revision appears promising, but I haven’t done a long enough look-back to be consummately confident. My findings are not unexpected– more transactions, lower overall accuracy rate– but during the most recent period, at least, considerably higher return. Clearly, added turnover makes the revision better for tax-deferred than taxable portfolios. And clearly you never want to base your methodology on an abnormal set of circumstances. So more investigation is required.

from the September 18, 2009 Weekly Update