Perhaps the most significant transformation of the asset allocation space, as advocated by Berkelaar and practiced by a number of more sophisticated institutional investors, has been the recognition that the traditional model1 of constructing a portfolio has several flaws:

• It fails to take into account the fact that expected returns, volatility, and asset correlations2 are not necessarily constant over time.
• The optimal portfolios produced by using the traditional model are highly sensitive to expected return assumptions such that a small change in the expected return, volatility or asset correlation will often result in a very different portfolio. Assumptions are merely beliefs about how an asset will perform and can deviate significantly from reality.
• The optimal portfolios created by the traditional model use constant weights to determine asset allocations (until the next time the optimization is performed), and that this constant weighting can lead to substantial fluctuations in volatility in the short to medium term.

While these flaws are important to recognize, in my view the strength of Berkelaar’s article lies in its recommendation to:

• Construct the portfolio from risk factors rather than asset classes
• Consider how best to create investment strategies that are tuned for different economic states of the world3
• Consider how to make asset allocation more dynamic to benefit from extreme variations in valuations4

To this I would add, drawing on what I have been observing in the literature emerging from several sophisticated sovereign wealth funds and from general reading in the financial markets (notably Mandelbrot, Taleb, and Buffett):

• Consider how to leverage the institution’s size and horizon to provide tail-risk insurance to investors with shorter horizons.5
• Consider what liabilities the institution has and the timing of those liabilities when constructing the asset allocation in order to manage the risk that the institution will be forced to liquidate positions to meet those liabilities.
• Consider that investment returns are non-normally distributed per Benoit Mandelbrot (they are fractal in nature), and that many mathematical models have adopted the normal distribution out of mathematical convenience.

Conceptually, these recommendations tend to support a level of active management at the asset allocation level, with portfolio managers rebalancing the portfolio in response to changes in long-term valuation signals and risk considerations. Per Berkelaar, this can take the form of a dynamic risk-parity approach, a constant-risk approach, or a risk-trigger based approach.

Now, let us look at the following matters in greater detail, with the aim of better understanding the components of a dynamic asset allocation model: risk factors, economic states of the world, and investment horizons and tail-risk.

Risk factors

Andrew Ang, a professor at Columbia Business School, has rather aptly described the relationship between risk factors and assets as:

Factor risk is reflected in different assets just as nutrients are obtained by eating different foods. In Table 1, peas, wheat, and rice all have fiber. Similarly, certain sovereign bonds, corporate bonds, equities, and credit default swap derivatives all have exposure to credit risk. Assets are bundles of different types of factors just as foods contain different combinations of nutrients. Just as, for example, rice contains both carbohydrates and fiber, an investor holding a corporate bond is subject to interest rate risk and default risk. This is the modern theory of asset pricing: assets have returns, but these returns reflect the underlying factors behind those assets.

Constructing a portfolio by using risk factors is a more granular, nuanced approach than the traditional asset allocation model of allocating among asset classes, which may have overlapping risk factors.

For example, bond excess returns may be decomposed into (not exhaustively): a credit risk premium that compensates investors for the risk of the bond issuer defaulting on the bond payments, a reinvestment risk premium that compensates investors for the risk that the bond may be prepaid by the issuer earlier than the stated maturity date, an illiquidity risk premium that compensates investors for the risk that they may not be able to readily liquidate their position at fair market value, an inflation risk premium that reflects exposure to the risk of inflation rising, etc.

In equities there are certain premia that are generally accepted as having some consistent empirical bases, such as: a value stock premium on stocks with low prices relative to fundamentals, an illiquidity risk premium to compensate investors for the risk that they may not be able to readily liquidate their position at fair market value, etc.

The benefit of a more nuanced approach is that it focuses the investor’s mind on the risks—the possibility that “bad times” will occur and wipe out returns and/or principal as Andrew Ang defines it or the probability of a “negative state of the world” occurring as I recall it from my corporate finance class here at Wharton—that the portfolio is taking and the appropriate amount of compensation it should receive in order to take those risks.

Economic states of the world

According to Berkelaar, the majority of institutional portfolios are bets on a single economic state of the world, generally “good economic growth”, however this may be defined by the portfolio manager. These portfolios generally suffer when economic state of the world differs from that predicted by the portfolio manager. A better approach would be to explicitly devise strategies that take into account different potential economic states of the world by selecting risk factors that provide payoffs in negative economic states of the world.

Investment horizons and tail-risk

Institutional investors differ in their investment horizons. For example, a sovereign wealth fund is often—but not always—a long-term investor with limited short-term liabilities and thus has the ability to hold an asset for an extremely long period of time without any pressure to liquidate that asset.6 By contrast, a university endowment fund has a long investment horizon but is constrained by substantial and regular short-term liabilities (the aggregate amount of investment income and/or principal that must be applied towards the operating budget of the university and financial aid to students). We can build similar profiles for other institutional investors, but these two examples should suffice for now.

Therefore, a sovereign wealth fund is often in a position to be a provider of tail-risk insurance, earning a premium during “good years” (which should be more frequent than “catastrophic years”) and absorbing a certain amount of losses during “catastrophic years”. The premiums earned during years of regular or good performance outweigh the payouts during “catastrophic years”.

Moreover, investment opportunities also differ in the time horizon needed to harvest returns. As an example, a venture capital, private equity, real estate, or infrastructure investment is a long time horizon investment, requiring at least five to ten years to harvest. An institutional investor that has substantial short-term liabilities should be cautious about the mismatch between the timing of the cash inflows from such investments and the cash outflows for its liabilities.