Guaranteed Annuity Options (1997-2003)
Guaranteed annuity options (GAOs) —options to convert the cash proceeds of a life policy into an annuity at a guaranteed minimum rate—represent a particularly difficult episode of post-war British life office history. The GAO story contains a number of the threads that have already been identified in our above discussions of post-war British life offices: long-term financial guarantees provided under the stress of competition with little initial actuarial consideration for charging and reserving requirements; a lack of actuarial interest in ‘modern’ thinking on economic risk measurement and hedging; an over-reliance on the assumption that the unique features of with-profits could mitigate whatever economic risk was created by guarantees; ultimately, in the face of actuarial prevarication, the intervention of external parties (regulators, and, in this case, the courts) to compel a resolution.
Forms of guaranteed annuity option were written by British life offices from the end of the Second World War, and, to a lesser degree, earlier. Longterm interest rates rose almost continuously between 1950 and 1973 (consol yields increased from 4 % to 17 % over this period) (Fig. 5.1). GAOs were valuable to the policyholder in low interest rate environments and they therefore did not create any financial stress for life offices over the third quarter of the twentieth century. From the outset, however, there was some awareness amongst actuarial thought-leaders that such long-term guarantees were inherently risky for the life office. A particularly notable critic was Dick Gwilt, who was President of the Faculty of Actuaries from 1952 to 1954 and the principal executive of Scottish Widows from 1946 to 1960. Speaking at a faculty meeting in 1948, Gwilt voiced his concerns:
Fig. 5.1 Consol yields (1970-2005)
Some of the guarantees given by offices at the present time  appear to me to be based on somewhat optimistic assumptions about the level of interest rates many years hence. I cannot see any justification for giving options on a basis which may involve offices in serious loss if interest rates are low but do not give a chance of profit if interest rates are high for, in that event, the policyholders take the cash and secure the benefit of the high rates. I feel we ought to keep this matter prominently before us.
A few years later at the 1952 Faculty meeting where Haynes and Kirton’s interest rate risk management paper was presented, Gwilt again voiced his concerns about these long-term options:
I think there is no doubt that many actuaries pay lip-service to the dangers of options, but, unfortunately the stress of competition leads them to grant terms which in the long run may prove to be a serious embarrassment and source of loss. I would go so far as to predict that the time will come when offices will bitterly regret some of the options which have been so light-heartedly granted in recent years .
Gwilt was not the only actuary with such worries about these product features at this time. In the Staple Inn discussion of Redingtons seminal immunisation paper, Haynes commented:
Reserves ought to be set up immediately to meet the chance that the options might become onerous and that the actuary’s assessment of current surplus should be reduced — in some cases very extensively reduced — by reason of the options having been granted ... to my mind, the fundamental answer was to restrict the granting of options to an absolute minimum.
These comments are particularly striking in the context of the life offices’ experience of guaranteed annuity options in the 1990s and 2000s, particularly as such options were written up until the mid-1980s and no reserves were held for them until the 1990s. But they also reflect a more general phenomenon that would recur in post-war British life actuarial experience: actuaries were struggling to keep up with the sales and product development functions of life offices. Whether it was guaranteed annuity options on with-profit policies, or maturity guarantees on unit-linked savings products, or perhaps even with-profit equity asset allocations, the actuarial profession found it difficult to adequately influence their firms’ decision-making in the face of the ‘stress of competition’. In all these cases, some actuarial thought-leadership was evident that could identify concerns and issues with the levels of associated risk being generated, but actuaries’ ability to influence and lead life offices’ corporate decision-making had started to wane.
It was the guaranteed annuity options written in the 1970s and first half of the 1980s that ultimately proved Dick Gwilt correct. At the time they were originally written, these options were generally far out-the-money. Indeed, they were viewed as so out-the-money that they need not be charged for or reserved for in any way. Even in the 1950s, actuaries were aware that out-the-money options had some value to the policyholder. Haynes and Kirton noted in 1952:
In practice, guaranteed options are always “valuable” to the policyholder, even though a margin is retained between the current rate of interest and the interest rate adopted in calculating the option.
This observation, however, did not influence the business practices of British life offices—GAOs were considered sufficiently worthless to the policyholder to be given without charge, yet also to be a product feature that was powerful in driving sales. Guaranteed annuity rates varied, but typically GAOs written in the 1970s and early 1980s provided a guaranteed annuity rate of around 11 % for a level annuity on a male aged 65. The a(55) mortality table typically in use in 1970s life offices implied a break-even long-term interest rate of 6-7 % for the GAO. Consol yields in 1975 were around 15 % and hence the options were deeply out-the-money. However, long-term interest rates had been below 6 % in the early 1960s so there was little reason for the possibility of rates again falling to these levels over a 25-year horizon to be considered completely outlandish.
Two unrelated phenomena emerged over the 1980s and 1990s that transformed GAOS from a deep out-the-money guarantee into a highly material liability. The heady rises in long-term interest rates of the third quarter of the twentieth century were reversed over the final quarter. Long rates fell inexorably from the early 1980s through to the end of the century (and beyond): the UK long-term interest rate fell from 13 % in 1981 to below 5 % by 1999. Secondly, the mortality assumptions of the a(55) table, which was designed for immediate annuitants in 1955, proved wholly inadequate for pricing annuities at the time these contracts matured. Pensioner longevity improved at an exceptional rate over the period from the 1960s to 1990s. For example, the one-year mortality rate for a 65-year-old male in the Permanent Male Assurances (PMA) 92 mortality table was less than half of that in the PMA 68 table. This was largely unanticipated by British actuaries. It had a material impact on GAO liabilities—the break-even interest rate of a 11 % guaranteed annuity rate increased from 6-7 % to 8-9 % when a(55) was replaced with a typical 1990s longevity basis.
What actions were taken by life offices and their actuaries as the doublewhammy of lower long-term interest rates and lower pensioner mortality rates emerged? Most offices stopped writing GAOs on new business during the late 1980s. By this time long-term government bond yields had fallen below 10 % and the prevailing market annuity rate approached the guaranteed annuity rate. Ceasing to write new GAO business could not, however, entirely stop the offices from accepting further GAO liabilities on to their books: the terms of the existing business applied the guarantees to future regular premiums and sometimes to one-off additional single premiums that the policyholder could choose to pay.
By 1993, the combination of further falls in long-term interest rates (by then below 8 %) and lighter pensioner longevity assumptions brought significant volumes of GAOs into-the-money for the first time. Some temporary respite was provided by the short-lived increase in long-term interest rates during 1994-1996 but by 1997 rates had fallen below their 1993 levels. UK insurance regulators were curiously inactive on the subject of GAOs over this period. The scale of life offices’ GAO exposures was perhaps not well- understood by regulators or indeed the life offices and their actuaries at this time. In January 1997, the Life Board of the Institute and Faculty of Actuaries established an Annuity Guarantees Working Party to establish the scale and nature of the potential GAO problem and how life offices were managing or planning to manage it. The terms of reference of the working party noted:
Currently there is no accepted practice for reserving for these guarantees and there is no published research to guide Appointed Actuaries in setting reserves. The DTI [Department of Trade and Industry, the UK insurance solvency regulator at the time] have not published any guidance or regulations specific to annuity guarantees.
The working party published their report in November 1997. It provided an insightful survey and intelligent commentary on the GAO s ituation at
British life offices and the prevailing actuarial practices around it in the key areas of reserving, hedging and bonus policy (including implications for PRE). But it stopped short of providing any recommendations. It was never published in the British Actuarial Journal or presented at an Institute or Faculty sessional meeting. With hindsight it seems a curiously understated contribution from the actuarial profession.
The working party found that around half of British life offices with GAO liabilities were ignoring them entirely in their statutory reserving at the time of writing. The other half reserved for the greater of the cash maturity value and guaranteed annuity value, as implied under the statutory reserving basis. This reserving adjustment was straightforward in the context of GAOs on unit-linked business, but the majority of GAOs were written on with-profits business, and incorporating the GAO into the net premium valuation was less straightforward. As the working party noted:
‘Adding in allowance for annuity guarantee reserves is another twist on top of a
series of potential adjustments to standard net premium reserves.’77
The statutory net premium valuation included a resilience test that required firms to calculate the capital required after some specified stresses to valuation assumptions such as changes in equity values and interest rates. But, as was noted in Chap. 3, the net premium valuation has a perverse sensitivity to interest rate changes. In particular, it understates the economic balance sheet impact of a fall in interest rates as a lower interest rate basis generates a higher net premium. As a result of this, many life offices found they had greater exposure to interest rate rises than falls under the net premium valuation. In those cases, no resilience reserve was required for the contingency that GAO costs would increase due to further falls in long-term interest rates (though that only arose because they were required to hold reserves that incorporated a loading for the imaginary fall in future regular premiums that would arise if interest rates increased). This was an archaic actuarial feature that reflected poorly on the profession’s ability to understand and manage financial risk. The originators of the net premium valuation never intended it to be used in solvency resilience testing, and it is an obviously flawed tool for analysing exposure to future changes in interest rates.
UK long-term interest rates continued to fall substantially over the late 1990s—from 8.0 % in 1996 to 4.9 % in 2000. GAOS were by this time unambiguously in-the-money options. Dick Gwilt’s prediction that such options would become a ‘serious embarrassment and source of loss’ had, half a century later, finally crystallised. GAOs costs were no longer a possible contingency. They had arrived and something had to be done. With-profit funds’ estates could meet some or possibly all of the losses but the general first preference of life offices in the late 1990s was to manage the GAO losses through reductions in with-profit terminal bonus pay-outs. This led to the much vexed question of whose terminal bonuses should be reduced to fund the GAO cost, and how this stacked up against PRE. The simplest approach would be for all current policyholders (both those with and those without GAOs) to receive the same reduction in terminal bonus that applied to the (pre-GAO) cash maturity value. This reduction could be calculated to fund all GAO costs, or the cost could be shared between current policyholders and future policyholders (i.e. the estate). Such an approach may or may not have been deemed consistent with PRE, perhaps depending on the specific constitutional arrangements of the given with-profit fund and the policyholder communications it had made.
However, some life offices, most notably Equitable Life, the venerable institution whose early history we discussed in Chap. 2, pursued a different approach. Equitable Life had become something of an outlier in terms of many life office practices in the second half of the twentieth century. Unlike virtually all other life offices, it reported on a gross premium rather than net premium valuation method; it did not sell any policies through independent advisors but only through a tied salesforce; its appointed actuary was also its chief executive through the 1990s until 1997. Most strikingly, and most importantly for our discussion, it had a policy of not retaining any estate. The idea was that all policyholders would receive an asset share-based pay-out, and hence there would be no need for the office to hold any additional assets beyond the assets accrued from the investment of existing policyholders’ premiums (notwithstanding that the with-profit policies included contractual guarantees whose ultimate costs could exceed the policyholders’ asset shares).
The Equitable believed that they had communicated their asset share-based bonus policy such that ‘the policyholder had effectively and knowingly mandated to the directors absolute discretion over each terminal bonus addition’. The Equitable used this ‘absolute discretion’ to make each individual policyholder bear the final cost of the GAO that they had been granted—that is, at an individual policyholder level, the terminal bonus was reduced so that the post-GAO value of the policy would be equal to the cash maturity value of the with-profit policy if it did not include a GAO. This meant the GAO was irrelevant to the value of the policy pay-out (except in the circumstance where the GAO cost implied the required equalising terminal bonus was negative). This was a fairly striking practice. Importantly from a PRE perspective, it was not a bonus policy that had ever been explicitly communicated to policyholders when the policies were sold, or indeed at any time prior to its implementation.
This bonus policy was questioned by some GAO policyholders through the pensions ombudsman and the Equitable sought clarification from the courts. The case went all the way to the House of Lords, the highest court of appeal in England at the time. In 2000, the law lords found that the Equitable’s ‘differential’ terminal bonus policy was unlawful. The Equitable immediately cut bonuses on all policies to zero in order to try to meet their GAO liabilities. But the Equitable had sailed too close to the wind and was now floundering. It put itself up for sale and closed to all new business in December 2000.
The House of Lords’ judgement and the closure of the Equitable to new business sent shockwaves through the British actuarial profession. Within weeks of the Equitable closing to new business, the Faculty and Institute established a committee of inquiry to review its implications for the role of actuaries in life offices and their actuarial professional guidance. It published its findings in September 2001. It called for a rewriting of key aspects of actuarial professional guidance, and for external peer review of the work of the appointed actuary.
The hedging of guaranteed annuity options through the purchase of derivatives from investment banks started in the late 1990s and received a fillip as the Lords judgement removed other management options. Such hedges were generally financed by with-profit funds’ estates, and they secured the financial safety of the leading British life offices. The hedges cost much more than they would have had they been purchased in 1997, but nonetheless have proven valuable investments in the seemingly interminable environment of falling long-term interest rates of the 2000s.
With the benefit of hindsight, how does the GAO story reflect upon the British actuarial profession? In 2003, David Wilkie co-authored a paper with Heriot-Watt Professor Howard Waters and the doctoral candidate S. Yang that developed some retrospective analysis. They considered how actuaries ought to have thought about charging and reserving for GAOs in 1985, given the state of actuarial thinking at that time. Specifically, they argued that the Maturity Guarantees Working Party paper of 1980, and the Wilkie model presented to the Faculty in 1984, collectively provided actuaries with an adequate conceptual and analytical framework to charge and reserve for GAOs in an appropriate way. Their analysis showed that in 1985 the Wilkie model would have implied that there was a c. 20 % probability of a GAO attached to a 25-year endowment maturing in-the-money. A probability-of- ruin reserving approach would have implied that a reserve of around 10 % of a single premium would be required at a 95 % probability level. Based on the guarantee charging approach used in the Maturity Guarantees Working Party (not an option price but the expected cost plus charge-for-cost-of-cap- ital approach), a charge for the GAO of around 5 % of the single premium would have been justified. There can be little doubt that had such reserving and charging parameters been in place, the result for GAOs would have been similar to that of unit-linked maturity guarantees: a lot less would have been written.
Wilkie, Waters and Yang were, of course, factually correct—a stochastic modelling approach to pricing and reserving for guarantees had been well- established in British actuarial thought by the mid-1980s. However, at this time, these concepts had only been rigorously applied within the domain of unit-linked business. In the 1980s there was no serious backing from the actuarial profession to apply these concepts to reserving for any of the guarantees in with-profits business. Some leading British life offices were undertaking internal stochastic analysis of their with-profit business in the 1980s, but few, if any, were suggesting that statutory reserving for with-profits business be undertaken in the way that had been implemented for unit-linked maturity guarantees.
In short, in the 1980s and 1990s there existed a misplaced confidence within the actuarial profession that with-profits business provided enough risk management levers to the actuary to make stochastic analysis unnecessary, difficult to apply and of limited value. Redington’s 1976 statement that the actuarial analysis of with-profits ‘was a countryside to explore on foot and not by fast car’ is the most colourful expression of this perspective. There is doubtless some truth in Redington’s view: it is easy to substitute computer modelling output for hard thought, and for insight to be lost amongst a plethora of ill-understood random numbers. But, by the 1980s and 1990s, was the actuarial profession doing enough walking of the with-profit countryside? And was the profession guilty, through hubris or fear, of rejecting a whole body of thinking that had emerged in the previous decades which could provide the key to the measurement and management of the increasing financial risks that it was charged with managing? Boyle and Hardy published a GAO actuarial retrospective in 2003 which forcefully concluded:
This entire episode should provide salutary lessons for the actuarial profession.
It is now clear that the profession could have benefited from greater exposure to the paradigms of modern financial economics, to the difference between diver- sifiable and non-diversifiable risk, and to the application of stochastic simulation in asset-liability management ... [this] would have enabled insurers to predict, monitor and manage the exposure under the guarantee.82
It is evident that, particularly during the 1980s and 1990s, there was a strong reticence on the part of the British actuarial profession to adopt or make use of financial economics in their thinking and best practices. With hindsight, these were partly wasted decades for the profession in this respect. On a positive note, after taking their direction from regulators, financial reporting standard-setters and judges, real progress was made in the 2000s in actuarial education, research and practices that started to rectify this damage.
-  Gwilt in Discussion, MacLean (1948), p. 318.
-  Gwilt in Discussion, Haynes and Kirton (1952), p. 213.
-  Haynes, in Discussion, Redington (1952), p. 331.
-  Haynes and Kirton (1952), p. 185.
-  Bolton et al. (1997), Appendix 1.
-  Bolton et al. (1997).
-  Corley et al. (2001), p. 12.
-  Corley et al. (2001).
-  Wilkie et al. (2003).
-  Boyle and Hardy (2003).