A consequence of this progress is an increasing need for a firmer grip on risk composition. To this end, ING Investment Management has developed the ALRA risk model, focusing on a maximum horizon of one year. The starting principle is the pension fund’s current investment portfolio and current liabilities. Ronald van Dijk, head of quantitative research equities at ING Investment Management, looks at the model in more detail.
FT Mandate: Why yet another new model and new report?
It is important for pension funds to understand the uncertainty in their funding levels and the expected growth path of this level. Another variable is the extent to which pension benefits are linked to inflation. The scheme’s cost and who pays what are also important.
Asset liability modelling (ALM) charts possible long-term developments of funding level, indexation percentage and pension contributions. Models currently used are based on long-term relationships between economic and financial data series. As these models adapt slowly to changing market conditions, they are less well suited to determining precise risks in the short term, for example on a quarterly basis.
There are other limitations inherent in ALM studies. The number of investment categories which can practically be included in the analysis is limited for statistical reasons. Current techniques cannot look much beyond major categories such as equities, fixed-income, real estate and derivatives. In the current investment climate, however, we need a firmer grip on risk composition. In addition, ALRA gives information on solvency ratios for decision-making. For example, an estimate of the required solvency to have a maximum 2.5 per cent risk of underfunding might lead to a particular asset mix and premium policy.
The model and all corresponding valuation and risk statistics are based on the market-value approach and the most recent market conditions. For Dutch pension funds the model complements the viscous standardised method of determining the required solvency as laid down by the De Nederlandsche Bank (DNB).
FTM: What does ALRA do?
ALRA stands for Asset Liability Risk Analysis. Our ALRA simulation model provides information about the pension fund’s risks and opportunities. It is a prudent addition to ALM studies and ALM Quick Scans. It provides a current view and additional information about what the solvency test aims to achieve: insight into risk of underfunding over one year. ALRA is not a fixed standard formula such as the solvency test, but a dynamic risk measurement and risk control instrument, taking into account changing market conditions and our house view on future developments.
The model’s major added value comprises:
• Measurement of short-term risk of underfunding and predicted indexation. Both the pension fund’s specific structure of investments and liabilities are included in the analysis.
• Insight into risks arising from decisions made in various phases of the strategic asset allocation process. The total risk is, for example, divided into risk arising from ALM, strategic asset allocation and manager selection. Figure one demonstrates how this is included in the ALRA report.
• Subdivision of the risks according to their sources, including equities, real estate, credit, alpha and beta risks. This is also shown in figure one.
The model’s most important technical features are:
• Changing market conditions are identified and dealt with quickly. Relatively high-frequency data is used. If the weekly data is unsuitable, then we will use monthly data. For other parts, we use the data at the moment at which the model is applied. The latest term structure of interest rates is an example of this.
• Account is taken of the fact that risks may be greater than revealed in the past. Simultaneous market crashes are allocated a positive probability in the ALRA.
• The interest rate models are arbitrage-free. As a result, interest rate derivatives can be included in the analysis. There is consistency between the interest rates which we apply to generate fixed-income returns and to simulate market values of liabilities, and the interest rates that we use to price the interest rate derivatives.
• All analyses are carried out using market values. Liabilities are valued on the basis of the term structure of interest rates. If applicable, a premium is added as a compensation for the lack of certainty in the liability structure.
• ING Investment Management’s views are used as well, because historic and current market data provide insufficiently reliable information on expected future developments. This is especially the case with expected returns.
FTM: When do you choose which risk analysis?
Several models are currently offered to pension funds, including ALM studies and ALM Quick Scans, and now ALRA too. Although all three model types chart risk of underfunding and indexation opportunities, there are great differences. An ALM study is the most comprehensive and should be able to answer general questions a pension fund has with respect to available policy instruments (indexation policy, premium policy and investment policy).
The horizon examined varies from one year to 50 years. In the ALM Quick Scan, liabilities are modelled in less detail but the process can be carried out more quickly. ALRA provides a relatively large number of details about underfunding risks and indexation opportunities. However the horizon ends where that of the ALM study and ALM Quick Scan begins, namely one year. At the same time, it does not just confine itself to the main categories and overlay strategies, but also models equity investment styles, alternative investment and fixed income instruments such as inflation linked bonds, senior bank loans, emerging markets debt, high-yield bonds and alpha risks.
ING Investment Management advises to carry out:
• an ALM study at least once every three to five years;
• an ALM Quick Scan every year;
• an ALRA analysis every quarter.
As an ALRA analysis can prove action is required by a pension fund, for example in the event of poor investment returns, this may be a reason to hold an ALM Quick Scan more frequently than annually or even to decide to do a new ALM study.
FTMandate: How does it work?
The ALRA model generates more than 10,000 possible future developments in the economy and markets. We aim to determine, with a reasonable degree of reliability, the funding level corresponding to one in 2000 scenarios. We take into account that not all investment instruments behave similarly. Some instruments have in the short term a return distribution, which does not statistically deviate from a normal distribution. Others display a distribution with such fat tails that a t-allocation with a low degree of freedom is a better choice. This distinction is important as otherwise we underestimate the real risk of underfunding.
Once the return allocation of the investment instruments has been estimated, we allow mutual cohesion by application of a Copula. In professional statistical and econometric literature, this is a relatively new but well-appreciated method of describing complex and non-linear links between return sequences. If this Copula is chosen well, we obtain more conservative estimates of extreme risks than by using normal distributions and the usual estimates for correlation coefficients and covariance structures.
Normal distributions have an undesirable property. With this frequency distribution, two extreme events can (almost) not occur. For example, the implied likelihood of an extreme decline in interest rates going hand-in-hand with an extreme stock exchange crash is zero in the limit. Most of us believe that this cannot be excluded in reality. Therefore we model that the correlation between the returns of various instruments can increase as a special event occurs. Another realistic example in which relations are not stable and correlation can change dramatically in exceptional periods concerns the relationship between real estate and equities. Although they usually have a low correlation, it is possible that in a crisis both will display a strong correlation. One advantage of a Copula is therefore that account can be taken of a strong relationship between two or more funds in extreme conditions. Another is that an event which has never actually occurred can turn up in simulations of the future.
FTM: Is ALRA suited to use by pension funds?
The ALRA model is suited to pension funds, which require a firmer grip on the short-term risks or are searching for a complement to a prescribed solvency test. ALRA offers the opportunity to compare the actual effect of various decisions on the total risk with the required or expected effect. Risk concentrations and inefficiencies can be traced and corrected as required. The total credit risk is measured and can, for example, be compared with the total interest rate risk. ALRA fits in the modern LDI paradigm. Risks arising from active policy (alpha) and risk as a result of exposure to a specific market (beta) can be compared. It can then be determined whether these are in the required proportion to each other.
Pension funds also have the option of developing this type of model themselves, but experience has taught us that this requires a reasonably large investment in financial expertise, econometric techniques, data collection and software development. The model also has to be maintained. Changing conditions on the financial markets and new financial instruments mean that maintenance is a continuous process. By offering ALRA, ING Investment Management enables access to a ‘carefree’ state-of-the-art risk model.
ING Investment Management
ING Investment Management is ING Group’s largest asset manager. We provide a full spectrum of investment solutions and administration services for institutional clients and we manage assets for the ING labels. We are a global asset manager with more than €350bn assets under management. Our three regional organisations (Europe, Americas and Asia Pacific) guarantee a detailed knowledge of local clients and local markets, while our global investment engine provides global investment opportunities. We have offices and investment professionals in more than 30 countries across the world, giving the organisation a global reach while maintaining a local focus.
