Diversification: A Look at Risk Factors

Garry Hawker examines some of the latest thinking around risk factors, 'return drivers' and whether they are a more effective means of diversification versus an asset class approach.

The principles of diversification are well understood. At the asset class level, the concept of diversifying away from a dependence on the equity risk premium for portfolio returns is accepted. This has largely been achieved through increasing the allocations to alternative assets or strategies.

However, it is generally accepted that diversification did not work as well as expected during the global financial crisis. The correlations between most risky assets increased significantly during the crisis. Institutional investors found out that, while their portfolios may have been well diversified for the good times, this was not the case in the bad times.

One reason for the failure of diversification during the crisis can be attributed to the fact that many of the additional asset classes introduced had similar underlying return drivers to the traditional asset classes. That is, some alternative vehicles simply repackage certain systematic return drivers (including the equity risk premium), often in much more expensive forms, including through leverage. In addition, while bond portfolios themselves have become more diverse over time as additional types of securities have been added to increase yields, bonds provided much lower diversification to the overall portfolio than they were probably intended to. The crisis highlighted just how closely linked the credit risk premium is to the strength of the economy and therefore to the equity risk premium.

Therefore, a potentially more robust approach worthy of consideration is to better understand the exposure of a portfolio to the underlying return drivers and to aim to diversify between the return drivers as opposed to simply diversifying between asset classes. This has been the subject of three recent articles. The purpose of this paper is to outline these recent developments and discuss the merits of this approach for portfolio construction.

In the first two of these papers by Ang et al and Bender et al, the authors examine approaches to portfolio construction that are based on allocations to various return drivers or risk premia that can be captured systematically via passive management. Bender et al include a comparison of how a conventional 60/40 equity/bonds portfolio would have performed historically against an equally-weighted portfolio of various return drivers (or risk premia). A third paper by Inker, as the title suggests, focuses on the additional risks that some of these types of portfolios might introduce.

This thinking is also being reflected in the way certain national funds express their asset allocation. Two large national funds have changed the manner in which they express their asset allocation from asset classes to risk classes or to recognize the underlying characteristics. For example, the ATP Fund in Denmark expresses it asset allocation in terms of exposure to what it calls five risk classes – interest rates, credit, equities, inflation and commodities. The Alaska Permanent Fund’s asset allocation is expressed in terms of allocations to companies (either debt or equity), interest rates, real assets, special opportunities and cash. We understand that other funds are currently considering re-expressing their asset allocations along similar lines.

Is this focus on risk factors a passing fad or does this approach have merit?

Mercer has been examining the influence of risk factors for some time, initially in relation to alternative asset classes. Our initial work was focused on an examination of the extent to which the returns of five alternative asset classes (private equity, commodities, real estate, infrastructure and hedge funds) can be explained by various risk factors – for example, equity risk premium, small cap premium, unexpected inflation, term premium and credit risk premium. Our subsequent analysis has extended the number of alternatives and the number of risk factors, while also applying the framework to traditional asset classes.

Mercer considers the approach of considering risk factors in the strategy setting process does have merit. In particular, it may assist institutional investors to better appreciate the risks and return drivers inherent in different asset classes and how these can be expected to behave in various environments. It also overcomes shortfalls that often arise in traditional risk management, where volatility of asset class returns is used as the sole risk metric.

Some risk factors can be quantified, in which case we can categorize them as return drivers. We can establish assumptions for the expected return associated with each return driver. However, some risks cannot be quantified but are still important to consider in order to form as complete a picture as possible of the risks associated with any investment strategy. For example, we can regard the equity risk premium or the credit risk premium as being return drivers since they can be quantified, while political or regulatory risk cannot be assessed in a quantitative framework but an investor needs to take these risks into account qualitatively in the sizing of their allocations.

It is also possible to estimate volatilities for the return drivers and to establish a correlation matrix from historical data. Having established volatility and correlation assumptions, an ‘optimal’ exposure to the preferred return drivers could be determined. However, this approach may be simply recasting the problem – that is, the correlations between the return drivers will not be constant over time in the same way that the correlations between asset classes are not. The same applies for volatilities. This approach also does not readily take into account the risks that cannot easily be quantified.

As such, we consider the primary benefit of a focus on risk factors is to assist investors in understanding the key risk exposures inherent in a candidate strategic asset allocation (“SAA”) in the strategy setting process. Rather than changing the strategy setting process from one focused on asset classes to one focused on return drivers or risk premia, we consider that the process can be enhanced by checking candidate asset allocations for their exposure to the underlying return drivers. To the extent an investor wanted more or less exposure to a certain return drivers (say illiquidity), then the SAA can be adjusted appropriately. Alternatively, if the investor is particularly concerned about particular risks, the framework assists in identifying which asset classes should be increased.

The following graphic summarizes this approach where we seek to qualitatively demonstrate the extent of the exposure that various asset classes or strategies have to return drivers or risk factors.

(See graphic here.)

For a fund with a conventional 60/40 allocation to equities and bonds, roughly 90% of portfolio volatility will historically have been attributable to equity risk. Diversification in a way that increases the exposure to other return drivers will reduce this percentage. However, it is still likely to be by far the dominant risk (or source of volatility) in the overall portfolio if a similar return to that of the 60/40 allocation is targeted, unless leverage were to be used to increase the effective exposure to other return drivers, which we expect would be unacceptable to many investors.

Volatility and risk are not the same thing and leverage changes the underlying risk characteristics of a portfolio. As discussed above, the credit risk premium and the equity risk premium are highly correlated and this correlation can be expected to increase in times of stress especially if leverage is being applied to increase the riskiness of the credit portfolio. While the term premium has historically had a negative correlation to the equity risk premium, sovereign bond yields are low and may well increase from current levels such that the term premium could reasonably be expected to be zero (or negative) over the next few years, thus potentially not being a risk factor that in the medium term would be desirable to leverage.

The framework we have developed can be extended to cover these types of “risk parity” asset allocation approaches, as they are being termed in the industry.

To conclude, we view a focus on risk factors and their associated return drivers to be an important evolution in the investment strategy setting process, with its primary benefit being to focus greater attention on how portfolios are structured for different investment environments. This has led some funds to change the way they express their strategy from asset classes to “risk classes”, while others are seeking to introduce leverage at the strategy level in an endeavour to achieve a better spread of exposures to various risk factors without compromising on expected returns. Many possible approaches exist, but some will potentially fall into the trap of equating risk to volatility such that insufficient attention is given to the risks that may not be quantifiable.

Garry Hawker
Director of National Funds Consulting
Mercer