An Analysis of the Pensions Regulator’s Review of impact on DB landscape following (the) LDI Episode: Part 2

Iain Clacher & Con Keating

This is the second of our blogs commenting on TPR’s Report to Work and Pensions Committee on LDI. The first can be found here:   This blog covers the summary, and sections 1 – 28 of the Report.

It is good to see some recognition of the limitations of TPR’s data in their report as it is often represented by many as being definitive and contemporaneous as opposed to lagged, and like all data has limitations and nuance:

“8. Whilst this analysis is likely to be the most comprehensive available, there are data limitations, caused by both legal and practical restrictions. TPR is the only organisation which collects data from all registered DB schemes subject to scheme funding regulations. Summarised scheme valuation data is submitted to us by schemes in deficit on a technical provisions basis at least every three years and membership numbers and asset splits are collected annually in the DB scheme return. However, the level of information provided within the restrictions does not enable us to provide the analysis and results disaggregated at the level requested.”

 And again, in the introductory conclusions of the Report there are explicit statements about the limitations of TPR’s own data.

“27. The analysis presented is based on the most complete data set available in respect of the DB universe, however it should be recognised that there are data limitations, caused by a combination of legal and practical limitations. That said, we are not complacent, and we continue to enhance the data that we collect and build working relationships with other bodies to understand how and where we can share data. We expect to obtain more information on leveraged investments in future annual DB scheme returns and improved information for scheme liabilities through proposed changes in scheme funding regulations. This will enable us to improve our ongoing monitoring and evaluation of scheme funding.”

 However, we are concerned by the presumption (emphasised below) of significant improvement in scheme funding, shown in section 9. There seems to be a dissonance between the nature of data that TPR holds and its acknowledged limitations, the strength of the statement below, and the scale of the LDI crisis. While we recognise that there is a burden and cost to reporting, the magnitude of the LDI crisis resulted in the Bank of England having to step in. It would seem to us that TPR have placed too much emphasis on their existing and imperfect data and models and not enough on being able to say more definitively where things stand today.

Much of this report is the result of modelling by TPR. In this series of blogs, we undertake an extensive review of both the data and the modelling based on that data. In undertaking this review we are concerned that TPR have modelled the world as they would like it to be, rather than as it actually is.

“9. To undertake such a detailed review would require a significant amount of data to be provided by trustees, and would require an extensive amount of resource from trustees, their advisors and TPR to undertake the analysis, such that the level of burden and cost would be disproportionate, especially in the context of the improvements in scheme funding presented in the remainder of this report.”

Section 11 of the report shows the table below:

As we have noted elsewhere, including in our previous note on TPR’s LDI report, the ONS survey of the amount of scheme assets differs markedly from those TPR.

We show the ONS estimates in the first row of Table 1 below. These are considerably different from the TPR numbers presented above. Early in the summer of 2023, TPR provided us with their estimates of the universe of schemes’ assets for the period December 2021 to March 2023; these are shown along with TPR’s asset estimates contained in the review. These are major downward revisions; the differences are shown in Table 2.

Table 1: Asset estimates TPR, PPF and ONS

In Table 2 (below) we show the differences between  PPF and TPR asset estimates, those between ONS and TPR, as well as those between TPR’s earlier estimates and those of the report (contained in TPR’s Table 1 above).

The revisions (Grey row in Table 2) to TPR’s estimates are all downwards and vary substantially in magnitude; from an initial £23 billion to a high of £73 billion in March 2022 and then to a low of £11 billion in December 2022. We can offer no plausible explanation of the pattern of these revisions.

Table 2: Differences in asset estimates

The differences between the ONS asset estimates and TPR’s are very substantial; £140 billion or £151 billion depending upon the TPR estimate considered. This difference continues to grow in the first quarter of 2023, the last period for which ONS estimates have been published. There is a pronounced trend for the difference to increase throughout 2022 and into 2023. This difference with TPR’s most recent value at December 2022 is 10.3% and 12.1% at March 2023.

These TPR asset estimates also vary markedly from the figures reported by the PPF, as shown above in Table 2. In the earlier case, only the March 2022 difference is positive; this appears to arise from a low PPF asset estimate in March 2022. The differences with the most recent TPR data show a pronounced and surprising trend away from the PPF estimate being above TPR by as much as £100 billion to being £90 billion below PPF at December 2023. We would expect the PPF value of assets to lie above TPR’s as buy-in insurance policies are more valuable to the PPF because buy-in insurance policies pay full benefits rather than the reduced benefits of the PPF.  We are at a loss to explain how scheme assets can command a £90 billion premium to their value to the PPF.

There is a further difficulty with liabilities. In early summer 2023, we were also supplied by TPR with their liability estimates for the period December 2021 to March 2023. We show both series for the year of 2022 in Table 3 below. Again, there are substantial differences, which we are at a loss to explain.

Table 3: ONS asset estimates

As can be seen above in Table 3, if we calculate the funding ratio using ONS asset estimates, and the liability estimates of presented by TPR in their LDI Report, the funding ratio scarcely improves – from 105.0% to 105.9%. If we repeat this with the ONS assets and the earlier liability values from summer 2023, we see an improvement in the funding ratio from 98.6% to 102.8%. Note that the revision to liability estimates improved the presented funding ratio by varying proportions, from as little as 3.1% to as much as 9.3%.

These two funding ratios imply that at the end of 2022, 52% of schemes were in balance or surplus based on the liability numbers provided to us in the summer, or 61% using the Report’s liabilities measure – this is very far from the 79% of schemes in surplus contained in the table published in TPR’s LDI Report, and repeated as 80% in the correction to the covering letter of TPR.

The more recent TPR liability estimates also, do not sit well with the PPF’s liability estimates, which fall by 38.8% over the year, while the earlier estimates decline by 35.2% and the most recent by 33.0%.

Table 4 below shows the asset estimates of TPR, PPF and ONS. It is notable that the ONS survey shows an asset decline of 32.5% while the TPR and PPF estimates are respectively declines of 23.8% and 22.5%.

Table 4: Asset values

In looking at the asset declines from TPR, PPF, and ONS: TPR estimate this to be £426 billion, PPF £409 billion, and ONS £591 billion.

This makes the assertions of section 12 suspect in extreme:

“12. At a high level, the results from our modelling show:

1. Scheme funding on a technical provisions basis improved significantly over 2022, with an overall increase in the aggregate funding level from 103% as at 31 December 2021 to 118% as at 31 December 2022, an improvement of funding levels by c.15%.”

 If we use the ONS asset estimates, and use TPR’s recent liability estimates, scheme funding improved from 105% to 105.9%, that is less than 1%.

“ii. This has led to a material change in the number of schemes estimated to be in surplus on a technical provisions basis, from an estimated 52% of the universe (by number of schemes) as at 31 December 2021 to an estimated 79% of the universe as at 31 December 2022.”

 In fact, using the ONS asset estimates, the number of schemes estimated to be in surplus would have only improved from 59% to 61%, if TPR’s most recent liability estimates are used, or from 49% to 52% if the earlier liability estimates are used. These are not the material improvements claimed.

“iii. The primary reason for the improvement in funding levels is due to the fall in the value of liabilities exceeding that of the value of assets.”

 Using the ONS data, assets fell by 32.5%, while liabilities fell by either the Report’s 33.0% or the earlier 35.2%. The improvement is marginal at best.

“iv. We estimate liabilities fell by c. £575 Bn, from an estimated value of £1,734 Bn to £1,161 Bn over 2022, a fall of 33%. The fall in the value of liabilities is primarily due to the increase in gilt yields, which itself is a function of the material fall in the value of gilts over 2022.”

It is indeed true that the present value of liabilities fell in this manner, however the actual liabilities, that is the amount of money schemes will need to pay to meet their pension obligations over time, rose due to the effects of higher than expected inflation.

 “v. We estimate that the value of the assets fell by of c.£425 Bn over 2022, equivalent to a 24% fall in the overall value of scheme assets. This fall in assets is primarily due to the loss in value of gilts, corporate bonds and property.”

It is interesting that there is no mention of leverage or the use of derivatives here. The ONS Financial Survey of Pension Schemes (FSPS) data include both repo and derivatives, which with their demands for collateral were material. This was the central aspect of the crisis spiral, and its omission here is surprising. The ONS estimate of assets will, because of its frequency and timeliness, capture these effects as well as operations such as portfolio rebalancing and sales, which were material over 2022 and into early 2023. For the year 2022, the ONS estimate is a loss of £591 billion. The difference between the ONS and TPR estimates, £166 billion, could fairly be described as a measure of the overall costliness of the LDI strategy over the year.

 “vi. The above figures imply that the overall hedging ratio of the scheme universe was less than 100% of the value of the liabilities, given that liabilities fell by more than the value of assets.”

 The change in the ONS numbers in fact suggest that the scheme universe was in aggregate close to 100% hedged (a 32.5% decline in assets versus the 33.0% decline in liabilities reported by TPR). However, this is the net position, with significant numbers of schemes being over-hedged (by design or otherwise) and some having no hedging at all.

Attempts to measure the degree of hedging from the asset allocation up are confounded by the fact that bonds are a significant part of the traditional diversified (60/40) investment portfolio. In other words, bonds were held by many schemes not pursuing LDI strategies.

The Report continues with:

“13. There has also been an improvement in funding levels on both a buyout and low dependency basis, with the percentage improvement in funding on these levels even greater due to the higher starting liability values and the greater impact that higher gilt yields has on reducing those liabilities. On a buyout basis c.40% of schemes were estimated to be fully funded as at the end of December 2022 compared to less than 10% at 31 December 2021.Further details regarding the alternative funding measures are set out in the section: Scheme Funding over the period 31 December 2021 to 31 December 2022.”

If schemes are funded at either of the TP levels, derived from ONS asset estimates, of 105.9% or 102.8%, that is 61% funded above TP or 52% above TP, it is most unlikely that 40% of schemes would pass the higher threshold needed for a buy-out funding level.

Early in 2023, various consultants reported that 20% or so of their schemes passed the buy-out funding level. It is also worth noting that the buy-out levels reported in the Purple Book at March 2023 are questionable, showing a decline in costs from £2,105.4 to £1,254.8 billion. That is a decline of  40.4%, taking place over a  year in which the estimated spreads above gilt yields have declined (and costs increased) by some 25 basis points for both pensioners in payment and deferred members.

“15. The LDI episode following the September 2022 mini-budget was a short-term phenomenon observed within the wider economic landscape of increasing gilt yields and increased global interest rates seen throughout 2022. As such, even without the LDI episode, our expectations are that much of the improvement of funding seen across the DB universe would likely have occurred over the timeframe covered in this analysis in any event. In other words, the outcome for gilt yields by the end of 2022 (and thereafter) it is supposed would have been broadly the same, irrespective of the LDI episode.”

 This is a novel assertion from TPR. It is almost surely true that gilt yields would have risen to the levels seen recently, but it should be recognised that the analysis of economic wellbeing requires consideration of the path taken to an endpoint. Timing may be all-important. There are many DC members who suffered material losses from the events during the crisis and its immediate aftermath, including some who had to postpone retirement. It should also be noted that investment grade sterling bonds, which are now trading at a yield of around 5.5% were trading at 7.0% during the crisis. There has been no recovery from that fire-sale. As well as this, many DC scheme members who are being life-styled as they approach retirement are also likely to find their retirement outcomes markedly different from their retirement expectations. It is also clear that schemes lost the £3.8 billion profit made by the Bank of England from its operations.

The assertion made by TPR with respect to their perceived improvements in funding amounts to a claim that the entire episode was costless to schemes. This is simply not true. Over the year, explicit leverage using repo had a realised cost of £51 billion and losses of similar magnitude were realised by pooled LDI funds. There were also significant losses from derivatives exposure, a minimum of £38 billion. Indeed, as we have noted before, it is possible to attribute the entire difference between TPR and ONS asset numbers of  £166 billion, to effects not captured by TPR’s models.

The report continues to consider Individual schemes performance over 2022 and states:

“16. Whilst the overall scheme funding universe improved in aggregate, the position will be scheme dependent, primarily based on a combination of the asset strategy adopted, including the level of hedging (and leverage) undertaken combined with the initial funding level.”

This is one of the few unconditionally true statements in the Report and from looking at individual schemes over the past 18 months as we have seen their valuations and annual reports emerge, there is huge variation in outcomes with many schemes experiencing a significant deterioration in their position over the year.

The Report shows the following table of scheme funding levels:

We restate this table using ONS asset values and our models below, as Table 5.

Table 5: Funding levels restated to ONS asset values

We have chosen to display the funding ratios and surpluses using TPR’s most recent liability figures and the ONS asset figures rather than the earlier liability values TPR provided us with, which would show all surpluses to be lower than those presented above in Table 5.

The contrasts with the figures in TPR’s LDI report are stark – a minor improvement in technical provisions. A doubling of schemes funded to buy-out rather than the five-fold improvement in TPR’s table. It is also worth noting that this doubling takes place in buy-out principally as a result of the cost of buyout falling due to interest rate and inflation expectations effects, not improvements in scheme funding. The absence of improvement in the number of schemes funded on a low dependency basis is the most significant effect. As an aside, it is TPR’s belief that 65% of schemes are funded to a self-sufficiency level that has produced the extremely low estimates of costs in the Impact Assessment for DWP’s proposed new Funding Regulations.[1]

“19. The vast majority (87%) of schemes are modelled to have experienced improved funding levels over 2022, with only 13% showing a funding level deterioration on a technical provisions basis. Of those that showed a funding level deterioration, nearly two thirds were either still in surplus at the end of 2022 or had a fall in the size of their deficit over 2022. Therefore, only 5% of the DB scheme universe had both a deterioration in their funding level and either an increase in their existing deficit or a movement from surplus to deficit over 2022.”

 We shall leave a fuller discussion of these assertions for a later blog, as the report further observes:

“23. Furthermore, there are some situations that will have occurred for individual schemes, for which we do not have the data required to model the impact. This includes circumstances such as any discounted selling of assets to meet collateral calls or where schemes lost and reapplied hedges (potentially at inopportune, or even at opportune times). Later in this report we have provided worked examples which set out additional details regarding these specific situations and furthermore explain why we do not believe it reasonable to undertake any further investigation on these specific issues.”

We will however draw attention to one simple arithmetic artefact, the denominator effect.

As discount rates rise the present value of liability estimates fall, which in the absence of asset declines will improve the funding ratio markedly. For example, if a scheme is 130% funded and liabilities decline by 33%, then the funding ratio will rise to 194% (an increase of 1.49 times the initial ratio) in the absence of any decline in asset values.

TPR’s belief that assets fell by just 24% and liabilities by 33% implies that an initial 100% funding ratio should rise to 113.4%. In contrast, under ONS figures that show asset declines of 32.5%, this would imply a funding ratio improvement to just 1.008%.

This section ends with an uncontested statement:

“24. Our modelling confirms that those schemes which were “on-risk”, i.e. those with higher levels of growth assets and lower levels of hedging, observed the greatest improvements over 2022. This is because these schemes have benefited from the significant fall in the value of liabilities linked to higher gilt yields whilst their asset values remained resilient.”

 The Report then moves to Conclusions, the first of which has been contested at length in this blog:

“25. The situation in late September 2022 and early October 2022 bought into sharp focus the size and scale of DB pension schemes’ overall investments in leveraged LDI. However, overall the movement in gilt yields over 2022 led to a significant improvement in DB universe scheme funding over the year.”

The Report correctly ends this section with:

“28. The true impact of the events of 2022 for the whole universe will not be known in full for several years, as each scheme goes through their tri-annual valuation process establishing the ultimate impact on their own scheme funding position. The first schemes required to undertake a valuation post September 2022 are due to complete this work shortly, i.e. primarily those schemes with a valuation date of 31 December 2022. TPR will start to receive valuations for those schemes required to submit a recovery plan to us. From a regulatory perspective, we will continue to engage with trustees on a scheme-by-scheme basis, for whom the valuation poses the greatest risk to savers. We will continue to reflect the updated data we receive from schemes throughout this time, be it through scheme returns, valuation submissions or through third party sources.”

 This expressly recognises the inadequacy of the current data possessed by TPR and PPF.  Against this background, it is difficult to provide a positive view for the position TPR has presented. Models based upon inadequate data can be expected to produce unreliable results. It would seem to us that the reality is that we are not much further along in understanding the impacts of the LDI crisis on DB pensions, and it will be many years and too far gone before we do.

[1] For a review of the Impact Assessment see: https://henrytapper.com/2024/02/10/keating-and-clachers-review-of-the-dwps-proposed-funding-regulations/