Asset Liability Modelling for Pension Funds Jon Exley

Asset & Liability Modelling for Pension Funds Jon Exley & Shyam Mehta June 2000

Asset & Liability Modelling for Pension Funds • • • Observations on current practice ALM in other financial applications Objectives Implications of Value Maximisation New ALM applications Conclusions

Observations • • • Different asset models, same conclusions Different jurisdictions, different conclusions Different liabilities, same output Minor assumption changes, different output Seeks high risk/return, not matching assets Accuracy is +/- 100%, not +/- 10% without implicit or explicit calibration

ALM in Other Financial Applications • Lack of resemblance is striking • Elsewhere focus is on matching – well defined output – very sensitive to liabilities – worthwhile to develop good model/techniques – results tested in real time • Asset model essential for options, in other applications may work direct with data.

Objectives • Clear objectives lead to clear conclusions ? • In practice, most objectives “maximise return subject to acceptable risk” • Arbitrary nature of “acceptable” allows results simply to follow convention • Value maximisation gives true clarity • Natural objective, contrast to “risk of ruin”

Implications of Value Maximisation • Advise parties separately • But, gain from equity investment for one party is loss to other • If both sides well informed/well served better to maximise aggregate and share

Implications of Value Maximisation: Members • Loss of value on default, gain on upside call • Prefer non systematic risk, efficiency useless • Optimal to match, use equity linked funding • Where does traditional ALM fail? – Inadequate focus on solvency events – Discretionary practices not modelled – Focus is not on value

Implications of Value Maximisation: Managers • Ideally serve interests of shareholders • In practice may support any ALM which serves own interests – Equity management more fun than admin – Do not bear downside risks personally, enjoy safety in numbers & like systematic risk – Substantial personal interest in pension benefits • Fee income is not proof of correctness

Implications of Value Maximisation: Owners • Opposite benefits to members from equities

Implications of Value Maximisation: Aggregate • Suppose focus on aggregate, then matching: – focuses management on value building – reduces agency costs – controls cash in business – improves transparency to members – reduced external costs – taxation gains, etc.

New ALM Applications • Identifying best match is valuable and challenging • There are risk/return trade offs but different magnitudes (cf realistic equity premium) – illiquidity premiums – disintermediation gains – tax • Direct placings need to be priced/evaluated

Conclusions • Existing ALM of questionable value • New ALM applications are possible in value -added framework • Advantages: – accurate liability modelling is valuable – good models/statistical methods worthwhile – judgement still important – coherent theoretical framework