Dynamic Portfolio Book
Posted: Fri Jul 02, 2004 7:10 am
I recently put a book, Dynamic Portfolio Theory and Management by Richard Oberuc, at the top of my reading list. Our investigations have sometimes been called market timing and dynamic portfolio allocation. Having read this book, I know the real thing. Those allegations against us are clearly reckless and misleading. I suspect that there is similarity between the approach taken in this book and the infamous Foolish Four.
The book has just enough mathematical notation so as to run many people off. It helps if the readers have at least a superficial understanding of some of the mathematical concepts involved. But the book has very little in the way of actual mathematics.
Dynamic Portfolio Theory and Management surveys several modeling approaches. Not knowing what was involved, I anticipated finding out something of value for us. But instead of measured, slow shifts in allocations, I discovered that the writer's time frame was usually one month and one year at the longest. Needless to say, that is the kind of hyperactive trading activity that has given market timing a bad name.
As it turns out, the author includes an excellent review both of modeling and market timing. He defines market timing as being 100% in a risky asset or 100% in a low risk asset such as cash. This definition differentiates market timing from dynamic asset allocation, which allows for partial allocations among any number of investments.
The academic case against market timing has been critically dependent upon high transaction costs. Sharpe's 1975 study assumed a 2% transaction fee. Jeffrey (1984) assumed a 1% transaction cost. Those investigations showed that an exceedingly high prediction accuracy would have been required. (Sharpe's numbers were 83% for matching the return of buy-and-hold and 74% to match the standard deviation. Jeffrey's number was 72%.) Later, Sy (1990) showed that Sharpe's study depended greatly on his selection of years. For example, adding 1929 to 1933 to the sample would have reduced the break-even point from 83% to 65%. Other investigators have looked at very short-term trading and low fees. Lam and Li (2002) showed that the break-even point is around 68% with monthly revisions for a transaction fee of 0.4%.
The writer presents the results of several studies that show the full return potential when assuming a perfect forecasting ability. Typical results with 0.5% transaction costs were 14% to 16% for annual revisions and 25% to 35% with monthly revisions. Adding a third investment choice brings the potential to 43% with monthly timing (with all allocations restricted to 0% or 100% at any given time). A fourth choice brings the potential to 59% with monthly timing.
As soon as prediction models are introduced, the advantage over long-term buy-and-hold drops to 2% to 3%. In terms of money managers who actually engage in timing, their advantage varies between 0% and 3% depending upon the time interval examined. Roughly 50% to 70% actually show significant market timing skill. The remaining 30% to 50% do not.
With dynamic asset allocation (which includes leverage and partial allocations), the greatest realistic advantage over long-term buy-an-hold increases to 4% (with stocks and T-bills) or 5% (with multiple asset classes).
Beyond the Numbers Themselves
I frequently mention that we need to look beyond numbers by themselves. I am more interested in the logic behind the calculations than in the mathematical details used in making the calculations. When we do this, technical training offers little advantage. Others can assist in identifying hidden flaws.
I think that John Mauldin has identified an important flaw in this approach. In one of his recent free email newsletters, he suggested that there is a finite amount of money that skill can extract from a market. With a large number of people following the same general approach with similar goals, the advantage gained by any one individual shrinks to a very low amount.
I believe that there is a more important flaw. It is what John Bogle has mentioned relating to the size of mutual funds. Money managers pay much higher the transaction costs than those studied. The costs are hidden, but they are real. Because money managers control significant amounts of money, their trading activity works against them. They cannot reallocate significant sums of money without changing prices to their own disadvantage.
Conclusions
I think that there is a potential advantage for small investors who time the market. It is small. It carries the risk of unforeseen factors and events.
The case against market timing has been based upon very short-term (one year or less), frequent trading with high fees. It breaks down when the fees are reduced.
I fault those who have applied the term Market Timing loosely just to gain a debating point. I single out John Bogle because he alone has credibility.
We don't do anything like what the book describes on this board. We change allocations gradually over long periods of time. We do not assume that we will generate exact numbers going forward. We include sensitivity studies.
Have fun.
John R.
The book has just enough mathematical notation so as to run many people off. It helps if the readers have at least a superficial understanding of some of the mathematical concepts involved. But the book has very little in the way of actual mathematics.
Dynamic Portfolio Theory and Management surveys several modeling approaches. Not knowing what was involved, I anticipated finding out something of value for us. But instead of measured, slow shifts in allocations, I discovered that the writer's time frame was usually one month and one year at the longest. Needless to say, that is the kind of hyperactive trading activity that has given market timing a bad name.
As it turns out, the author includes an excellent review both of modeling and market timing. He defines market timing as being 100% in a risky asset or 100% in a low risk asset such as cash. This definition differentiates market timing from dynamic asset allocation, which allows for partial allocations among any number of investments.
The academic case against market timing has been critically dependent upon high transaction costs. Sharpe's 1975 study assumed a 2% transaction fee. Jeffrey (1984) assumed a 1% transaction cost. Those investigations showed that an exceedingly high prediction accuracy would have been required. (Sharpe's numbers were 83% for matching the return of buy-and-hold and 74% to match the standard deviation. Jeffrey's number was 72%.) Later, Sy (1990) showed that Sharpe's study depended greatly on his selection of years. For example, adding 1929 to 1933 to the sample would have reduced the break-even point from 83% to 65%. Other investigators have looked at very short-term trading and low fees. Lam and Li (2002) showed that the break-even point is around 68% with monthly revisions for a transaction fee of 0.4%.
The writer presents the results of several studies that show the full return potential when assuming a perfect forecasting ability. Typical results with 0.5% transaction costs were 14% to 16% for annual revisions and 25% to 35% with monthly revisions. Adding a third investment choice brings the potential to 43% with monthly timing (with all allocations restricted to 0% or 100% at any given time). A fourth choice brings the potential to 59% with monthly timing.
As soon as prediction models are introduced, the advantage over long-term buy-and-hold drops to 2% to 3%. In terms of money managers who actually engage in timing, their advantage varies between 0% and 3% depending upon the time interval examined. Roughly 50% to 70% actually show significant market timing skill. The remaining 30% to 50% do not.
With dynamic asset allocation (which includes leverage and partial allocations), the greatest realistic advantage over long-term buy-an-hold increases to 4% (with stocks and T-bills) or 5% (with multiple asset classes).
Beyond the Numbers Themselves
I frequently mention that we need to look beyond numbers by themselves. I am more interested in the logic behind the calculations than in the mathematical details used in making the calculations. When we do this, technical training offers little advantage. Others can assist in identifying hidden flaws.
I think that John Mauldin has identified an important flaw in this approach. In one of his recent free email newsletters, he suggested that there is a finite amount of money that skill can extract from a market. With a large number of people following the same general approach with similar goals, the advantage gained by any one individual shrinks to a very low amount.
I believe that there is a more important flaw. It is what John Bogle has mentioned relating to the size of mutual funds. Money managers pay much higher the transaction costs than those studied. The costs are hidden, but they are real. Because money managers control significant amounts of money, their trading activity works against them. They cannot reallocate significant sums of money without changing prices to their own disadvantage.
Conclusions
I think that there is a potential advantage for small investors who time the market. It is small. It carries the risk of unforeseen factors and events.
The case against market timing has been based upon very short-term (one year or less), frequent trading with high fees. It breaks down when the fees are reduced.
I fault those who have applied the term Market Timing loosely just to gain a debating point. I single out John Bogle because he alone has credibility.
We don't do anything like what the book describes on this board. We change allocations gradually over long periods of time. We do not assume that we will generate exact numbers going forward. We include sensitivity studies.
Have fun.
John R.