Risk vs. structure based asset allocations

2008 reminded us that diversification can fail spectacularly when risky assets are clumped together. This is because when panic sets in, there is a sort of hoarding activity in the markets as investors jump ship from risky to less risky assets. By doing so, the correlation between risky assets spike, leading to simultaneous losses from investments of different asset classes. This is why, for example, at the peak of the crisis, high yield and emerging bonds and commodities followed the exact same path as equities at the very moment when the benefits of diversification were most needed. What use to be diversified suddenly becomes concentrated.

How can we avoid or at the least minimize such serious pitfalls? One method would be to implement a risk based asset allocation, grouping together investments according to their risk characteristics.

The traditional approach to asset allocation has been to classify assets according to their structure. Genotype/phenotype are good biological analogies to help illustrate the differences. Genotype describe hereditary characteristics (structure based classification) whilst phenotype are the observed properties (risk based classification). As an example, grouping together fixed income investments comprised of bonds with characteristic cash flow streams that include coupons and a final payment at maturity would belong to a traditional structural classification. The problem with the traditional classification is that the investment landscape has changed substantially over the last decade. The risk spectrum of investments with a similar underlying structure has broadened substantially. Today, for example, bonds can include low risk, short duration treasuries as well as high yield/emerging debt that share risk characteristics that are comparable to that of equities.

There are effectively two ways to tackle this problem:

• Managing the risk internally, within the asset class by implementing strict guidelines
• Defining risk at the asset class level

The first method would involve limiting the choice of underlying investments for more effective risk control. One such approach would be to implement a core/satellite approach in which the "larger" core would comprise of less risky investments in contrast to the satellites.

The second method is best explained as a target based approach in which the asset classes are defined as targeting specific objectives. We could very well imagine, for example, a portfolio comprised of four "asset classes" with the following defined objectives:

1. Longer term high returns (equity like risk/reward characteristics)
2. Hard assets (namely for inflation protection)
3. Liabilities matching (high quality low duration fixed income)
4. Pure excess return (de-correlated to the rest of the markets)

This method has the advantage of providing greater transparency and better risk control, at the same time allowing for a greater degree of specificity in matching investor objectives and constraints. A person close to retirement, for example, would typically allocate a large portion of their wealth to liability matching instruments in contrast to someone at the prime of their career, in which case the attributions would probably be more balanced between the four classes.

The two proposed methods aim for the same objectives, the big difference is in the approach.