MODELING ASSET CLASS RISK
Learning Outcomes
explain how asset classes are used to represent exposures to systematic risk and discuss criteria for asset class specification
explain the use of risk factors in asset allocation and their relation to traditional asset classābased approaches
Asset classes are one of the most widely used investment concepts but are often interpreted in distinct ways. Greer (1997) defines an asset class as āa set of assets that bear some fundamental economic similarities to each other, and that have characteristics that make them distinct from other assets that are not part of that class.ā He specifies three āsuper classesā of assets:
Capital assets. An ongoing source of something of value (such as interest or dividends); capital assets can be valued by net present value.
Consumable/transformable assets. Assets, such as commodities, that can be consumed or transformed, as part of the production process, into something else of economic value, but which do not yield an ongoing stream of value.
Store of value assets. Neither income generating nor valuable as a consumable or an economic input; examples include currencies and art, whose economic value is realized through sale or exchange.
EXAMPLE 4
Asset Classes (1)
Classify the following investments based on Greerās (1997) framework, or explain how they do not fit in the framework:
Precious metals
Precious metals are a store of value asset except in certain industrial applications (e.g., palladium and platinum in the manufacture of catalytic converters).
Petroleum
Petroleum is a consumable/transformable asset; it can be consumed to generate power or provide fuel for transport.
Hedge funds
Hedge funds do not fit into Greerās (1997) super class framework; a hedge fund strategy invests in underlying asset classes.
Timberland
Timberland is a capital asset or consumable/transformable asset. It is a capital asset in the sense that timber can be harvested and replanted cyclically to generate a stream of cash flows; it is a consumable asset in that timber can be used to produce building materials/ packaging or paper.
Inflation-linked fixed-income securities
Inflation-linked fixed-income securities is a capital asset because cash flows can be determined based on the characteristics of the security.
Volatility
Volatility does not fit; it is a measurable investment characteristic. Because equity volatility is the underlying for various derivative contracts and an investable risk premium may be associated with it, it is mentioned by some as an asset.
Greer (1997) approaches the classification of asset classes in an abstract or generic sense. The next question is how to specify asset classes to support the purposes of strategic asset allocation.12 For example, if a manager lumps together very different investments, such as distressed credit and Treasury securities, into an asset class called āfixed income,ā asset allocation becomes less effective in diversifying and controlling risk. Furthermore, the investor needs a logical framework for distinguishing an asset class from an investment strategy. The following are five criteria that will help in effectively specifying asset classes for the purpose of asset allocation:13
Assets within an asset class should be relatively homogeneous. Assets within an asset class should have similar attributes. In the example just given, defining equities to include both real estate and common stock would result in a non-homogeneous asset class.
Asset classes should be mutually exclusive. Overlapping asset classes will reduce the effectiveness of strategic asset allocation in controlling risk and could introduce problems in developing asset class return expectations. For example, if one asset class for a US investor is domestic common equities, then world equities ex-US is more appropriate as another asset class rather than global equities, which include US equities.
Asset classes should be diversifying. For risk control purposes, an included asset class should not have extremely high expected correlations with other asset classes or with a linear combination of other asset classes. Otherwise, the included asset class will be effectively redundant in a portfolio because it will duplicate risk exposures already present. In general, a pairwise correlation above 0.95 is undesirable (given a sufficient number of observations to have confidence in the correlation estimate).
The asset classes as a group should make up a preponderance of world investable wealth. From the perspective of portfolio theory, selecting an asset allocation from a group of asset classes satisfying this criterion should tend to increase expected return for a given level of risk. Furthermore, the inclusion of more markets expands the opportunities for applying active investment strategies, assuming the decision to invest actively has been made. However, such factors as regulatory restrictions on investments and government-imposed limitations on investment by foreigners may limit the asset classes an investor can invest in.
Asset classes selected for investment should have the capacity to absorb a meaningful proportion of an investorās portfolio. Liquidity and transaction costs are both significant considerations. If liquidity and expected transaction costs for an investment of a size meaningful for an investor are unfavorable, an asset class may not be practically suitable for investment.
Note that Criteria 1 through 3 strictly focus on assets themselves, while Criterion 5, and to some extent Criterion 4, involve potential investor-specific considerations.
Asset Classes Should Be Diversifying
Pairwise asset class correlations are often useful information and are readily obtained. However, in evaluating an investmentās value as a diversifier at the portfolio level, it is important to consider an asset in relation to all other assets as a group rather than in a one-by-one (pairwise) fashion. It is possible to reach limited or incorrect conclusions by solely considering pairwise correlations. To give an example, denote the returns to three assets by X, Y, and Z, respectively. Suppose that Z = aX + bY; a and b are constants, not both equal to zero. Asset Z is an exact weighted combination of X and Y and so has no value as a diversifier added to a portfolio consisting of assets X and Y. Yet, if the correlation between X and Y is ā0.5, it can be shown that Z has a correlation of just 0.5 with X as well as with Y.
Examining return seriesā correlations during times of financial market stress can provide practically valuable insight into potential diversification benefits beyond typical correlations that average all market conditions.
In current professional practice, the listing of asset classes often includes the following:
Global public equityācomposed of developed, emerging, and sometimes frontier markets and large-, mid-, and small-cap asset classes; sometimes treated as several sub-asset classes (e.g., domestic and non-domestic).
Global private equityāincludes venture capital, growth capital, and leveraged buyouts (investment in special situations and distressed securities often occurs within private equity structures too).
Global fixed incomeācomposed of developed and emerging market debt and further divided into sovereign, investment-grade, and high-yield sub-asset classes, and sometimes inflation-linked bonds (unless included in real assets; see the following bullet). Cash and short-duration securities can be included here.
Real assetsāincludes assets that provide sensitivity to inflation, such as private real estate equity, private infrastructure, and commodities. Sometimes, global inflation-linked bonds are included as a real asset rather than fixed income because of their sensitivity to inflation.
Emerging Market Equities and Fixed Income
Investment practice distinguishes between developed and emerging market equities and fixed income within global equities. The distinction is based on practical differences in investment characteristics, which can be related to typical market differences including the following:
diversification potential, which is related to the degree to which investment factors driving market returns in developed and emerging markets are not identical (a topic known as āmarket integrationā);
perceived level of informational efficiency; and
corporate governance, regulation, taxation, and currency convertibility.
As of mid-2016, emerging markets represent approximately 10% of world equity value based on MSCI indices.14 In fixed income, investment opportunities have expanded as governments and corporations domiciled in emerging markets have increasingly issued debt in their own currency. Markets in local currency inflation-indexed emerging market sovereign debt have become more common.15
āAsset classesā are, by definition, groupings of assets. Investment vehicles, such as hedge funds, that apply strategies to asset classes and/or individual investments with the objective of earning a return to investment skill or providing attractive risk characteristics may be treated as a category called āstrategiesā or ādiversifying strategies.ā When that is the case, this category is assigned a percentage allocation of assets, similar to a true asset class. Economically, asset classes contrast with āstrategiesā by offering, in general, an inherent, non-skill-based ex ante expected return premium.16
Effective portfolio optimization and construction may be hindered by excessive asset class granularity. Consider Exhibit 6.
Exhibit 6:
Examples of Asset Classes and Sub-Asset Classes
As more and more sub-asset classes are defined, they become less distinctive. In particular, the sources of risk for more broadly defined asset classes are generally better distinguished than those for narrowly defined subgroups. For example, the overlap in the sources of risk of US large-cap equity and US small-cap equity would be greater than the overlap between US and non-US equity. Using broadly defined asset classes with fewer risk source overlaps in optimization is consistent with achieving a diversified portfolio. Additionally, historical data for broadly defined asset classes may be more readily available or more reliable. The question of how much to allocate to equity versus fixed income versus other assets is far more important in strategic asset allocation than precisely how much to allocate to the various sub-classes of equity and fixed income. However, when the investor moves from the strategic asset allocation phase to policy implementation, sub-asset class choices become relevant.
EXAMPLE 5
Asset Classes (2)
Discuss a specification of asset classes that distinguishes between ādomestic intermediate-duration fixed incomeā and ādomestic long-duration fixed income.ā Contrast potential relevance in asset-only and liability-relative contexts.
Solution:
These two groups share key risk factors, such as interest rate and credit risk. For achieving diversification in asset riskāfor example, in an asset-only contextāasset allocation using domestic fixed income, which includes intermediate and long duration, should be effective and simple. Subsequently, allocation within domestic fixed income could address other considerations, such as interest rate views. When investing in relation to liabilities, distinctions by duration could be of first-order importance and the specification could be relevant.
Any asset allocation, by whatever means arrived at, is expressed ultimately in terms of money allocations to assets. Traditionallyāand still in common practiceāasset allocation uses asset classes as the unit of analysis. Thus, meanāvariance optimization based on four asset classes (e.g., global public equity, global private equity, global fixed income, and real assets) would be based on expected return, return volatility, and return correlation estimates for these asset classes. (The development of such capital market assumptions is the subject of another reading.) Factor-based approaches, discussed in more detail later, do not use asset classes as the basis for portfolio construction. Technically, the set of achievable investment outcomes cannot be enlarged simply by developing an asset allocation by a different means (for instance, using asset classes as the unit of analysis), all else being equal, such as constraints against short selling (non-negativity constraints).17 Put another way, adopting a factor-based asset allocation approach does not, by default, lead to superior investment outcomes.
There are allocation methods that focus on assigning investments to the investorās desired exposures to specified risk factors. These methods are premised on the observation that asset classes often exhibit some overlaps in sources of risk, as illustrated in Exhibit 7.18
Exhibit 7:
Common Factor Exposures across Asset Classes
The overlaps seen in Exhibit 7 help explain the correlation of equity and credit assets. Modeling using asset classes as the unit of analysis tends to obscure the portfolioās sensitivity to overlapping risk factors, such as inflation risk in this example. As a result, controlling risk exposures may be problematic. Multifactor risk models, which have a history of use in individual asset selection, have been brought to bear on the issue of controlling systematic risk exposures in asset allocation.
In broad terms, when using factors as the units of analysis, we begin with specifying risk factors and the desired exposure to each factor. Asset classes can be described with respect to their sensitivities to each of the factors. Factors, however, are not directly investable. On that basis, asset class portfolios that isolate exposure to the risk factor are constructed; these factor portfolios involve both long and short positions. A choice of risk exposures in factor space can be mapped back to asset class space for implementation. Uses of multifactor risk models in asset allocation have been labeled āfactor-based asset allocationā in contrast to āasset class-based asset allocation,ā which uses asset classes directly as the unit of analysis.
Factor Representation
Although risk factors can be thought of as the basic building blocks of investments, most are not directly investable. In this context, risk factors are associated with expected return premiums. Long and short positions in assets (spread positions) may be needed to isolate the respective risks and associated expected return premiums. Other risk factors may be accessed through derivatives. The following are a few examples of how risk factor exposures can be achieved.
Inflation. Going long nominal Treasuries and short inflation-linked bonds isolates the inflation component.
Real interest rates. Inflation-linked bonds provide a proxy for real interest rates.
US volatility. VIX (Chicago Board Options Exchange Volatility Index) futures provide a proxy for implied volatility.
Credit spread. Going long high-quality credit and short Treasuries/government bonds isolates credit exposure.
Duration. Going long 10+ year Treasuries and short 1ā3 year Treasuries isolates the duration exposure being targeted.
Factor Models in Asset Allocation
The interest in using factors for asset allocation stems from a number of considerations, including the following:
The desire to shape the asset allocation based on goals and objectives that cannot be expressed by asset classes (such as matching liability characteristics in a liability-relative approach).
An intense focus on portfolio risk in all of its various dimensions, helped along by availability of commercial factor-based risk measurement and management tools.
The acknowledgment that many highly correlated so-called asset classes are better defined as parts of the same high-level asset class. For example, domestic and foreign equity may be better seen as sub-classes of global public equity.
The realization that equity risk can be the dominant risk exposure even in a seemingly well-diversified portfolio.
Last updated
