Practical flaws in the liability-driven theory

It has been a few years since the term liability-driven investment was praised by pension consultants and fund managers as the next big thing in the defined benefit pension industry. The sudden interest started around 2005, but the concept has been around for many years.

In the past, pension scheme sponsors assumed a high tolerance for risk, then later found it difficult to fund for pension shortfalls at times of economic difficulties. The pension's funding position should be taken into account in the assessment of risk tolerance levels.

For years, large pension sponsors and trustees pondered the question of how they could protect their schemes from unwanted surprises. Pension schemes are similar to large financial institutions - characterised by small capital (surpluses in pension terms) relative to the size of their assets and liabilities. Because financial assets and liabilities are both sensitive to different underlying economic factors, small fluctuations in either assets or liabilities can cause substantial yet undesirable volatility to the entity's capital.

The idea of asset/liability modelling (ALM) is to construct asset portfolios that can minimise the risks of mismatch movements between the assets and liabilities and thereby protect the capital from market risks such as interest rate risk.

The traditional ALM process starts with creating an asset/liability model simulating the cash flows and market value movements of various asset classes and pension liabilities, and finding an optimal range of asset allocation among various asset classes that can minimise the unwanted risk. Investment managers are then appointed to manage portfolios whose performances are measured against a specific benchmark portfolio. Here is the problem. The process requires constant review and rebalancing, as cash flows from pension schemes can rarely be predicted with precision, unless the liability portfolio is a closed group of pensioners with fixed pension payments.LDI is an extension of ALM using the scheme liability as a tailored performance benchmark with an aim to minimise unwanted risk of mismatched cash flows between assets and liabilities (known as 'tracking error‘).

This new tool is being adopted by schemes as sponsors and trustees rediscover that their liabilities are the benchmark to beat. The approach can enable sponsors and trustees to become risk managers focusing on delivering the scheme's liabilities and separating the risk matching and risk/return decisions.

It typically involves a mixture of bond portfolios and a portfolio of interest rate swaps that can extend the term and reduce the mismatch timing of cash flows. Bond matching portfolios are classic examples of ALM with a custom portfolio of bonds matching the liability cash flows. However, it has proved difficult to match all timings of cash flows without sacrificing yields.

Another LDI approach is to overlay the bond matching portfolio with a portfolio of interest rate swaps to redistribute the timing of asset income to match the liability cash flows better.

A more advanced approach is to use swaps to match the benchmark liability payments exactly, leaving the cash for investments and allowing investment managers to produce returns in excess of the swap interest costs. In short, investment managers are being measured to match the benchmark liability cash flows while making excess returns.

However, there is an underlying flaw in these approaches. Pension cash flows are incredibly hard to forecast with precision. All the underlying factors - including wage inflation, retirement ages, mortality rates and withdrawal rates - are stochastic (random) processes down to an individual level. Moreover, the cash flows for lump sum schemes are much less predictable than those of mandatory pension payments. Some events, such as legislative changes, scheme changes and redundancy programmes, simply cannot be modelled. One wonders whether the benchmark cash flow projections are relevant beyond a few years into the future. Nevertheless, LDI would be effective in very large pension schemes with a sizeable fixed group of pensioners with fixed pension payments, such as civil servants' pension schemes. However, many of these public pension schemes are considerably under-funded so there may not be sufficient assets to produce effective results.

By using more exotic interest derivatives, investment managers can extend the term of investments given a ‘fixed' cash flow schedule. Manager performance can probably be better measured by separating their investment performance and tracking error. Since the benchmark liability cash flow is only notional, the tracking error measurement is somewhat questionable. Consequently, the effectiveness of LDI in minimising mismatch risks may be far less credible than has been claimed.

A more realistic view is to accept that it may not be possible to eliminate the asset/liability mismatch risk entirely unless the liabilities are completely settled through settlement or insured arrangements. Thus, if we take the scheme liability as a whole, rather than focusing on matching cash flows exactly, LDI can be an effective tool in managing asset/liability mismatch risks.

It is, however, possible to segregate the asset portfolio between an ALM component and a risk-taking component. The objective of the ALM portfolio is to minimise the interest rate risk on an aggregate level and to protect the level of surpluses. The other portfolio can take a more aggressive approach with a single purpose - optimise risk-adjusted returns, or simply lock into an alpha or beta portfolio.

The proportion of the ALM portfolio to the risky portfolio would be highly dependent on the funding level of the schemes and the demographic of their members.