Risk sharing and efficiency
The recent consultation on the proposed extension of the Solvency II insurance regulatory Directive to pensions has thrust the issue of the relative efficiencies of differing forms of pension provision design into the limelight. There are material differences in efficiency of pension production arising from the myriad of variants in coverage and structure that exist. Inexplicably this proposed attempt to apply a common regime to insurance and occupational arrangements fails to consider this aspect, which should be the central concern of national and international policy makers.
If we consider a pension simply as an income for life in retirement and use as our measure of efficiency the simple (normalised) ratio of inputs to outputs, the differences in efficiency are stark. Under current circumstances, if it costs £1 in contribution to produce one unit of pension under an assured book-reserve DB arrangement, then to produce one unit using the orthodox UK funded arrangement costs £2, and this same pension unit under individual defined contribution (DC) costs £4.
These figures are perhaps surprising. The scale of some differences may be simply illustrated. For example, as noted earlier, to remove the dependence of a funded DB scheme on the employer sponsor the scheme needs to be funded to the ‘buy-out’ value – currently around 150% of the best estimate of the present value of liabilities. Beyond this, DC has further impediments which are costly; the absence of risk-pooling and risk-sharing among members and economies of scale and scope. The advantages of collective risk-pooling are well known. Funding also brings with it, the costs of management and maintenance of these asset portfolios. Plausible estimates place these costs at between 1% and 2% for collective institutional funds and even higher for individual DC management. If investment returns are of the order of five percent annually, these costs amount to a reduction of 20% – 40% in the capital value of these assets, which will be progressively reflected in their values. This has failed to consider the quality of management of the scheme assets and its effect on experienced investment returns. Nor does it consider the direct costs of regulatory compliance, the additional valuations, covenant reviews and PPF levies.
The prudential regulation of funded DB is remarkably costly. As currently practised and proposed, it even introduces an entirely spurious source of risk, interest rates. However, the principal source of expense is the focus upon scheme funding. This is misguided in that this risk is conditioned on sponsor insolvency and only strictly relevant after that event. This distinction is also very relevant for the EC consultation – for insurance companies, regulation operating on assets and imposing solvency capital requirements is operating on the insurer’s likelihood of insolvency, while for pension schemes it operates on the consequence and increases the primary, unconditional risk source the employer sponsor’s likelihood of insolvency. The relative inefficiency of the proposed EC approach will have costs for the shareholders of sponsor employers which are greater than for the shareholders of insurance companies.
As funded DB and DC enjoy the same tax concessions, their relative efficiency is important. If DC were ever to be offered at the scale necessary to replace the pensions payable under historic DB, the tax cost would be twice as much. With that in mind the complacency with which the decline of DB has been met by the authorities takes some explaining.
The costs of schemes
Over the first decade of the 21st century, employer sponsors and employees made contributions amounting to £340 billion into their defined benefit pension schemes, yet the deficits reported by these schemes are larger than ever. (See Exhibit 9) Asset allocations were radically revised. UK equity now accounts for less than 20% of funds’ assets, but the correlation of fund returns to the FTSE all-share index is almost unchanged.
It seems that few stones have been left unturned in the search for ways to reduce or limit liabilities. We have seen buy-outs, closure to new members and future accrual, pension income exchanges, enhanced transfer value offers and shifts in the basis of inflation indexation from retail to consumer price indexation. Even wage growth has been benign from the employer standpoint, if very uncomfortable for the employee. Notwithstanding all this, reported pension liabilities have been growing at 5.5% p.a. in nominal terms or 2.3% p.a. in real terms.
Some would have us believe that this should be attributed to increasing longevity, but that cannot explain this magnitude of discrepancy. The much-cited two year increase in life expectation in a decade is a growth rate of just 1.8% p.a. (see also Exhibit 21 later)
Employee contributions also present a challenge. Schemes have undoubtedly been closing to new members and future accrual on a large scale. A recent Association of Consulting Actuaries survey reported 91% closed to new members and 37% to future accrual. However, since 1992, employee contributions have fallen, by only 15% in real terms and 29% in average earnings adjusted terms.
Of course, the poor performance of investment funds over the period has been a material contributor to today’s problems. Over the period 2003 – 2010, total assets increased by £211 billion but additional contributions in this period totaled £289 billion. The investment performance was truly dire. (see Exhibit 22 later)
Even more sophisticated and complex investment strategies, such as liability driven investment, have had little observable effect. Liability driven investment itself provides a conundrum; its effect should have been to increase the correlation between changes in asset and liability values, but for the past two years, and more, this has been statistically indistinguishable from zero. (see Exhibit 10) The use of derivatives has increased markedly and they now account for more than 7% of scheme assets, but their net position is very small, accounting for just 0.6% of scheme assets at maximum in 2008. If these are being used in hedging strategies, they have had little effect.
Exhibit 10: Assets, liabilities and rolling 12 month correlation of changes in asset and liability values (PPF, Author’s calculations)
These poor financial market results are in sharp contrast with the performance of the UK private sector, where private non-financial companies have consistently earned net profits greater than 10%, even after the payment of these inflated pension contributions. See Exhibit 4 earlier. Pension contributions are a significant drain on the private sector – some £37 billion in 2010 against net profits of £159 billion. Indeed, the lack of relation between the returns on financial market investments and PNFC earnings, though surprising to many, is perfectly explicable and to be expected. It is also notable that the volatility of listed equity returns is an order of magnitude higher than the volatility of PNFC earnings. The accounting standard now in force brings this volatility into the corporate balance sheet and is a material source of concern for the finance directors of these companies.
An examination of the relation between contributions and pension liabilities is an appropriate analysis. Exhibit 11 shows the evolution of contributions in 1992 pounds, together with an index of pension cost. This index represents the factor cost of pension provision, making allowance for changing longevity, wage and price inflation and the other factors, such as cash commutations, which determine the ultimate pension payable, which in this figure is a two thirds final salary pension with dependent’s benefits. It uses pensioner population longevity rather than national figures and makes allowance for varying levels of wage inflation in differing categories of active employee. Interest rates are not a parameter of this model since they are not a determinant of the pension ultimately payable. This factor cost model is calibrated by reference to the cash pensions actually paid in a period. Exhibit 11 also shows the present value of the factor cost index calculated using an AA corporate bond rate as the discount function
It is evident from examination of this figure that contributions tracked the factor cost rather than the present value of pensions closely until the year 2002. However, in the period since, which coincides
with the introduction of the new FRS17 / IAS 19 accounting standard, it is evident that contributions now reflect the present values rather than the factor costs of pensions. These present values are, of course, dominated by the choice of the discount function.
Regulations are based upon these discounted values rather than the ultimate cost, which makes it worthwhile to consider how much of the increase in pension contributions can be attributed to regulation. In 2010 special contributions, which are almost exclusively a product of regulation amounted to £15.6 billion, some 35% of the total. The more difficult question to address is how much of the increases in ordinary contributions results from regulatory concerns. Exhibit 12 shows the evolution of the PPF liability series together with the evolution of the factor cost index at current prices.
This figure also shows the underlying factor cost of the stated PPF liabilities, that is to say, PPF stated liabilities adjusted for the AA discount rate factor. It is evident that, as the Pension Act 2004 institutions, the Pensions Regulator and PPF, took effect, the reported underlying liabilities of pension schemes have departed from and risen above the factor cost of pension provision. This departure is material amounting to £110-£120 billion in 2010.
A central doctrine of the Pensions Act 2004 is that trustees should be ‘prudent’, but leaves that term undefined. The Regulator, by contrast, asserts that ‘prudent’ means that estimates and assumptions should have a safety margin added or subtracted from the best estimate of those parameter values. It seems that this ‘prudence’ manifests itself in these departures from the true factor cost of pension provision. It is also possible to interpret the aberrant behaviour of 2008 as ‘prudence’ run riot, with trustees taking the opportunity afforded by high corporate bond yields in the crisis to ‘kitchen-sink’ their liabilities. This interpretation is confounded by the liquidity issues discussed earlier but there is supporting evidence for this ‘prudence’ hypothesis in as much as use of a gilt-based discount rate at this time, when it was depressed by a flight to safety, would have rendered the PPF reported liabilities almost equal to the factor cost index estimate.
It is clear that ‘prudence’ in setting scheme parameter assumptions can have effects of the order of magnitude evident. Exhibit 13 shows the cash-flow projections of a scheme under best estimate and ‘prudent’ assumptions. The specific differences between these assumption sets are shown as Table 5. The difference between these cash-flow projections amounts to 14.5% of the best estimate, a little above 10% in present value terms.