MODEL 1: TRADITIONAL ASSET ALLOCATION MODEL

What trustees should be aware of

• Centered on a quasi-static policy portfolio (“policy beta”) controlled by trustees with modest value-add (“alpha”) within each asset class.
• Longer time horizon needed to harvest equity and liquidity risk premiums while a tight link between policy and actual portfolios imposes rigidity.
• Lack of transparency regarding underlying risk drivers and sources of skill-based value creation, especially for privately traded assets.
• Easy to implement and to outsource in separate sequential steps. Limited delegation by the board. Easy to monitor.

We qualify this model as the middle of the road, the “let's do what the neighbor does” model: it is light on beliefs, and is often implemented via external managers who always promise but hardly ever deliver outperformance. Costs can be relatively low given efficient implementation, while higher costs generally signal a governance problem. This would suggest pension funds with average qualified trustees, a very small investment management organization, and/or overly active investment consultants.

The traditional asset allocation model is one of the processes through which a pension fund aims to realize the returns targeted by the fund. Pension funds usually aim to generate a specified net return over liabilities at the lowest possible risk. Therefore, four ingredients tend to drive investment decision-making:⁶ (i) matching assets and liabilities, (ii) risk management through portfolio diversification; (iii) generating investment returns on top of the matching return, and (iv) cost management. The sequence of actions is usually implemented in a decision-making hierarchy.

The choice of strategic asset allocation is paramount here: the underlying assumption is that the choice of policy portfolio (asset allocation and resulting benchmarks) accounts for the bulk of the variability in returns,⁷ although this is not undisputed.⁸

Important in determining the strategic asset allocation is firstly the share of risky vs. low-risk assets. The board has to describe the preferences of its participants, and translate their risk aversion into a risk appetite statement; how much risk are they willing to take, and under which circumstances? The outcome of this discussion is reflected in the share of matching versus return, or bonds versus equities. Next, the translation of risk appetite into quantifiable measures forms the framework for the strategic asset allocation, which is in turn further specified in terms of investment styles, and made explicit into a number of mandates that can be handed out to internal or external investment managers. Generally, the board of trustees of a pension fund, endowment or foundation establishes the strategic asset allocation across major asset classes. Strategic asset allocation then represents the institutional investor's investment policy.⁹

MODEL 2: ENDOWMENT MODEL

What trustees should be aware of

• Centered on a simple policy portfolio allocated to a limited number of core asset classes.
• Often has an extreme focus on returns.
• Entrepreneurial, human centered and skill based.
• The quality of the governance and the investment committee seems to be a factor contributing to the success of endowments.
• Extensive delegation to and investment management organization by the board. A highly skilled investment committee is necessary to keep the investment management organization reined in.
• Core belief—liquidity comes at a price, in the shape of lower returns. This leads to a relatively heavy exposure to alternative, illiquid asset classes such as private equity.
• Superior endowment performance may be due not just to asset class allocation, but also to the selection of assets within each class, as well as deal orientation.¹⁰
• Alternative investments require a staff with considerable skill and experience, engaging in early sourcing of strategies and rigorous self-evaluation.
• Typical size is $10–$25 billion: large enough for a very highly qualified staff, small enough to be flexible and have room to maneuver in markets and deals.

The Endowment Model can be characterized roughly as an investment model where the natural advantages of a long term investor are leveraged. This model was developed in the 1980s by Yale Endowment,¹¹ with pension funds drawing inspiration from the different approaches that endowments took towards portfolio construction and implementation. Endowments are important institutions in the academic world and in investment society. In the former, they play a role in funding, thereby maintaining academic research and excellence among universities. Regarding investment society, endowment funds operate under less regulatory restrictions than pension funds, allowing them to pioneer and adopt new insights in investment management at an earlier stage than pension funds. Endowments have received much attention recently for their superior investment returns compared with other institutional investors;¹² it has also been observed that these superior investment returns are difficult for other institutional investors to duplicate. We will now briefly review a few reasons for the endowment's superior investment returns.

Similarly to the traditional asset allocation model, much of the endowment's performance is achieved through allocation decisions between asset classes that have different risk and return characteristics. Investment can be separated into five major asset classes: equities, fixed income, real estate, alternatives (which include hedge funds, commodities such as oil and timber, as well as private equity funds), and cash. While the 1970s and 1980s saw a gradual move away from fixed income securities and cash and into equities by endowments, the 1990s and 2000s in turn witnessed a shift away from equities and fixed income towards sophisticated, often illiquid alternative assets.

Secondly, the superior endowment performance may be due not solely to asset class allocation, but also to the selection of assets within each class. The skills and experience of investment managers appear to play a substantial role in the success of endowment investments, with the staff able to identify and source strategies and asset classes at an early stage. The staff of successful academic endowments have considerable experience and have often worked together for many years. Finally, the staff of successful university and college endowments have an academic orientation, which leads to a process of periodic self-evaluation. The staff of many of these funds occasionally stop to consider the processes that led them to make investments that proved particularly successful or problematic, and also engage in an active dialogue with their peers.

A third distinguishing element with regard to the superior endowment performance is investment governance. Top-performing endowments have active investment committees, generally drawn from the ranks of alumni. These bodies see their role not as micromanaging the decisions of the investment staff, but rather as setting broad policy and serving as an informed sounding board. The contrast with public pension funds is particularly stark here.

Riding on the success of Yale Endowment and Harvard Management Company, many boards have been advised by consultants to adopt elements of the endowment model and make major commitments to “alternative” investments in hedge funds of various types, private equity, real estate, venture capital, etc. The long-term record achieved at leading endowments may seem compelling. However, pension funds have at best struggled to emulate this success, and in some cases have even lagged behind compared to traditional bond/equity asset allocation benchmarks.

Mimicking the asset allocation strategies of the best endowments may not lead to the same stellar results, for the simple reason that markets are adaptive and the rewards of mimicking are steadily eroded.¹³ The alternative investment markets on which successful endowments have relied are particularly sensitive to inflows of capital. A new or niche alternative investment market often has a limited number of opportunities, so additional capital tends to result in the purchase of securities at higher prices and, therefore, ultimately lower returns. This effect is particularly relevant because the strategies of the elite endowments are being scrutinized and imitated as never before. Investing in alternative asset classes may prove to be risky, because these classes are often opaque and harder to understand. The recent struggle for performance of the endowment model can also be explained by the correlation of an endowment's portfolio with stock markets. When a financial crisis occurs, the performance of endowment funds deteriorates; for example, endowment funds were hit hard during the 2008 crisis. Finally, the scale of the fund could affect returns as well. Ninety-one percent of the funds under $1 billion in assets performed worse than a simple 60/40 holding strategy in every period since records were maintained.¹⁴ Vanguard offers an explanation for this by stating that the larger endowments have better relationships and leverage for negotiating fees.¹⁵ The endowment model is a long-term model. Institutions and investment committees that are not well prepared to maintain a long-term perspective would do well to study the endowment model carefully and adopt the model only to the extent that they have the necessary staff capabilities, financial disciplines, and internal understanding.

MODEL 3: FACTOR ALLOCATION MODEL

What trustees should be aware of

• Premised on the belief that asset returns can largely be explained by a parsimonious set of common factors, e.g. real interest rate, growth, inflation, credit spreads and volatility.
• By design, factor models provide higher transparency on systematic risks, but suffer from investability and stability issues.
• Because of the complexity, it is only suitable for boards with very high skill levels.
• The use of factors also complicates communication with stakeholders, hence its slow adoption rate—even among the converted.
• There is no fund that we know of that seriously puts this into practice across the whole asset mix except, perhaps, ATP.

Traditionally, portfolio diversification was an intuitive, sensible choice for an investor to make, and researchers have steadily added important theoretical ideas, including Mean Variance Optimization, Capital Asset Pricing Model (CAPM) and Asset Pricing Theory (APT), which have shaped our understanding of the meaning of diversification. Investment portfolios today may look quite different from portfolios of a decade ago, but they still share diversification as a central tenet. The factor allocation model takes a different approach to one of the central tenets of the traditional asset allocation model, namely diversification. It is far more important to diversify over a number of risk factors that the assets may have in common than to diversify over assets.

Institutional investors warmed to new forms of diversification in the 1990s, when advanced tools such as Asset Liability Management (ALM) and new innovative concepts in risk management allowed investors to expand the scope of possible investments. Asset categories, hitherto only available to a few savvy investors, slowly entered the portfolios but only really made headway from the 2000s onwards. Investors increasingly became aware that an overreliance on the traditional asset allocation to equities and bonds was vulnerable to high volatility and a lack of sufficient diversification, particularly during periods of crisis. After the 2001–2002 financial meltdown, there was an accelerated shift in asset allocation towards alternative asset classes such as private equity, hedge funds and commodities. These new allocations, however, provided only part of the expected diversification. Alternative asset classes have turned out to be more correlated with traditional equities and bonds than previously thought. Private equity provides diversification if it consists of high-alpha funds, but much of the return is structurally dependent on equities and bonds. Similarly, the majority of long–short equity hedge funds produce returns that are highly correlated with the equity market.¹⁶

In hindsight, many new alternative investment strategies incorporated in portfolio construction by boards that were inspired by the successes of the endowment model suffered from two problems. First, the success of many new alternative strategies was boosted by the same factors: low interest rates and robust economic growth. Some alternative investments share more factors; private equity gives investors exposure to the same kind of risk as publicly traded equity, albeit with added leverage. When prospects deteriorated, investors with liquidity constraints were forced to sell those alternative asset classes simultaneously. Secondly, some of the asset classes are rather small. For some investors, illiquidity can be attractive, given that it can offer higher returns. As more investors get involved, however, the market becomes more liquid, and the higher return is eroded. But when everyone tries to sell, illiquidity again rears its ugly head—there are no buyers to be found, and prices tumble. This is especially unpleasant when pension funds have to value their investment at mark-to-market.

Armed with this sobering knowledge, institutions began to reconsider what purpose diversification serves within the investment process and what the most important factors should be to drive portfolio returns.¹⁸ Exhibit 6.2 presents a list of what the drivers of some asset classes are. Within this framework, which is rooted in the APT developed in the late 1970s, any asset can be viewed as a bundle of factors that reflect different risks and rewards. Therefore, it makes perfect sense to focus on these factors, allowing investors to identify them and understand what really drives asset returns. These insights help to develop portfolios that realize the required risk profile not only before, but also during volatile periods.

The novel proposition is that investors should consider redesigning the investment process by directly allocating to factor premiums within the strategic asset allocation as a way forward. A traditional portfolio would first be diversified across asset classes, followed by allocation, for instance to regions and sectors and then to (external) fund managers. A factor-based portfolio is diversified across premiums, such as low-volatility, small-cap, value and momentum premiums. Investment management practice has developed methods of incorporating one or more of these premiums, either by developing investment strategies solely to capture one of the premiums, or by embedding them as tilts in the overall strategic asset allocation. The Norway Pension Fund views this as “harvesting risk premia from multiple sources [which] can be seen as broadly consistent with risk parity investing, which should result in more effective diversification and avoid a heavy reliance on the equity premium.”¹⁹

MODEL 4: OPPORTUNITY COST OR REFERENCE PORTFOLIO MODEL

What trustees should be aware of

• Reference portfolio comprising liquid assets acts as a realistic, cheap-to-replicate passive alternative.
• Investments outside of the reference portfolio (such as real estate, private equity and infrastructure) are funded from best proxies (“projections”) of reference betas, which enhance risk and return transparency after adjusting for leverage and illiquidity.
• Residual risks are controlled by the active risk budget and provide a raison d'être for investing in alternatives.
• The investment management organization has the goal of generating a certain excess return over and above the return to the reference portfolio, usually on a rolling three-to-five-year horizon.
• Depending on the size of the risk budget, quite a wide degree of freedom is delegated by the board. In the Canadian pension funds, the modus operandi will typically be quite entrepreneurial, and there will be a substantial allocation to private asset classes such as private equity, real estate and infrastructure.
• In order to control the risk effectively, excellent risk budgeting and monitoring should be in place. This should be a point of attention for the board.

The Reference Portfolio of Opportunity Cost Model was developed by Canada Pension Plan Investment Board (CPPIB) in 2006. The main elements of this governance model, with some variations reflecting specific circumstances, have been adopted by GIC Private Limited (GIC), one of the sovereign wealth funds of Singapore, the New Zealand Superannuation Fund, and a number of other funds. When developing a strategic benchmark, a board could encounter a few dilemmas. If the board were to refine the benchmark into all sorts of sub-asset classes or investment styles, it would be in control, but run the risk of “over-engineering” and losing sight of what the benchmark should do, i.e. optimally reflect the risk/return trade-off needed to achieve the goals. Moreover, no matter how detailed and complete, there would always be investments that are not included but might fit the overall profile of the fund, and the investment management is well placed to identify such opportunities. Therefore, the starting point is that the board decides on a passive investment strategy or reference portfolio that, given the contributions to the fund, can reasonably be expected to meet the liabilities of the fund.

The Opportunity Cost Model does not specify asset classes. Portfolio construction starts with the specification of a Reference Portfolio, which reflects the desired or necessary level of systematic risk required to achieve the fund's objectives in a clear and simple way. The investment manager then considers each investment in the context of its exposure to the underlying factor drivers captured in the Reference Portfolio. The board also specifies an “active risk” appetite, which is the degree to which the fund can diverge from the composition of the Reference Portfolio, through means such as limits on deviations taken from benchmark (conventional “tracking error”), rebalancing requirements, and other risk measures. The asset manager is then responsible for determining asset allocation and approving investment programs within the defined risk appetite constraints. In practice, the fund manager would likely define a Policy Portfolio to reflect its asset allocation decisions and corresponding approvals for investment programs.

A hallmark of the Opportunity Cost Model is that the fund manager is free to invest in assets that are not included in the Reference Portfolio, but all investments are benchmarked against it. The Reference Portfolio could be implemented by a small team of 10–12 people, at almost no cost. The fund manager is deemed to have added value if it covers its own costs and beats the Reference Portfolio. For example, CPPIB's Reference Portfolio consists of 85% global equities and 15% Canadian bonds,²⁰ with further breakdowns for domestic and international assets. GIC's Reference Portfolio is also a 65% equities and 35% fixed income portfolio. These portfolios are simple, highly scalable, and can be implemented very cheaply.

An underlying principle is that the risk and return attributes of individual assets can be represented by some distinct combination of the factors in the Reference Portfolio (equity and bond exposures). The nature and degree of certainty of the cash flows, as well as the capital structure of each asset determine the mix of asset exposures used to fund that asset. Investments such as real estate, private equity, high-yield bonds and infrastructure can have idiosyncratic characteristics not necessarily captured in the funding assets, which may in turn provide beneficial risk-adjusted returns for the portfolio.²¹

Perhaps the most attractive feature of the Opportunity Cost Model is the clear delineation of accountability. The asset owner makes the most important decisions concerning return objectives and the required level of systematic risk by approving the composition of the Reference Portfolio and specifying other risk appetite parameters. The board ensures that the manager has the requisite capabilities in place before commencing investment programs, monitors the execution and results of those programs, and approves appropriate internal benchmarks. The fund manager then has the latitude to make investment decisions and shape the actual composition of the portfolio subject to the active risk limit determined by the asset owner. It can be argued that this places accountability with the party best able and positioned to make informed decisions. It is the fund manager's responsibility to justify costs of active management and to outperform the Reference Portfolio. Thus, the opportunity cost of the Reference Portfolio can be used for all alternative assets as their benchmarks, rather than having to specify separate benchmarks for each alternative asset class. This also provides more flexibility for the fund manager to take into account the time-varying nature of risk premiums by not having fixed allocations to asset classes. The manager has appropriate incentives to make the investments with the best marginal contribution to risk and return for the overall portfolio, rather than the best available investment within each asset class.

The Opportunity Cost Model is only suitable for long-horizon investors. Inevitably there will be performance differences in any short time period between a real estate asset, for instance, and the mix of equity and bonds used to fund it. Indeed, in the case of real estate for example it could take as long as 7–10 years for the Total Portfolio investment thesis to be realized.

A final governance challenge is that the board—or similar governing body of the investment manager—needs to have the skills to oversee, evaluate, and monitor the systems created by the investment manager in order to implement the Total Portfolio Approach. The board must also be conversant with the construction of—and the concepts behind—the Reference Portfolio.

MODEL 5: DUAL STRATEGIC/OPERATIONAL BENCHMARK MODEL

What trustees should be aware of

• A generic, strategic benchmark is provided by the asset owner, the fund. This strategic benchmark captures the overall risk appetite, adequately reflecting the risks in the pension agreement.
• A policy benchmark is then made more specific by the investment management organization in order to increase diversification and to benefit from a number of systematic sources of return such as value and small cap. The difference between the strategic and policy benchmark reflects the value of the strategic advice from the investment management organization.
• An operational benchmark is then made, refining the policy benchmark into investable mandates. The difference between the policy and the operational benchmark reflects the value of the selection and monitoring of the mandates, including the value of internally managed mandates.
• The ownership of the strategic benchmark lies with the board. Policy and operational benchmarks should also be approved or set by the board.

The Operational Reference Portfolio (ORP) is an approach to improving diversification without losing the reference to broadly accepted standard benchmarks. The ORP was introduced in 2011 and is an internal benchmark constructed by the Norwegian Fund. It is a tailored benchmark with non-market capitalization weights, and takes as its starting point the benchmarks set by the Norwegian Ministry of Finance. The ORP takes advantage of the characteristics of the fund: a large investor with no fixed liabilities and no immediate liquidity requirements, and unconstrained by domestic currency considerations.²²

The Norwegian Fund uses the ORP for three purposes, to:

1. Diversify more widely than standard benchmarks;
2. Take on systematic factor risk exposure;
3. Implement smart rebalancing.

We discuss each of these in turn, and explain how each component requires a different verification horizon for risk management. The ORP is internally owned by Norges Bank Investment Management (NBIM) and only takes on a limited set of risk factors that have been specified by the Ministry of Finance. A part of NBIM's mandate regarding fiscal strength in the sovereign bond portfolio is not, however, quantitatively expressed in an off-the-shelf benchmark. This mandate is incorporated by NBIM in the ORP.

Alternative investments may have become the mainstream, but not in the ORP of the Norway Fund. Here, the Government Pension Fund Global keeps 60% of its assets in publicly traded equity, 35% in bonds and up to 5% in real estate. Hedge funds, private equity and venture capital have failed to gain a foothold. In other words, it is a clean-cut portfolio that focuses on the main asset allocation choices, with a limited amount of active management.

There is a pragmatic reason for this approach. The Norway Fund is simply too large to invest in alternative types of strategies. It would be too cumbersome and costly to build up a portfolio of illiquid assets that would amount to anything meaningful. Additionally, the Fund operates very much under the public eye. The investment policy is therefore articulated and clear, with limited maneuvering scope for grasping “exciting” new investment opportunities. And yet, the Norwegian fund has fared quite well over the past years. As might be expected, its apparent simplicity could be misinterpreted as an “anyone can do it” approach, i.e. simply by shedding active management and alternative investments. However, trustees should be aware that investment “rules” become important in order to discipline decision-making and combat behavioral biases. The Fund is disciplined in its asset allocation. When stocks decline, the Fund buys enough to get the 60% equity target back; conversely, it buys bonds when that proportion falls below 40%. The Fund rebalances in this way to eliminate emotion from its decisions, compelling it to buy assets with a low valuation and higher expected compensation for the risk premium and vice versa.

“Liquid” is still long term. Therefore, the choice and construction of the indices to achieve the right exposure becomes a big deal. Which risk premiums can reasonably be earned in liquid markets with a long-term horizon? Norway takes on smaller stocks and value stocks, which may be trading at lower prices in the short term, but this is setting them up for higher performance in the long term—which research interprets as decades, not a matter of a few years.

MODEL 6: RISK PARITY MODEL

What trustees should be aware of

• In a risk parity model, different sources of return, such as the equity risk premium, the credit risk premium, and the term premium, are all scaled in such a way that they contribute the same amount of risk to the portfolio.
• This is done by applying leverage to the less-risky return sources.
• The expected return per unit of volatility risk increases as a result, because the diversification over the different sources of return is much better than in traditional portfolios.
• The successful implementation of risk parity requires continuous rebalancing of the different positions. This is operationally challenging and demands deep and liquid markets.
• In practice, few—if any—asset owners use risk parity as their governance model.

The idea behind risk parity is that a portfolio based on traditional asset allocation techniques may look roughly balanced, but from a capital allocation standpoint rather than from a risk perspective. For example, equities have contributed far more to total portfolio volatility than the equity allocation would suggest: between January 1997 and July 2017, 93% of the volatility in a 60% equity + 40% bonds portfolio could be attributed to equities.²³ The solution to this disproportionate risk allocation to stocks is to reduce the weight of stocks according to risk parity and increase the weight of bonds, so that equities and bonds have a similar volatility impact on the portfolio. However, the overall volatility of the portfolio will drop. If the investor is comfortable with the initial volatility of the 60/40 portfolio, proponents of risk parity suggest that leverage should be used to achieve the required risk level. Overall, average returns would have increased with decreased risk. The risk parity–based portfolio construction process is agnostic in the sense that it balances the portfolio so that each asset class contributes the same potential for losses, so-called “risk parity.”

The assumption of this risk parity is that an investor will always be rewarded over the long term for taking risks rather than holding cash. Pension funds can develop risk parity, or an “all-weather portfolio,” in a two-step process. First, the risk budget is allocated into four compartments (low or high economic growth and low or high inflation). Second, within each risk budget compartment, asset classes are suballocated (also on an even basis), based on the exposure to the growth–inflation dimension. Here, the “purer” the asset class or strategy, preferably based on one factor premium, the more effective this approach is.

Pension funds are increasingly adopting a matching-return approach. This investment model divides the portfolio into a matching portfolio based on fixed income instruments that either partially or fully replicate the cash flows resulting from the pension liabilities. The return portfolio, on the other hand, takes investment risk to build up regulatory reserves and funds for indexation. Theoretically, the risk parity model could also be applicable in a matching-return approach.²⁴

The best-known pioneer of this approach is Ray Dalio's Bridgewater hedge fund, which is based on the idea that a portfolio built on the allocation of risk should be more resistant to market downturns than a traditional portfolio. This is an appealing concept for pension funds and their trustees. Ben Inker, director of asset allocation at GMO, singles out three points of criticisms²⁵ for this approach. His main objection is that volatility and risk are not the same things. For example, leverage allows investors to boost returns, but also introduces path dependency. While an unleveraged investor can wait for a process to converge certain valuations, a leveraged investor may not have this luxury. Leveraged investors cannot expect to maintain the same amount of leverage during high periods of volatility, but rather are forced to reduce leverage and sell securities—at the risk of forgoing positive future returns due to overreaction and mean-reversion after periods of financial distress. Another issue relates to the selection of factors. While risk parity may suggest a valuation-neutral approach, in reality it is not neutral and investors might end up leveraging risks without associated positive returns. For example, the term-structure premium is a factor behind fixed income. A risk parity approach would significantly increase exposure to this factor, but does this make sense in today's monetary environment? Looking at a risk distribution pie chart does not give the investor information as to whether the portfolio is optimally diversified. For example, while country risk might be dealt with in the risk distribution, it still matters which country's default risk is included. In other words, for pension funds considering looking into the risk parity approach, one consideration could be that risk parity improves the insight of the investor into the portfolio's risks and sensitivity to scenarios. On the other hand, in-depth insight into the portfolio is still needed, alongside knowledge of the intended or unintended interaction between leverage and risk appetite.