The bear market in equities from 2000 to 2002 exposed an underlying weakness in the traditional approach to pension fund management. Conventional pension portfolios characterised by substantial allocations to public equities (often with a large domestic bias) and intermediate duration fixed income proved costly. The confluence of falling equity prices and declining interest rates led to a major deterioration in the asset/liability ratios of many pension plans (see figure 1).

In response, many plan sponsors and trustees sought to reduce or eliminate risk, while satisfying their long-term benefit obligations. Recent accounting and regulatory changes have created further incentives to eliminate risk, as it appears that the changes will lead to increased volatility in financial statement values and required contributions. Consultants, asset managers and investment banks, in turn, responded to the concerns of plan sponsors by developing a variety of so-called ‘liability-driven investment’ programmes.

Generally speaking, LDI programmes took their lead from asset/liability management techniques employed by insurance companies to match the interest rate sensitivity of their asset portfolios to the interest rate sensitivity of their liability cash flows. The tools employed in many LDI programmes are fairly limited, focusing mostly on longer-duration bonds and interest rate and/or inflation swaps.

 

The problem

The need to generate returns - often high returns - remains, yet many low-risk (and likely low-return) LDI strategies cannot satisfy this need. Additionally, pension fund managers must cope with improving life expectancy, low interest rates, and potentially lower risk premiums, along with the aforementioned accounting and regulatory changes.

There are problems with the current array of very-low to no-risk LDI strategies as proposed solutions to these issues. One is that, while such programmes can often achieve the laudable goal of greatly reducing interest rate risk, they often reduce long-term return potential in the process. Another problem is the difficulty in matching very long-term benefit cash flows (40-50 years in the future) with the existing supply of long-term fixed income investments. Many LDI strategies also neglect important aspects of pension plans, such as plan sponsor commitments to make contributions, the dynamic nature of plans that are still open to new members, and the ability of some plan sponsors to reduce the indexation of benefits to inflation.

More fundamentally, with their primary focus on fixed-income type investments, these LDI programmes deprive pension funds of one of their principal strengths: the ability to achieve equity-like returns by collecting risk premiums over a very long time horizon. What is needed is a portfolio design that enables a pension fund to capture equity-like returns, while buffering the portfolio against short-term financial market conditions that produce volatility in both financial statement values and required contributions.

 

A proposed solution

We believe there is a smarter way to address this need for higher returns, one that calls for better allocation of risk so as to improve long-term returns, while reducing asset/liability risk and potentially reducing asset-only risk. We refer to this risk allocation methodology as integrated asset/liability management (IALM). The IALM investment methodology takes advantage of a full array of investment strategies, as well as risk management and financial engineering techniques that, to date, have not been widely used in pension fund management. Unlike typical LDI programmes, it is not a “single solution” approach. IALM is a customised integration of strategies that allows pension fund managers to re-adjust their “risk budget,” as it were, to achieve their long-term return objectives, yet maintain a buffer against adverse equity markets.

A key tenet of IALM is not to minimise risk, as is the case with other LDI programmes. Rather, the goal is to achieve a more efficient allocation of risk exposures, so as to maximise return per unit of risk. IALM seeks to lower the pension fund’s asset/liability risk (or “surplus risk”) and potentially lower the plan’s asset risk, while maintaining or increasing its return potential. This can be accomplished with a better allocation of market risk exposures (commonly referred to as the beta portfolio) and active management exposures (commonly referred to as the alpha portfolio). The strategy avoids a concentration of risk exposures, particularly an over-reliance on domestic public equities. IALM uses a variety of financial techniques to facilitate the optimal allocation of risk exposures, so that the portfolio’s risk/reward balance can be improved and desired returns can be achieved over time.

Asset/liability risk

One way to engineer a better trade-off of risk and return is to reduce the contribution of interest rate risk to asset/liability risk. For illustration purposes, let’s consider the typical allocation for a corporate pension fund, as shown in figure 2. Assume that the duration of the liabilities is 12 years, and that the pension plan is 100% funded.

From an asset-only perspective, most of the risk is equity-related. However, from an asset/liability perspective, interest rate risk is quite significant, as it is more than one-third of the asset/liability risk. This significant interest rate risk exposure largely explains why asset/liability risk is much higher than asset-only risk. In a worst-case scenario (defined as having a 5% probability that it will occur), asset-liability risk is 5% greater than asset-only risk (22% versus 17% - see figure 3).

Asset/liability risk should be the focus, since it is a true measure of pension fund risk, namely a future shortfall of assets with which to pay benefits. In addition, it is asset/liability risk, not asset-only risk, that creates uncertainty in regulatory required contributions. Lastly, as accounting rules are revised to incorporate mark-to-market approaches, asset/liability risk will be translated into additional volatility in financial statement values. The significant interest rate risk exposure is a result of the duration of the assets being considerably shorter than the duration of liabilities. The duration of assets is approximately 1.35 years based on a 27% allocation to US fixed income, which is assumed to have a five-year duration, and compares to a liability duration of 12 years.

In other words, the pension fund is assuming a substantial amount of reinvestment risk. To better measure and manage this component of asset-liability risk, pension funds should consider creating a custom benchmark that mirrors the duration characteristics of the liabilities. Typical fixed income benchmarks are not appropriate.

In our example, we assumed the equity duration to be zero. There is a strong theoretical argument that the equity duration should be positive over long periods: when interest rates fall, equities tend to rise in value. Positive equity duration would reduce the duration gap and thus mitigate interest rate risk exposure; however, equity duration is not stable over shorter time periods. Equity duration may be negative during equity bear markets that coincide with a recessionary environment. This is a worst-case scenario for most pension funds. Over long periods of time, yield curves typically slope upward, so extending duration often enhances returns. Even if we maintain a view that interest rates are more likely to rise than fall, we still want to reduce the large duration gap somewhat, as the probability of being consistently correct with forecasting interest rates is likely not high enough to justify the risk assumed.

One way to build a better risk/return tradeoff is to redesign the beta portfolio, or asset allocation. While there is no single best way to construct a beta portfolio, certain principles and actions are essential:

n
Minimise large market exposures that have low expected return per unit of risk.

n Diversify across a broad set of asset classes.

n
Search for structural inefficiencies in illiquid asset classes.

n
Focus manager selection efforts on asset classes where the largest manager alpha opportunities exist.

n
Consider the correlation of the beta exposures to the plan sponsor’s business .

n
Consider a more dynamic approach to implementing the portfolio by developing forward-looking assumptions for expected returns, standard deviations of return, and correlations of returns.

 

n Minimising exposures A simple way to design the beta portfolio is to divide asset classes into two groups. The first group consists of asset classes that demonstrate risk characteristics embedded in the liability benchmark; with IALM, the first group consists of nominal rate bonds and inflation indexed bonds. These asset classes can be used to reduce the pension plan’s large exposure to interest rate risk and/or inflation risk - ie, risk exposures that may provide a low expected return per unit of risk. The second group consists of asset classes that are less correlated with the liability benchmark, particularly over shorter periods, but on average provide an excess return above the liability benchmark.

 

n DiversifyA second action would be to increase the diversification of the portion of the beta portfolio designed to provide an excess return over the liability benchmark. Asset classes that provide an excess return over the liability benchmark and also have a positive correlation to the liability benchmark are particularly attractive. If the correlation is positive, it reduces asset/liability risk; if negative, it increases asset-liability risk. Most of the asset risk in the typical pension fund consists of developed market equities, with a large domestic bias. In our example, over 85% of the risk is public equity risk and 65% of the risk is domestic equity risk. We would submit that this is a sub-optimal use of a risk budget. Greater diversification can be achieved by reallocating the risk budget to incorporate increased allocations to international equity risk, particularly small cap equities and emerging market equities, or to credit risk. Further dimensions of risk diversification can be exploited through real estate, private equity, infrastructure, commodities, timberland and other non-traditional asset classes. Diversification is the only free lunch in investing. It also helps to avoid common investor pitfalls, namely overconfidence in one’s ability to predict the best performing asset classes and the anchoring of return expectations based on recent history.

 

n Illiquid asset classes Private asset classes, such as private equity, private real estate and distressed debt provide additional diversification opportunities. Pension funds have long time horizons and illiquid, relatively deterministic obligations (and can usually anticipate receiving future contributions). Thus, such illiquid investments represent “natural fits” for the pension fund portfolio. Since these asset classes are not employed by many other institutional investors that are constrained by shorter time-frames, liquidity needs, or binding regulations, pension funds can benefit by capturing liquidity premiums derived from structural inefficiencies that exist in private asset classes. In other words, pension funds can achieve returns in excess of those predicted by equilibrium models (such as CAPM). Private real estate also provides an additional diversification benefit as real estate returns are driven more by local supply and demand fundamentals than are equity returns.

 

n Manager selectionPension fund investment professionals should consider focusing manager selection efforts on less efficient markets with better alpha opportunities, where higher information ratios can be achieved. Also, the variation in manager performance is often quite large in less efficient markets, particularly in private markets, so manager selection is critical. The potential upside from manager selection for private equity is illustrated below, revealing that there has been a 10% difference in returns between top quartile and median managers.

On the other hand, exposure to more efficient markets can often be obtained inexpensively, via pure index strategies, enhanced or structured index strategies, exchange-traded funds (ETFs) or derivatives.

 

n Correlation There are two sources of asset growth: contributions and investment returns. Due to the cyclicality of the plan sponsor’s business, the sponsor’s ability or willingness to make contributions may be highly correlated with asset class returns. Thus, it is important to understand how market risk exposures also impact the sponsor, whether in the form of variability in tax revenues and entitlement programme payments, as is the case with public pension plans, or in the form of variability in operating cash flows, as with corporate pension plans.

n More dynamic portfolio constructionAssumptions for asset class expected returns, standard deviations of return, and correlations of returns typically are scrutinised only every three years, during which time changes can occur in asset class valuations or risk levels. A disciplined process, whereby asset class assumptions are estimated more frequently, may help identify opportunities due to changes in valuation. In practice, investment professionals could tilt the portfolio towards undervalued asset classes and use the pension fund’s long time horizon to wait for the presumed mean reversion in valuation.

 

The alpha portfolio

Traditionally, active management risk has been a small proportion of a pension fund’s overall risk. In our example, almost the entire asset risk budget is devoted to market risk, and market exposure provides most of the excess return above the risk free return.

 

Expected asset return 8.53%

Risk free return 4.75%

Market exposure return 2.92%

Active manager return (alpha) 0.86%

 

Asset risk* 10.3%

Market risk 10.2%

Active manager risk (alpha) 1.6%

 

*Asset risk is defined as the annualised standard deviation of returns

 

Since active management risk has a low correlation to market risk (beta portfolio), increasing the proportion of active risk can improve the overall portfolio structure and produce a higher return per unit of risk. The opportunity set is also much larger with the alpha portfolio.

The problem with the beta portfolio is that the universe of asset classes is limited and many asset classes are highly correlated. Also, several asset classes may be correlated with the plan sponsor’s business. On the other hand, the range of alphas is less limited - at least theoretically - and the alphas can be selected to have low correlations to the beta portfolio and the plan sponsor’s business. An opportunity to improve the pension fund’s risk-return trade-off derives from the fact that most of the expected excess returns from active management are obtained from the most efficient markets. The traditional approach in pension fund management is to determine the beta portfolio, or strategic asset allocation, then to identify the best managers in each asset class. Thus, the alphas have been sourced from the same asset classes that provide the beta exposures. Developed market, long-only large-cap equity has typically been the largest beta exposure.

The alpha portfolio can be designed using funded strategies, such as hedge funds, or unfunded strategies, particularly overlays that rely heavily on derivatives. The alpha portfolio can also include the practice of removing constraints on long-only asset managers. For instance, equity managers can use so-called “active extension” approaches, whereby they are allowed to short a specified percentage of the equity portfolio. For example, a portfolio that permits a 30% short position would maintain 100% exposure to the equity market via long positions of 130%. Several issues exist in constructing the alpha portfolio. One, alphas are negative, on average, because managers experience transaction costs and fees in executing their strategies, and because skill is a zero sum game in the capital markets. The variation around this average is large, so the penalty for selecting poorly is also large. In addition, a manager’s out-performance is often described as alpha, but may be due to residual beta. Moreover, the strategies used to produce true alphas are not free of risk and may include non-linear return distributions.

Thus, risk measurement can be challenging. Certain alphas also may prove transitory, because the strategies used to produce them can be replicated easily and widely by others. Finally, building a diversified alpha portfolio, consisting of numerous alpha sources with low correlations, involves a considerable time commitment. Even so, this time commitment has a large potential payoff in the form of a higher information ratio: the ratio of the excess return to the volatility of the excess return. The improvement in information ratio is dramatic as more uncorrelated alphas are added. When the correlation between alphas is positive but low, the improvement in information ratio is still significant, though the improvement is small once the number of alpha sources exceeds a threshold of around 15 to 20.

The problem of residual market risk exposures
can also be managed. One simple approach is to tilt the alpha portfolio toward strategies that are unlikely to possess the market risk exposures embedded in the beta portfolio. For instance, a beta portfolio with a large equity exposure can be combined with an alpha portfolio that does not use equity securities, such as fixed income relative value and commodity trading strategies.

In the past decade, the proliferation of financial engineering, via derivatives, structured products, and securitisation technology, has blurred the distinction between asset classes. Risks can now be unbundled and repackaged in new forms.

A pertinent example for pension funds is the ability to separate the interest rate and credit risk components of a bond, via interest rate, inflation, and credit derivative markets. Pension funds can use interest rate swaps to reduce the interest rate duration gap versus the liabilities, while at the same obtaining credit risk, via credit default swaps or structured credit products. Derivatives can also be used to separate beta and alpha exposures. Thus the search for alpha can be delinked from the optimisation of the beta portfolio. For instance, total return swaps or futures contracts can be employed to efficiently obtain developed market public equity exposure. In brief, pension fund investment professionals should consider obtaining risk exposures synthetically when using derivatives is inexpensive, and investing pension capital in asset classes and hedge funds when using derivatives is expensive.

The overall improvement from such reallocations can be seen in figure 4.

 

Conclusion

Pension fund investment managers face an enormous challenge: achieve high returns over time to meet future obligations, while reducing asset/liability risk. We believe this challenge can be met by employing the full array of investment strategies, risk management, and financial engineering techniques currently available in the capital markets. There is no single solution to this challenge, as pension fund structures and needs vary widely due, in part, to regulations, funding status, sponsor financial strength, size of pension fund relative to the sponsor, the cyclicality of sponsor cash flows, the demographics of the plan participants,the policy for contributions, and the benefit formula. We do not believe that common LDI products can deliver the desired returns given their predisposition toward lower risk and lower reward strategies.

IALM, on the other hand, can assist plan sponsors, trustees, and investment staff in constructing portfolios that will achieve their long-term return objectives, while minimising negative impacts, such as contribution volatility, that result from shorter-term adverse financial market conditions. We have described in detail the components of the framework and its rationale, but the basic framework is quite simple and can be condensed into a single diagram, as shown in figure 5.