05 julio 2007

An Unacceptable (But Instructive) Model

I've frequently noted that investors need not take market risk at all times (or even a majority of the time) to capture the long-term return of the stock market. What is needed most is the avoidance of deep losses. It also helps to be willing to accept market exposure when valuations are depressed and other investors are bearish - particularly when yield pressures are falling. Indeed, there are numerous investment approaches that would have outperformed the market over time with less risk than a buy-and-hold strategy - the work is to find those that have good theoretical underpinnings and aren't just hindsight.

To advance that case without going into anything proprietary, here is a very simple (and definitely simplistic) model that I would never advocate using in practice, but I think it is instructive nonetheless. It substantially outperforms the market in historical data since 1965, and has sidestepped every major bear market, with a maximum drawdown of less than 15% at its worst point, compared with a maximum loss of nearly 50% for the S&P 500 (in the 73-74 bear, and again in 2000-2002).

The model goes long when the S&P 500 P/E is below 18 (based on the highest level of trailing net earnings to-date), the 10-year Treasury yield is below its level of 6 months earlier, and the Investors Intelligence figures show bearishness above 30% (measured as bears plus half of the “correction” camp). It sells on the reverse conditions, when the P/E is above 18, the 10-year Treasury yield is above its level of 6 months earlier, and bearishness is below 30%. It stays in its existing position until all three factors shift one way or another.

In my view, this is an unacceptable model for several reasons. First, it is a “timing” model – it is either in or out, on a “buy signal” or a “sell signal,” which misses a great deal of potential returns that can be earned by varying market exposure in proportion to the return/risk profile of the market (which is what we do in practice). Second, it requires certain indicators to “ring a bell” and hit very specific “magic” levels. Such models are not very robust, in that the indicators can just miss such levels, and in the process, miss important moves altogether. While this model, for example, has sidestepped bear markets since 1965, there's no particular reason to believe that it won't miss a future one. A good investment model uses specific levels of an indicator only for fine-tuning – not to generate an all-or-nothing investment position.

Third, despite its reasonable long-term performance (though nothing exciting), the model has sat out of the market for 90% of the past decade, including much of 2003, which was quite an acceptable period to take market exposure. Though the model outperforms the market since 1965, it would have been invested only about 30% of the time. While that seems like a good thing, it also means that it is too sensitive to specific conditions being met, and leaves far more on the table than it should. The model has to compensate for that by allowing the possibility of 100% exposure at relatively high P/Es near 18, while good models can be more selective at rich valuations while capturing a larger portion of returns at lower valuations. Overall, a good model may very well be defensive for long periods of time, but taking an exposure only 30% of the time is far too conservative and creates far too much tracking risk relative to the major indices.

Despite its shortcomings, there are several reasons to present this model. First, it offers some insight into the considerations that separate a practical, useful investment model from one that simply looks good on paper. Second, it underscores the fact that an investor can sit out an enormous amount of time from the market and still capture all of its long-term return – provided the time spent out of the market is generally characterized by rich valuations, rising yield pressures, and high bullish sentiment, and that the investor generally does take exposure during periods of favorable valuations, falling yields, and high bearish sentiment.

In any event, moving to a defensive investment stance on rich valuations, rising interest rates, and high bullish sentiment has not historically resulted in missed long-term returns. To the contrary, the main outcomes an investor would have missed were the major bear market declines of the past half century.

Given current rich valuations, rising interest rates, and high bullish sentiment, all of this is clearly worth keeping in mind. Suffice it to say that I have no concern that our recent defensiveness will result in missed long-term returns. Though I have no strong opinions regarding short-term market direction here, I expect that the present speculation of investors will be quite poorly rewarded over time

2 comentarios:

F.J. dijo...

lo malo de estos modelos es que funcionan... hasta que un día dejan de funcionar. Pero es un poco del tipo del contrarian investing, es decir, meter en empresas solventes con PER bajo y sin perspectivas especialmente negativas o que ya tengan descontada la parte negativa.

Fernan2 dijo...

El propio autor lo califica de inaceptable... pero yo no lo veo tan malo.