How does one manage an €18bn pension fund with only three staff? Most people would say the three staff have to work very hard. While that is correct, one then asks, how this fund manages to rank amonge the best returns within the industry with a strong funding position? The bias would be to assume large allocations towards risky assets and/or just pure luck. Wrong! The answer is almost too simple to be true, and it will work for all funds, regardless of size and complexity; namely ‘dynamic alpha and beta management’ which leads to proactive asset-liability management (ALM).
Managing pension asset portfolios to match liabilities has always been a challenge for many reasons, but the simplest challenge faced by clients is that most pension liability cash flow projections are made annually at best, while assets have to be managed on a day-to-day basis. Clients conduct asset-liability studies to establish strategic asset allocations (either static or dynamic) to meet to a broad group of objectives for a given set of liabilities, but these are based on simulated (or optimised) expectations of future returns and inflation forecasts. Often these simulations cannot capture the practical realities of managing portfolios. So clients need to migrate away from the old practice of doing ALM analyses and focus on a more innovative ‘asset only’ management approach (ie, beat the strategic asset allocation through active management strategies)2.
Liability indices can be developed (using daily available swap indices) to allow for an intra-year tracking of liabilities, especially in a regime where pension funds are required to mark liabilities to market3. Once investable liability indices are developed, pension funds can adopt innovative investment approaches to create appropriate investment strategies, both beta and alpha to effectively track liabilities on an intra-year basis. This ensures the solvency and funding ratio of their pension funds. We leverage the work done at the Dutch €18bn PME pension fund to clarify our views. Our company has developed a software platform to manage this framework.
Portfolio construction normally proceeds along the following path:
o Step 1: Previously, pension funds would hire ALM consultants to evaluate the optimal long-term strategic asset allocation (SAA) to various broad asset classes. These ALM consultants would use cash flow projections from the actuaries to develop a broad investment policy that would meet the complex objectives of the pension funds, such as minimise contributions while ensuring funded ratios in excess of 100%, but also minimising peer comparison risk.
Good investment policies would include detailed allocations to broad asset classes (domestic stocks, foreign stocks, bonds, commodities, etc) as well as specific benchmark indices to which these assets could be measured. Very often, the ALM policy analysis would also include some naïve rebalancing recommendation (eg, rebalance the actual portfolio to the policy portfolio once a year or rebalance the portfolio when any asset drifted more than +/-3% away from the target mix).
More sophisticated versions of these ALM studies would include dynamic asset allocations whereby future annual target allocations depend on the funded status of the plan at the start of the period;
o Step 2: The pension funds would then invest these assets either internally or hire a series of external managers and further would extend investment categories. For example, domestic stocks could be broken down further into large cap stocks, small cap stocks and further tiered into value stocks and growth stocks depending on their price-to-book characteristics. This activity actually involves two separate and distinct steps.
Firstly, portfolio structuring, where the asset classes are further broken down into country/size/style sub-classes (what we call static beta), and secondly, investment with internal or external active managers (static alpha). Historically, many pension funds have combined these activities or let portfolio structuring evolve implicitly as a result of the manager selection process. Further, manager selection really involves two decisions, namely whom to hire and how much of the portfolio to allocate to them. If the plan sponsor hires even two managers per investment area, this client is often managing 30 external managers;
o Step 3: Most pension funds are dynamic in that there are cash flows constantly coming into the pension funds from contributions of participants or cash flows from maturing bonds or venture investments, and cash flows exiting the pension funds to pay benefits or meet capital calls for alternative assets. These raise a number of challenges for clients trying to meet the liability obligations (with the most important one being the change in the market value of liabilities), but also present interesting opportunities for the more sophisticated clients. This aspect of portfolio management as dynamically managing beta and alpha.
Figure 1 shows the typical investment decision process (IDP) of a European pension fund where the manager selection decisions are denoted by the arrows in the picture.
The old paradigm put forward that most funds’ returns derived from static beta and alpha, and fund managers were well advised to consistently ‘rebalance’ their portfolio to benchmark allocations rather than attempt any sort of ‘market timing’ or dynamic beta and alpha management. Also, implicit in this approach was the assumption that dynamic alpha and beta management could not consistently produce superior returns and hence the low tracking error approach of rebalancing was preferred.
There are a few shortcomings with approach:
o If the portfolio was constructed to have truly diversified assets, the low correlation between these assets/asset classes would dictate that at any given time some of them would underperform while other investments would outperform, as that is what by definition low correlation means. Most clients have built portfolios of assets with low correlations to achieve long-term diversification, without recognising the short-term risks of low correlations, namely that a statically rebalanced portfolio could have very bad drawdowns;
o Every cash flow decision made by a plan sponsor, whether an investment or a withdrawal of funds, amounts effectively to ‘market timing’ as it is taking a view that impacts portfolio performance. In light of this, it is more important to make these decisions in an informed and disciplined way (ie, analysed in the context of the investment objectives) than to be naïve about these decisions and hope for the best outcome. As our mentor Professor Franco Modigliani used to say: “You cannot pay pensions with low tracking error but only with returns.”;
o While the SAA could drive a majority of the portfolio returns (or their variability), it does not preclude the investor from trying to extract every potential return source and hence ignoring the returns or risk improvement potential from areas like dynamic alpha and beta management can leave valuable returns on the table. Also, inevitably, there will be cycles where each of these sources of return will dominate the portfolio performance; to ignore them is to condemn the portfolio to an unattractive cyclicality;
o As shown in figure 1, dynamic beta management that shifts the allocations at the asset class level according to market conditions and intelligence affects the fund’s entire assets and should commensurately have a greater impact on total fund returns than decisions that affect only a small portion of the portfolio like a single manager selection/allocation (the arrows in figure 2). Therefore, the cost of ignoring dynamic beta decisions on the higher level assets could be costly. Funds may get a bigger impact from allocating internal resources to ensuring that their asset allocation decisions are appropriate than on manager selection (though they are not mutually exclusive)4.
In ensuring good investment practices, sophisticated pension funds will focus attention on both dynamic beta management (ie, smart decisions on which assets/styles to favour) and alpha management (ie, which managers to favour and when).
Innovative pension plans that protect their funded positions effectively will be those that migrate from the old pension paradigm to the newer paradigm of focusing on dynamic beta management and dynamic alpha management (figure 2).
Moreover, making good pension fund structuring decisions may be a more cost effective method of achieving effective liability matching than investing in some liability-driven funds, as the yield on fixed income instruments globally is below the expected liability return. In any case, such decisions should be carefully researched and analysed, and the potential performance implications fully understood for each funds specific structure before committing to a particular course of action.
The liability benchmark of the client can be replicated by a portfolio of swaps at key maturities (two year, five year, 10 year, etc). Effectively, this portfolio of swaps captures the interest rate risk of the liabilities
and in effect, if interest rates decline, the value of these liabilities will rise over time.
On the basis of an ALM exercise, the optimal long-term strategic asset allocation for this hypothetical fund was fixed income 53%, equities 32%, real estate 10% and alternatives 5%. If the client had implemented this SAA, hired managers and implemented simple rebalancing strategies, their performance relative to their liabilities would have led to a decline in the surplus over the January 2003 - March 2005 period. Instead, by adding smart dynamic beta and alpha management strategies, a potentially declining surplus profile can be converted into an increasing surplus
profile and, in turn, a lower risk of being below some funded ratio threshold.
In a similar manner, external managers have investment profiles that are cyclical or reflective of their asset class, and dynamic alpha management would involve being intelligent about when such managers are funded or assets withdrawn to meet cash flow needs of the pension fund. As pointed out earlier, all pension funds are effectively dynamic because of rebalancing policies and cash flows. Following a good process of capturing the impact of factors affecting asset classes and managers can lead to dramatic improvements in the ALM profile of a fund.
In the case of this hypothetical client, the annualised liability return was 8.6% over the Jan 2003 - March 2005 period. Against this benchmark, the static SAA would have detracted -1.36% annualised and adding a managers who in aggregate generated positive excess returns would have reduced the annual underperformance to -0.72%. As is typical with some pension funds, rebalancing ranges are set at +/-5% of the strategic allocation and this would have lowered the annualised underperformance of the asset portfolio relative to liabilities further to -0.58%. As the table shows, the risk of having a funded ratio under 105% has declined from 47% to 33%, but the annualised growth in the surplus
is negative. Using the IDP shown
in chart 1 and developing some intelligent rules between stocks/bonds/ cash, and within equities between euro equities and UK equities, and within fixed income between government bonds and high yield, the perspective changes dramatically. The annualised growth in surplus turns positive to 0.48%.
This is easily understood if one examines the recommended allocation to bonds (and stocks), which increased dramatically in a period in which bond performance would have been attractive and hence liabilities increasing in value (through mid 2003), while tapering off gradually and overweighting stocks when the reverse occurred and stock performance has been strong.
Similarly, in the interest of simplicity, we developed some intelligent rules across the five managers handling assets in the European government area and this could have added additional returns of 0.18% at the total fund level, thereby making the annualised growth in surplus decidedly positive. The trade-off is that the volatility of the growth rate of the surplus could increase, but
not sufficiently to deteriorate the ALM risk statistic of being under the 105% funded threshold, which is now just 16%.
With changes in regulations, especially in the Europe, there is pressure to implement risk ‘measurement’ systems, which requires better decision-making going forward.
This note demonstrates that the old paradigm of, (a) doing ALM studies and then managing assets against the SAA, could be greatly improved by creating liabilities indices that can be marked to market periodically, and (b) the old approach of hiring managers on a static basis and naively rebalancing to certain target weights, can be dramatically improved by using market intelligence and the dynamism of the portfolio to create intelligent beta and alpha strategies, leading to positive annual growth in the surplus.
In addition, establishing smart dynamic beta management strategies can have a more dramatic impact on the overall ALM profile because these decisions are made on the entire assets under management. Manager selection and dynamic alpha management can be a complement to such beta management strategies, but may have a lower impact on the management of total ALM risk because these decisions reside lower in the IPD.
Innovative clients will recognise that in a lower projected return future, every investment decision impacts returns and risks and it is critical for pension funds to measure, manage and monetise all their investment decisions and focus their efforts on decisions that have the greatest potential impact.
The proof of the pudding is in the eating: with only three investment staff since implementation, the PME scored amazingly well.
1The authors are Sanjay Muralidhar, CEO; Rahul Rauniyar, investment research analyst; and Arun Muralidhar, chairman of Mcube Investment Technologies, LLC.
We would like to thank Roland van den Brink, managing director of Investments PME, Paul van Gent, head of external
managers, PME and Patrick Groenendijk, head of investment, Pensioenfonds Vervoer. Thanks to Benne’ Bette’ for assistance.
These are the personal views of the authors.
2A Muralidhar (2000). Innovations in
Pension Fund Management. Stanford
University Press, Chapters 1-5.
3J W Stuivenberg and A Muralidhar (2005). ‘Devising an Investable Index,’ Investments and Pensions Europe, October 2005,
4A mistake made commonly in the industry is to ask whether it is better to extract
alpha from manager selection (external managers) or asset allocation (beta), and it usually comes down to the glib comment that the information ratios are higher for security selection. As Figure 2 shows, these apply at different levels of the fund and smart funds will do both to diversify their risks.
5This is meant to be approximately what is being required by the Dutch regulators.