DB Pensions Data and Policy Analysis

Iain Clacher & Con Keating

The new Funding Regulations laid before Parliament have an impact assessment which relies upon analysis provided by the Pensions Regulator (TPR). Similarly, the current consultation on options for DB schemes relies heavily on TPR supplied analysis. In this note, in addition to those two sources, we also draw on TPR’s report on LDI to Work and Pensions Select Committee and TPR’s Occupational defined benefit (DB) landscape in the UK 2023.

Both TPR and the Pension Protection Fund (PPF) use data collected by TPR which serve as inputs for their modelling of the current funding position of the universe of DB schemes. TPR has acknowledged that its data is between 2 and 4 years out of date and that its models are applied to this to produce its current estimates. The Purple Book produced by the PPF derives its native input data from that same source.

There can be material differences in the estimates produced by TPR and PPF modelling. For example, TPR reports scheme assets of £1,518 billion at December 2023 but the PPF reports scheme assets of just £1,428.2 billion at that time. TPR’s asset value is £90 billion higher than the PPFs (6.3%). In the period since December 2021, the relative values of TPR’s asset estimates to PPF’s have shown a range from a 6.3% premium to a 6.5% discount (£90 billion higher to £100 billion lower).

Our concerns are not limited to asset values; there also appear to be material differences in liability estimates. For example, the PPF’s purple book reports total buy-out liabilities at March 2023 of £1,255 billion, while TPR, at this date, reports buy-out liabilities as £1,441 billion. This is a monetary difference of £186 billion, 13% of the buy-out liability values. Over the period from December 2021, TPR’s buy-out liability estimates as a proportion of their Technical Provisions (TP) estimates have fallen from 132.3% to 119.8% in March 2023 and 116.0% in December 2023. The PPF’s buy-out liability estimate at March 2023 would be just 4.3% above TPR’s estimate of TP liabilities, which seems highly improbable.

Looking at TPR’s estimate of low-dependency liabilities, which are a critical element of the impact assessment costings, this estimate shows a fall from £2,036 billion in December 2021 to £1,265 billion in December 2023, a 37.9% decline, while TP liabilities have fallen from £1,734 billion to £1,192 billion in December 2023, a decline of 31.3%. The ratio of low-dependency liabilities to TP liabilities has fallen from 117.4% to 106.1%. In monetary terms, the excess cost of low-dependency has fallen from £302 billion to just £73 billion. If the single equivalent discount rate (SEDR) applied to low-dependency projections is gilts plus 50 basis points, and schemes have an average 15-year duration, then schemes’ TPs must have an average SEDR of gilts plus 90 basis points. The additional or excess cost of buy-out has, by TPR’s estimation, fallen from £258 billion to £118 billion.

Unfortunately, these data inconsistencies are not just between TPR and other bodies such as PPF, they extend to differences between TPR’s own estimates reported in different documents.

For example, in TPR’s “Occupational defined benefit (DB) landscape in the UK 2023”, the proportion of schemes in surplus, on a technical provisions basis, are reported as 50.1% at March 2022 and 72.7% in March 2023. However, in TPR’s LDI report, these are shown as 57% in March 2022 and 80% in March 2023.

We report below, as Table 1, TPR’s estimates of liabilities for technical provisions (TPs), low dependency (Lo-Dep) and Buy-Out (BO).

Table 1: TPR liability estimates for period Dec 2021 to September 2023

Source: TPR, Authors’ calculations

We find these figures extremely surprising. To reproduce the relative pricing of buy-out and low-dependency to TPs of December 2021 implies a TP spread of +126 basis points over gilts and a duration of 22 years, while to reproduce the 2023 relative prices implies a spread over gilts of +100 basis points and a duration of 16 years. The combination of such high durations and changes in them, together with a 26 basis point cut in technical provisions spreads, are highly implausible as a representation of the economic reality of schemes.

We show as table 2 below, the excess costs of buy-out and low-dependency above the technical provisions costs.[1]

Table 2: Excess cost of buy-out and low-dependency relative to Technical Provisions

Source: TPR, Authors’ calculations.

The DB options consultation also contains a scene-setting section which we have presented below in italics below, with our commentary underneath the relevant sections.

“24. The Pensions Regulator (TPR) analysis [footnote 4] suggests that of the approximately 5,000 DB schemes, over 3,750 are in surplus on a low dependency basis [footnote 5], with a further 950 schemes approaching surplus [footnote 6] on a low dependency basis. For those schemes in surplus, there was in excess of £225 billion in aggregate surplus assets on a low dependency basis across the DB universe [footnote 7], roughly 17% of total DB assets.”

And the footnotes cited are

“4. Figures provided by TPR with a calculation date as at 30 September 2023.

5. For these purposes low dependency liabilities are calculated using a generic approach of gilts + 50 bps. Other bases might be suitable.

6. Schemes funded over 75% are considered as approaching surplus levels.

7. Based on the net funding position of schemes in surplus, on a buyout basis. Purple Book 2023.”

It is worth noting that the scheme numbers are based on TPR’s low dependency estimate, which has a premium of just 6% relative to Technical Provisions. The £225 billion low-dependency surplus references the PPF Purple Book buy-out estimates, which is odd. The Purple Book does not provide any estimates of schemes in surplus on a buy-out basis and does not cover low-dependency at all. As noted earlier, the Purple Book reports total buy-out liabilities as £1,255 billion, £186 billion lower than TPR’s buy-out liability estimate of £1,441 billion. The aggregate surplus at March 2023 on buy-out using TPR’s estimates of assets and liabilities is a deficit of £26 billion. Using the Purple Book’s estimates, it is a surplus of £160 billion.

TPR has answered some questions on these issues:

“Whilst there are a variety of factors that impact on the movement in the calculation of the estimated liability measures over time, the two key factors over the period from 31 December 2021 to 31 December 2023, which have the greatest impact on our estimates are:

• The relative size of the liabilities, which has had the biggest impact on the differences in £ amounts between the three bases over time due to the large increase in gilt yields; and

• Changes in the premia applied to gilt yields for the different valuation bases changing over time. In particular, we made changes to the discount rate premium used for the calculation of technical provisions and to the premium used to estimate the cost of buyout. These changes impacted the relative differences in the value between the three bases.”

However, these responses do not tell us what changes were made, and do not cover low-dependency at all. Changes to gilt-relative spreads observed over this period would have the effect of increasing buy-out liabilities by some 2.0% to 2.5%.

Other estimates of scheme assets

For some time we have been drawing attention to the lower estimates of pension fund assets arising from the ONS’s quarterly survey of scheme assets, the Financial Survey of Pension Schemes (FSPS). As we have said in many previous commentaries, the results of the FSPS is far more timely than TPR’s base data. There are further indicators that the ONS data is the more likely to be accurate. We show below in Table 3, TPR and ONS estimates of scheme assets for the period December 2021 to September 2023. TPR’s analysis contained within the DB options consultation documents is based on their September 2023 estimates.

Table 3: TPR and ONS private sector DB asset estimates

Source: TPR, Authors’ calculations

There is a near monotonic increase in the differences between TPR and ONS estimates of assets held by schemes. ONS values have declined by £698 billion while TPR’s figures have declined by just £419 billion, a difference of £279 billion. The lower decline in asset values reported by TPR is most surprising when we consider that schemes were leveraged in excess of 1.25 times.

The steady increase in difference would be consistent with our hypothesis that TPR’s figures are failing to capture the effects of leverage, the rebalancing of portfolios and losses realised as a result of the LDI crisis, while these effects are captured and reported in the ONS surveys.

Table 4 offers some basic analysis of the asset and liability values using TPR’s estimate of scheme liabilities. Rather than the dramatic improvement in funding ratios shown by TPR, from 103.3% to 127.4%, the ONS figures show a small decline in funding ratio from 105.0% to 104.3%.

The number of schemes in surplus under the ONS values falls marginally, by around 50 schemes to around 2880. TPR’s figures have this rising from around 2600 to 4,400. The differences by value or by number of schemes are materially different.

Table 4: TP funding ratios and schemes in surplus

Source: TPR, ONS, Author’s calculations

Our earlier reservations with respect to TPR’s liability estimates are also relevant here. The cutting of the discount rate implicit in TPR’s estimate, will tend to overstate the liability value, by an amount of the order of 5.0%. Adjusting the liability values to reflect this would increase the ONS funding ratio to 110%, and the number of schemes in TP surplus to 3,530, 70% of schemes. Our hypothesis here is that TPR’s discount rate of 1% is lower than that being used by schemes in practice.

We next consider buy-out values, in Table 5.

Table 5: Buy-out values.

Source: TPR, ONS, Authors’ calculations

The principal difference we see in buy-out funding levels, relative to the earlier TP changes, is that the improvements are larger; a 30% improvement on buy-out rather than 23.3% for TPs based on TPR’s estimates. However, using ONS assets, this is an 8.8% improvement for buy-out funding, rather than the 0.7% decline of TPs. The improvement in the number of schemes in buy-out surplus is accordingly far less pronounced using ONS asset estimates figures. TPR’s figures have the number of schemes in buy-out surplus increasing from around 400 to some 2,800, while the ONS improvement is from around 450 to around 1,400. This latter increase is sufficient to explain the very substantial increase in buy-out pricing activity which has been widely reported. Note that while TPR reports an overall buy-out surplus of £99 billion, the ONS is still reporting an aggregate buy-out deficit, of £150 billion.

Finally, we consider the low dependency position in Table 6.

Table 6: Low dependency funding levels

Source: TPR, ONS, Authors’ calculations.

Here we see similar effects to those of buy-out, a 33.8% improvement in the funding ratio using TPR’s figures, but just 7.7% by the ONS. Schemes in low-dependency deficit improve from 77% to just 20% using TPR’s figures, but the improvement is far less pronounced using the ONS asset estimates, from 75% to 54%. If TPR is correct there are just 1,000 schemes in low-dependency deficit, however, using ONS, this is  would be 2,700 schemes. These differences are neither small by value nor by number.

If TPR were correct, then there would be effectively no cost to the adoption of low dependency as a regulatory requirement, as is suggested by the impact assessment accompanying the Funding Regulations. In the recent DB options consultation, TPR indicates that schemes in surplus at September 2023 had a surplus of £225 billion, which when combined with the aggregate surplus of schemes above, of £206 billion, would imply that schemes in low-dependency deficit had a deficit of just £19 billion. All else equal, the deficit of schemes in deficit under the ONS figures would be £268 billion.

Conclusions and policy implications

TPR’s analysis was fundamental to the impact assessment of the proposed new Funding regulations. In light of the ONS survey results, there can be no confidence in that assessment. The costs of the legislation look to be extremely high. If the ONS figures prove accurate, this will place a highly significant new funding requirement on sponsors. Clearly, this would run counter to the long-term productive investment policy of government and the Mansion House agreement.

Our own modelling suggests that this cost will be of the order of £150 -£200 billion, rather than the £268 billion above.

The cost of corporate taxes foregone would be of the order of £30-£40 billion. With pressure to shorten deficit repair contributions, spread over five to eight years, this would have an annual cost to the Exchequer of between £3.5 billion and £8 billion.

With this in mind, it would be sound and rational risk management to defer passage of the new funding regulations until such time as the uncertainty in the aggregate funding position of defined benefit pension schemes has been resolved.

Appendix 1

If we first consider the effect of a rise in the gilt yield from 1% to 5% with a fixed margin of 50 basis points, and pensions having a duration of 15 years, we see declines in price (or cost) of the order of 44%, similar to those reported by TPR. (See light purple cells in Table A below.) The table also shows the contributions to the overall discount function from the gilt yield and the spread.

However, a movement in gilt yields of this magnitude would also shorten the duration of the liabilities materially. If the duration of liabilities falls from 15 years to 13 years, then the declines in buy-out price (or cost) are considerably smaller – of the order of 38%. (Dark purple)

Table A: Discount rate effects

Authors’ Calculations

Note the highly significant shift in the relative importance of the spread in the total discount of the liabilities; at yields of one percent, it accounts for 31 percent of the total, while at yields of 5% it accounts for just 6% of the total. One interpretation of this would be that spread is a very poor indicator of prudence.

We would also note that the earnings margin for an insurer declines as gilt yields rise.

For simplicity of presentation here, we have not considered there to be any difference in gilt yields between 13- and 15-year durations. If the 13-year were 10 basis points lower than the 15 year, as might be considered usual, the declines would be respectively 36.9% and 37.7%.

[1] Appendix 1 illustrates the effects of changes in gilt yields with fixed spreads.

Iain Clacher and Con Keating