| |
 May 2004 Blend and Change Introducing Portfolio 6: A Dynamic Quantitative Portfolio A Primer on Asset Allocation Strategies
To be sure, asset allocation isn't nearly as sexy as market timing and stock picking, yet studies have shown it's the major determinant in long-term performance. Isn't it ironic that most investors spend most of their time trying to time the market or find the latest hot stocks? For many, asset allocation is no more than an afterthought -- if even that.
We addressed this issue several years ago when we introduced Portfolio 5. It's an all-stock model that's designed to take advantage of the shifting performance among styles and capitalizations. But most investors also hold bonds and/or cash. Even P5 had a rough go in 2002 and early 2003 when stocks were still suffering through the bear market. Portfolios with a fixed income element held up much better. Until now, all of our models were pure stock portfolios. With the introduction of Portfolio 6, we're offering greater diversification through the inclusion of bonds as well. Keeping it SimpleWhat we're talking about here is a broad-market model. There are various classes of assets (e.g. large cap growth stocks, foreign stocks, intermediate term government bonds, high quality corporate bonds, etc.) not to mention thousands of individual securities that could be included. To make this workable, it's necessary to narrow down the possibilities yet still provide diversification.We wanted to create a model that individual investors could comfortably use so we limited the universe of potential holdings to those commonly found in their portfolios. This reduced the broad categories to five, a very workable number. Of the five, two are bond (intermediate government bonds and a combination of government, agency, and corporate bonds of various maturities and credit quality) and three stock classes (large cap domestic, small cap domestic, and developed foreign country stocks). In designing and backtesting the model, we used their proxies, the Lehman Brothers Intermediate Government Bond Index, the Lehman Brothers Aggregate Bond Index, the S&P 500, the Russell 2000, and the MSCI EAFE Index, respectively.
Emerging market stocks have lower correlation with domestic stocks and could provide a greater degree of diversification. Junk bonds or real estate investment trusts (REITs) are also good diversifiers, but many individual investors avoid these classes. In essence, the classes we included in P6 aren't necessarily the best representatives of the broad market, but they are the ones most investors know and use. In an additional simplifying step, P6 doesn't use individual stock or bonds, but rather exchange traded funds (ETFs). This is the same approach used in Portfolio 5. ETFs are similar to index mutual funds in that they are designed and managed to track major market benchmarks. Unlike mutual funds, however, they're listed on an exchange (generally the AMEX) and can be bought and sold throughout the day. Like stocks, you have to go through a broker to buy and sell them, and each transaction carries a commission. Since these costs vary by brokers and levels of service, we've ignored them in our calculations, relying solely on gross returns. Aside from trading costs, most ETFs have relatively low management expenses making them excellent proxies for their target benchmarks. The ETFs used in P6 are given in Chart 2. With the exception of SPY, the S&P Depositary Receipt for the S&P 500, all the rest are iShares from Barclay American.
The universe of potential holdings for P6 is therefore limited to five different issues. Since each is an ETF, this model is cost efficient as well as diversified. You may have noticed that cash is not one of the available asset classes. Although a staple in many individual investors' portfolios, cash is not an appropriate holding for the long-term -- not even when it's rebalanced every quarter. Static vs. Dynamic AllocationThe next step is to come up with the allocation model. Once that's determined it's just a matter of running the data through a mean-variance optimizer to arrive at the optimum mix.We wanted P6 to actually be forward-looking, not simply the result of past performance -- we've already considered that. Instead, the goal for P6 is to be predictive based not only on historical market activity, but current conditions as well. For this to work, P6 has to be a dynamic model rather than a static one. In other words, it needs to change the mix of assets as market conditions dictate. This is not simply a process of rebalancing back to a specified allocation but rather the creation of new mixes best suited to current conditions. There are fundamental differences between dynamic and static approaches to asset allocation, not only in their construction but in their return profiles as well. It's worth taking a few minutes to consider the differences. As you've probably surmised, there's more than one way to allocate assets. With the exception of those whose portfolios are simply conglomerations of previously hot stocks, most investors use an allocation process -- whether they know it or not. Indeed, the simplest form of asset allocation is buy-and-hold. This is popular with many so-called "conservative" investors who shun the risks of trading. In essence, they create a portfolio of securities and simply hold them for the long-term. With little (or no) trading, this is clearly a "passive" asset allocation approach. But to say it's passive is not to say there aren't any changes. While the number of shares of stock or par value of bonds may remain the same, the actual allocation is constantly changing. The investor isn't responsible for the changes, the market is. The allocation varies right along with market fluctuations. A static or "constant-mix" allocation sets a specific percentage for each asset class and then periodically rebalances back to it. For example, many investors use 60% stocks/40% fixed income as their standard. If the stocks rally and the equity portion rises to 65%, they'll sell 5% and reinvest it back in fixed income to bring the portfolio back to the original mix. This strategy forces the investor to "sell high and buy low" by trimming appreciated assets and reinvesting the proceeds back into the classes that have underperformed. Dynamic asset allocation calls for periodic rebalancing, too, but not to a constant standard. Instead, dynamic approaches attempt to rebalance in such a way as to take advantage of current market conditions. If stocks are rising and it looks like they will continue to rally, this approach may actually increase the equity portion rather than reinvesting in underperforming asset classes. If done properly, dynamic asset allocation models can exploit short-term market trends rather than relying solely on long-term results. Different Strategies, Different PayoffsAs you've probably already realized, each allocation strategy provides a distinctly different set of returns under changing market conditions. You can clearly see this in Chart 3's payoff diagram. It graphs the total portfolio value on the vertical axis and the level of the stock market on the horizontal axis.
This diagram assumes each strategy starts with 60% stocks and 40% T-Bills and that the stock market starts off at a level of 100. As the market moves (in this case stocks since T-Bills are considered constant over the short-term), the buy-and-hold investor makes no changes, the static asset allocator rebalances back to the 60/40 mix, and the dynamic allocator adds 5% to the initial equity allocation if stocks are rising and subtracts that amount if they are falling. The black line on Chart 3 illustrates the various payoffs of the buy-and-hold approach. Although it may not be obvious from the chart, the payoffs actually fall in a straight line since the value of this portfolio moves in direct proportion to the level of the stock market. The minimum value will never be less that $40 since this will be the value of the T-Bills even if the stock holdings go to zero. That's not the case for the green line that represents the static asset allocation strategy. Why? Because as the stock market falls, equities will be the underperforming asset so it forces the investor to sell bills and buy stocks to maintain the 60/40 mix. Notice that this also causes it to underperform the buy-and-hold approach when stocks are rising, too. Rather than allowing the winning asset class to run, it forces the investor to continually trim that position and reinvest in bills. That doesn't happen with the dynamic approach. As the red line shows, it outperforms the other two strategies when stocks are either trending up or down because it adds to the winning class at the expense of the trailing one. But don't be misled by Chart 3: The dynamic approach is not always destined to yield the best results. The reason it looked so good in Chart 3 was because it outperformed at the extremes, when the stock market was trending. That's not always the case as there have been long periods of time -- even years (think the 1970s) -- when the market has been virtually directionless or at the very least, didn't move as expected. Payoffs from the three strategies are decidedly different then.
To see why, consider Chart 4 which illustrates what happens in a trendless market. As before, it begins with the stock market at 100 and each strategy with 60% in stocks and 40% in T-Bills. Now assume the stock market falls to 90. Since each strategy started with the same mix, each will suffer the same losses, ending with a value of $94. it's what happens next that's significant. The buy-and-hold strategy calls for no changes -- it never does. The static allocation strategy will rebalance the portfolio to have 60% in stock and 40% in bills. In this case, there will be $56.40 (60% x $94) in stock and $37.60 (40% x $94) in bills. Since stocks fell, the dynamic strategy will actually increase T-Bills to 45% and cut the stock position to 55%. This will leave it with $51.70 (55% x $94) in stock and $42.30 (45% x $94) in bills. But now suppose stocks don't continue falling but rather reverse course and once again end at 100. This is what happens in a trendless market. In this case, the buy-and-hold portfolio will once again be worth $100. It's right back where it started and that's exactly what you'd expect given that the only changes in the mix are the result of the market. Payoffs of this strategy move along a straight line, remember? That's not the case for the other two strategies. Each of them altered the mix left by the market. In rebalancing back to the 60/40 blend, the static approach had to increase stocks relative to where the market left them. As a result, when the market reversed and went up, it was 27 cents ahead of its original value ($100.27 vs. $100.00). In this case, buying low and selling high worked out. The exact opposite happened to the dynamic approach. It cut the stock position and added bills just as stocks began to recover. As a result, this portfolio would only be worth $99.74.
At this point, the dynamic strategy would again add to stocks, bringing them back to 60% of the portfolio or $59.85 (60% x $99.74). T-Bills would be reduced to $39.89. The static mix would also rebalance back to the 60/40 mix, but starting with its higher value it would have $60.16 (60% x $100.27) in stocks and $40.11 in T-Bills. The buy-and-hold strategy would, of course, make no changes. If the market again reverses and falls back to 90, the dynamic strategy would again be the loser, falling to a value of $93.76. Once again it increased the wrong asset class on the expectation that the market would trend instead of reverse. The static approach fared much better given the fact that it reduced stocks and increased bills prior to the market reversal. The buy-and-hold approach moved back to where it was before when the market was at 90. The final row in Chart 4 shows the values should the market head back to 100. You, no doubt, already see the pattern which is illustrated in Chart 5. Throughout the market oscillation, the buy-and-hold approach simply moves back and forth in a straight line. The slope of this line (remember 8th grade algebra?) is determined by the stock/bill mix and since this strategy initiates no changes, it remains constant. That's not the case for the other two strategies since each time they rebalance, they alter the slope of the payoff curve. The static approach actually benefits from a trendless market by buying high and selling low. The longer the market fluctuates, the better it does. The dynamic strategy suffers since it's designed to benefit from market trends and in this case there are no market trends. Instead, it gets whipsawed by making bets on the market just before a reversal. Returns continue to suffer until a lasting market trend emerges. Why P6 is DynamicBy now, a few conclusions should be fairly obvious:
Which brings up two final observations regarding the previous examples. First, only one decision rule was demonstrated for both the static and dynamic strategies. Changing the rule would alter the specific payoffs, but not their relative performance. Regardless of the rule, static approaches will dominate in fluctuating markets and dynamic will outperform in trending ones. That is, of course, as long as the dynamic decision rule works (guesses?) properly. In the trendless market example, the dynamic approach suffered because the market moved against it however similar results can be obtained if the strategy's decision rule simply leads it in the wrong direction.
And that brings us back to Portfolio 6. In attempting to create an asset allocation model appropriate for most investors, the dynamic approach is arguably the most appropriate. As you can see from Chart 3, it affords some degree of downside protection by reducing equity exposure in a falling market while also allowing the portfolio to take on additional risk as wealth and stocks are rising. These characteristics should appeal to individual investors. This would also seem to fit with their risk profiles since most would at least consider cutting back on stocks as the size of their portfolios dwindled. Of course the buy-and-hold approach also offers this benefit, but not to the same extent. It also lacks the potential to outperform in an upwardly trending market. As far as the decision, rule, we've never been in favor of trading for trading's sake. To a great extent, the rebalancing decision is left to the quarterly mean-variance optimizing process, but P6 will only be rebalanced if at least one of its asset classes requires a change of 5% or more. Small changes may have impact on return, but trading costs definitely affect net returns. Two BenchmarksSo how do you measure results? There are no true broad market benchmarks to use as a yardstick, and with a combination of both stocks and bonds a pure equity or fixed income benchmark would be inappropriate.Instead, the most meaningful comparison would be to a buy-and-hold and static asset allocation approach. That's precisely what we're proposing to do. In March 2004 we looked back over the past 25 years to determine what the best static asset allocation model would have been. It turns out to have been a blend of 25% domestic government, agency, and corporate bonds, 48% domestic large cap stocks, 21% domestic small cap stocks, and 6% foreign stocks of developed nations. This finding will provide the benchmarks for P6. As in the earlier research, P6 is based on the past 25 years of historical market results. This data is augmented by projections of interest rate levels and equity risk premiums to create a forward-looking optimization. Inputs, therefore, are quite compatible in these static and dynamic models. P6's results will be compared to two different benchmarks, one static and one buy-and-hold. The static benchmark will use the 25/48/21/6 mix above. It will be rebalanced in the first month of every calendar quarter on the same day as P6. The buy-and-hold benchmark will start as of January 1, 2004 with the same mix of assets as the static benchmark. Just as you'd expect, there will be no changes to this blend other than those caused by the market. We're using January 1, 2004 as the starting date for the buy-and-hold portfolio since that is the official date when we took P6 out of sample and began live testing. The model and the benchmarks will therefore have the same start date. P6's returns are updated every week at the bottom of the Home Page. Beginning in May 2004, Historical Performance will detail its performance and that of its two benchmarks. Finally, research and observations regarding this model will periodically appear on future updates of this page, since like P3-P5, P6 is still a work in progress. Search this site! Just enter you key word or words:
Get current quotes or follow your own custom portfolio,
courtesy of E-Line Financials:
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
EAL ESTATE AGENTS WILL TELL YOU the secret to success in their field is location, location, location. The secret to successful investing is quite similar: allocation, allocation, allocation.
