FS1/22 – Potential Reforms to Risk Margin and Matching Adjustment within Solvency II

1: Introduction

This Feedback Statement provides a summary of the responses received to the Prudential Regulation Authority’s (PRA) Discussion Paper (DP) 2/22 ‘Potential Reforms to Risk Margin and Matching Adjustment within Solvency II’. The DP, published in April 2022, set out the PRA’s views at that time on some key aspects of potential reforms to Solvency II covered in HM Treasury’s (HMT) April 2022 consultation on the Solvency II Review (the Review).

The PRA supports the objectives of the Review, and worked closely with HMT on potential reforms to the matching adjustment (MA) and risk margin (RM).

The PRA attaches importance to gathering data and other evidence to inform policymaking proposals, including on the Review. To inform this work, the DP sought views on:

• a potential formulation for a Credit Risk Premium (CRP), an allowance for uncertainty around credit risk in the MA framework to correct a weakness in the current design;
• a potential new suggested design and calibration of the RM; and
• the extent to which combined potential reforms to the MA and RM might result in a package that is compatible with the PRA’s statutory objectives.

The aim of this statement is to provide an overall summary of the responses received to the DP, which helped to inform the PRA’s position. It does not seek to provide the PRA’s views on the merits of the different comments in the responses received, but does provide commentary in some cases where the responses appeared to indicate a potential misunderstanding of the points set out in the DP.

This statement summarises the responses to DP2/22 as far as they concern the matters raised in that DP. Respondents’ comments on other policies that are the responsibility of the PRA or the Bank of England that were not discussed in the DP are not covered in this statement. Responses sent to HMT’s separate consultation were subsequently shared with the PRA, but those responses are also not included in this statement.

The responses described in this statement are presented in an anonymised way. Where information about the number of responses is presented, it reflects the PRA’s judgements about how the responses can be grouped into a small number of categories.

The content of this statement is primarily of relevance to the annuity sector. It will also be of interest to other insurance and reinsurance undertakings and industry stakeholders.

The PRA took into account the feedback received to DP2/22 in its discussions with HMT on potential reforms. HMT has since announced its decisions on the Solvency II Review.

Background

DP2/22 explored options for reforming the design and calibration of the RM and the MA. The PRA welcomed comments on all aspects set out in DP2/22. There were deliberately no specific questions laid out in the DP; respondents were able to comment on any aspect of the discussion.

Respondents

The PRA received 22 responses to DP2/22 from PRA-regulated firms, trade bodies, and a range of other stakeholders. These responses – particularly where supported by specific data and evidence – helped to inform the PRA’s further analysis and consideration of the issues covered in the DP. The PRA would like to thank all respondents for engaging with the DP and for their contributions, both in writing and in discussions.

Structure of DP2/22 and FS1/22

Chapters 2, 3, 4, and 5 of DP2/22 related respectively to the MA, the RM, validating the reform packages against transfer values, and the impact of reforms. A separate technical annex explored the MA and reforms to the Fundamental Spread (FS) in more detail. This FS describes the responses largely in that order, but groups the responses to the technical annex with those in chapter 2.

2: Matching Adjustment and reform of the Fundamental Spread

Chapter 2 and the technical annex in DP2/22 covered the PRA’s position on MA and potential reforms to the FS. The DP set out a number of issues that the PRA has with the current design and calibration of the FS, and included some suggestions as to how these concerns could be addressed, in a way compatible with its statutory objectives of safety and soundness and policyholder protection, particularly for the life insurance sector.

On the MA generally, some responses agreed with the PRA’s concerns and challenged the soundness of insurers’ use of the MA as a concept, noting the way in which it allowed insurers to recognise profits before they were earned, and the risks it poses to policyholder security if such profits are not ultimately realised.

Some respondents supported the PRA’s assessment that the FS should be made more risk-sensitive, and agreed with the suggestion to introduce a specific allowance for credit risk uncertainty. Of these, some went further to argue that the PRA’s suggested method to address the concerns it had set out did not go far enough, or that the MA should be removed entirely given the risks associated with recognising unearned profits upfront.

Other responses, particularly those from UK life insurance firms, were generally opposed to reform of the FS. Many of these responses argued the PRA had not provided sufficient evidence to support its concerns. Others acknowledged the case for some reform to the FS, recognising there were limitations in the current approach, but considered any weaknesses in the current FS design and calibration to be less material than the PRA had set out in DP2/22.

One respondent was strongly supportive of moving to an approach to the FS based on credit spreads as a way of reflecting more accurately the credit risks retained by insurers. However, most respondents considered that the method suggested in DP2/22 – the ‘index-spread’ or ‘X/Z’ approach for determining the FS – would be a significant change which would be complex and impractical to implement, or included features which were likely to have adverse impacts on new business and investment. Nevertheless, it was acknowledged by some respondents that the impact on pricing and investment will depend on many factors including risk appetite, investment strategy, yields, matching potential, and availability of assets.

One respondent also questioned whether assets such as infrastructure investments were suitable to back annuity business and stated (although did not provide evidence) that these assets are higher risk and tend to support projects that are often late and above budget. Another respondent considered the MA to incentivise investment decisions that are designed to specifically take advantage of the MA benefit and its ability to improve the immediate solvency position of the firm, but did not include examples of where this had occurred in practice. Related to this, some respondents expressed the view (without accompanying quantitative analysis) that the current FS would not generally incentivise productive investment and there were wider concerns that Solvency II reforms would not unlock capital for productive uses.

Some respondents also raised additional concerns around the appropriateness of the current MA calibration or specification, for example to ensure the long-term nature of investment to back annuities is adequately recognised in the regime, notably (as one respondent suggested) for assets that could become stranded as part of a transition to net-zero.

Where respondents’ comments to the annex on FS reforms disagreed with the suggested approach, these were under the following themes:

• the need for FS reform;
• the suggested index-spread approach;
• the consequences of a material change to the design and calibration of the FS;
• potential alternative solutions; and
• interaction with the Solvency Capital Requirement (SCR).

The need for FS reform

In DP2/22, the PRA expressed three concerns regarding the current FS construct:

• the FS does not capture all retained risks which insurers face and as such its level (in basis points) is generally too low. This is because the FS does not fully and explicitly allow for uncertainty around credit risk;
• the FS is not sensitive to differences in risks across asset classes for a given currency, sector, and Credit Quality Step (CQS); and
• the FS does not adjust to reflect structural shifts in the credit environment over time, unless there are actual defaults or downgrades.

A number of the responses received either did not address these concerns directly or stated that they were unfounded or immaterial without necessarily providing evidence to support this. The majority of respondents – particularly those from life insurers – did not agree with the PRA’s case for reform, with at least one respondent considering the concerns set out to be more theoretical in nature rather than practical issues. For these respondents, the key themes were:

• the current FS was noted as already being materially higher than historic defaults. Many felt it contained sufficient margins for uncertainty and considered the regime as a whole was working well. However, respondents did not generally address the point made in DP2/22, that the FS is not intended to represent a given percentile of the distribution of historic observed defaults nor other points around more recent default experience being somewhat dampened due to actions taken by public bodies;
• the case for the introduction of a Credit Risk Premium (CRP) into the FS was deemed weak. Even if considered appropriate for other investors, it was felt that its measurement would need to be based on the level of uncertainty facing buy-to-hold investors, which was argued to be typically lower than for other investors. That said, some responses highlighted that buy-to-hold investors, including insurers, may nevertheless need to sell assets to rebalance their portfolio, given their investment strategy over the longer term; and
• the academic literature used by the PRA to support its views within DP2/22 was not considered definitive and it was argued that other available work did not support the PRA’s view. One response focused in particular on one of the papers used by the PRA where alternative key judgements and correction of an inconsistency could potentially reduce estimates of an appropriate CRP based on that study. However, responses tended not to go on to consider whether such lower levels of CRP – even if potentially supported by some academic literature – could also be supported by other forms of validation mentioned in DP2/22 such as comparisons to CDS pricing or the 1930s default and downgrade experience.

Conversely, some respondents were strongly supportive of the concerns set out by the PRA in DP2/22 and considered the case for reform to be merited, with some suggesting there was potential for reforms to go further and potentially remove the MA from the regulatory regime, although these respondents did not address the benefits of a well-designed MA to provide incentives to match long-term assets and liabilities.

The suggested ‘index-spread’ approach

To address its three concerns, the PRA considered in DP2/22 that (i) the level of the FS should increase to reflect uncertainty around expected future default losses, (ii) the FS should capture differences in risk and uncertainty between different asset classes, and (iii) the FS should be more sensitive to structural changes in credit conditions without introducing excessive volatility.

DP2/22 set out the case for renaming the ‘Probability of Default’ element of the FS as ‘Expected Loss’, and introducing an explicit CRP component to capture the credit risk uncertainty faced by insurers. On the assumption that asset prices reflect risks and returns, and changing market expectations of those, DP2/22 suggested that one possible way to define the CRP would be as the sum of (i) a proportion (‘X’) of a reference index spread, averaged over a suitable period and subject to appropriate caps and floors and (ii) a proportion (‘Z’) of an individual asset’s excess spread. This suggested CRP design was termed ‘index-spread’ and, although a calibration was not suggested in DP2/22, the PRA provided evidence to support a minimum CRP equivalent to 35% of asset spot spreads on average over time.

Several respondents agreed that spreads could be a useful indicator of credit risk, but stated that spreads also contain a lot of ‘noise’ that is not relevant to a buy-to-hold investor. One respondent considered that insurers are exposed to asset price movements, arguing that these movements reflect the market perception of changes in risk. However, a number of respondents also argued that academic papers did not conclusively show a link between spreads and historical defaults. A direct link to spreads – even with the suggested averaging periods, caps, and floors in the index-spread design – was also viewed by many respondents as introducing too much volatility and potential procyclicality into insurers’ balance sheets which would affect firms’ risk management and investment decisions.

Some respondents argued that an approach based on X solely, or with a small value of the Z parameter, might be an acceptable way of addressing some of the PRA’s concerns. Another respondent argued for a higher Z where spreads are driven by different credit features. But most respondents argued strongly to retain an FS design which did not include a direct link to credit spreads.

Other concerns raised about the index-spread approach included that the:

• X term was over-calibrated;
• Z term penalised non-traded assets when a lot of the excess spread on such assets is true illiquidity. These respondents argued that the introduction of a material Z term might act against the government’s investment objectives, and there may be no suitable reference indices for a number of assets meaning the Z term would introduce volatility in balance sheets over time; and
• averaging period should be longer than the suggested five years to reduce volatility and adverse impacts.

The consequences of a material change to the design and calibration of the FS

Responses on the likely impact of the possible design and calibration of the FS focused largely on: annuity provision and prices; investment in productive assets; and use of reinsurance.

In respect of annuity provision and prices, the majority of respondents shared the view that an index-spread based approach could compress new business volumes due to increased capital intensity. They contended that prices would increase as a result of the higher capital costs, which may result in individual annuities and pension scheme buy-outs becoming more expensive, reducing pension security, and representing a cost to the wider economy. However, one respondent noted that the existence of the MA in itself made annuities appear artificially more affordable, by supporting them with higher-yielding but riskier assets.

The majority of respondents also considered that investment in productive assets would reduce under the suggested ‘index-spread’ approach as a result of both reduced capacity due to lower volumes of new business and reduced investment appetite for them, because penalising higher-spread assets reduces their relative attractiveness. However, some respondents appeared to conflate productive and illiquid assets, and did not explain why it is appropriate to equate productive investment and assets with high spreads. Limited evidence was provided to support that either no change to the current FS would be conducive to productive investment or that any capital released more widely as part of the reforms would be automatically directed by insurers towards more productive assets.

Finally, some respondents considered that the possible FS changes would lead to an increase in the demand for offshoring of credit risk through the use of funded reinsurance.

Potential alternative solutions to address the FS concerns

Some respondents recognised some or all of the PRA’s concerns about the FS set out in DP2/22 and summarised above, even if they did not support the potential change to an ‘index-spread’ design and calibration of the FS suggested in DP2/22. Instead, a number of these respondents suggested alternative solutions to address the PRA’s concerns. The wider themes from these responses are set out below with examples where appropriate.

The most frequently suggested solution, to help address the PRA’s concerns on the current FS design and calibration, was for the PRA to increase the graduations in ratings (notching), as well as to recognise the positive and negative ratings outlooks published by ratings agencies, when setting the FS. Respondents noted that this would increase the granularity of the FS design beyond the current differentiation by CQS, and would also improve its risk sensitivity and responsiveness to changes in credit risk conditions, thereby helping to address two of the PRA’s concerns.

Another frequent suggestion was to recalibrate the existing FS framework, increasing the FS somewhat to ensure it reflected a more appropriate allowance for long-term credit risk. Some examples given were a more in-depth calculation for recovery rates, a revised Long Term Average Spread (LTAS) floor and a higher FS during periods of stress for liquid assets triggered by systemic spread widening. Other suggestions were to include additional allowances for ratings uncertainty, valuation uncertainty, and transition volatility.

The increased use of risk management tools was another commonly-suggested approach. Options included greater use of Pillar 2 (supervisory review process), the Prudent Person Principle, the Own Risk and Solvency Assessment, enhanced asset eligibility criteria (with a sharper focus on internally-rated assets), and the use of additional tools such as MA capital add-ons.

One respondent also suggested that as part of the solution, the RM could be adjusted to capture risks associated with potentially unearned MA.

Interaction with the SCR

In DP2/22, the PRA set out some high-level comments on the SCR, noting that while any change to the FS design would necessitate revisiting the SCR methodology and calibration, the base FS and the SCR need to be considered in terms of whether they are each meeting their own objectives.

The PRA also suggested, in DP2/22, that comparing the allowance for credit risk implied by the FS against actual observed historic experience can be a useful form of validation. The 1930s was cited as a valuable reference point, given it was a period where there was material adverse credit experience without the significant dampening effects of public authority actions that have been witnessed in more recent adverse periods.

DP2/22 set out the PRA’s view that under an ‘index spread’ approach, a CRP of 35% to 55% of the credit spread in base gave a credible coverage of the 1930s experience on the base balance sheet. Some respondents interpreted this incorrectly to imply that the PRA was expecting the base FS to cover 100% of the 1930s default and downgrade loss and were concerned that there would be a double-count with the SCR if that were the case. Although specific numbers were not provided, the reference to ‘credible coverage’ was not intended to imply that 100% coverage of the 1930s was considered appropriate in the base FS. Given the extent of the 1930s default and downgrade losses, validation against this period would generally be expected to cover base and stress in total and in concluding the coverage in base was ‘credible’ the PRA had also taken into account the likely overall level of coverage of the 1930s, ie the likely coverage in total across base and SCR.

In addition to the point on the 1930s, respondents raised two other points in relation to the SCR:

• The SCR should reduce if the base FS is strengthened.
• Changes to the design and calibration of the FS approach in base would also necessitate potentially material changes to internal models that would be resource-intensive and could take considerable time.

On the former, DP2/22 set out that the base FS and the SCR perform different roles and it is important to ensure that they are separately meeting their objectives.

Respondents that referred to the time and resource required to update internal models generally made these points in the context of disagreeing that material changes to the FS in base were required. They also considered the internal model implications to provide further evidence of the challenges that a change to the base FS would cause. These points are particularly relevant to the cost benefit considerations around any potential reforms.

On a more operational level, some respondents noted that there would be a need to ensure adequate time to implement necessary model changes before any reforms went live, and also that there was consistency between the treatment of Standard Formula and Internal Model firms.

3: Reform of the risk margin

Chapter 3 in DP2/22 outlined the PRA’s suggestion for reforming the RM through the modification of the existing cost-of-capital approach,^{footnote [1]} and set out its technical view on the reduction in RM that could form part of a future proposed package of reforms. DP2/22 noted that a calibration of a reformed risk margin design to the edge of what the PRA’s technical analysis could support could result in a reduction to the risk margin of around 60% for long-term life business, under economic conditions prevailing at the time of DP2/22. For non-life business (where the problems with the current design are less pronounced), DP2/22 noted that the PRA saw justification for a reduction of around 30%.

Respondents were generally supportive of modifying the existing cost-of-capital approach. Reasons for this included: ease of implementation; maintaining comparability with the EU regime; and the reasonableness of outcomes for both shorter and longer-term business.

Views were split on the appropriate reduction in the RM, and the impact this would have on policyholder protection. Some respondents noted concerns about the impact on financial resilience of a 60% reduction in the RM for life business, and the consequential impact on policyholder protection.

Other respondents considered that, given other features in the regime that support policyholder protection, there was scope to reduce the RM further than indicated in DP2/22. For example, some respondents noted that protection is provided by: an SCR that is calibrated at a 1-in-200 level; capital management buffers; the UK supervisory framework and process; and the existence of insurer recovery and resolution plans. Respondents also noted that a reduction in the RM would support policyholder protection by reducing the sensitivity of the balance sheet to interest rates and making the RM easier to hedge.

One respondent noted that for Lloyd’s syndicates, any reduction in the RM would have only a second-order effect on capitalisation because the ‘to-ultimate’ basis that they use offsets the RM against the other items in the capital stack to avoid double-counting.

Respondents made comments about the calibration of the level of the RM for life and non-life insurance business. These included comparison to the RM reforms being considered by the EU (from a competitiveness perspective), and suggestions for combinations of parameters to use within the RM formula. Some of these suggestions would lead to a greater RM reduction than was set out in DP2/22; the rationale to support these parameter choices referred back to material that was previously submitted under HMT’s Call for Evidence.

In terms of the tapering parameter, one respondent stated that this should also be applied to non-life business. One respondent put forward some technical considerations for how the value of the tapering parameter might differ by risk type. This respondent considered that the assumptions needed to achieve an RM reduction of 60% or more were unrealistic. One respondent supported the idea of a floor on the tapering parameter so that the RM could be tailored to different business lines.

Some respondents stated that, given recent levels of activity in the longevity reinsurance market, longevity risk should be considered a hedgeable risk^{footnote [2]} and so should be excluded from the RM calculation. However, there was no view expressed around whether the market cost of hedging the longevity risk should be incorporated into the technical provisions.

Several respondents active in the longevity reinsurance market noted that a 60% reduction in the RM would not be sufficient to change their incentives to use longevity reinsurance. Some stated that longevity reinsurance would still be capital-efficient after such a reduction. Respondents further mentioned that decisions on reinsurance are also influenced by firms’ risk management strategies and appetite for retaining longevity risk, as well as by competitive considerations in the annuity market.

4: Validating reform packages against observed transfer values for long-term life business

Chapter 4 in DP2/22 outlined the PRA’s ideas around validating reform packages against data on transfers of insurance business.

A significant number of respondents commented that the RM and MA serve different purposes within the regime and that reforms to these should not be considered in aggregate, but independently.

Some respondents noted that transfer values can at least partly be a function of regulatory requirements, and that this introduces an inherent circularity when seeking to validate regulatory requirements against the value of transfers between insurers.

Some respondents commented that the current design of the RM has a conceptual gap in that it does not include any allowance for the risks associated with the MA assets.

One respondent supported longevity reinsurance prices as being relevant for validating the level of the reformed RM.

5: Impact of reforms on overall capital levels, safety and soundness and investments

Chapter 5 in DP2/22 outlined the PRA’s estimation of the potential capital impact of the package of reforms suggested in the DP, and considered how any capital released from these reforms might be used by insurers.

Respondents generally considered that the PRA should measure the capital impact of the reforms taking into account the effect of these reforms on firms’ transitional measures on technical provisions (TMTP). Some respondents also commented that the PRA’s measurement of a phased in capital impact overstated the RM benefit, due to the assumption that the proportion of reinsured business in firms’ portfolios will remain at year-end 2020 levels, rather than this proportion increasing over time.

A number of responses outlined concern at the potential for a reduction in financial resilience associated with a capital release for insurance firms.

A significant number of respondents commented that, for annuity providers, the combined impact of RM and MA reforms would not result in a capital release in the short term once TMTP was taken into account.

Some respondents also stated that, should a capital release arise from the reforms, deployment of this would be based on firms’ existing capital utilisation frameworks (ie comparing the return on capital across various possible actions). Some respondents specifically asserted that any capital released would pass through to improved pricing of annuity business and greater new business appetite, rather than being distributed to shareholders or used to increase remuneration. Some respondents stated that annuity firms would seek to reinvest in new business to support the long-term growth of the company.

The responses received did not provide detailed quantitative evidence to explain why, in the absence of FS reform, a capital release for life insurers would be targeted to increase productive UK investments rather than the other high spread assets in which life insurers currently invest.

One respondent commented that there would be no assurance that any capital release would be used by life insurers to increase productive investment. They argued that it might instead be used to increase insurers’ dividends or remuneration.