The Wrong Lessons

The experts have let us down. Too many are teaching us the wrong lessons.

Assertion: Valuations are Meaningless

If you restrict yourself to short timeframes and ignore confidence limits, you can easily conclude that the price to earnings ratio P/E has no predictive power.

The single year confidence limits of stocks are of the order of plus and minus 40% to 50% (for two standard deviations). The single year nominal return is of the order of 10%. The single year real return is of the order of 6.8%.

[Keep this in mind the next time that you see experts arguing over 0.2% to 0.5% adjustments to the Gordon Model.]

Predicting only one or two years into the future is difficult.

Making matters worse, single year earnings are highly volatile. Using a single year price to earnings ratio is difficult.

When you extend the timeframe to a decade, the uncertainty drops dramatically, even faster than we would expect if each year’s return were independent of the past. When we average the earnings over several years as Benjamin Graham recommended, the price to earnings ratio becomes meaningful.

If you wish not to see anything, you can construct an analysis method that shows nothing. If you wish to know the truth, read the findings of David Dreman, James O’Shaughnessey, Lowell Miller and others.

Assertion: Monte Carlo Models Are Right (TIPS Volatility)

Typically, Monte Carlo models show a spread of Safe Withdrawal Rate probabilities for TIPS. According to the rhetoric, this is the effect of capital gains and losses. In reality, it is the effect of modeling error.

I have never seen words about setting TIPS volatility as an intended input into a Monte Carlo model. Rather, it shows up as an unexpected and unexplained output.

The fact is that TIPS history is far too short for us to have meaningful statistics of capital gains and losses. If we were to use what we have, we would credit TIPS as an outstanding investment class, vastly superior stocks. An all-TIPS portfolio would be the gold standard.

During their short history, TIPS interest rates have fallen dramatically. If we were to extrapolate the past into the future, we would credit TIPS with consistent double digit returns.

There is an actual effect that causes a spread in the data in historical sequence models. It is caused by the slight timing mismatch between when inflation occurs and when it shows up in the TIPS principal. [Retirees can counter this uncertainty by waiting until they have received their TIPS interest payments before making withdrawals, similar to how they handle other cost of living adjustments.]

Assertion: Investment Workarounds Are Simple

How can we increase Safe Withdrawal Rates? One solution is to increase returns. Another is to reduce volatility.

To an excellent approximation, both work as advertised.

Investing in Small Capitalization Value stocks fails to solve all problems because it is a widely accepted solution. Popularity reduces returns. Even then, whether there really has been a strong advantage with this asset class is in doubt. David Dreman has written about this in detail.

The “Dogs of the DOW” is a good example. It worked very well until it became popular. Then it was implemented as poorly as is imaginable, with high fees and encouraging front running. It underperformed for three or four years before being declared dead. More recently, it has regained its status as a superior investment strategy.

Almost all new asset classes go through an early stage similar to the “Dogs of the DOW.”

Diversification with rebalancing is offered as a solution. The problem is that diversification usually reduces the overall return. If you can identify which investments are most likely to perform best, you are better off selecting the best investments. Rebalancing produces superior returns only if you are unable to identify superior investments. Otherwise, rebalancing DRAGS down the overall return much more than it reduces volatility, especially over longer time periods.

Rebalancing is a fad based on a false premise: that it is impossible to measure valuations in a meaningful manner. Adding asset classes can backfire. New asset classes seldom come with a meaningful history.

Assertion: Short Histories are Sufficient

Our investigations using the Sortino Ratio made this clear. Great statistics based on short time intervals remove the effect of valuations. To predict what will happen at Year 10, you need statistics that are based on what has happened over periods of ten years. Monthly statistics, annual statistics and two or three year statistics don’t do the job.

Using the longer histories, valuations show an enormous effect.

Given this conclusion based on the S&P500 index, is it any wonder that almost all projections made from short history data are in question?

Keep this in mind when looking at new asset classes.

Have fun.

John Walter Russell
September 24, 2006