| |
 September 2004 History Isn't Enough
Or does it? On the face of it, asset allocation is nothing more than a portfolio's division between stocks, bonds, and cash. More sophisticated models may incorporate any number of subclasses such as domestic large cap growth stocks, foreign value stocks, municipal bonds, and even real estate or gold. Not surprisingly, the first decision is the selection of potential asset classes (or subclasses). The second decision involves rebalancing. As soon as any asset allocation is put in place, the various component classes will immediately begin trading with differing results. As a result, the initial asset allocation will change very quickly. Do you simply let it go, rebalance back to the original mix, or create another one?
The first alternative is the well-known "buy-and-hold" approach. It saves on transaction costs since no rebalancing is ever needed, but it leaves the portfolio at the whims of the market. The second approach is called "static asset allocation" and tends to work best when the market is relatively directionless. The third option is "dynamic asset allocation" since it doesn't rely on just one set mix, but instead tries to rebalance based on current market conditions. It works best when the market has a definite trend. The third asset allocation decision only applies to the static and dynamic dynamic methods since it involves the timing of rebalancing. Do you realign the portfolio at specific time intervals or when the asset classes have strayed a specified percentage from the original allocation? The answer to this will not only have a bearing on returns, but transaction costs as well. The introduction of Portfolio 6 contained a detailed comparison of these asset allocation approaches including their different return patterns and rebalancing mythologies, so it won't be necessary to repeat it here. The one thing it did not address was the significance of the fourth asset allocation decision. The Fourth DecisionAlthough the fourth decision may apply to the creation of buy-and-hold and static asset allocation strategies, it lies at the core of dynamic asset allocation. This is a question of inputs: Is the asset allocation based on entirely historical returns or does it also incorporate predictive elements?For many investors, history is enough. Buy-and-hold investors may study return patterns from the past and construct portfolios based on long-term trends. This is appropriate since buy-and-hold is by definition, a long-term approach. The underlying belief is that there's little need to try to time the market since financial markets can be expected to behave as they have in the past. Investors using static asset allocation can take this same approach. What they seek is the mix that worked the best in the past since it's likely to show similar results in the future. That's precisely what we did when we created the static benchmark for Portfolio 6 and searched for the best asset allocation of the Twentieth Century. Those using dynamic asset allocation can do so as well, but at the risk if undermining its whole purpose. After all, the reason you use dynamic asset allocation is to take advantage of current market conditions but that's not really possible with purely historical data. The problem stems from the fact that recent data points -- the most recent months or quarters -- are overwhelmed by older data. This is true even when present conditions are rapidly changing. That's why historical data works much better for the longer time horizons of buy-and-hold and static asset allocation. The time horizon for dynamic asset allocation is considerably shorter, in fact it's actually equivalent to the rebalancing frequency. In other words, if a new allocation is created every quarter, its time horizon is three months. 123 or 126, Not Much DifferenceTo illustrate, consider Portfolio 5. This is a multicap all-equity portfolio based on market factors dating back to January 1, 1994. It's rebalanced in the first month of every calendar quarter.P5 was rebalanced in mid-July using market data through June 30, 2004. At that point, there were were 126 data points. The prior rebalancing occurred in mid-April. Since that was three months earlier, there were three less, or 123 data points.
Do you think the model changed much during that time? Did an additional three months of data really make a difference? It didn't on a strictly historical approach. In fact, the recommended asset allocation for the third quarter was the exact same one for the second: 7% Large Cap Growth and 93% Midcap Value. But think back to what happened in the second quarter of 2004. Investors suddenly shifted gears from worrying about job growth and the sustainability of the recovery and turned their attention to the prospect of higher interest rates. As April began, they feared growth was too tepid but by the final days of June the Fed was beginning a tightening campaign to prevent it from overheating. That's quite a shift in just three months. Many investors spent the second quarter realigning their portfolios in preparation for the coming changes. Wouldn't you think an all-equity portfolio -- arguably one of the riskiest types -- might also need to undergo some modification? It did indeed when short-term assumptions were added to the historical information. As you can see from Chart 1, it made a value-weighted move into midcaps: 67% Midcap Value and 33% Midcap 400 Index. Instead of completely relying on how markets reacted over the past ten and half years, the "Building Blocks" model incorporates current market factors as well. This is a procedure developed by Ibbotson and Siegel to enhance the model's predictive ability. According to Ibbotson, By combining current expectations with historical risk premia, you take into account current market conditions (the economic expectations of investors) and historical market returns. The use of risk premia versus a pure historical return, increases the predictive power of the model since historical risk premia are more stable over time than the pure historical return of an asset class...With the Building Blocks model, the expected return of an asset class represents the sum of the current risk free rate and one or more historical risk premia or building blocks. To generate the third quarter allocation for P5, we used a risk-free rate, and equity risk premium, and a small stock premium. With interest rates hovering near generational lows and not likely to rise quickly, we used a lower than average risk-free rate. Historical values were sufficient for the equity risk and small stock premiums although there are times in the market cycle where adjustments are appropriate there, too. Frontier LifeNot surprisingly, the resulting asset allocation -- and indeed, the entire frontier of all efficient portfolios -- differed considerably from those of the historical model. Chart 2 illustrates the differences. As you you'll notice, both efficient frontiers are shown. (The "efficient frontier" is simply the collection of all asset mixes that produce the highest return for a give level of risk.) The Building Blocks frontier is represented by the red line while the historical one is given by the thinner black line.
Obviously there's not just one efficient frontier. It's a function of its underlying asset classes and the time period. Even when the same data series are used, the frontier will shift over time. As the number of data points rises, the less impact each additional one will have. That's again why the three extra months of results resulted in very little change to historical asset mix. That's also why the historical frontier has a distinctly different shape than the short-term one. Remember it's based on the time period extending from January 1994 through June 2004. While it includes the three-year bear market (2000-2002), for the most part this was a period of abnormally high equity returns. All stocks rose, led by the S&P 500's largest of the large -- typically some of the least volatile issues. With very little reward for additional risk, the resulting historical efficient frontier is quite steep and the returns across all domestic equity classes are above historical norms. On the other hand, the Building Blocks approach yields a flatter efficient frontier. It covers a much greater range of risk (standard deviation on the horizontal axis) ranging from 7% to just under 19%. This means it offers greater returns for additional risk than does the historical frontier. Across the board, returns are also lower, even though they are still somewhat higher than the norm. At present this is quite reasonable since few if any investors expect to see a return of the 20%+ returns of the 1990s. In fact, with interest rates poised to continue rising, even the Building Blocks efficient frontier may be too optimistic. Both efficient frontiers are based on the nine domestic equity subclass returns combining growth, value, and core along with large, mid, and small capitalizations. Most investors are familiar with this from the Morningstar Stylebox (see Think Inside the Box). Each is represented by the corresponding S&P index.
Chart 2 plots the expected risk and returns for each series based on the Building Blocks model. For the historical approach they would be higher up the graph and further to the right. Again this is the difference between short-term assumptions and actual historical results. Any point along the efficient frontier represents an asset mix that produces the greatest return for the given level of risk. As you move along the frontier, the allocation changes. Those asset classes falling nearer the frontier will enjoy greater representation although low correlations may at times bring in the others as well. All portfolios lying on both the Building Blocks and historical efficient frontiers are captured graphically in Chart 3. Each of the colors represents one of the nine asset classes shown on Chart 2. The positions on the horizontal axis represent the percentage of the left-to-right distance along the frontier. The percentages on the vertical axis can be used to judge the asset class weighting at any given level of risk. When we created Portfolio 5, we based it on the "Point 75 Portfolio", the asset mix that's 75% of the way along the efficient frontier when moving from left to right. This is the point used whenever P5 is rebalanced. That portfolio is highlighted by the red line on both graphs in Chart 3. Again the difference between the short-term Building Blocks and long-term historical methodologies is abundantly clear. Both heavily rely on Midcap classes, but the Building Blocks portfolios are more diversified. Indeed, on the historical frontier, Midcap Value starts at a 45% weighting at position 0 and moves up from there. On the other hand, it doesn't reach 45% for the Building Blocks until position 60. So it's easy to see why the two approaches offer differing allocations for P5 -- they differ at all points along the frontier. Of course that's what you'd expect when the two frontiers (as shown on Chart 2) have such different characteristics. That's also why the Historical Point 75 Portfolio lies beneath the Building Blocks efficient frontier while falling on the historical one. Although not illustrated on Chart 2, the Building Blocks Point 75 Portfolio suffers a similar fate when plotted against the historical efficient frontier. Again, all of this is attributable to the fact that the historical frontier is based upon the performance of the indexes in the 1990s while the Building Blocks approach incorporates present market and economic conditions. In essence, it doesn't anticipate the oversized equity returns of the past decade. Does It Really Make a Difference?Speaking of returns, that's the one remaining thing to be considered. It's obvious that the two methods' efficient allocations differ, but what about their returns?On the face of it, one might think they would be quite similar. As you can see from Chart 1, both rely heavily on Midcaps, and even more specifically Midcap Value.
There's no precise way to answer this question since this is being written during the quarter in which these allocations are recommended. There are, however, other means of comparison. First of all, how would these mixes have fared last quarter when economic and market conditions were quite similar? Obviously this isn't a prediction of expected results for this quarter but simply a means of comparing returns. For the period April 1, 2004 - June 30, 2004, the Building Blocks Point 75 Portfolio rose 1.12% while the historical counterpart added 1.30%. That may seem like a small difference, but a quarter is a short period of time. In percentage terms, the historical allocation was actually 16% higher. That won't always be the case, but at least in this instance, the difference was significant. A second way to compare the allocations is to consider expected returns. Obviously no prediction has absolute certainty, but you can derive meaningful estimates based on probabilities. Chart 4 shows the relevant statistics as well as the return probabilities for each mix. All of these figures assume a one-year holding period that is somewhat longer than the quarter we normally employ. Nevertheless, they illustrate differences that would apply in the shorter as well as longer periods.
The top of the chart compares return, risk (standard deviation), and risk-adjusted return (Sharpe Ratio). You may have already gotten a sense of the allocations' risk and return from Chart 2. Both are roughly at the same distance up the vertical axis (return) but the historical mix is much further along the horizontal axis (risk). Chart 4 quantifies this, showing only a 0.06% difference in expected return, but a considerably greater difference between standard deviations. In other words, both mixes have roughly the same return, but the historical allocation requires greater risk to achieve it. This difference is also captured by the respective Sharpe Ratios that measure risk-adjusted return. They are calculated by subtracting the risk-free rate of return from the allocation's expected return, and dividing that result by the allocation's standard deviation. The greater the ratio, the higher the risk-adjusted return. In this case, the Building Blocks mix is 12.1% greater than the historical alternative. The return probabilities at the bottom of Chart 4 are the projected likelihood that each portfolio would achieve each level of return over the one-year holding period. In each instance, the Building Blocks allocation has a greater probability. Of course these are just projections. In fact, there's no guarantee these projected annual results will even be partially approximated over the usual three-month holding period. Indeed, while all of Chart 4's statistics suggest a greater efficiency for the Building Blocks' allocation, it would have underperformed its historical counterpart over the prior quarter. Still a ChoiceNo matter how quantitative your approach may be, statistics and probabilities can only point you in the direction of what might happen, not necessarily what will occur. In any given period -- especially over the short-term -- results may deviate from expectations.So the best you can do is to get the odds in your favor. As Charts 1 - 3 so vividly illustrate, there is a substantial difference between the allocations resulting from purely historical data and those generated with current and forward-looking projections. Chart 4 quantifies these differences, but doesn't necessarily favor one approach over the other. The fact that the current Building Blocks allocation compares favorably to its historical counterpart does not mean this will always be the case. It is not necessarily an endorsement for the approach, just this particular result, and even then there's no guarantee of outperformance. The only thing it does establish is that the different approaches yield different results. The Building Blocks method is only one of many alternatives to the historical approach. The choice to use the former rather than the latter (or any of the other alternatives) is simply that -- a choice. We've elected to use the Building Blocks method for both P5 and P6 in an effort to make them more dynamic. As economies and financial markets grow more global and complex, volatility has increased. It's our belief that a more dynamic approach is better enabled to capture this changing environment. But that's our opinion. Others may, and indeed do, differ. That's why there are so many different models and why there's such a divergence in managers' returns even when they have the same stated objective. Assumptions and inputs really do matter. Search this site! Just enter you key word or words:
Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
SSET ALLOCATION IS ONE OF THE major determinants of an investor's long-term return. It's a good thing it doesn't require a lot of thought.
