|
About IndexInvestor.com |
Privacy Policy |
Transaction Policy |
Legal Disclaimers |
Contact Us |
My Account |
Home
|
|||||
|
|
|
|
|
||
Reviewing your portfolio's performance and rebalancing it back to your target asset class weights is one of those ideas that sounds great in theory, but in practice can get you into trouble if you're not careful.
Consider two extreme approaches. In the first case, the portfolio is rebalanced to equal asset class weights on a daily or weekly basis. In the second case, the portfolio is never rebalanced; its asset class weights simply vary over time in line with changes in their respective market capitalizations (a strategy that research suggests is implicitly used all-too-often in many 401k accounts!). In effect, the essence of the first strategy is to constantly sell winners and buy losers (call it "extreme value"), while in the second it is to do just the opposite (call it "extreme momentum"). Intuitively, most people sense that the best approach is probably somewhere in between these two. To put it a bit more formally, most people sense that both value, (i.e., the idea that asset returns tend to revert over time to their long term average), and momentum (i.e., the idea that returns tend to continue in the same direction) both contribute to long-term asset class returns, though to varying degrees from year to year.
Academic research supports this view. For example, in their recent paper "Momentum and Mean Reversion Across National Equity Markets", Balvers and Wu find considerable evidence that both mean reversion and momentum affect security returns, though over different time periods. Typically, mean reversion (also known as contrarian or value strategies) operates over time horizons of roughly three years, while momentum is a shorter term phenomenon, working over a one year horizon.
When you think about it, these findings intuitively make sense. Mean reversion in financial markets reflects the operation of real economic processes at the firm level, where companies often earn high returns following innovation, which are subsequently eroded as competitors either copy them at lower cost or introduce even newer offerings themselves. In contrast, momentum in financial markets is solidly grounded in individual investor psychology. It is widely recognized that it takes far less information to form an initial opinion than it does to change it. Moreover, once formed, an initial opinion affects not only the information to which we pay attention, but also the weight we give to it. Both of these factors cause us to be overconfident about the accuracy of our forecasts for the future (e.g., a belief that a security or market will continue to deliver high returns).
Given these findings, portfolio rebalancing is essentially a risk management tool, in which your are trying to manage this balance between value and momentum so as to maximize your return per unit of risk. Unfortunately, rebalancing tends to be costly, in terms of both the transaction fees involved and tax payments it can trigger (if your investments are held in a taxable account). So not only will very frequent rebalancing limit one's exposure to momentum driven gains, but the taxes and transaction fees involved will also further reduce returns. On the other hand, very infrequent rebalancing will minimize transaction fees and taxes and maximize your exposure to the momentum factor, but will also minimize your exposure to the value factor (that is, to the gains from mean reversion).
Given the relative unattractiveness of these two extremes, the real challenge is how to maximize your exposure to gains from both momentum and mean reversion, while limiting transaction costs and taxes. To answer this question, we analyzed the potential benefits from a number of different rebalancing strategies.
These strategies are based on an investor's decisions on a number of sub-issues. The first is whether rebalancing should be triggered by the calendar (e.g., do it every December), or by the extent to which actual portfolio weights have diverged from their targets. In the case of the former, the second issue is how often one should rebalance -- e.g., every year, every other year, or some other period. Where rebalancing is determined by portfolio weights, the second issue is the level at which one should set the trigger -- e.g., when one or more asset class is five percent above or below its target weight, twenty percent away form it, or somewhere in-between. The third issue is what one should do when one rebalances -- should you simply return your portfolio to its target weights, or should you try to take advantage of mean reversion by rebalancing overweight asset classes to slightly below their target weight, while rebalancing underweight asset classes to slightly above their target weights?
We tested all of these questions in our analysis, as well as the impact of transaction costs. We evaluated time based rebalancing using one, two, and three year frequencies. We evaluated portfolio weight strategies that were triggered when the actual weight for one or more asset classes was either 5%, 10%, or 20% above or below their target weights We also looked at the impact of rebalancing the two most "misaligned" asset classes back to 2.5% and 5.0% above or below their respective target weights (above target weights for the worst performing asset class, and an equivalent amount below the target weight for the best performing asset class). Finally, we varied the transaction cost involved in executing the rebalancing, from a low of .35% of the value of the transactions, to a high of 1.35% (to reflect both trading commissions and bid/ask spreads).
We did not, in our initial analysis, include the cost of capital gains taxes potentially triggered by the sale of above target weight assets held in taxable accounts. In a subsequent analysis, we took them into account by raising the effective transaction costs involved. Doing this did not significantly affect our conclusions. Moreover, in many cases capital gains triggered by rebalancing can be offset by other transactions. For example, assume you have to realize $100 in capital gains due to your rebalancing driven sale of a European equity index fund. Further assume that you will be redeploying these funds into U.S. equities, where the market value of your current position (an index fund that tracks the Wilshire 5000 index) is $100 below your target weight. One way to offset your capital gain on the sale of the European index fund shares could be to sell your $100 holding of the Wilshire 5000 index fund (which would generate a capital loss), and then invest $200 into another U.S. equity index fund that tracks a different broad based U.S. equity index, such as the Dow Jones Total Market Index or the Russell 3000 index. A number of writers have noted that the change in the underlying index that is tracked appears to exempt this type of transaction from the "wash sale" rule, and therefore make it an attractive tax management tool. However, given that few regulations seem to remain constant in the world of taxes, one should always confirm with a tax advisor that this interpretation still applies before executing this type of transaction.
We ran our rebalancing analysis on our U.S. dollar 3%, 5%, and 7% target real return portfolios, and their respective asset allocation weights. Our criterion for evaluating the impact of different rebalancing strategies was the extent to which they increased the probability of achieving the portfolio's target rate of return over a twenty year holding period. Our analysis was based on a simulation optimization approach. We first started with a candidate rebalancing strategy (e.g., time or trigger based, time period or trigger level, and adjustment factor, if any), and then simulated how well it performed over 10,000 simulations of different annual returns each year for each asset class and different costs for rebalancing transactions. We used two different distributions for future asset class returns. The first (which we used in 67% of the simulations) was based on historical asset class returns. The second (which we used in 33% of the simulations) was based on our estimated future asset class returns, as described in our recent asset allocation review. To be consistent with our previous work, the 67%/33% split was the same one we used when formulating our recommended model portfolios. The 10,000 simulations produced a distribution of expected outcomes for the candidate rebalancing strategy. We then switched to a new rebalancing strategy, and repeated the process. After we had evaluated all the possible rebalancing strategies, we chose the one which maximized the probability of achieving the target real compound annual rate of return over the twenty year holding period. To put it slightly differently, our goal was to identify the most "robust" rebalancing strategy -- that is, the one which would maximize the probability of achieving the target rate of portfolio return under a wide range of possible future asset class return and transaction cost scenarios.
Our results were very interesting. Let's start with the 7% target real return portfolio. Our base case was rebalancing every year back to the portfolio's target weights. We found that the optimum rebalancing strategy for this portfolio (that is, the one which maximized the probability of achieving the real target rate of return over twenty years) was to set the trigger at 20%, and, when one or more asset classes reached this point (i.e., were either 20% above or below their target levels), to rebalance the asset class furthest above its target to 2.5% below it, the asset class furthest below its target to 2.5% above it, and all the others to their respective target weights. Compared to the base case, this rebalancing strategy raised the probability of achieving the target real portfolio return by just under three percent.
Our analysis of the 5% target real return portfolio found that the same rebalancing strategy produced the best results -- again increasing the probability of achieving the target real portfolio return by just under three percent. As was the case with the 7% target return portfolio, we observed that, due to the impact of diversification, the simulated asset class weights in the portfolio rarely reached the 20% trigger point over 10,000 simulations. This minimized transaction costs, while also allowing the maximum gains from momentum effects. Moreover, given the relative rarity with which the 20% over or under rebalancing trigger level was reached in these simulations, the use of the 2.5% adjustment factor usually enhanced the portfolio benefit when asset class returns reverted toward their long term means.
The 3% target real return portfolio proved to be a different case. Here the incremental benefits of different rebalancing strategies (compared to the baseline strategy of rebalancing every asset class every year back to its target weight) were very slight -- delivering at most, a .50% (that is, one half of one percent) improvement in the probability of achieving the target real rate of return over twenty years. The reason for this was the fact that most of the 3% target real return portfolio is allocated to asset classes with relatively low standard deviations of returns. In other words, if the returns on the asset classes in which you've invested aren't likely to vary much from year to year, the additional benefit from not using the base case rebalancing strategy is likely to be minimal, and probably not worth the time involved.
Finally, we note that these rebalancing strategies generally agree with another important finding from asset allocation research. As we have frequently noted, because of unavoidable estimation errors in their assumptions about future asset class returns, risks, and correlations, the results produced by all asset allocation models are themselves only statistical estimates of the "true" optimal weights for different asset classes in a portfolio. This means that portfolios whose asset class weights differ by five to ten percent from each other could, statistically speaking, be equally optimal. On the other hand, the transaction costs and taxes incurred by rebalancing are very real. In the past, this led us to reach the qualitative conclusion that frequent rebalancing was, at best, of dubious value (the exception being those very rare situations when an asset class is obviously and substantially either over or undervalued). We now have a quantitative analysis which supports this view.
But there is more to the rebalancing story than statistics. We also have to recognize that psychologically, it can be a very difficult strategy to apply in practice.
Rebalancing asks you to sell assets that have performed relatively well in the past (and about whose performance you may have bragged a bit), and to buy those that have done just the opposite. Not surprisingly, being human, many people find this very difficult to do. Consider this example. Last year, after much careful research and consideration, your friend Susan decided on the asset allocation that was right for her, and implemented it through a mix of mutual and exchange traded funds. However, when reviewing her portfolio's performance twelve months later, she saw that while her U.S. equity and U.S. bond asset class allocations had increased in value, her international equity and bond allocations had suffered a loss. What do you think she feels, and how do you think she will react?
Psychologists have identified the concept of "prospect theory", which helps to answer these questions. Prospect theory basically has three elements. The first is the observation that for many people, the emotional pain of a loss is about twice as severe as the enjoyment one experiences after an equivalent gain. The second is that losses and gains are usually determined relative to some reference point. And the third is that following losses, people become risk seekers, willing to take chances to relieve the painful feelings they are experiencing, while following gains people become risk averse, and less willing to take chances that might result in the pain and disappointment associated with losses.
Now let's apply this to our friend Susan. Prospect theory suggests that a key issue is the nature of her reference point. Unfortunately, there is no obvious answer to this question. Let's look at two alternatives. The first reference point is the best investment performance achieved by Susan's friends over the past year (assuming they're telling the truth about their portfolios' performance, which isn't always the case!). As consumer research has shown, people increasingly don't want to be "just average." The disappearance of traditional "mass market" products and stores, and their replacement by "mass luxury" items and outlets is testimony to the power of this trend. Given this, it is easy to see how Susan could experience feelings of loss if the performance of her portfolio appeared to lag behind that of one or more of one's peers.
In this situation, Susan's feelings of loss could also be compounded by something called "hindsight bias", which is a normal human tendency to believe that the events that occurred in the past were more likely than our previous foresight originally had estimated them to be. Consider a the findings from a typical experiment in this area. On January 1st, you may estimate that there is a 50% probability that U.S. bonds will be the best performing asset class this year, a 25% chance that it will be U.S. stocks, a 20% chance that it will be foreign stocks, and a 5% chance that it will be foreign bonds. Assume that by year end, it turns out that foreign stocks have delivered the best performance. If you are then asked to recreate from memory your January probability estimates, it is almost certain that the probability you assign to foreign stocks will be higher than your original 20% estimate. This is the essence of hindsight bias. In practice, hindsight bias probably adds to Susan's feelings of frustration at not having matched her best performing peer's investment results, and tempts her to adopt a riskier asset allocation in the hope of outperforming her peers the following year.
But what if Susan's reference point isn't the performance of her friends' portfolios? What if it is simply whether or not at year end her investments are worth more or less than they were twelve months earlier? Frankly, there isn't much difference at all. Prospect theory suggests that if Susan has lost money, she'll be tempted to change her asset allocation toward a riskier stance, while if she has made money, she will be tempted to adopt a more conservative one.
The moral of the story is this: the whole purpose of diversifying your portfolio across different asset classes is to reduce its overall risk, and to maximize your long term returns by avoiding big short term losses. A disciplined approach to rebalancing is a key part of your diversification strategy. In practice, diversification means that from year to year, different asset classes will have positive and negative returns. Susan needs to keep in mind that the benefits of her portfolio diversification strategy will only become apparent (and get bigger, to boot) over time. Given this, the biggest challenge for Susan is to avoid overreacting when one of the asset classes in her portfolio, or even the portfolio itself has a bad year. As we have seen, human nature makes this level of self-control a very difficult challenge to master. And the more frequently we review our portfolio's performance, the greater the psychological pressure we put on ourselves to deviate from our rebalancing strategy. Besides resisting the urge to evaluate your performance too often, it also helps to be explicit about the reference point you are using when you do evaluate it, and to choose one that is long-term and internally focused (e.g., am I on track to achieve my long term goals), and not short term and externally focused (e.g., did I outperform my Uncle Carl last year?).
Finally, we should cover one special rebalancing situation that is occasionally encountered: how to implement a major change in your portfolio's asset allocation. In many cases, the investor considering such a change has substantial amounts of unrealized gains and losses that could trigger significant tax consequences were they to substantially change either their asset allocation and/or their holdings within different asset classes (e.g. switching from actively managed to index funds). Under these circumstances, professional tax advice is a critical input when deciding whether, when, and how much change to undertake. To be sure, one should keep in mind some general guidelines on asset location and tax efficiency. And one should also keep in mind the trade-offs between index mutual funds and exchange traded index funds (ETFs). For example, the former are usually much better for people who are gradually changing their position, because the commission costs charged on ETF purchases are avoided. On the other hand, index mutual funds offer less control over the realization of future taxable capital gains distributions than do ETFs. However, to reiterate the main point that applies in these situations, because of the complexity of many investors' tax situations, these general principles must be integrated with (and occasionally traded off against) the specifics of an individual's tax situation.
So, to sum up, it is important to keep these points in mind: