Modern Thought on Pension Scheme Valuation and Investing (1990-1997)

The clash of new and traditional thinking on the funding, valuation and investment of defined benefit pension funds grew in intensity over the 1990s, representing one of the more turbulent episodes in British actuarial thought. The 1990 Journal paper, ‘Actuaries, Pension Funds and Investment’, by T.G. Arthur and PA. Randall can be viewed as the start of this period.[1] The paper was not as radical or provocative as those that would follow over the next few years, but it unmistakably advocated a further move away from traditional actuarial valuation approaches to those that attached greater legitimacy to market values and fully embraced their use in pension fund valuation.

The central idea set out in the paper was that changes in the market values of assets that did not match liabilities should be fully reflected in the change in the actuarial assessment of surplus in the pension fund. They argued that to do otherwise was ‘implicitly taking an investment view that the market was wrong’ and that ‘this is not our [actuaries’] job when constructing [pension fund] valuation models’.[2] They further elaborated ‘there is no actuarial reason to suppose that any relative changes in market values that have taken place in the past, whether recent or otherwise, will be reversed in the future’.[3]

This perspective implied a significant narrowing of the pension actuary’s professional input. It was a departure from the traditional actuarial tendency to attempt to ‘look through’ irrational short-term market volatility to the underlying intrinsic worth of the pension fund investments on a going concern basis. For the previous 40 years, pension actuaries had felt quite comfortable placing values on pension fund investments that could differ quite substantially from their market values. This was partly justified on the basis of consistency with an off-market and stable liability basis, but underlying these choices was a philosophical view that short-term market value movements were of no consequence to an open pension fund’s long-term funding: as a going concern, it would never have to sell its assets, and asset income was more stable than was implied by market value movements. More widely, actuarial asset models such as Wilkie’s implied that future risky asset returns were significantly influenced by past returns (mean-reversion in equity returns, for example). This gave a technical manifestation to the actuarial convention that market values varied more than could be justified by changes in assets’ expected future cashflows, and a justification for removing this variation from actuarial valuations. Arthur and Randall argued that all of this had no place in actuarial methods for pension fund valuation.

A natural corollary of their logic was that liabilities should be valued at the market value of those assets that matched them. In the terminology of financial economics, this was the fundamental concept of the replicating portfolio. Irrespective of what pension actuaries chose to call it, it was essentially what S.G. Warner had called for in 1921. Whilst the Thornton and Wilson 1992 paper discussed above still advocated the use of the traditional discounted income approach to asset valuation using parameters that need not be market- based, the actuarial proponents of the market value approach were undoubtedly growing in number and influence.

The final implication of the Arthur and Randall perspective was that pension fund investment strategy should be set at a scheme-specific level that reflected the specific liability cashflow profile of the pension fund. This was a fairly novel idea in 1990 that only became well established in the 2000s under the jargon of liability-driven investing. Their vision of the actuarial role in pension funds removed subjective investment views from actuarial valuations and made the valuation independent of the actual asset strategy of the pension fund. But it suggested a greater role for actuaries in pension funds’ strategic asset allocation—it made actuarial estimates of liabilities central to pension fund investment strategy.

In the Staple Inn discussion there were inevitably voices who were uncomfortable with Arthur and Randall’s modern perspective and its implied diminution of the role of actuarial judgement in pension fund valuation. There remained amongst some senior actuaries a strong view that the pension fund valuation should be an estimate of the amount of assets they expected to be required to meet the liability cashflows given the actual asset strategy of the fund, rather than be based on the market cost of a theoretical risk-free matching strategy. In Dr Sisson’s speech to open the discussion, he argued:

The valuation ought increasingly to reflect the realities of life. Included in this would be a recognition of the investments expected to be held by the trustees in the future and the assumed returns thereon.[4]

This encapsulated the fundamental difference of opinion as to the purpose of the pension fund valuation: was it to assess the current market cost of matching the liabilities (which should reflect the cost of transferring the liabilities to a third-party immediately), or to make a best estimate of the average amount of assets required to meet the ultimate payment of the liability cashflows if a particular set of investment risks (and corresponding expected rewards) were borne by the fund until its extinction? Both of these quantities were, to borrow Dr Sisson’s phrase, ‘realities of life’. Neither was wrong, they were merely answers to different questions. But this clarity was generally missing from the debate.

Arthur and Randall’s call for a market-based valuation approach to pension fund assets and liabilities received further support in a paper by Dyson and Exley that was published in the British Actuarial Journal in 1995.[5] Dyson and Exley emphasised that, by this time, pension fund valuations had a wider set of applications than their traditional use in the assessment of the longterm contribution rate required to fund expected liability outgo. And whilst the assessment of the contribution rate expected to fund long-term expected liability outgo necessarily required assumptions about asset risk premia, a number of other valuation purposes had emerged that did not. In particular, the assessment of whether the fund had sufficient assets to fund the transfer of the accrued liabilities to a third-party—the discontinuance solvency assessment—was fundamentally an assessment of a market cost. It did not require subjective actuarial estimates of long-term asset risk premia. Such applications were naturally more suited to market-based approaches to asset and liability value. The traditional approach of using a discounted income asset valuation coupled with a valuation interest rate based on a long-term expectation for investment returns that was developed in the early 1960s did not answer the discontinuance solvency question.

Dyson and Exley also expressed misgivings about the use of the discounted income approach even in the context of long-term contribution rate assessment. In particular, they highlighted that the equity valuation produced by the discounted income approach was extremely sensitive to the actuary’s long-term dividend growth assumption, and that this was a parameter around which there was significant uncertainty—‘the level of confidence in any estimate of m [the long-term rate of real dividend growth] is likely to be very low indeed’.[6] But there was no escaping this—the estimation of the expected amount of contributions required to meet ultimate liability outgo in the presence of a risky asset strategy was an inherently subjective, difficult-to-estimate judgement.

Dyson and Exley also emphasised that the existence of index-linked gilts (from the early 1980s) permitted a market-consistent approach to the valuation of inflation-linked liabilities that was not available to actuaries in the 1950s and 1960s era when discounted income methods were pioneered. They argued that pension liabilities should be valued using the nominal and real term structures of interest rates implied by gilt and index-linked gilt prices at the given valuation date. Predictably, these proposals led to discomfort for some actuaries who saw their valuation expertise being reduced to mechanically plugging in the market’s latest bond prices. In opening the Staple Inn discussion of the meeting, G.J. Clark suggested:

We must not allow ourselves to become so mesmerised by market information that we exclude one very valuable source of information — the professional judgement of the actuary.[7]

Interestingly, whilst market-based approaches to valuation had increasingly been advocated in actuarial thinking in recent years, their proposals were still viewed as quite drastic. R.C. Urwin commented:

It [the paper] has quite a radical conclusion which is iconoclastic to current actuarial practice.[8]

Concerns were also raised as to the wisdom of relying on index-linked bond prices when the volume of issuance was low relative to the scale of pension fund assets. C.D. Daykin, the Institute President of the time, stated:

It has been indicated by several people that the real return on index-linked gilts is not necessarily a very good indicator [of the market expectation of the real rate of return] ... the low liquidity and the way in which they are influenced by a particular sector of the investment market means they are of less value than they might be in determining the market value of liabilities.[9]

But the biggest protest against the extensive use of market prices was a philosophical one. Many actuaries retained the view that market prices were irrationally volatile, and that actuarial judgement should be able to look through this. S.J. Green rhetorically asked:

What new piece of information became known on 19th October 1987 which was not known three days before? It must have been very significant to lead to falls of more than 25% in a few days, and that is the answer to the author’s point where they say that the market provides rational prospective expectations.[10]

The influential Professor Wilkie struck a similar chord in defence of discounted income valuation approaches:

Using this methodology [discounted income approach], the actuary is, indeed, saying that the market has temporarily got it wrong, but that, in due course, it will get it right, especially by the time the liabilities fall due.[11]

Wilkie’s view of the role and breadth of actuarial judgement in asset valuation was the polar opposite of that advocated by Arthur and Randall five years earlier. But, as noted above, these differences were perhaps not as irreconcilable as they appeared: different assumptions were required to answer different questions. A rigorous discontinuance solvency assessment would be market- based; an assessment of the amount of contributions required to, on average, meet liability cashflows as they fell due would inevitably require someone somewhere to make an assumption about the prospective long-term risk premia offered by the pension fund’s asset strategy. The two were not theoretically inconsistent or mutually exclusive, they were merely distinct.

Throughout their paper, Dyson and Exley were careful to position their arguments in as inoffensive a manner as possible to actuaries schooled in traditional methods. They were quite prepared to concede that the discounted income methods were appropriate or even impressive at the time they were conceived. Their position was that recent circumstances dictated it was time for the actuarial profession to continue to improve and modernise its methods. An important paper by Exley, Mehta and Smith that followed in 1997

adopted a markedly different tenor.[12] By contrast, its tone barely concealed a relish in smashing all actuarial shibboleths in sight.

Like Dyson and Exley before them, Exley, Mehta and Smith again highlighted the broader range of contemporary applications of pension fund valuations and how these applications were better served by market-based valuation approaches. This paper developed some further sophistication around the market-consistent valuation method for pension liabilities, tackling complexities such as the valuation of the complex non-linear inflation exposures of deferred member liabilities that arose in the form of limited price indexation. The paper also presented some empirical evidence from UK economic data to argue that national real salary growth was fairly stable. This supported the argument that index-linked bonds could be regarded as a good hedge for active liabilities (and hence helpfully avoiding the need to consider equities or other risky assets as a hedging asset within liability matching and valuation).

The main contribution of the paper was, however, in the analysis of the investment strategy of a defined benefit pension scheme from the perspective of the corporate sponsor. Little had been written in actuarial journals on the management of pension funds from the specific perspective of the sponsor. Since the 1980s, however, consulting pension actuaries often formally advised sponsors as well as pension fund trustees. Exley, Mehta and Smith pointed out that from an economic perspective, the sponsor of a defined benefit pension scheme should, in theory, be indifferent to the pension fund’s investment decisions. The assets of the scheme could be considered as off-balance sheet assets of the corporation and, from a shareholder value perspective, bonds with a market value of ?1 had the same economic worth as a portfolio of equities that had a market value of ?1. They argued that shareholder wealth therefore could not be created or destroyed by merely switching from one to the other, in much the same way that Modigliani-Miller showed that shareholder wealth could not be created or destroyed by changing the corporate financial structure of the firm.

However, this conclusion came with some caveats. Firstly, it only strictly applied in the case of a ‘pure’ defined benefit scheme. If, for example, the members participated in discretionary benefits that would be increased in the event of a pension scheme surplus arising but which could not be reduced beyond a floor in the event of a deficit arising, then in economic terms the shareholder was writing a call option to the member. The greater the volatility of the pension fund’s assets relative to liabilities, the greater this option would be worth. In such circumstances, the shareholders’ interest would be served by minimising this volatility, which implied matching the pension liabilities as well as possible.

The second complication was that their argument ignored the pension fund members’ exposure to sponsor default—if the sponsor went bust when there was a deficit, the shareholder would never make good on the pension promises. In economic terms, this could be viewed as a put option that the pension fund member had written to the shareholder. It was a complicated option because its value was a function of the sponsor default probability and the correlation between the default event and the size of the pension fund sur- plus/deficit (as well as the volatility of the pension fund surplus). Nonetheless, it was an option that would increase in value (to the shareholder) as the volatility of the pension fund surplus increased. An increasingly mismatched asset- liability position in the pension fund would therefore increase this option’s value to the shareholder. These two effects were therefore impacted in opposite directions by changes to the pension fund’s degree of asset-liability matching. The specific circumstances of the pension fund and its sponsor would determine which effect dominated at a particular point in time.

The Staple Inn discussion of the paper unsurprisingly included much defence of the traditional actuarial approaches to the management of final salary schemes. A.J. Wise argued that discretionary benefit features such as early retirement pensions remained an important plank of scheme benefits, and more important than suggested by the authors. The ability to reduce such benefits in times of poor equity performance meant that ‘for shareholders ... the downside to equity investment is less onerous when that equity is in the pension fund’[13] (though presumably the shareholder’s exposure to the upside would be similarly dampened).

Thornton pointed out that the shareholder value analysis was not a priority for trustees who were responsible for managing the pension fund assets. He also could not accept the removal of the equity risk premium from the active members’ liability discount rate, though he could not articulate a rational reason why not:

I am suspicious of their arguments that pension fund trustees should not be investing so heavily in equities, and that valuation bases should factor out any excess return expected. This conflicts with economic reality![14]

As ever, ‘economic reality’ meant different things to traditional pension actuaries and the modernisers. The traditionalist view was that a realistic measure of liability cost meant the expected amount of (risky) assets required to fund the cashflows as they fell due; to the moderniser, economic reality was the current market cost of transferring the liabilities to a third party. This basic difference in perspective got lost in a debate of increasing passion that could verge on animosity. Nonetheless, it appeared that the use of market values for asset valuation had now reached close to universal acceptance amongst the actuarial profession. As L.P Tomlinson noted in closing the discussion:

I was astounded by the uniform acceptance of the use of market values. Speaker after speaker endorsed the principle of using market values for assets, and then valuing the liabilities consistently. My actuarial background was in the 1970s when no actuary would endorse market values, so we have moved a very long way since that period.[15]

  • [1] Arthur and Randall (1990).
  • [2] Arthur and Randall (1990), p. 5.
  • [3] Arthur and Randall (1990), p. 6.
  • [4] Sisson, in Discussion, Arthur and Randall (1990), p. 28.
  • [5] Dyson and Exley (1995).
  • [6] Dyson and Exley (1995), p. 489.
  • [7] Clark, in Discussion, Dyson and Exley, p. 543.
  • [8] Urwin, in Discussion, Dyson and Exley, p. 544.
  • [9] Daykin, in Discussion, Dyson and Exley, p. 552.
  • [10] Green, in Discussion, Dyson and Exley (1995), p. 546.
  • [11] Wilkie, in Discussion, Dyson and Exley (1995), p. 549.
  • [12] Exley, Mehta and Smith (1997).
  • [13] Wise, in Discussion, Exley, Mehta and Smith (1997), p. 942.
  • [14] Thornton, in Discussion, Exley, Mehta and Smith (1997), p. 952.
  • [15] Tomlinson, in Discussion, Exley, Mehta and Smith (1997), p. 955.
< Prev   CONTENTS   Source   Next >