Risk management is becoming increasingly important for Dutch pension funds. Trends like more intense competition, more transparency and tighter legislation are putting pressure on funds to control the investment performance.
Risk budgeting within Mn Services is integrated at two levels in the investment process. The absolute return level is managed using asset-liability management (ALM). The outcome of the ALM study is an optimal asset mix given a maximum absolute risk: the risk of underfunding. The ALM asset mix is subsequently translated into a yearly benchmark portfolio, which contains several indices. The investment department is expected to outperform this benchmark portfolio within specified constraints. One of these constraints is the relative-risk budget: the risk versus the benchmark portfolio.
In this article we focus on the four-step approach that Mn Services is using to manage this relative risk
budget (see figure 1).
First step: Setting the overall risk budget
The first step is to determine a risk budget for the total portfolio, including non-financial assets like direct real estate and private equity. This budget is specified as a tracking error (TE). The TE is the standard deviation of the excess returns relative to a pre-defined benchmark portfolio. For example, a TE of 2% indicates that we expect the return relative to the benchmark to be in the range of –2% to +2% in two out of three years.
The appropriate level of TE is set in close cooperation with the client. It depends on the relationship between the risk appetite/aversion of the particular client; the target outperformance; and the ability of the investment department to outperform the benchmark (information ratio). It is also important that the TE is chosen in such a way that the absolute risk level is not disturbed.
The client and investment department agree upon a target and maximum TE . The target (or average) TE can be viewed as the appropriate risk profile. The maximum TE is a constraint. The reason for this is that the investment department should be able to increase risk above the target whenever market opportunities arise. Nevertheless, it is not allowed to exceed the maximum. Over the space of a year, the total portfolio should have a TE that approximates the average. Otherwise the expected outperformance will not be achieved.
The TE is an ex-post measure of risk1. To manage risk upfront we need to translate the TE into an ex-ante measure of risk: relative value at risk (VAR)2. This indicates the expected loss versus the benchmark with a certain probability over a certain time horizon. We use relative VAR with a 5% probability over a three-month horizon. According to the above example, an annual tracking error of 2% corresponds to a three-month relative VAR budget of 1.65% three of the total portfolio. Every risk measure has its limitations4. Therefore, we view VAR not as the ‘truth’ but as a valuable instrument in our investment process.
Second step: Allocating VAR budget
After determining the VAR budget, the budget is allocated to several assetclasses (that is, funds) and thereafter to regions (that is, units). This creates a so-called ‘fund-unit’ structure.
The allocation of the VAR budget is a two-phase process. It starts with a top-down approach, whereby the budget is divided equally throughout the fund-unit structure, to ensure a balanced risk profile. For example, assume a VAR budget at top level of 1.5% of the total portfolio value and three asset classes. Due to diversification between the asset classes each asset class could be allocated more than one third of the total portfolio risk.
Equally weighted risk allocation is rational if you have no idea on future information ratios (IRs), or if IRs are very volatile. It has the advantage of enhancing the possibility of outperformance. An assumed majority of positive bets should outperform the minority of negative bets. This is unlikely to be the case if, for example, one unit has as much risk budget as five other units. When the high-risk unit underperforms, it will cause an overall underperformance.
The second phase consists of bottom-up adjustments based on three principles. Each fund/unit is evaluated according to its size compared to the VAR budget; the possibility of adding value through taking active risk; and the number of alphas (that is, managers). Taking active risk is based on the notion of exploiting market inefficiencies which are most common to high yield, small cap and emerging markets. In short, more VAR budget will be allocated to units which invest in such markets compared to, say, a large US unit cap containing a large number of alphas.
Third step: Implementing risk budgets
The third step in the risk-budgeting process is implementing the risk budget for each unit. As figure 1 explains, the last two steps are an iterative process which have to be implemented in conjunction with each other. The primary aim is an excellent IR – the risk budget is only a parameter and not a goal in itself.
We believe that there are managers with a positive alpha over a longer period in time and, more importantly, that we have the skill to select them. We seek managers who are superior in five criteria of their asset category .
We determine the combination in which the selected managers should be implemented in our structure. Figure 2 shows that this active structure is the basis of the unit structure. Complemented with a passive portfolio, the risk budget can be put in place. By adding or withdrawing money from the active side to the passive side, or vice versa, the risk of the total unit can be influenced.
Fourth step: Monitoring risk
Over the course of a year, the risk management department is to monitor the VAR positions. It will focus on several important events. It will determine if the maximum VAR budgets are exceeded and when a maximum VAR budget is exceeded, an escalation procedure will commence. There will be a decision on whether to hold or adjust the position in line with the budget based on the rational of the position.
The risk management department will also monitor whether VAR budgets are used. The risk managers will advise on when the average risk positions deviate too much from the appropriate VAR budgets. The department will also calculate upfront the risk position for a possible major change in the portfolio of a manager. Lastly, it will concentrate on any changes in the VAR position and try to explain them, in order to understand better the underlying risk.
1Tracking error measured over a one-year period using 12 monthly returns.
2Riskmetrics – Technical Document, JP Morgan/Reuters, fourth edition, 1996.
3This is calculated as follows: (yearly TE/ square root of time periods)* confidence interval = (2% / 4^0.5) * 1.65.
4 For example, VAR assumes a normal return distribution, in practice fat tails occur.
This article was put together by Olof Könst, Marco Ruijer, Pim van Santen and Joav de Waard, at MN Services in the Netherlands. The co-operation is acknowledged of Jan Bertus Molenkamp, senior risk manager