Surviving Market Crashes, Part 2: The Dot-com bust

In the previous article on surviving market crashes (“Surviving market crashes,” August 23, 2011) we looked at an innovative modification to Modern Portfolio Theory (MPT) that addresses one of its major drawbacks—that of keeping the investor fully invested during bad times as a well as good. We saw that MPT by itself can lead to substantial losses during severe market downturns, losses that can take years if not decades to recover.

To solve this problem, we applied an optimized market timing scheme to the MPT model. This not only resulted in superior returns, it did so at much lower risk. Since the results were so dramatic, there were many questions left in the minds of readers. This article is the result of those questions which will hopefully bring clarity to the issues that were raised.

It was decided that the best way to show how a portfolio can survive market crashes using our model was to examine one in detail. The bursting of the dot-com bubble was chosen because it was the first severe market decline in recent history which most people remember (though many may not remember it fondly).

The mechanics of the market timing model are given in greater detail in the previous article which the new reader is strongly encouraged to read. For those of you who have read it or just want to skip it, a brief summary of our approach follows.

Nine of the most commonly used asset classes form the basis of the our investment portfolio: large-cap stocks, small-cap stocks, long-term investment grade corporate bonds, long-term government bonds, intermediate-term government bonds, real estate investment trusts (REITs), international stocks, international bonds, and T-bills (or some other risk-free asset class such as an insured money market). Gold and other precious metals are not included for reasons given in the previous article.

From these asset classes, a portfolio is determined per the required desired compounded annual return. The examples used in the previous article were based on a 10% required return and that 10% return will again be used here for sake of comparison.

Portfolios are determined according to the classic MPT model and also to our market timing model which we call Modified Modern Portfolio Theory, or MMPT for short. The MMPT approach uses an oscillator to determine when one should be in or out of a particular asset class. When the oscillator says to get out, the percentage that MPT would have allocated to that asset class is instead put into the risk-free asset class (Treasury bills or an insured money market). Further details on how the timing aspect of the MMPT model works are given at the end of this article.

Portfolios are rebalanced every month as soon as new asset class performance data is published. In this article as well as in the previous one, portfolio allocations for both the MPT and MMPT models are the output of our portfolio allocation tool, the SMC Analyzer.

The collapse of the dot-com bubble
Those with stock-heavy portfolios, especially in stocks of that newfangled invention called the internet, were thrilled to see their nest-eggs balloon during the dot-com bubble, but they were probably more distressed to see them deflate during the dot-com bust, those agonizingly turbulent months between September of 2000 and October of 2002. During that time, the S&P 500 (representing the Large Company Stock asset class) lost almost 50% of its value—not very pretty!

During the collapse, an MPT portfolio with a 10% required annual compounded return would have been heavily invested in the stocks of large companies (the S&P 500). (See Table 1.) But the MMPT equivalent model would have taken investors out of that asset class as early as the beginning of November of 2000 and reallocated those funds into the safety of Intermediate-Term Government Bonds and T-bills. (See Table 2.)

Because classic MPT does not consider trends but rather looks at all times as an average, the allocation recommendations kept the investor in Large Company Stocks through a period of sharp decline. In comparison, because MMPT can react quickly to changes in market direction it got the investor out of Large Company Stocks as soon as it was apparent the decline was not to be short lived.

Table 1. Classic MPT Historical Allocations During the Dot-Com Bubble Bust for a Target 10% Compounded Annual Return

Table 2. MMPT Historical Allocations During the Dot-Com Bubble Bust for a Target 10% Compounded Annual Return

You can see from the above tables that during this period MMPT placed the investor heavily in the safety of U.S. Treasury Bills and Intermediate-Term Government Bonds. But for some of the time there was a substantial allocation in Small Stocks. This can be understood when looking at their behavior (Figure 1). While the Large Stock asset class was collapsing (Figure 2), the Small Stocks actually hung in there despite fluctuations in value.

Figure 1. Total return: Small-cap stocks

Figure 2. Total return: Large-cap stocks

Table 3 below summarizes the performance of each asset class during this period. It’s interesting to note that Long-Term Corporate Bonds fared the best but MMPT did not allocate funds to it instead preferring the lower volatility and risk of Intermediate-Term Government Bonds. It’s important to note that in MMPT the oscillator strategy only applies to equity based asset classes and not bonds. Bonds are treated just as they are in classic MPT since they are by nature less volatile than stocks.

Table 3: Annualized total returns of all asset classes during the bust

Comparing models
During the months of the dot-com bust, the MMPT portfolio returned 8.3% compounded annually (green line) while the classic MPT portfolio lost money at an annual rate of -12.7% (magenta line). Further, the MMPT portfolio was much less volatile. It experienced a standard deviation of only 4.9% while the classic MPT investor was being whipsawed to the tune of 13.6%. Wouldn’t you have liked to make money during that time instead of losing it?

The reason that MMPT was unable to achieve the desired 10% return is because in the short term there is never a guarantee that asset classes will perform up to their historical norms. Also, figuring in the lower return of T-bills and the slightly negative return in Small Stocks explains why MMPT fell short of its goal during this two year time span.

Figure 3. Comparison of results
(green = MMPT)
(magenta = MPT)

We have shown that an MMPT investor during the dot-com bust was able to reap a decent return while the classic MPT investor, left heavily invested in large-cap stocks at the wrong time, suffered. The reason is that a judiciously and properly applied market timing approach injects an element of nimbleness and reactivity to a portfolio while still benefiting from historical experience. This is of tremendous value especially in today’s uncertain markets.

An even better reason to consider a market timing approach has to do with the increasingly positive correlation among asset classes. Historically, bonds have moved independently of stocks meaning that these asset classes were uncorrelated. Many commodities also moved contrary to stocks but lately all of this has changed.

The correlations among stocks, bonds, and commodities are becoming increasingly positive as national economies become more interdependent. Investors today are unsure as to where to place their nest egg. Every day the news concerning the deterioration of the global economy is increasing volatility across most of the traditional asset classes, and it’s for this reason that market timing approaches should be given their proper due. I believe we have shown with this case study that the MMPT approach combines the best of Modern Portfolio Theory with the desirability of a viable market timing strategy.

Appendix: The MMPT methodology using the SMC Analyzer
The SMC Analyzer uses an oscillator to identify when to stay in an asset class and when to stay out of it. Each component asset class has its own oscillator. Each oscillator is based on an averaging period, and it is this averaging period that is optimized every month according to new input data.

When the oscillator goes from positive to negative, that asset class is exited and the funds that would have been allocated to it are instead put into the risk-free asset class. Conversely, when the indicator turns positive, funds are placed back into the asset class according to the current allocation. Only equity-based asset classes (stocks) are subject to this treatment; bond classes are treated exactly alike in both the MMPT and MPT models since their values typically don’t fluctuate much if at all.

It is important to note that the time history of performance of an asset class in an MMPT allocated portfolio will differ from that of an MPT allocated one. This seems obvious but the point here is that you will not get even close to the same results by applying an oscillator externally to MPT-derived allocations alone. The reason is because MPT bases its current allocations on the entire time history of data for each asset class, and so to get the proper allocation you’ll need the MMPT modified time series that incorporates those periods of time where funds were not placed into a particular asset class.

For this article and in the previous article we used the CCI (Commodity Channel Index) with an optimized averaging period as our market timing oscillator. This is the indicator we use to generate our monthly subscriber email reports because we have found that this one works the best in general. That is, it produces the lowest levels of variability in portfolio value, i.e. lowest risk, while still achieving the targeted returns. Professional money managers, quants, and mathematically sophisticated investors will find two others in the professional software version of the SMC Analyzer along with many other useful features including the ability to add your own asset classes.

For further information on how you can save your nest egg from the devastation of market crashes such as the one we’re in right now, please visit our website.

Leave a Reply

You must be logged in to post a comment.