When an average investor thinks of Asset Allocation he or she will generally think of some percentage of stocks, bonds, and cash. Some of the ostensibly advanced investors will probably include other asset classes like real estate, precious metals, TIPS (Treasury Inflation Protected Securities), and possibly some other, esoteric asset classes that may be largely unavailable to the investing public. The idea being that the larger number of uncorrelated asset classes one can add to an asset portfolio the greater the diversification benefit one achieves. There are mountains of data supporting the benefits of diversification and it is largely accepted that proper asset allocation is more important than choosing the right assets within the asset class (i.e., deciding whether the portfolio will have 25% or 75% equities is more critical than picking Microsoft over Proctor & Gamble). The author agrees that asset allocation is critical to the construction of an efficient portfolio.
While the purpose of this article is not to go into the gruesome details of asset allocation, it is important to touch on some of the more recent developments in the capital markets that have cast some doubt on the effectiveness of the standard mean variance approach to asset allocation. In our standard mean variance model used to produce an optimal asset allocation, it is assumed that correlations between asset classes are static. Correlation, for the non-statiticians out there, essentially measures whether Asset A zigs when Asset B zags. If the two assets move exactly in the same direction over time (i.e., when Asset A increases by 1% then Asset B increases by 1%) then they would have a perfect positive correlation, which would be reflected in a correlation of "1". On the other hand, if Asset A goes down by 1% when Asset B goes up by 1% then the two assets have a perfect negative correlation, which is reflected by a correlation of "-1". Anything in between shows up between -1 and 1 with 0 representing a situation in which the movement of Asset A provides no information as to the movement of Asset B.
Earlier we stated that the correlations between assets are static in our asset allocation model. Generally the correlations assumed in the model are based on historical correlations over a given time period. For example, if you were to examine the annual returns of large cap equities (represented by the Standard and Poor's 500) and gold over the past ten years, the data would show a mildly negative correlation between the two assets. Given this result, it would assumed that this result would hold true over the investing time horizon and the asset allocation model would base its results off of this assumption. One of the problems with this is that in times of severe market stress it appears that many of these historical relationships between assets breakdown. In many of the trading sessions of the current bear market, days with sharp equity sell-offs saw simultaneous drops in gold prices. This clearly causes problems for an investor that based his or her asset allocation on the assumption that historical assumptions would continue over the investing horizon.
What has caused this break with historical data and what is the proper way to adjust for it? On theory to explain this market behavior is that there is a "Margin Effect" taking place. The idea behind a Margin Effect is that in times of market stress many of the participants, especially highly leveraged hedge funds, are going to receive margin calls on days of sharp equity sell-offs. One could think of margin connecting different assets classes in the same way an aquifer might connect ground level lakes that are ostensibly independent of one another. Given the high levels of leverage utilized by the hedge funds which are becoming a larger and larger driver in the market, this Margin Effect may explain some of this increase in correlation that has shown up in recently.
Now that we have a hypothesis to explain what is causing the increase in correlations in times of market stress how does it affect our asset allocations? One possibile solution is the use of stress tests to see how the portfolio will perform when correlations between assets are increased. Clearly, an increase in the correlations to mimic these periods of stress will increase the modeled risk. The model could be used to reoptimize the portfolio assuming the higher correlations and then one could take a weighted average of the original portfolio (based on historical correlations) and the reoptimized portfolio (based on the stressed correlations). Keep in mind that the increases in correlations seen recently have not been limited to gold and equities. There have been days in which is seems like the only asset known to man that was up were the US Treasuries.
Now we have come to the Liability part of Asset Liability Management. While Asset Liability management is widespread on the corporate level, very few individuals ever give the liability side of the equation a thought. Here is scenario that the average investor should consider. Let's pretend that we have two investors (A and B) who are both very conservative individuals when it comes to their financial lives. Both investors have $1MM each sitting in a savings account with their bank. Both investors decide to purchase a vacation home in Texas for $500,000. We did say that they are conservative investors didn't we? They could have bought the same house on one of the coasts and paid three times as much...that is a digression for a different post. Investor A decides to lock in a fixed rate mortgage so he will always know what his payments will be. Investor B decides to take out a variable rate mortgage because his broker talked him into it.
Which investor is in a less risky position? To most investor's surprise, Investor B has made the less risky choice. How can this be though? His payments could fluctuate over time and he is old enough to remember the early eighties so he should know better. The problem with this line of thought is that it is leaving out the other side of the equation which are the assets. Savings deposits also fluctuate over time but Investor B will have partially hedged his deposits by matching his liabilities via a variable rate mortgage. When short-term interest rates are high, Investor B will pay more on his mortgage than Investor A but he will also earn more on his deposits. On the other hand, when short-term rates are low Investor B will pay less on his mortgage and receive less on his deposits. In that scenario, Investor A will still be paying the same high fixed mortgage rate but he receives the low deposit rate.
Of course this example only worked because it was assumed that the investors had $1MM lying around in a savings account. This will not be the case for the majority of investors but the point is for individual investors to consider both assets and liabilities when making allocation decisions. Otherwise, a conservative investor may find that he or she is actually taking on additional aggregate risk by making conservative allocations to liabilities without taking into consideration assets.