Market Timing Buries 'Buy and Hold' in Asset Allocation

The recent collapse in the financial markets was the final nail in the coffin for the traditional buy-and-hold strategy. Not only do I believe it, but so does virtually every talking head in the financial media—even Jim Cramer thinks it’s a trader’s market. To some, the term “trader’s market” conjures up java-jolted post-teens on their Wii stations buying and selling equities with a holding period measured in milliseconds. That’s day-trading to the max and not what is meant by a “trader’s market.”

The term is used by most to convey a holding period of days to months, or as long as current market conditions remain valid. On that point, I wondered how applying a market timing scheme to the tenets of Modern Portfolio Theory would affect portfolio risk and return.

To that end, I appealed to Professor Pat whom my readers may remember as contributing many articles to the Stock Market Cook Book last year on the topics of Modern Portfolio Theory (MPT) and Post-Modern Portfolio Theory (PMPT). Pat also wears another hat—that of ace website designer and expert programmer. Who better to tackle this problem?

And that he did. The result of which is the SMC Analyzer, a powerful tool that can be used not only for portfolio analysis but also for market research. Over the next couple of weeks I’ll be using the Analyzer to provide you with market insights that have never been published before, to the best of my knowledge.

Today I’m going to show you how a market timing scheme beats the pants off of the traditional strategy of buy-and-hold.

Bye-bye Buy-and-Hold
The past ten years in the market have been brutal on many retirement accounts where the concept of buy-and-hold reigns supreme. Witness the two nasty downturns in the S&P 500 (chart below). The bursting of the dot-com bubble in 2000 and the mortgage meltdown in 2007 both slashed market capitalizations roughly in half.

Where would that have left the investor? Let’s compare the damage it would have done on two portfolios where the assets are allocated according to MPT guidelines: one that employs a market timing scheme versus a buy-and-hold strategy.

Market timing criteria
In developing a market timing scheme, we found the Commodity Channel Index (CCI) to be an excellent timing oscillator. When the CCI turns positive, the market is trending up. When it turns negative, the market is trending down. (The Analyzer also has the capability of using the rate-of-change (ROC) and the moving average convergence/divergence (MACD) indicators as timing tools. I prefer the CCI because it tends to give the best results over a wide range of inputs.)

The Analyzer uses monthly data for each of these nine asset classes: Large-company stocks (e.g., the S&P 500), small-company stocks, long-term corporate bonds, long-term government bonds, medium-term government bonds, Treasury bills, Real-Estate Investment Trusts (REITs), international stocks, and international bonds. A separate CCI indicator is computed for each asset class and the averaging period is re-optimized every month. (This is a computationally intensive process, by the way.) The result is that when the indicator is positive, the investor would be long the percentage of that asset class as determined by MPT; and if the indicator falls below zero, the investor has the choice of either shorting (if appropriate) or putting that portion of the portfolio into a risk-free asset such as T-bills.

There is only one user input to the MPT part of the asset allocation calculation: the desired average annual portfolio return. Needless to say, the higher the desired return, the greater the associated risk. Let’s take a look at how applying a market timing scheme to an MPT-generated portfolio compares with a buy-and-hold strategy.

Market timing kicks buy-and-hold butt!
The chart below compares a buy and hold portfolio with one using a market timing scheme we call Long/T-bills. That is, when the CCI indicator for that asset class is positive, we’re long in that class according to the portfolio percentage as calculated by MPT. When the CCI turns negative in that asset class, we move into T-bills according the percentage specified by MPT.

Let’s look at the case where we want an average annual 10% compounded return, and let’s see how we would have fared through the previous two market downturns.

Beginning in 1990 with a $1 in each portfolio, the one based on market timing did in fact give us close to the 10% desired return whereas the buy-and-hold portfolio only returned a measly 4.2%, not even half of what we wanted. Even more surprising is how much riskier a buy-and-hold strategy is compared with a timed one; the risk parameter, as measured by the standard deviation of actual returns, is more than double (15.8% compared with 6.9%). Wow!

What this graph is showing us is how very detrimental a major downturn can be to a buy-and-hold portfolio. The moral of this story is that it pays to cut your losses when they’re small before they get to be so large that it would take many, many years to get back to previous peak values.

Conclusion
I’m really glad my thesis paid off in the creation of the SMC Analyzer because now there’s an analytical tool to help us realize our portfolio goals at reduced risk. I believe this to be a major advance in asset allocation theory. Some of the results from the Analyzer are mind-blowing, especially seeing how different market strategies impact portfolio returns. We’ll be discussing these in upcoming blogs. These next few weeks will be exciting indeed!

For further reference:
To learn more about Modern Portfolio Theory, please refer to Professor Pat’s excellent articles found in the blog archives published on these dates: 4/21-24/08 and 6/2-3/08.

Comments

[Jasper M:]

A beautiful piece of heresy, like recognizing there are irregulaities on the Moon, that we can now recognize is True. And actually say it out loud.

May 13 04:43 AM

[CM in MA:]

Dr. Kris, thanks for the interesting article. I do have a few questions, though.

- Were transaction costs and taxes taken into account when calculating the returns?

- What is meant by the "averaging period is re-optimized every month?"

- Does it worry you that two assets with the following monthly Hi-Low-Close data (55, 45, 50; 100, 20, 30) yield the same "typical price" input into the CCI calculation? If these assets then had following months of (50, 40, 45; 70, 25, 40), the "typical prices" would again be the same for that month. Carry this on long enough and the deviations from the moving averages also become the same. Likely, perhaps not. But it raises the possibility that a very volatile asset can masquerade as a far less volatile asset. Would this affect your strategy?

Thanks!

May 13 09:22 AM

[Roger Schre...:]

Good stuff Dr. Kris.

We are a market timing firm and our investors had double-digit POSITIVE returns in 2008. We have been doing market timing and sector rotation for twenty years, and we do NO PASSIVE investing whatsoever of our clients - the risk is just too high. Every major market in the world had declined 75% - 90% during the last century. No one - with more than two brain cells - can honestly believe that buy-and-hope is a conservative way to invest!

May 13 09:49 AM

[Dr. Kris:]

To CM in MA:

Thanks for your insightful comments. Here's a response to the points you brought up:

- Were transaction costs and taxes taken into account when calculating the returns?

No. These are, of course, different for everyone. Within a mutual fund family the occasional movement in and out of an asset class can be done without charge. As to taxes, if you are holding your investments in an IRA there are none. Outside of an IRA you will be paying taxes on realized gains at a rate that depends on many things but that is still preferable to loss of principal. Taxes must be paid sometime. It's one of the two things you can be certain about.

BTW, the SMC Analyzer does have the capability of adding these parameters along with others such as margin costs and account interest.

- What is meant by the "averaging period is re-optimized every month?"

This refers to the portfolio growth graph which is essentially a backtest of following the market timing versus the buy and hold strategies each month and then reaping the results of following the recommendations. Every month with the addition of new data the averaging period of the CCI for each asset class is re-optimized to maximize the effectiveness of its use in terms of portfolio growth.

- Does it worry you that two assets with the following monthly Hi-Low-Close data (55, 45, 50; 100, 20, 30) yield the same "typical price" input into the CCI calculation? If these assets then had following months of (50, 40, 45; 70, 25, 40), the "typical prices" would again be the same for that month. Carry this on long enough and the deviations from the moving averages also become the same. Likely, perhaps not. But it raises the possibility that a very volatile asset can masquerade as a far less volatile asset. Would this affect your strategy?

First of all, we are using asset classes here composed of many individual components hence the volatility intra-month should theoretically be dampened. Nevertheless, the analyzer uses only closing prices as a substitute for typical price. You do bring up an important point however. That sampling the performance of an asset class only at periodic intervals has a nonzero probability of misunderstanding its true characteristics. With 80+ years of monthly data available however there is a high confidence that the volatility of the asset classes used in the analysis is properly represented.

May 14 07:31 PM

[CM in MA:]

Dr. K,

Thank you for the replies to my questions. I appreciate the time you took to provide the extensive answers.

CM


On May 14 07:31 PM Dr. Kris wrote:

> To CM in MA:
>
> Thanks for your insightful comments. Here's a response to the points
> you brought up:
>
> - Were transaction costs and taxes taken into account when calculating
> the returns?
>
> No. These are, of course, different for everyone. Within a mutual
> fund family the occasional movement in and out of an asset class
> can be done without charge. As to taxes, if you are holding your
> investments in an IRA there are none. Outside of an IRA you will
> be paying taxes on realized gains at a rate that depends on many
> things but that is still preferable to loss of principal. Taxes must
> be paid sometime. It's one of the two things you can be certain about.
>
>
> BTW, the SMC Analyzer does have the capability of adding these parameters
> along with others such as margin costs and account interest. <br/>
>
> - What is meant by the "averaging period is re-optimized every month?"
>
>
> This refers to the portfolio growth graph which is essentially a
> backtest of following the market timing versus the buy and hold strategies
> each month and then reaping the results of following the recommendations.
> Every month with the addition of new data the averaging period of
> the CCI for each asset class is re-optimized to maximize the effectiveness
> of its use in terms of portfolio growth.
>
> - Does it worry you that two assets with the following monthly Hi-Low-Close
> data (55, 45, 50; 100, 20, 30) yield the same "typical price" input
> into the CCI calculation? If these assets then had following months
> of (50, 40, 45; 70, 25, 40), the "typical prices" would again be
> the same for that month. Carry this on long enough and the deviations
> from the moving averages also become the same. Likely, perhaps not.
> But it raises the possibility that a very volatile asset can masquerade
> as a far less volatile asset. Would this affect your strategy?<br/>
>
> First of all, we are using asset classes here composed of many individual
> components hence the volatility intra-month should theoretically
> be dampened. Nevertheless, the analyzer uses only closing prices
> as a substitute for typical price. You do bring up an important point
> however. That sampling the performance of an asset class only at
> periodic intervals has a nonzero probability of misunderstanding
> its true characteristics. With 80+ years of monthly data available
> however there is a high confidence that the volatility of the asset
> classes used in the analysis is properly represented.
>

May 15 07:56 AM

[Ken - Break...:]

Great article, thanks! Buy/hold died 10 yrs ago and even then it was mediocre at best.

May 21 05:38 PM