Technical Practices Survey 2014

Transcription

1 FINANCIAL SERVICES Technical Practices Survey 214 Risk and Capital Management kpmg.co.uk

2 Contents 1 Foreword 1 2 Executive summary 2 3 Introduction 3 3.1: Objectives 3 3.2: Survey methodology 3 3.3: Topics of interest 4 3.4: Interpretation of the results 4 4 Profile of Respondents 5 4.1: Profile 5 4.2: Size of actuarial function 9 4.3: Which Peak is biting? : Which Pillar is biting? 12 5 Assets : Investment Strategy: Current vs. future Solvency II : Asset transition plan for Solvency II : Strategy for issuer options in light of the Matching Adjustment under Solvency II : Pillar 3 readiness for assets : Challenges faced when investing in alternative assets : Total assets and alternative assets : Rationale for investing in alternative assets 23 6 Solvency I : Base mortality assumptions : Mortality improvement model : Choice of CMI model : Long term improvement factor (CMI model) or underpin (cohort model) : Liquidity premium 29 7 With-Profits 3 7.1: The with-profits landscape : With-profits capital strength : Solvency II and with-profits : With-profits risk management 37 8 Risk Capital : All Risks : Market Risk Modelling : Interest Rate Risk : Credit Spread Risk 57 B

3 8.5: Equity risk : Equity, Property and Interest Rate Volatility : Market Risk Diversification : Credit Default Risk : Insurance Risk : Mortality and longevity risk : Morbidity Risk : Persistency Risk : Mass Lapse Risk : Expense Risk : Liquidity risk : Operational Risk : Aggregation : Capital Fungibility Modelling : Modelling platforms : Economic Scenario Generators : Projecting the balance sheet : Projecting new business Solvency II : Transition from ICA to Solvency II : Solvency II Approach : Profit and Loss Attribution Financial Reporting : Analysis of change : The most important metric to the Board : Experience Analysis : Solvency monitoring : IFRS 4, phase II : Current Embedded Value Reporting : Embedded Value Reporting in the Future : Key differences between Embedded Value Reporting and Solvency II in the Future Tax : Deferred Tax on the Stressed Balance Sheet : Future position and tax modelling : Tax Methodology Acknowledgements List of participants 189

4 1 Foreword As always, it gives me great pleasure to present our report on this year s Technical Practices Survey. Now in its ninth year, our Technical Practices Survey continues to be highly regarded by individuals, respondents and indeed other consultancies as a guide to the range of practices adopted in various areas of UK life actuarial work. John A Jenkins Partner, KPMG LLP The main focus of our Technical Practices Survey remains on ICA, to identify how practices have changed since last year. However, with the greater clarity around the Solvency II implementation date, we have expanded this section to explore a number of current issues relating to Solvency II. Our approach aims to build on the experience of past surveys and deliver more insights into the UK life industry s approach to both the ICA and Solvency II. Working in a top advisory firm, some of the most common questions we have been asked by our clients over the last 12 months relate to ICA and the transition towards Solvency II. Therefore, specific questions on these areas were included, focusing on what other respondents are doing, how they approach certain problems and what best practice (and the range of practices) on certain items appears to be. The survey requires a large investment of resources on our part an investment that we think is well worth the time and effort because of the usefulness of the results. We are grateful to all the respondents who found the time in their busy schedules to take part and would like to extend our thanks to all of you once again. In Section 14 of this report, we have listed the 32 respondents who have contributed to this survey. As I am sure you will agree, the range of firms involved by size and type makes the results set out an excellent indication of the UK life industry s approach to ICA and Solvency II. I hope that if you have not been able to take part in the survey this year that you would be able to do so in 215. I would like to extend a very special thank you to all my colleagues for their hard work in carrying out the survey and compiling this report (details of whom can be found in Section 13), whilst at the same time carrying out their client service responsibilities. I would also like to extend particular thanks to Kim Owen, Simon Hogley, and Jane Parker for their hard work in managing the survey. I believe that you will find this report useful and interesting and look forward to receiving any comments or suggestions you may have on how we can make the questions, analysis or report even more useful or relevant to you. Regards John A Jenkins Partner, KPMG LLP 1

5 2 Executive Summary The purpose of our Technical Practices Survey is to enable UK life insurance firms to identify where the key technical issues lie within the industry, and the range of methodologies and approaches that have been used. In addition to the insight gained from the responses to each of the distinct survey questions, we have observed some overall themes from our analysis of this year s responses. With regard to the calculation of risk capital under the ICA, we have observed that firms approaches are broadly consistent with last year. Other than the specific points noted below, the calibration of stresses and modelling techniques in use has been largely stable. The standard industry approach for applying longevity stresses under the ICA has been to use a run-off approach. As we move towards Solvency II, the industry has been considering whether this approach is still appropriate. The Solvency II Risk Margin should cover the risk from time 1 onwards and so potentially the run-off approach results in double counting the capital requirement. We have observed that a number of respondents are now using an approach based on a 1-year stress, and increased use of this approach is consistent with our expectation as the implementation of Solvency II approaches. We can understand both approaches and we believe that there are challenges to overcome with both of them. We observed movement in firms aggregation methodology, with an increased use of simulation and copula approaches. This movement is also to be expected as firms align to Solvency II, however it is interesting to see this borne out by results. We have increased the granularity with which we asked questions relating to interest rate and credit spread risk, and also relating to correlations. In particular, our questions relating to interest and credit risk have been split by duration in addition to by credit rating. This year is the first time that we have asked respondents to populate a template correlation matrix. We believe that these examples of additional granularity represent an improvement to the survey and will be more useful for firms benchmarking requirements. With regard to Solvency II, we observe that larger Internal Model or Partial Internal Model firms are focussing on Pillar 3 and IMAP, while the focus of Standard Formula firms remains on Pillar 1 valuation and the ORSA. There is still uncertainty over the final form of the regulations, in particular for the Matching Adjustment and Volatility Adjustment, and we have observed that many firms are not intending to apply for the Matching Adjustment or use the Volatility Adjustment. In terms of the overall Solvency II impact, half of respondents indicated that they are worse off under Solvency II (including 11 firms who stated they will be materially worse off). We have observed some movement in firms use or intended use of embedded value. Compared to last year, there are now more firms adopting full Market Consistent Embedded Values or Market Consistent European Embedded Values. Of the firms who currently produce embedded value results, a quarter of firms intend to continue reporting embedded value, just under a quarter of firms intend to discontinue reporting embedded value, and just over half of firms are undecided. 2

6 3 Introduction 3.1 OBJECTIVES The KPMG Life Actuarial team has been carrying out the Technical Practices Survey since 26. This year our aim has been to provide detailed analysis on how the UK life insurance industry has approached the year end 213 ICA process and calculation, and also to provide insight into the main technical challenges firms are facing with regard to Solvency II implementation. 3.2 SURVEY METHODOLOGY The data for this project was collected through a survey that was sent out to respondents for completion in May and June. In order to make the data as representative as possible, almost every UK life office with an internal actuarial function was invited to participate. We attempted to keep the survey to a reasonable length and were hopeful it was not too onerous to complete. For data protection and commercial confidentiality reasons, individual responses have been, and will continue to be, treated with strictest confidence. For the purposes of this report, the results have all been presented in an aggregate format or have been made anonymous. The survey primarily contained multiple choice or numeric response questions. Multiple choice questions typically are quicker to answer than open response questions and so we have used the multiple choice format wherever possible (with a suitable other or not applicable option) so that the survey could be completed in a time-efficient manner. Each year we review the feedback we receive in order to improve the options for these questions. Note that where results are presented as percentage of firms, there are instances where the sum of the separate components may not total exactly 1% due to rounding. 3

7 Introduction We wanted to highlight any common issues that respondents may be having, as well as provide a reflection of the variety of approaches adopted within the industry. 3.3 TOPICS OF INTEREST The survey questions were designed to address the issues that clients have raised over the past 12 months. This year s survey has kept the main area of focus on the ICA to enable comparisons with last year. However, to reflect the greater clarity around the Solvency II implementation date, we also expanded this section to cover some of the main technical challenges firms are facing with regard to Solvency II implementation. As with prior years, the survey also covered other areas where firms have experienced some difficulty with actuarial techniques. We asked questions relating to alternative assets, with-profits, Solvency I, other reporting metrics, and deferred tax assets. We wanted to highlight any common issues that respondents may be having, as well as providing a reflection of the variety of approaches adopted within the industry. 3.4 INTERPRETATION OF THE RESULTS Our survey was targeted at over 6 UK life offices with internal actuarial functions, and we are delighted to have had 32 responses this year. The true test of the survey is that it continues to retain a significant level of interest. It is pleasing to see that this year we have a high response rate despite the burden presented by reporting requirements and their developments. We believe that this reflects that firms participating in the survey continue to find the results to be very useful and relevant. While the number of responses should be considered as very healthy for this type of survey, the response rate alone does not convey all the information about the representative nature of the survey. For example, some individual responses were completed on behalf of all the separate insurance businesses within an individual group. The responses have been from firms of varying sizes which operated in a wide range of markets. Additionally, respondents varied in structure and have included some that were part of larger (often multi-national) groups; others are large in their own right and listed on various European exchanges. When presenting our analysis we have also provided the context for our findings by including a profile of the respondents. A graphical representation of the respondents can be found in Section 4, and a full list of participant is in Section 14 of this report. Most of the major UK life offices have taken part in the survey. 4

8 4 Profile of Respondents In order to set the context for the findings of the survey, this section outlines the profile of survey participants.the profile captures respondents attributes such as size of in-force liabilities in terms of Peak 1 insurance liabilities plus WPICC, ownership status, composition of in-force business by product class and reporting basis. We have also included in this section an analysis of the number of actuarial staff working in first and second line functions. 4.1 PROFILE Our survey received participation from a wide spectrum of respondents in the UK life insurance market, ranging from small to large businesses in terms of their in-force liabilities. We have categorised respondents by size, with reference to the size of their Peak 1 insurance liabilities plus WPICC. Throughout this year s survey, we use the following definition when referring to the size of the respondent: Small: Peak 1 liabilities plus WPICC totalling less than 5m Medium: Peak 1 liabilities plus WPICC totalling more than 5m, but less than 5bn Large: Peak 1 liabilities plus WPICC totalling more than 5bn 5

9 Profile of Respondents Graph 4.1.1: Respondents by size of Peak I liabilities plus WPICC as at 31 December 213 Small Medium 16% Large 53% 31% Graph shows that out of 32 respondents, 16% are small, 31% are medium and 53% are large. The profile of respondents is very similar to that in the 213 survey, hence making it possible to compare trends in responses over recent years. Graph 4.1.2: Respondents by ownership status Listed Company Has listed parents Unlisted company Part of unlisted group Mutual 34% 22% 3% 3% 38% The respondents represent a diverse group by ownership status, as indicated in Graph Most respondents are listed, either directly (22%) or through a parent company (38%). Mutuals comprise 34% of respondents while just 6% of respondents were unlisted. 6

10 Profile of Respondents Graph 4.1.3: Composition of in force business of respondents in terms of Peak 1 liabilities plus WPICC Conventional with-profits business 1% Unitised with-profits business Non-linked non-profit business 1% 13% Unit-linked business Annuities Reinsurance accepted 18% 11% Other 11% 38% Graph shows the average composition of the in-force business by product class in terms of Peak 1 insurance liabilities plus WPICC. Unit-linked business remains the largest product class, as it was in 213, accounting for over a third of the total. The next largest class is annuities, which accounts for 18% of the total. This is again very similar to previous years; however, we might expect to see a gradual decline in this proportion in the future following the changes to retirement income announced in the budget this year. In total, with-profits business comprises around 24% of all business. Other liabilities are largely made up of index-linked products. 7

11 Profile of Respondents Graph 4.1.4: Composition of in force business of respondents in terms of Peak 1 liabilities plus WPICC (split by size of firm) Conventional with-profits business Unitised with-profits business Non-linked non-profit business Unit-linked business Annuities Reinsurance accepted Other 1% 9% 8% 7% 6% 5% 4% 13% 11% 11% 38% 11% 27% 22% 15% 7% 14% 35% 11% 16% 4% 5% 3% 2% 18% 4% 19% 1% % 1% 1% 16% 1% Overall 1% Small Medium Large 1% Further analysis of the results in Graph shows that the composition of liabilities in medium and large firms is broadly consistent with the overall picture in Graph above. However, the composition of Peak 1 liabilities for small firms is significantly different, with none of the small firms responding having any unit-linked or unitised with-profits business, instead having more non-linked non-profit business and reinsurance accepted. 8

12 Profile of Respondents 4.2 SIZE OF ACTUARIAL FUNCTION We asked respondents to provide their number of FTE actuarial staff, by area, ignoring line 1 and line 2 differences and internal structure. Graph 4.2.1: Number of FTE actuarial staff by size of firm Mean number of FTEs Small Medium Large The relationship between the size of the 32 respondents and the number of actuarial staff employed by them is shown in Graph As expected, this shows that, on average, large firms have considerably larger actuarial teams than small and medium sized firms. In the analysis above we have excluded two respondents from the small firms category. These were both reinsurers with significantly larger actuarial teams than other firms in the same category. Reinsurers typically have larger actuarial teams than other firms, reflecting the importance of pricing and portfolio analysis for these firms. Including these responses, the average actuarial team size for a small firm would have been 22 staff, the same as for medium-sized firms. 9

13 Profile of Respondents Graph 4.2.2: Distribution of Actuarial Team Size Number of FTEs Small Medium Large 13 Note that the graph shows as a box and whisker plot the distribution of team size. The box represents the inter-quartile range, the whiskers represent the minimum and maximum survey responses, and the dot represents the median or 5th percentile. Graph shows the range of responses by size of firm and includes all respondents (i.e. including the 2 reinsurance firms that were excluded from Graph 4.2.1).This shows that, although the median team size is marginally larger for small-sized firms than medium,medium-sized firms have a much broader range of team sizes, with the maximum team size for medium-sized firms being 86 actuarial staff, compared to 5 for small firms. The range of team sizes for large firms was by far the widest, with 2 of the 14 large firms having teams in excess of 2. It is worth noting that the wording of this question has been changed significantly since the 213 survey, asking respondents to state the number of actuarial staff by function rather than select from specified ranges covering the actuarial function as a whole. We also explicitly asked respondents to consider staff working in first and second line functions this year. 1

14 Profile of Respondents Graph 4.2.3: Team size by business area Pricing Valuation and financial reporting (including pillar 2) Financial and risk management 1% 8% 2% 48% 27% 16% Other including business support Average split 6% 4% 45% 13% 35% 38% 7% 47% 15% 2% % 22% 6% 28% 21% 12% Overall Small Medium Large When considering the split of resources by function, valuation and financial reporting was the largest area overall, accounting for 45% of actuarial employees on average. However, for small respondents, most staff were employed in pricing activity. This may reflect the nature of the business of these firms as highlighted in Graph 4.1.4, small firms have a greater proportion of reinsurance and annuity liabilities than medium and large respondents. Also of interest is that large firms have a greater proportion of staff working in financial reporting and risk management than smaller firms. This may be a reflection of the resources deployed by some large firms on the development and validation of Internal Models for Solvency II. 11

15 Profile of Respondents 4.3 WHICH PEAK IS BITING? Graph 4.3.1: Which UK PRA Pillar 1 Peak was biting at the 213 year end valuation? UK PRA Peak 1 UK PRA Peak 2 12% N/A (Regulatory reporter) 41% 47% Graph splits respondents by reporting basis and depicts which peak bites for respondents who report on a realistic basis. Of the 32 respondents, 19 reported on a realistic basis and 13 on a regulatory only basis at year end 213. Out of the realistic basis life firms, Peak 1 bites for 4 firms (21%) with Peak 2 biting for the remaining 15 firms (79%). The 213 survey showed that Peak 1 was biting for around 4% of the realistic basis respondents, compared with 21% this year. This is in part due to different respondents between the two surveys, although two respondents have indicated that they switched from Peak 1 biting at the end of 212, to Peak 2 at the end of WHICH PILLAR IS BITING? Graph 4.4.1: Which Pillar will bite as at year end 213? Pillar 1 Pillar 2 41% 59% As shown in Graph 4.4.1, respondents were split 41% to 59% on whether Pillar 1 or Pillar 2 ICA capital requirements were more onerous as at year end 213, respectively. This shows little change from the 213 survey when 6% of firms had Pillar 2 biting. 12

16 5 Assets In this section we consider assets, in particular exploring alternative assets held by insurers as well as investigating asset strategies and planning for a Solvency II environment. Firms who participated in this section hold a combined total of 347bn assets of which 12bn are alternative assets. We are revisiting alternative assets which were a theme of the 213 survey. Insurers have shown a growing interest in these asset types in order to obtain higher yields and diversification in a low interest environment. We explore firms appetite for various alternative assets and how this compares to the results of the 213 survey. Preparation for Solvency II will include considering asset data flows for Pillar 3 reporting requirements and reconsidering asset investment strategies to optimise capital positions, in particular, considering admissibility of assets and eligibility of assets in order to apply for the Matching Adjustment. 13

17 Assets 5.1 INVESTMENT STRATEGY: CURRENT VS FUTURE SOLVENCY II Graph 5.1.1: When setting the investment strategy for your business in the current / future Solvency II environment, rate the following metrics in terms of their importance? Profit return metric: Current 34% 29% 24% 13% Solvency II 37% 24% 25% 14% Capital Measures: Current Solvency II 16% 22% 27% 15% 14% 26% 34% 45% Risk adjusted performance metrics: Current Solvency II 48% 17% 41% 21% 1% 14% 24% 24% ALM / liquidity coverage: Current Solvency II 16% 31% 32% 23% 25% 3% 21% 22% Stress tests: Current Solvency II 24% 41% 31% 31% 28% 31% 7% 7% Projected metrics: Current Solvency II 47% 33% 39% 4% 16% 3% 16% 6% % 1% 2% 3% 4% 5% 6% 7% 8% Percentage of firms 9% 1% Not Considered Considered to a small extent to inform or verify investment strategy in parts of the business Widely used measure for setting investment strategy across most business units Essential measure for setting investment strategy across all business units Graph shows an aggregated view of responses to this question, by grouping the different metrics into higher level headings. In general we can see that metrics used for setting investment strategies have the same level of importance in the current environment when compared to a Solvency II environment. This suggests that firms aren t planning significant changes to the use of their metrics when setting investment strategies in a Solvency II environment. Capital measures are an important investment measure for the majority of firms (on the graph this category has combined responses for regulatory solvency measures, economic capital and risk appetite). In particular, risk appetite had most firms ranking it as important, with 17 out of 29 respondents ranking it as essential and 6 out of 29 ranking it as a widely used in the current environment. A slightly greater level of importance is expected in the Solvency II environment, which may be driven by the encouragement of firms to manage their business using a risk based view. Projected metrics appear not to be an important investment measure for the majority of firms (on the graph this category has combined responses for projected future capital, profit, risk appetite and liquidity metrics). In particular 14 out of 29 respondents say they do not consider projected liquidity requirements and a further 12 say they only consider it to a small extent in the current environment. Only 3 ranked it as a widely used or essential measure. It is no surprise that when asked, these three firms are annuity providers and have a higher proportion relative to their peers invested in alternative assets, which tend to be illiquid by nature. Firms may want to understand their liquidity exposure better so they can explore options in purchasing higher yielding assets via the liquidity premium and avoiding forced asset sales at inopportune times. 14

18 Assets 5.2 ASSET TRANSITION PLAN FOR SOLVENCY II Graph 5.2.1: What is your current position in terms of developing an asset transition plan for Solvency II? Not applicable - we do not intend to change our asset strategy, so no transition plan is required We are awaiting further guidance and/or the outcome of our internal analysis of the Solvency II regulations (e.g. transitional rules) before we decide a formal transition plan 7% 7% 2% We are currently developing our Solvency II asset transition plan We have already developed a draft transition plan 27% Other 4% Firms may want to reconsider their asset strategy in order to comply with Solvency II admissibility requirements and also for the eligibility of assets in order to apply for the Matching Adjustment. The highest proportion, 4% of respondents said that they are awaiting further guidance and/or the outcome of their internal analysis of Solvency II regulations before deciding a formal transition plan. Out of the 2% of firms who do not intend to change their strategy, none of them hold alternative assets. This may suggest they are holding simpler assets types and hence no asset transition plans are required. Only 7% of firms have already developed a draft transition plan. The firms who selected other said that they were already managing their assets based on a Solvency II world but that they will review and reassess when further clarification on Solvency II guidance is issued. 15

19 Assets 5.3 STRATEGY FOR ISSUER OPTIONS IN LIGHT OF THE MATCHING ADJUSTMENT UNDER SOLVENCY II Graph 5.3.1: Please select your firm s current thinking on the approach it will take to allow the Matching Adjustment to be recognised Not applicable - we do not currently and do not intend to hold assets with issuer options Not applicable - we intend to sell our assets with issuer options before Solvency II fully implements No action - we expect that the ongoing lobby / discussions with the regulatory bodies will mean that we will be able to recognise the illiquidity premium without restructuring when Solvency II fully implements No action - we will accept the lower free capital business implications of not being able to recognise the illiquidity premium 3 We intend to create a SPV structure to allow us to recognise the matching adjustment Other 5 1 Under current Solvency II rules, assets with issuer options are not eligible for a Matching Adjustment. As the Matching Adjustment is only applicable to annuity business, we have restricted the results in this graph to the 18 firms who responded and who are also annuity providers. Seven firms responded not applicable as they do not hold or intend to hold assets with issuer options before Solvency II implements. Five firms said they would take no action and will accept the lower free capital business implications of not being able to recognise the Matching Adjustment. Surprisingly four out of these five are large annuity providers with over 5m annuity business each. The cause for this is not clear, but it could possibly be that assets with issuer options are not significant for these firms. Only a small proportion of firms are planning to take action, with 2 respondents saying they intend to create a Special Purpose Vehicle structure to allow them to recognise the Matching Adjustment. If choosing this option, firms will need to weigh up whether the extra capital required in order to get a rating for a Special Purpose Vehicle is at least offset by the capital benefit gained from recognition of the Matching Adjustment. Although only two respondents have said that they plan to take action, we have seen a large increase in companies considering this option in the recent months. Those who selected other said that they were awaiting further guidance from the PRA or they will use a combination of the options in Graph above; i.e. for some assets they will take no action, for others they will use a Special Purpose Vehicle. It is interesting to see that where action is being taken, a Special Purpose Vehicle strategy has been the only one used. Firms will have to be careful if they plan investment restructuring such as setting up a Special Purpose Vehicle structure, in order to obtain a portfolio of eligible assets for the Matching Adjustment. This is because they will also need to demonstrate their compliance with the Directive s requirements for risk management and with the Prudent Person Principle. The latter requires firms to be able to identify, measure, and manage risks within their asset portfolios, to invest in the best interest of all policyholders and beneficiaries, and to only use derivative instruments where they genuinely contribute to a reduction in risk or facilitate efficient portfolio management. 16

20 Assets 5.4 PILLAR 3 FOR READINESS FOR ASSETS Graph 5.4.1: What is the current status of your Solvency II Pillar 3 Reporting for Assets? Overall Developed methodology for separating investment returns between risks Developed process for collection of data Developed plans / framework to ensure control, quality and ownership of data Developed delivery plans to set out process to ensure able to meet tighter reporting deadlines Engaged with external data providers to remedy the gaps Developed tactical and strategic solutions for the sourcing / remediation of data Data gap analysis on the asset forms Number of firms Completed, no further work required Planned for 215 Currently underway Planned for later this year Other Not applicable not relevant for our business The recent confirmation that Solvency II will be introduced on 1 January 216 and the publication by EIOPA in September 213 of its proposals for the preparation of Solvency II, has put renewed pressure on firms to further develop their Pillar 3 plans in the preparatory period and, ultimately, when Solvency II goes live. The first key reporting milestone for firms will be the preparatory QRTs and narrative reporting, with Solo reports due end June 215 and Group reports due in the second week of July 215. The overall readiness bar in graph is calculated as the average of the readiness of firms to the seven listed asset development areas in this graph. Overall, very few firms have completed the Pillar 3 asset development areas mentioned in the question. Only one firm, which is a large firm, has responded to having already completed all areas, whilst no small firms have completed any of the areas mentioned in the question. Only one firm has stated they will start all tasks in 215. The task which is furthest behind in completion is developing methodology for separating investment returns between risks. Only 6 out of the 24 firms for whom this task is applicable have completed or started the task. Seven plan to start later in the year and a further 6 to start in

21 Assets 5.5 CHALLENGES FACED WHEN INVESTING IN ALTERNATIVE ASSETS Graph 5.5.1: When assessing the viability of investing in Alternative Asset Classes and/or performing due diligence on the investments, please select the level of difficulty/challenge the following areas currently pose to your business Developing asset selection criteria 2% 2% 47% 33% 8% 8% Availability of asset / market data Understanding / developing asset modelling and methodologies 12% 4% 46% 18% 6% 7% 12% 47% 22% 1% 6% 1% Regulatory constraints / concerns Developing risk monitoring and governance processes 9% 5% 18% 24% 1% 47% 38% 4% 11% Developing ALM strategy to balance profit vs capital volatility 6% 29% 53% 6% 6% Hedging unwanted risk elements Political and Reputational risk considerations 3% 47% 24% 6% 6% 18% 3% 32% 29% 6% 26% % 1% 2% 3% 4% 5% 6% 7% 8% 9% 1% Political and reputational risk considerations Some significant issues exist Some limited issues but not to the extent they would stop us investing Not an issue, we have a fully developed process We have not yet developed our thinking on this issue Not applicable - we outsource this element Not applicable - we do not invest in assets where this is an issue Graph shows an aggregated view of responses to this question, by grouping the different challenges into higher level headings. Out of the 28 respondents to this question, 11 firms indicated that all of the above areas were not applicable as they do not invest in assets where it is an issue. As such this left 17 firms for which the rest of the analysis is based on. For a given area, roughly half of the firms experience some level of issue, whether limited or significant. The aggregated categories causing the greatest issues are availability of assets / market data and regulatory constraints / concerns. In particular, regulatory constraints / concerns causes an issue for the largest number of firms, with 1 firms indicating they are experiencing limited issues and 2 firms indicating significant issues. As noted above, the analysis in Graph is shown at a grouped level. At an ungrouped level (not shown on the graph) the response for which firms indicated they had the greatest number of significant issues is obtaining relevant market data to assess credit risk calibration, with 3 out of 17 firms selecting this option. We have seen firms having a number of different issues with credit risk calibration data, including the use of US data to model UK exposures, length of the dataset to use (i.e. credibility vs. appropriateness), the extent to which data should be split by duration, and the extent to which data should be split by sector. 18

22 Assets 5.6 TOTAL ASSETS AND ALTERNATIVE ASSETS Total conventional and alternative assets held Graph 5.6.1: Please provide the amount of total assets (conventional and alternative) held in relation to each of the funds Total assets Total alternative assets Bn Bn Bn 8.9 Bn 2.8 Bn 38.1 Bn Annuity business With-profits Protection business Shareholder funds.1 Bn 32 Bn.3 Bn When compared to the 213 survey results, we observe that annuity business and with-profits business remain the funds with the most amounts of alternative assets held. However, out of the 2.8bn of alternative assets held within with-profits funds, 41% of this is held by one firm. As per the 213 survey results, a very small proportion of alternative assets are held within protection business and shareholder funds. This is not surprising since protection business is non illiquid and guaranteed in nature. Alternative assets within the annuity fund and with-profits fund represent 8.6% and 1.7% of total assets within each of these funds respectively in 214. This compares to 6.7% and 2.1% in 213. This suggests that the proportion of alternative assets held has remained fairly stable over the year. However, it should be noted that the sample of respondents for 214 does not entirely match that of

23 Assets Relative amount of alternative assets held Graph 5.6.2: Distribution of alternative asset types held within annuity funds PE/venture capital 3% Infrastructure bonds Infrastructure loans 25% Social housing bonds Social housing loans Covered bonds Hedge funds Residential mortgages Relative % of alternative assets held (%) 2% 15% 1% 15.4% 23.6% 26.7% Commercial mortgages Other structured/securitised assets Other alternative assets 5% %.% 6.9%.5% 9.6% 1.1% 4.5%.% Asset category 11.7% Most alternative assets invested by insurers are held within the annuity funds. We have therefore provided analysis of the amount of different alternative assets held within annuity funds only. As per 213, the most widely held alternative assets within the annuity fund are other structured / securitised assets and other alternative assets, which include equity release and sale and lease back type assets. None of the respondents hold investments in PE / venture capital and hedge funds. In fact, respondents who participated in both the 213 and 214 survey have since switched out of PE / Venture capital assets since 213. Investments in commercial and residential mortgages appear to be more popular. When comparing the firms who responded to both the 213 and 214 survey, investment in these asset types have increased from 347m and 585m to 1,42m and 1,368m respectively. This could be due to firms looking to benefit from a higher yield through holding more illiquid asset types. 2

24 Assets Investment of new money in alternative assets Table 5.6.3: Number of firms who plan to invest a proportion of new money in alternative assets within each fund over the next 12 months Nil % < x 5% 5% < x 1% 1% < x 25% 25% < x 5% 5% < x 1% Annuity Business With- profits 17 2 Protection business 19 Shareholder funds There were a total of 19 respondents to this question. All of these respondents indicated that they did not plan to newly invest in alternative assets within the protection fund and sixteen said the same for the shareholder fund. This is consistent with the low investments in alternative assets within protection and shareholder funds that we are currently seeing. Surprisingly, a high proportion, seventeen out of nineteen said they do not plan to newly invest in alternative assets within the with-profits fund. Out of the six respondents who said they currently hold alternative assets in the with-profits fund, two did not respond to this question, two said they did not plan to newly invest in alternative assets and two said they would. These results suggest there may be less investment in alternative assets within the with-profits funds over the next 12 months. Around half (1 out of 19) of the respondents plan to newly invest in alternative assets within the annuity funds. Seven out of these ten firms said they will invest up to 1% of their new assets within the annuity fund on alternative assets. Only one firm said they will invest more than 5%. Of these ten firms, two do not currently hold alternative assets and have said that they wish to invest 1% and 15% of new money into alternative assets. Interestingly, there are a further three firms who currently hold alternative assets and do not plan to invest new money in these asset types. 21

25 Assets Graph 5.6.4: Likelihood of firms investing in different alternative asset types backing annuity business over the next 12 months PE / Venture Capital Infrastructure Bonds Infrastructure Loans 3 1 Social Housing Bonds Social Housing Loans Covered Bonds Hedge Funds Residential Mortgages 2 2 Commercial Mortgages Other Structured / Securitised Assets Equity release Commercial Loans Number of firms Certain Very likely Slightly likely Moderately likely Table shows that the vast majority of assets to be invested in alternative assets are within annuity funds hence our analysis focuses on the ten respondents who said they plan to newly invest in alternative assets within the annuity fund over the next 12 months. None of the respondents plan on investing in PE/Venture Capital and hedge funds. The most likely alternative assets for insurers to invest in are infrastructure loans and commercial and residential mortgages. Newly invested money in commercial and residential mortgages would be consistent with the investments we are seeing in the current environment. In fact, 7 out of the 1 firms are at least considering investing in commercial mortgages, with 4 of them saying they are certain to invest in this asset type. However, currently a small proportion of total alternative assets (.5%) are currently held in infrastructure loans. Out of the three firms who said they are certain to invest in infrastructure loans, only one of them currently holds this asset type. This may be demonstrating an increased appetite for the purchase of highly illiquid assets in trade off for a higher yield. One firm indicated they were certain to invest in equity release and another one firm was certain to invest in commercial loans. These firms already have current holdings in these asset types and the choices to invest in them may be due to these firms overall strategy as opposed to the attractiveness of these asset types. 22

26 Assets 5.7 RATIONALE FOR INVESTING IN ALTERNATIVE ASSETS Graph 5.7.1: Where appropriate, please indicate the rationale(s) for investment in your alternative assets Yield pick-up 2 Risk/return trade-off Portfolio diversification ALM - duration matching 15 ALM - cash flow matching Expected capital benefit Number of firms We asked insurers to indicate the rationale for their investment in alternative assets and there were 19 respondents to this question. The most popular rationales for investment in alternative assets are portfolio diversification and risk/return trade off, with 16 and 15 respondents respectively selecting these options. Firms with long term illiquid liabilities such as annuities may already have a high exposure to corporate bonds and will not consider liquidity risk as a material one. Hence, there is no surprise these are the most popular rationales since insurers are able to further diversify their portfolio by investing in new alternative assets that also offer higher yields via the liquidity premium. The least widely used rationale for investing is expected capital benefit, with only 6 out of 19 respondents selecting this option, suggesting that the cash flow properties of alternative assets are more valuable to insurers than capital considerations. Five firms selected all six options as a rationale for investing in alternative assets. 23

27 6 Solvency I While the main focus of our Technical Practices Survey is the ICA and Solvency II, we recognise that clients continue to find comparisons of key basis items for Peak 1 and Peak 2 to be very useful. As such we have continued our questions around the Pillar 1 mortality assumptions and introduced a new question on the assumed level of liquidity premium under Peak ANNUITY BASE MORTALITY ASSUMPTIONS We asked respondents which base annuitant mortality table they use for their most material annuity business. Of 28 respondents who answered the question on the Peak 1 basis, 23 (82%) said they use the tables, with the other 5 respondents saying they use other tables. All respondents who answered tables for their base annuity table use gender-specific mortality tables. Of the 17 respondents who answered the question on the Peak 2 basis, 15 (88%) said they use the tables, with the other 2 respondents saying they use other tables. All respondents who answered tables for their base annuity table use gender-specific mortality tables. We also asked firms what table multipliers they apply to their base mortality tables. We received 23 responses to this question, of which all 23 covered Peak 1 and 12 covered Peak 2 bases with the results shown overleaf. For both the Peak 1 bases and the Peak 2 basis, the majority of respondents use table multipliers lower than 1% (both for males and females). On the Peak 1 basis, the average table multiplier is 91.5% for males and 8.5% for females. On the Peak 2 basis, the average table multiplier is 93.7% for males and 9.9% for females. We also compared table multipliers between Peak 1 and Peak 2. Of the 12 respondents on the Peak 2 basis, 8 respondents use higher table multipliers, 2 use the same table multiplier and the remaining 2 use lower table multipliers than their Peak 1 basis. We note that of those respondents who use higher table multipliers for the Peak 2 basis, 5 use lower mortality improvement factors, 2 use the same mortality improvement factors and one respondent said they use higher mortality improvement factors for the PRA Peak 2 basis. 24

28 Solvency I Graph 6.1.1: Mortality table multiplier - Peak 1 Valuation Male Female Number of firms < 8% [8%, 85%) 1 [85%, 9%) 1 [9%, 95%) [95%, 1%) [1%, 15%) [15%, 11%) [11%, 115%) 1 [115%, 12%) 1 12% Graph 6.1.2: Mortality table multiplier - Peak 2 Valuation Number of firms < 8% [8%, 85%) [85%, 9%) [9%, 95%) [95%, 1%) [1%, 15%) [15%, 11%) [11%, 115%) [115%, 12%) 12% 25

29 Solvency I 6.2 MORTALITY IMPROVEMENT MODEL We asked respondents about the mortality improvement model used to calculate their liabilities under Peak 1 and Peak 2. Graph 6.2.1: Peak 1 - Mortality Improvement Models Long cohort CMI 29 12% 4% 4% CMI 211 CMI 212 CMI % Other 28% 28% Graph 6.2.2: Peak 2 - Mortality Improvement Models Long cohort CMI 29 7% 6% CMI 211 CMI 212 2% CMI % Other 4% For both Peak 1 and Peak 2 the majority of respondents use the CMI model. In the 213 survey the majority of respondents used the 211 model; however we can see that over the past year many respondents have moved to either the 212 or 213 models. Three respondents answered other, two of which said they use their own model and the other saying they use average of medium and long cohort models. 26

30 Solvency I 6.3 CHOICE OF CMI MODEL This question considers only those firms who use the CMI models as shown in section out of 19 respondents use the Core CMI mortality improvement model to value their Peak 1 liabilities, and 5 use the Advanced model. 1 out of 14 respondents use the Core CMI mortality improvement model to value their Peak 2 liabilities, and 4 use the Advanced model. Graph 6.3.1: Peak 1 - Core / advanced version of the CMI mortality improvement model Core Advanced 26% 74% Graph 6.3.2: Peak 2 - Core / advanced version of the CMI mortality improvement model 29% 71% 27 These results are in line with our expectation as the Core model is easier to use than the Advanced model given the complexity of calibrating the various parameters in the Advanced model. As expected, out of the five respondents who used the Advanced model, 4 are large insurers and the remaining one is a medium-size reinsurer. The results are similar to those from the 213 survey.

31 Solvency I 6.4 LONG TERM IMPROVEMENT FACTOR (CMI MODEL) OR UNDERPIN (COHORT MODEL) Close to half of the 19 respondents who used the CMI model used a 1.76% to 2% improvement factor for male mortality in their Peak 1 valuation. The responses also suggest that lower improvement factors are typically used for females, ranging from 1.51% to 1.75%. This is in line with the answers received to our survey last year, however we note an overall increase in mortality improvement rates from prior year (15 out of 2 respondents this year use mortality improvements rates for males higher than 1.75%, compared to 14 out of 24 respondents last year). Half of the 12 respondents who used the CMI model for their Peak 2 valuation used a 1.51% to 1.75% improvement factor for male mortality, and close to half of respondents for female mortality. As expected, on average the improvement rates used for the Peak 2 valuation are lower than those for the Peak 1 valuation. Graph 6.4.1: Mortality Improvement Factors - Peak 1 Valuation Male Female Number of firms <1% [1%,1.25%) [1.25%,1.5%) [1.5%,1.75%) [1.75%,2%) 2% Graph 6.4.2: Mortality Improvement Factors - Peak 2 Valuation Number of firms <1% [1%,1.25%) [1.25%,1.5%) [1.5%,1.75%) [1.75%,2%) 2% 28

32 Solvency I 6.5 LIQUIDITY PREMIUM We asked firms what proportion of their credit spread was attributed to the liquidity premium in their valuation interest rates for Peak 1 reporting, by credit rating. We received 22 responses to this question. Of these, 8 applied a flat percentage across all ratings, and 14 applied a percentage that differed by rating. Graph illustrates the range of responses received. Graph 6.5.1: For PRA Peak 1 reporting, what percentage of the spread did you attribute to liquidity when determining your valuation interest rate for the following bond ratings? 1% 8% Percentage of spread 6% 4% 6% 58% 53% 51% 5% 51% 2% % AAA AA A BBB BB Flat% Note that the graph shows as a box and whisker plot the distribution of percentage of spread. The box represents the inter-quartile range, the whiskers represent the minimum and maximum survey responses, and the dot represents the median or 5th percentile. The median percentage of yield spread attributable to the liquidity premium generally decreased by credit rating. For stronger credit ratings (AAA and AA) the lower interquartile was lower than that for weaker credit ratings (A and BBB), except BB. For firms where the responses differed by rating, a broad range of responses was observed. For example, for AAA rated bonds the minimum and maximum percentage were respectively 21% and 95%. For firms with a flat percentage across all ratings, responses were between 21% and 6%. 29

33 7 With-profits Since publication of the last Technical Practices Survey we have seen firms continuing to re-evaluate their strategies with respect to with-profits. The impacts of FSA PS12/4, the Retail Distribution Review and the reduction in customer demand for with-profits products over recent years have caused a number of firms to stop writing new business. The vast majority of with-profits funds are now closed and a significant volume of business is now reaching maturity, driven by the peak in mortgage endowment sales in the late 198 s. This run-off may be accelerated further for funds with significant amounts of pension savings business if customers choose to take advantage of the increased benefit flexibility announced in this year s Budget. These factors mean that the past twelve months have seen a number of firms take further actions to help ensure an efficient run-off of their with-profits business. For example, a number of firms have undertaken transactions to transfer annuity business out of their with-profits funds, either to the firm s non-profit fund or to another provider. Without such measures the ratio of with-profits business to non-profit business within the fund would change rapidly over future years, potentially giving rise to risk exposures that are inconsistent with the expectations of with-profits policyholders and constraining the fund s ability to distribute the estate as it would like. It is likely that firms will continue to explore such options and that, as the withprofits run-off accelerates across the industry, further consolidation will take place between providers. The past year has also seen firms increasingly turn their attention to the treatment of their with-profits business under Solvency II. This poses a number of challenges, in terms of the reporting requirements of the QRTs, the impact of ring-fenced funds and the extent to which management actions are allowed for within the balance sheet. This year s Technical Practices Survey provides excellent coverage of the UK with-profits market. 21 of this year s respondents have with-profits funds and these include all but three of the UK s realistic reporting firms. Given the nature of with-profits business and the definitions used in this survey, the respondents with with-profits funds are primarily large firms (14 respondents) with 6 medium firms and only 1 small firm. 3

34 WITH-PROFITS 7.1 THE WITH-PROFITS LANDSCAPE Number of with-profits funds Graph 7.1.1: How many with-profits funds are there in your firm? Number of funds 1 8 Number of firms >1 Graph shows the number of with-profits funds of each of the respondents. In total, 21 of the 32 participants in this year s survey have at least 1 with-profits fund. Throughout the remainder of this section the analysis considers only these 21 firms. 13 of the 21 respondents with with-profits business have more than one with-profits fund, and on average each firm has 2.9 with-profits funds. This demonstrates the extent to which the with-profits market has already consolidated, with many of the firms with multiple funds having acquired at least one of these through acquisition activity. A small number of firms have been the most active in this area pursuing consolidation strategies, with one respondent having 13 separate with-profits funds. Excluding this respondent, the average funds per firm would have reduced to

35 WITH-PROFITS Number of open and closed with-profits funds Graph 7.1.2: How many of your with-profits funds are closed to new business and in run-off, or open to new business? Closed to new business and in run-off Open to new business 2% 8% Graph shows the number of open and closed funds covered by this year s survey. Only 2% of the funds currently remain open to new business. While this appears to be a significant fall since last year, this is primarily due to a difference in the information requested between the two surveys. The 213 survey did not ask firms to specify how many of their funds were open or closed, asking only whether the main with-profits fund remained open. 32

36 WITH-PROFITS 7.2 WITH-PROFITS CAPITAL STRENGTH Pillar 2 basis Graph 7.2.1: How many of your with-profits funds are self-supporting on a Pillar 2 basis? Do not cover their realistic liabilities Fully cover their own realistic liabilities, but not their ICA & ICG Fully cover their own realistic liabilities, ICA & ICG 7% 1% 83% Graph shows that 83% of the with-profits funds covered by the survey are fully self-supporting in that the assets of the fund are sufficient to cover their own realistic liabilities plus the fund s ICA and any ICG that applies. 17% of the funds covered (1 funds) are not self-supporting. 6 of these cover their realistic liabilities but not the capital requirements while 4 require external support to cover their realistic liabilities. 9 of the 1 funds that are not fully self-supporting are proprietary funds, and therefore are reliant on capital support from other shareholder owned funds. Only one mutual with-profits fund was unable to meet its capital requirements and was receiving capital support from another fund. Due to the timing of the survey it is likely that some respondents will have answered this question based upon year-end 212 results while others will have answered with respect to year-end 213 results. 33

37 WITH-PROFITS Solvency II framework Graph 7.2.2: How many of your with-profits funds do you expect to be selfsupporting under the Solvency II framework? Funds do not cover their BEL Funds fully cover their own BEL, but not their Risk Margin or SCR Funds fully cover their own BEL and Risk Margin, but not their SCR Funds fully cover their own BEL, Risk Margin and SCR 8% 6% 7% 79% Under Solvency II Pillar 1, 48 of the 61 with-profits funds are fully self-supporting and cover their own BEL, Risk Margin and SCR. Of the 13 funds that do not, the split is broadly even 5 do not have sufficient assets to cover their best estimate liabilities, 4 cover their BEL but not the Risk Margin and the remaining 4 cover their Risk Margin but not the SCR. This is broadly the same as the pattern observed in the 213 survey. A comparison with Graph shows that slightly more funds are expected not to be fully self-supporting under Solvency II compared to the current ICAS regime. This is perhaps to be expected given some of the Solvency II requirements. However, it is worth noting that not all of the funds requiring additional capital under the ICAS regime also require capital support to meet their Solvency II capital requirements there are two funds that currently support their Pillar 2 realistic liabilities but not their ICA or ICG which are fully self-supporting under Solvency II. All 13 of the funds requiring additional capital under Solvency II are owned by proprietary firms and 12 of these belong to large firms, the remaining fund belonging to a small firm. It is unclear the extent to which respondents have made allowance within their current Solvency II estimates for the revised long-term guarantees (LTG) package. Furthermore, it is possible that the recently published PRA consultation paper CP 16/14 may also change the position shown in graph

38 WITH-PROFITS 7.3 SOLVENCY II AND WITH-PROFITS Presentation of with-profits funds under the prospective approach required under Solvency II Graph 7.3.1: Are you currently able to present your with-profits funds under the prospective approach required under Solvency II? Yes No - some further development is required 6% No - significant further development is required 44% 5% Graph shows the proportion of respondents that are able to present their with-profits funds under the prospective approach required under Solvency II for QRT reporting. This is a complex area with firms being required to value first the future guaranteed benefits and then any discretionary benefits in excess of this on a prospective basis. The nature of this differs fundamentally from the current Realistic Balance Sheet approach where the liabilities are based on the retrospective asset share plus a value for the cost of guarantees. In addition, firms will need to explicitly value and report items such as future premiums and expenses and show liabilities gross and net of reinsurance. Current models may not readily provide these outputs and, in the case of reinsurance, may implicitly allow for this by considering only the net liability. Half of the respondents indicated that they are currently able to meet the prospective reporting requirements, with 44% recognising that some further system and process development is required to satisfy the QRT requirements. 6% of respondents require significant further development in this area. 35

39 WITH-PROFITS Consideration of with-profits run-off plan as part of the ORSA Graph 7.3.2: Will you be considering your with-profits run-off plan as part of your ORSA going forward? Yes No - we will consider it outside of the ORSA process 37% 63% 16 of the respondents indicated that they have one or more run-off plans in place that will need to be considered and reviewed on an ongoing basis. The analysis shows that there is not a consistent approach across the market with approximately one third of respondents intending to consider this as part of their ORSA process and the remainder intending to address this separately. This is a noticeably different picture to that provided by the 213 survey when 56% of respondents stated that they would consider their run-off plans as part of their ORSA, suggesting that this is an area where firms are continuing to develop their thinking. 36

40 WITH-PROFITS 7.4 WITH-PROFITS RISK MANAGEMENT Structural derisking options Graph 7.4.1: Which of the following structural derisking options are you considering in respect of your with-profits funds? Outsource all/part of the cost base 42% 5% Remove conventional non profit business 6% 41% Reinsurance 28% 6% Buy out any pension scheme deficit 9% 18% Fund restructuring to allow NP sales without WP sales 22% Remove unit linked non profit business 21% Remove subsidiaries from the with-profits fund 11% Other 17% 33% % 5% 1% 15% 2% 25% 3% 35% 4% 45% 5% % of firms with with-profits funds Already undertaken Under active consideration Graph shows the structural derisking options that firms with with-profits funds have either already taken or are considering. This graph excludes any respondents who stated that a particular action was not applicable to their with-profits funds. The most widely used structural action is outsourcing all or part of the with-profits cost base, either to a third party or an internal service company. 42% of respondents have already taken this action with another 5% giving active consideration to this. This is another example of the effects of with-profits run-off, with those responsible for managing with-profits funds using such arrangements to protect with-profits policyholders from the effects of spreading overhead costs across a rapidly reducing in-force book. The second most common action overall is removing non-profit business from the fund. 47% of respondents have either taken this action (6%) or are actively considering this (41%). Although the question did not specifically ask about the type of non-profit business, much of this is likely to be annuities which, given their rate of run-off, otherwise would remain in-force within the fund after the withprofits liabilities have all expired. Fewer firms (21%) were considering removing unit-linked business from the with-profits fund. This perhaps reflects the uncertainty around the value that will be generated by such business, particularly in light of the recent budget changes which mean future persistency experience on unit-linked pensions business is much more of an unknown quantity, and the likely faster run-off compared to annuities. A number of firms either already have or are actively exploring buying-out the withprofits fund s share of any pension scheme deficit. Such actions are likely to increase as the with-profits run-off accelerates and those responsible for managing the fund seek to limit the fund s exposure to what can often be a material risk. 37

41 WITH-PROFITS Finally, 22% of respondents indicated that they are considering changing their fund structure to allow the continued sale of non-profit business, should new sales of withprofits cease. This question was only relevant to mutuals. The actions potentially include making use of the rule waiver outlined in PS 14/5: Response to CP12/38 Mutuality and with-profits funds: a way forward, which would allow mutuals to recognise some of the inherited estate as members capital and to use this to fund other commercial activities. This is an interesting development since, as yet, no firm has used this new option. A number of firms indicated that they have taken, or are considering, other options. These included reviewing existing estate distribution mechanisms and converting with-profits business to a non-profit basis. Number of management actions Graph 7.4.2: Average number of management actions assumed for with-profits business BEL 7 SF SCR 6 IM SCR ORSA ICA EC Average number of management actions Graph shows the average number of management actions that each firm is considering in respect of their with-profits liabilities and capital requirements. As expected, respondents expect to consider more management actions within their capital calculations than within their best-estimate liabilities. Internal model firms expect to use a larger number of management actions in their SCR than those using Standard Formula, reflecting the greater extent to which the Internal Model can reflect the specific features and behaviour of the fund s liabilities. For Partial Internal Model firms, respondents indicated separately the management actions that are allowed for within their Standard Formula and Internal Model SCR components. Ignoring any Partial Internal Model firms, the average number of management actions allowed by Standard Formula firms within their SCR would reduce from 5. to 3.3. Graph shows some differences between the number of management actions that are assumed within each capital measure. However, these primarily reflect differences in the respondents providing information in respect of each measure for example some firms provided responses in respect of their ICA but not economic capital or ORSA. Where individual firms responses covered more than one capital measure, the management actions that were taken account of were typically consistent across metrics. However, 3 respondents indicated that they do not intend to allow for all of the management actions currently assumed within their ICA, when calculating their SCR. Only 1 respondent intends to allow for more management actions in their SCR than they do presently in their ICA. 38

42 WITH-PROFITS Number of management actions Graph 7.4.3: Most common management actions assumed for with-profits business Changes in final bonus rates Market value reductions Changes in regular bonus rates Changes to equity backing ratio Remove misc surplus/planned enhancements Introduce/change guarantee charge Day to day ALM decisions Changes to DB pension scheme Increase in charges for insurance benefits Change in hedging strategy Increases in administration charges Change in new business levels Implement/change outsourcing agreements Other No management actions % 1% 2% 3% 4% 5% 6% 7% 8% 9% 1% Percentage of firms EC ICA ORSA IM SCR SF SCR BEL 39

43 WITH-PROFITS Graph shows the most common management actions indicated by respondents, with those used in each particular component of their liabilities and capital requirements shown separately. Allowing for changes to final bonus rates is the most common management action with all respondents intending to allow for this in their SCR, ORSA and economic capital. Only 1 respondent does not allow for this in their ICA, and all but 3 intend to reflect this in their BEL. Changes to MVRs and regular bonus rates followed closely behind this, with respondents assuming identical treatment of these in their economic capital, ICA, ORSA and SCR. One respondent allowed for MVR reductions but not changes to their regular bonus rates in their BEL, but other wise the responses were identical. Most firms allow for changes to the EBR, but the allowance for this varies between components. It is most commonly used in Internal Model SCRs (82%) and economic capital (73%), with just 44% of respondents allowing for this within their BEL. Respondents allowed for the removal of past miscellaneous surplus or planned enhancements and the introduction of guarantee charges in similar ways for most measures. Around 65% of respondents allowed for each of these in their economic capital, ORSA and SCR, with 55% reflecting these in their ICA. Firms were more likely to allow for the removal of past miscellaneous surplus or planned enhancements within their BEL (39% of respondents) than increases to guarantee charges (22% of respondents). Around 4% of respondents allow for day-to-day ALM decisions such as rebalancing the asset mix or duration matching liabilities within their BEL, Standard Formula SCR, ORSA & ICA, with around 55% reflecting this in their Internal Model SCR and economic capital. Respondents indicated that a number of other management actions were also considered but to a much lesser degree. These included increasing administration and risk benefit charges, limiting new business levels and changing their hedging strategy. 4

44 8 Risk Capital 8.1 ALL RISKS Method used to determine 1-in-2 marginal stresses Graph 8.1.1: Which method do you primarily use to determine your 1-in-2 marginal stresses for the following risks? Fit distributions directly to internal historical data Fit distributions directly to external historical data External advice Judgement Industry data Other Percentage of firms % 61% 4% 7% 14% 11% Equity/ Property 56% 4% 4% 15% 22% Interest rate 52% 4% 8% 16% 2% Credit spread 52% 11% 19% 19% Defaults 43% 11% 14% 18% 14% Mortality 19% 19% 4% 26% 19% 15% Longevity 39% 4% 4% 29% 18% 7% Persistency 24% 41% 17% 17% Expense 17% 3% 55% 3% 21% Operational 15% 8% 38% 8% 31% Liquidity 41

45 RISK CAPITA L We asked firms to state the methods they use to determine their 1-in-2 marginal stresses for a number of risks. As expected, most firms relied primarily on external data to set their market 1-in-2 stresses. This was in part due to the wide availability of good quality external data for market risks. Other responses also indicated the use of internal as well as industry data, expert judgement or other methods to set their 1-in-2 stresses for the market risks. The other responses included the use of forward looking economic scenarios to calibrate the interest rate stresses, the use of the modified Merton model for default risk and the use of a stochastic model with parameters determined primarily using expert judgement. A number of firms also use the Standard Formula to set their 1-in 2 stresses for market risks. For the non-economic stresses, judgement plays the most significant part in the calibration. A large number of firms also use historic data, mainly their own internal experience data, to derive the stresses. As expected, external historical data is less widely used, as such data is less readily available (other than for mortality and longevity where population statistics are available). The other responses included using the QIS5 stresses and scenario analysis for expense risk. 42

46 RISK CAPITAL Allocation of final risk capital Graph 8.1.2a: What percentage of your final risk capital is allocated to the following risks pre-diversification? Market (including credit spread risk) Credit default Insurance Operational risk Liquidity Other Percentage of final risk capital 1% 9% 8% 7% 6% 5% 4% 3% 48% 3% 36% 21% 1% 74% 46% 8% 29% 48% 3% 35% 2% 1% % 7% 6% All Small 3% 11% 7% Medium 7% 7% Large We asked firms to provide their pre and post-diversification capital risk allocation. Please note that Graph 8.1.2a reflects the aggregate allocation of risks across the industry. Pre-diversification, the largest proportions of the final risk capital requirement were allocated to market risk (48%) and insurance risk (36%). This is not unexpected given the fundamental nature of the protection and savings products. The other risks (6%) indicated by respondents included group, counterparty, new business and closure, basis, pension scheme and tax risks. Some insurers have large pension schemes and hence pension scheme risk may be a significant contributor to other. However, it is possible that some firms have captured the impact of various risks in the pension scheme within the risks themselves. Operational risk formed a significant component of firms capital requirements, with medium size firms having the largest proportion, while small firms have the smallest proportion. Liquidity risk amounted to less than.1% of total risk capital. The majority of firms take the view that liquidity risk is better mitigated by means other than holding risk capital. Only 2 out of 25 firms indicated that they held capital for liquidity risk. The difference between the small, medium and large firm groupings is likely due to the mix of business written in the specific firms as well as their size. Looking at the size of the respondents, small firms allocated more risk capital to insurance risk (75%) and market risk (21%) and less to the other risks. Medium sized firms allocated less risk capital to insurance risk (29%) relative to small and large firms. 43

47 RISK CAPITAL Graph 8.1.2b: What percentage of your final risk capital is allocated to the following risks post-diversification? Market (including credit spread risk) Credit default Insurance 11% 1% 9% 16% Operational risk 8% 2% 45% Liquidity Other Percentage of final risk capital 7% 6% 5% 4% 3% 2% 1% % -1% 59% 5% 24% 7% 5% All 7% 81% 31% 1% 3% 6% -2% Small Medium Percentage of firms 61% 4% 23% 7% 5% Large Please note that Graph 8.1.2b reflects the aggregate allocation of risks across the industry. Post-diversification, the allocation of risk capital to the various risk drivers broadly mirrors the pre-diversification figures. Overall, the results show an overall reduction of insurance risk capital and an increase of all other risk capital. This suggests that respondents observed less diversification of market risks than they observe with insurance risks. It is interesting to note that the proportion of capital held in respect of operational risk has not changed substantially post diversification. It is also interesting to note that for small insurers the other risks had a small negative contribution to the post diversification capital requirement suggesting negative correlation between the capital requirements between other risks and the market, credit, insurance, operational and liquidity risks for some respondents. The firms with a negative contribution from other risks did not indicate what these risks represented. Please note section 8.17 (Aggregation) on the correlation assumptions used by firms responding to this survey. 44

48 RISK CAPITA L Asset stressing Graph 8.1.3: Which of the following best describes where you perform your asset stressing? Within aggregation model Within actuarial valuation model Provided by asset management tool 3% 13% Calculated in a spreadsheet Other 17% 6% 7% We asked firms to state the primary tools they used to perform their asset stressing. The use of spreadsheets is the prevailing method (6%) used to perform asset stresses, followed by performing the stresses in the actuarial valuation model (17%). Not surprisingly, smaller firms are more likely to use spreadsheets to perform their asset stressing than their larger counterparts. All small respondents use spreadsheets to perform their asset stressing, while the percentage of respondents using spreadsheets for medium and large firms is 78% and 47% respectively. 45

49 RISK CAPITA L Challenging risks to model In all of our surveys since 211, we asked participants which of the risks they found most challenging to model. As in previous years, respondents were asked to select all that apply. Table 8.1.4: Which risks are you finding the most challenging to model? Number of % of % of respondents respondents respondents (214) (214) (213) Market Spread risk 16 53% 49% Interest rate risk 12 4% 43% Credit default risk 8 27% NA Currency risk 2 7% 9% Swap-Gilt Spread risk 1 3% NA Equity risk 1 3% 6% Property risk 1 3% 3% Concentration risk 1 3% 11% Implied interest volatility risk % NA Implied equity volatility risk 1 3% NA Implied property volatility risk % NA Life Longevity risk 12 4% 29% Lapse risk 7 23% 29% Expense risk 4 13% 2% Catastrophe risk 3 1% 9% Mortality risk 1 3% 6% Disability / morbidity risk 1 3% 14% Revision risk % 3% Health SLT Expense risk % 3% Mortality risk % % Longevity risk % % Revision risk % % Health Non- Catastrophe risk 1 3% 6% SLT Lapse risk % 3% Premium and expense risk % % Non-life Premium and reserve risk % % Reserve risk % % Catastrophe risk % % Other Operational risk 14 47% 54% 46

50 RISK CAPITA L Consistent with the observations in last year s survey, the risks that firms find most challenging to model are operational risk, spread risk, interest rate risk and longevity risk. This reflects the use of new and more complex methodologies as firms align their ICA and their Solvency II methodologies, in particular for Internal Model firms. Responses for other risks were consistent with responses from the 213 survey, with marginal improvements perhaps indicating that a few more firms were overcoming the challenges they previously faced in modelling these risks. In general the responses indicated that market risks were more onerous to model than life insurance risks. Operational risk was also singled out as a particularly challenging risk to model. Market Risk The market risk category includes interest rate, equity, property, credit spread, currency, concentration and counter cyclical premium risks. In line with the last year s survey, spread risk and interest rate risk were listed as the most challenging to model with 53% of respondents selecting spread risk and 4% selecting interest rate risk. For the interest rate stress, most firms use principal component analysis (see section 8.3 on methods used to model interest rate risk) in order to set the shocks. We have seen firms apply expert judgement to the principal component analysis calibrations, as stresses calibrated to historical data might not be severe enough to produce appropriate shocks. Expert judgement overlays may be difficult to calibrate, which may be a source of challenge for many firms. In addition, extrapolation methods around the interest rate curve have received some attention recently, prompted by the Long Term Guarantees Assessment and adding to the challenges faced by the respondents. 27% of respondents stated that they found credit default to be challenging to model. This was a new question in this year s survey. Insurance and Operational Risk The proportion of firms indicating longevity risk as challenging to model has risen to 4% compared to 29% in last year s survey. 12 firms indicated that they found longevity risk challenging to model compared to 1 in last year s survey. See section 8.1 on the discussion of industry debate around longevity risk, which provides some indication on why firms may be finding longevity risk challenging to model. The percentage of firms finding operational risks difficult to model is consistent with previous findings, and is marginally reduced from 54% last year to 47% this year. We saw a large increase in the percentage of firms finding operational risk modelling challenging in last year s survey. This reflected the firms adoption of more complex operational risk modelling methodologies in order to overcome some of the limitations of more simple approaches. The responses to the other questions in this section were either consistent with last year s responses, or showed a decrease in the proportion of firms finding other insurance risks challenging to model. 47

51 RISK CAPITAL 8.2 MARKET RISK MODELLING Approach to modelling market risk losses Graph 8.2.1a: What technique best describes your approach to modelling market risk loss on assets? Curve fitting Replicating portfolios Least squares Monte Carlo 23% 1% 3% Direct evaluation Other 64% Graph 8.2.1b: What technique best describes your approach to modelling market risk loss on liabilities? Curve fitting Replicating portfolios Least squares Monte Carlo Direct evaluation 23% 23% Other 6% 3% 45% 48

52 RISK CAPITA L We asked firms to state their primary approach to modelling market risk losses on assets and liabilities. Of the firms responding to the survey, 45% report using the direct evaluation approach for their liability modelling. This remains the most widely used method of evaluating the values of liabilities under stress and in the main reflects the proportion of firms using the stress and correlation matrix approach for the ICA. All of the firms modelling using the direct evaluation approach for liability modelling also use the direct evaluation approach for the assets. A significant proportion (32%) of respondents reported using either curve fitting, replicating portfolio or least squares Monte Carlo for liability stressing. Of these firms, 9% used curve fitting for their asset stressing, 16% used the direct evaluation method, 3% used Least Squares Monte Carlo, and 3% reported using other methods. Of the remaining firms, 18% reported other for both assets and liabilities, while 3% reported other for liabilities, and direct evaluation for assets. The range of approaches used by firms for modelling their market risks reflect the general approaches they are now using for their ICA. Internal model firms, who are usually the larger firms, have adopted simulation based methods for calculating their ICA. They largely use either curve fitting, replicating portfolio or Least Squares Monte Carlo for their capital calculation. 49

53 RISK CAPITA L Tools used for replicating portfolios, curve fitting or Least Squares Monte Carlo (LSMC) Graph 8.2.2: If you use replicating portfolios, curve fitting or Least Squares Monto Carlo as part of your model, what tools do you use? RiskAgility Internally developed Excel based tool 15% Externally developed Excel based tool Algorithmics 23% 46% 16% We asked firms to state which tools they primarily used for replicating portfolio, curve fitting or for Least Squares Monte Carlo. RiskAgility is the most widely used proxy model tool used by the respondents. All but one firm (which uses RiskAgility for replicating portfolios along with an internally developed Excel tool) use it for curve fitting. One firm uses RiskAgility for both curve fitting and Least Squares Monte Carlo. A key component of the Least Squares Monte Carlo technique is the ability to produce nested stochastic scenarios during calibration, so typically we would expect to see firms using a fitting / calibration tool alongside suitable Economic Scenario Generator software, such as the B&H Liability Proxy Generator. Of the two firms using Algorithmics, one uses it for replicating portfolios, and the other uses it for curve fitting. Internally and externally developed Excel based tools are also widely used amongst respondents, with four firms using them for curve fitting, one using alongside RiskAgility for replicating portfolios (as also described above) and one using alongside SMART for replicating portfolios. One of the challenges for firms using internally developed software, particularly when considered in the context of Solvency II regulations, is ensuring that robust controls are in place. Under Solvency II, users of externally developed software will need to demonstrate a deep understanding of the models which will include the development of their own documentation. Whilst we understand that a number of firms have established processes in place and have already chosen their modelling frameworks for Solvency II, it will be interesting to see the regulatory challenges applied dependent on the type of modelling framework that firms have adopted (in particular internal vs. external). 5

54 RISK CAPITA L 8.3 INTEREST RATE RISK Use of a relative or absolute shock Graph 8.3.1: Do you calibrate a relative or absolute shock for the purposes of your interest rate stress calibration? Relative Absolute 1% Other 45% 45% We asked firms whether they calibrated absolute or relative shocks for the purposes of their interest rate risk calibration. Of the 29 responses, 13 firms stated that they calibrate an absolute shock, and 13 firms stated they calibrate a relative shock. In the current low interest rate environment, an advantage of using the relative approach is that it does not produce negative interest rate assumptions for downward interest rate stresses. Of the three firms who responded other, one uses a principal component analysis based method to calibrate a relative interest rate stress, with an overlay of expert judgement applied to uplift the stress to take into account the prevailing low interest rates. 51

55 RISK CAPITA L Interest rate stress methodology Graph 8.3.2: When determining your interest rate stress, what methods do you use? Principal component analysis (PCA) Term dependent shift Constant shift Other 11% 4% 53% 32% We asked firms what methods they used to determine their interest rate stress assumption. Of the 28 responses to this question, 15 firms stated that they used the principal component analysis method for setting the interest rate risk stress assumption. Additionally, 9 firms stated that they adopted a term dependent shift approach. Three firms stated they used a constant shift approach, and two firms described other approaches. Firms adopting the principal component analysis method tend to be larger organisations: 73% of these firms have Peak I liabilities above 5 billion, compared to 56% of the survey population. Compared to last year, there has been an increase in the proportion of firms using the principal component analysis method and a decrease in the proportion using the term dependent shift approach. The proportion of firms using the constant shift or other approaches has decreased. The change in profile implies a general move by the market to adopt more complex interest rate stress methodologies. Of the firms using the principal component analysis approach, 8% modelled 3 principal components. 52

56 RISK CAPITA L Expert judgement overlays Graph 8.3.3: Do you apply any expert judgement overlays that influence the level of the yield curve under stress? Ye s No 46% 54% We asked firms to state whether or not they applied an expert judgement overlay to their yield curve stresses. There is a fairly even split between firms that do and don t apply an expert judgement overlay to adjust the yield curve under stress. There is no relationship between the size of the firm and whether or not it applies expert judgement. 53

57 RISK CAPITA L Discount basis Graph 8.3.4: Discount basis used in the calculation of best estimate liability (excluding liquidity premium) Gilts Gilts +1bps Swaps 3% 3% Swaps -1bps Swaps -35bps Other 28% 35% 7% 24% The majority of respondents (55%) calculate their best estimate liability using swaps with or without adjustments, while the rest largely use gilts, with or without adjustment. The adjustments that firms are making to the swap spreads lie in the range of the adjustments for credit risk that the Solvency II requirements specify. Swaps minus 1bps is the most widely used basis with 28% respondents indicating that they use this. Only 3% indicated that they are using a 35bps adjustment, which is the maximum allowed under current Solvency II rules. We note that 42% of firms use a discount rate based on gilts, with 7% making a +1bps adjustment to the gilt discount rates. We note that the use of the discount rate based on gilts is not consistent with Solvency II requirements which require the use of discount rates based on swap rates for calculating the technical provisions. Historically gilts have been considered risk free, e.g. for the ICA and MCEV. However we are seeing a general trend towards using swap rates for market consistent calculations. 54

58 RISK CAPITAL Level of the 1-in-2 yield curve stresses Graph 8.3.5: 1-in-2 yield curve stresses in basis points as the difference from the base curve Term 1 - up stress 244 Term 5 - up stress 226 Term 1 - up stress 197 Term 15 - up stress 17 Term 2 - up stress 15 Term 1 - down stress -5 Term 5 - down stress -127 Term 1 - down stress -142 Term 15 - down stress -14 Term 2 - down stress Basis Points Note that the graph shows as a box and whisker plot the distribution of yield curve stresses. The box represents the inter-quartile range, the whiskers represent the minimum and maximum survey responses, and the dot represents the median or 5th percentile. We asked firms to state their yield curve stresses at terms of 1, 5, 1 and 15 years. Graph shows that overall, the magnitudes of the interest rate down stresses tend to be lower than the interest rate up stresses. This likely reflects the current low interest rate environment as well as the application of floors to the post-stress interest rates resulting in smaller stresses in comparison. Overall, we observed a slightly higher magnitude of yield curve stresses this year compared to last year. The inter-quartile ranges for the stresses are fairly tight which suggests a convergence in market practice. We note that for some firms the specified extreme downward stress may result in negative interest rates. Hence, it is possible that some firms have stated the whole stress in their response rather than just the stress above the base curve. We note that a large proportion of firms apply a floor to the stressed yield to avoid negative rates being applied in practice (see Graph 8.3.6a). 55

59 RISK CAPITAL Application of a floor or a cap on the yield curve in the 1-in-2 interest rate stress Graph 8.3.6a: Do you apply a floor to the yield? Yes No 48% 52% Graph 8.3.6b: Do you apply a cap to the yield? Yes No 4% 96% 48% of firms responding to this question indicated that they applied a floor to interest rates, while only 4% (1 out of 25 firms) indicated that they also applied a cap to interest rate stresses. Of the firms applying interest rate floors, all but one applied a % floor to the stressed level, while one indicated that they applied a 1bps floor to the interest rate. One firm indicated that they capped their interest rate stress at 15bps. 56

60 RISK CAPITAL 8.4 CREDIT SPREAD RISK BEL calculation: Spread between government bond yields and corporate bond yields Graph 8.4.1: In your BEL calculation, what was the spread between your government bond yields and your corporate bond yields? AAA 5 years AA 5 years A 5 years 118 BBB 5 years BB 5 years AAA 1 years Max: 129 AA 1 years 91 A 1 years BBB 1 years BB 1 years AAA 15 years 74 AA 15 years 91 A 15 years 139 BBB 15 years BB 15 years AAA 2 years 66 AA 2 years 9 A 2 years 118 BBB 2 years BB 2 years Basis Points Note that graphs to show as a box and whisker plot the distribution of spread and the proportion of spread attributed to the liquidity premium under the BEL and stress calculations. The box represents the inter-quartile range, the whiskers represent the minimum and maximum survey responses, and the dot represents the median or 5th percentile. Note that where whiskers have been truncated, a label indicates the maximum value taken. We asked respondents to provide the basis point spread they assume between government and corporate bonds for the calculation of the BEL as at year end 213. In this year s survey, we asked for more granular information on the spreads at different terms, therefore these results cannot be directly compared with last year s survey. 57 However, in general, the credit spreads were lower this year compared to last year, which is broadly consistent with market movements observed over 213.

61 RISK CAPITAL BEL calculation: Percentage of the spread attributed to the liquidity premium Graph 8.4.2: In your BEL calculation, what percentage of the spread did you attribute to the liquidity premium when determining your valuation interest rate? AAA 5 years 6% AA 5 years 54% A 5 years 51% BBB 5 years 5% BB 5 years 5% AAA 1 years 6% AA 1 years A 1 years 54% 51% BBB 1 years BB 1 years 5% 5% AAA 15 years 6% AA 15 years A 15 years 53% 5% BBB 15 years BB 15 years AAA 2 years AA 2 years A 2 years BBB 2 years BB 2 years 5% 5% 5% 51% 5% 5% 5% 1% 2% 3% 4% 5% 6% 7% 8% 9% 1% Percentage of spread Graph shows that most firms assume a significant proportion of the spread to be attributed to the liquidity premium, with the average proportion reducing as the rating of the corresponding bonds reduce. However, as Graph shows, there is not much variation by term. Only 6 of the 16 firms responding to this question indicated that they attributed different proportion of the spread to liquidity at different terms. There was no clear relationship between the size and type of firm and their approach to attributing the credit spread to a liquidity premium in the base assumptions. 58

62 RISK CAPITAL Stress calculation: Spread between government bond yields and corporate bond yields Graph 8.4.3: Under your 1-in-2 credit spread stress, what was the spread between government bond yields and the yields on your corporate bonds? AAA 5 years 188 AA 5 years 265 A 5 years 315 BBB 5 years BB 5 years AAA 1 years AA 1 years A 1 years BBB 1 years BB 1 years AAA 15 years AA 15 years A 15 years BBB 15 years BB 15 years AAA 2 years AA 2 years A 2 years BBB 2 years BB 2 years Basis points We can observe the following clearly from Graph 8.4.3: As expected, the credit spread stress increases the lower the bond is rated; and Credit spread stresses show some decline as the term of the bond increases. All of the firms responding to this question differentiated their spread stress by term. 59

63 RISK CAPITAL Stress calculation: Percentage of the spread attributed to the liquidity premium Graph 8.4.4: Under your 1-in-2 credit spread stress, what percentage of the spread did you attribute to the liquidity premium when determining your valuation interest rate? AAA 5 years AA 5 years A 5 years BBB 5 years BB 5 years AAA 1 years AA 1 years A 1 years BBB 1 years BB 1 years AAA 15 years AA 15 years A 15 years BBB 15 years BB 15 years AAA 2 years AA 2 years A 2 years BBB 2 years BB 2 years 5% 54% 5% 5% 5% 5% 54% 5% 53% 5% 51% 54% 54% 53% 5% 5% 5% 54% 5% 5% 1% 2% 3% 4% 5% 6% 7% 8% Percentage of spread Graph shows there is a range of practice in the allowance firms make for the liquidity premium (or the Matching Adjustment) when determining the discount rate under a 1-in-2 credit spread stress. Part of this variation will reflect the composition of the corporate bond portfolios the firms hold. The graph shows that firms are more likely to attribute a larger spread to the liquidity premium for the higher rated bonds, hence the tendency for higher 75th percentile proportions for AAA and AA. Of the 16 firms responding to this question, four indicated that they attributed different proportions of the spread to liquidity premium by term. The same four firms also attributed different proportions of the spread to liquidity premium by credit rating. There was no clear relationship between the size and type of firm and their approach to attributing a proportion of credit spread to liquidity premium under stress. The proportions attributed to liquidity premium are largely similar to those under the base scenario, with the majority of firms having the same base and stress assumptions. 6

64 RISK CAPITAL Application of a floor or cap on the credit spread widening or liquidity premium Graph 8.4.5a: Do you apply a floor or cap on the credit spread widening? Yes, biting Yes, not biting No 4% 8% 88% Graph 8.4.5b: Do you apply a floor or cap on the liquidity premium (credit default allowance)? Yes, biting Yes, not biting 4% No 96% We asked firms to state whether they applied a floor or a cap on the credit spread widening or on the liquidity premium. Most firms indicated that they do not apply caps or floors on the credit spread widening or liquidity premium in their capital calculations. Of the three firms applying caps or floor to the spread widening, 2 indicated that the caps or floor were not biting for year end Only one firm applied a floor or a cap to the liquidity premium, and this cap was not biting at year end 213.

65 RISK CAPITAL 8.5 EQUITY RISK Size of shock to equity and property market values Graph 8.5.1: For the most material asset type and/or geography, what is the size of shock to equity and property market values under your 1-in-2 stress? 6% 5% Percentage reduction 4% 3% 4% 41% 3% 2% 1% % UK equity Overseas equity Property Note that the graph shows as a box and whisker plot the distribution of equity and property stresses. The box represents the inter-quartile range, the whiskers represent the minimum and maximum survey responses, and the dot represents the median or 5th percentile. We asked firms to state the size of their 1-in-2 equity and property shocks. The responses to this survey question indicate that firms model UK equity stresses in a fairly small range between 39% and 48%. The differences may be due to a number of factors, including the firms equity investment strategies which may restrict investment in different risk categories e.g. in higher or lower equity risk categories. There is a wider range of responses on the allowance for overseas equity stresses, with reductions in market values between 25% and 5%. Firms assumptions on the appropriate 1-in-2 stresses will reflect the different overseas markets that firms are invested in as well as the investment strategies that the firms may follow. The equity stresses compare to the 46.5% and 56.5% Standard Formula stress under Solvency II as at year end 213 for type 1 and type 2 exposures. The Standard Formula stresses comprise the base stress of 39% and 49% for type 1 and type 2 exposures respectively, and a symmetric adjustment of 7.5%. 62

66 RISK CAPITAL 8.6 EQUITY, PROPERTY AND INTEREST RATE VOLATILITY BEL calculation: Approach to modelling equity, property and interest rates volatility Graph 8.6.1: In your BEL calculation, how do you model the volatility of equity, property and interest rates? Stochastic volatility model Term dependent volatility Fixed level of volatility Not modelled rms Percentage of fi 1% 9% 8% 7% 6% 3% 18% 21% 4% 4% 15% 61% 5% 25% 4% 3% 3% 2% 1% 26% 36% 32% % Equity volatility Interest rate volatility Property volatility We asked firms whether they modelled equity, property or interest rate volatility. The majority of firms responding indicated that they modelled volatility for equity, property and interest rates in the BEL calculations. For equity volatility, 2 out of 27 firms responding to this question indicated that they modelled equity volatility in their BEL calculations, which is more than last year. Of these, eight modelled a fixed level volatility, four modelled a term dependent volatility structure and eight used a stochastic volatility model. This was the first time we asked firms whether they used stochastic models to model the volatility risk. For interest rate volatility, 18 out of 28 firms responded that they modelled this in their BEL which, again, is more than last year. Of these, seven modelled a fixed level volatility, six modelled a term dependent structure and five use a stochastic volatility model. For property volatility, 19 out of 28 firms responded that they modelled this in their BEL, which is also a higher proportion than last year. Of these, 17 modelled a fixed level volatility, one modelled a term dependent volatility structure and another used a stochastic volatility model. 63

67 RISK CAPITAL BEL calculation: Extrapolating to longer term volatilities Graph 8.6.2: In your BEL calculation, how do you extrapolate the following out to longer term volatilities? Flat 1% Change linearly to long term rate Change exponentially to long term rate 8% ntage of firms Perce 6% 4% 1% 1% 1% 2% % Interest rate volatility Equity volatility Property volatility All firms indicated that they extrapolated out to longer term volatilities by assuming the volatility curve was flat beyond the last point at which they had volatility data. For interest rate volatilities, this point is not necessarily related to the last liquid point, and we would expect this term to be shorter than the last liquid point for the relevant interest rate. This is a new question in this year s survey and there is no comparative information from last year. 64

68 RISK CAPITAL Base volatility assumptions Graph 8.6.3: What assumptions do you use for base volatility levels at various terms? Equity 5 years 189 Equity 1 years Equity 15 years Equity 2 years Interest 5 years 17 Interest 1 years 155 Interest 15 years Interest 2 years Property 5 years Property 1 years Property 15 years Property 2 years Basis points Note that the graph shows as a box and whisker plot the distribution of base volatility levels. The box represents the inter-quartile range, the whiskers represent the minimum and maximum survey responses, and the dot represents the median or 5th percentile. The volatilities in the base assumptions vary significantly between different firms. As Graph shows, the assumptions used by the respondent show a fairly wide spread for equity and interest rate volatility. In contrast, the majority of firms responding to this question had the same volatility assumption for property volatility, and the inter-quartile range was equal to zero at all terms. Graph also indicates the presence of significant outliers in either direction for the base equity and property volatility assumptions. The median volatilities increase with the term for equity, while they are constant for property. This reflects the different approaches taken by firms for these volatilities. More than half of firms modelling equity volatility use either term dependent volatilities or a stochastic volatility model. In contrast, only one firm indicated using a stochastic volatility model for property volatility, with the rest of the firms modelling a flat volatility term structure, and the majority of firms using the same volatility assumption. The inter-quartile ranges show that there is more variation in volatility at shorter terms, with the variation falling as the term increases. 65

69 RISK CAPITAL Stress calculation: Approach to modelling equity, property and interest rates volatility Graph 8.6.4: Do you apply a 1-in-2 equity volatility, property volatility or interest rate volatility stress? If so, how is this determined? Yes, stochastic volatility model 1% 4% Yes, shift is term dependent 9% Yes, shift is a fixed level 8% No, volatility stress 7% 52% 12% 8% 4% 48% f irms (% ) Percentage of 6% 46% 5% 4% 3% 2% 44% 35% 48% 1% % Interest rate volatility Equity volatility Property volatility We asked firms whether they modelled equity, property or interest rate volatility risk, and if they did, what their approach to allowing for the risk was. Of the firms modelling a stress for equity, property or interest rate volatility, the majority modelled a fixed level stress across all terms. About 2% of firms used more sophisticated approaches for equity volatility, with 12% using a stochastic volatility model while approximately 8% used a term dependent volatility stress. About 4% of firms use a term dependent interest rate volatility stress while 4% use a stochastic volatility model for their property volatility stress. 66

70 RISK CAPITAL Stresses applied to base volatility levels Graph 8.6.5: What level of 1-in-2 stress do you apply to base volatility levels at various terms? Equity 5 years 125 Max: 487 Equity 1 years 125 Max: 3215 Equity 15 years Equity 2 years Interest 5 years Interest 1 years Interest 15 years Interest 2 years Property 5 years 788 Property 1 years 788 Property 15 years Property 2 years Basis points Note that the graph shows as a box and whisker plot the distribution of volatility stresses. The box represents the inter-quartile range, the whiskers represent the minimum and maximum survey responses, and the dot represents the median or 5th percentile. Note that where whiskers have been truncated, a label indicates the maximum value taken. We asked firms to state the level of stress they applied to base volatility levels for their ICA. In contrast to last year s survey, we asked firms to provide more detailed information on the volatility stresses at different terms and therefore the responses are not directly comparable with last year s results. Most firms volatility stresses lie within a fairly narrow range. However, there are a few significant outliers, in particular for the equity volatility stresses, where one firm models much larger equity volatility stresses at the 5 and 1 year terms than the other firms in the comparison. 67

71 RISK CAPITA L 8.7 MARKET RISK DIVERSIFICATION Approach to allow for diversification benefits within market risks Graph 8.7.1: How do you allow for diversification benefits within your market risks? Single correlation matrix for all risks 1 Correlation matrix for market risk 9 Copula approach for market risks 8 41% 44% Copula approach across all risks 7 Implicit in Economic Scenario Generator f firms 6 75% 31% 13% Percentage o % 19% % 56% 19% 1% 25% 19% 1% All Small Medium Large Firms have adopted different approaches to allowing for diversification between market risks and reflect their overall ICA methodology. We received 29 responses to this question. Of the 29 respondents, 2 indicated that they use a correlation matrix approach to allowing for diversification between market risks. Of these, 8 allow for diversification between market risks separately and presumably then allow for diversification between market and non-market risks as an additional step. 12 firms use a single correlation matrix to allow for diversification between market risks, and between market and non-market risks. Only 6 firms are using copula approaches to allow for diversification and, of these, 4 use the copula approach across all risks, while 2 allow for diversification separately between market risks. All of the firms using the copula approaches were large. 68

72 RISK CAPITA L 8.8 CREDIT DEFAU LT RISK Approach to modelling credit default risk Graph 8.8.1: What technique best describes your approach to modelling credit default risk? Econometric or time series approach Structural model (e.g. Merton) % 25% 14% 25% Actuarial, or reduced form intensity model Not applicable Other firms (%) % 11% 14% 14% 25% age of Percent 5 15% 4 75% 3 57% 13% 13% % 25% All Small Medium Large We asked respondents about the techniques they used for credit risk modelling, and we provided the following answer categories: An econometric approach. This is a technique where the probability of default is assumed to be dependent on user-defined economic factors. Under this approach the probability of default for each counterparty is driven by the random values that the macroeconomic factors take. A structural approach (Merton approach). The risk of default or migration is captured in terms of the random change in a firm s assets relative to its liabilities. Default is assumed to occur when the simulated value of a firm s assets fall below its liabilities. A reduced form approach. Key inputs (probability of default, loss given default) are assumed to follow certain specified distributions. Under this approach, it is possible to calculate portfolio losses analytically. The econometric and the structural approaches are the more sophisticated and tend to be used by larger firms. The structural approaches require a lot of counterpartyspecific data in order to model the default. Reduced form models are simpler to calibrate and use than the econometric and structural approaches, and medium size firms are more likely to use these models. 5 out of 27 respondents stated that they use the econometric or time series approach. Three respondents stated that they use an actuarial or reduced form intensity model, and four respondents stated that structural models are used. Firms that use structural models typically model the credit spread and credit default risks together (see Graph 8.8.2) respondents stated that they use an approach other than those specified above. Of these, five described that they based their approach on the Solvency II Standard Formula calculation. One further respondent described an approach that is broadly similar to the structural approach described above, and four respondents described approaches broadly similar to the reduced form approach described above.

73 RISK CAPITA L Modelling of credit spread risk and default risk together or separately Graph 8.8.2: For traded securities (e.g. corporate bonds, ABS), do you model credit spread risk and default risk together or separately? Together Separately % Not applicable % 63% 56% ms (%) Percentage of fir 6 5 5% % 25% 38% 1 11% All 25% Small 13% Medium 6% Large We asked whether or not respondents model the credit spread and credit default risks separately. Of the 28 responses, 15 stated that they model these two parts of credit risk together and 1 stated that these are modelled separately. 7

74 RISK CAPITA L Capital held for default on sovereign debt Graph 8.8.3: Does your firm hold capital for the default (on either local or foreign currency denominated debt) of any sovereign debt in the following categories? All - UK, US and other sovereigns rated AA or above 14% 82% 4% All - Other sovereigns with a rating below AA 36% 29% 36% Small - UK, US and other sovereigns rated AA or above 1% Small - Other sovereigns with a rating below AA 67% 33% Medium - UK, US and other sovereigns rated AA or above 11% 89% Medium - Other sovereigns with a rating below AA 11% 56% 33% Large - UK, US and other sovereigns rated AA or above 19% 75% 6% Large - Other sovereigns with a rating below AA 56% 6% 38% % 2% 4% 6% 8% 1% Percentage of firms Yes No Not applicable (no exposure) We asked firms whether they held capital for default on any sovereign debt. 82% of respondents indicated that they did not hold capital for default on UK, US and other sovereign debt rated AA or above with 14% indicating that they held capital against the risk. Three out of four of the respondents modelling default risk on AA and above rated sovereign debt were large firms with operations in multiple geographies and who likely have exposure to overseas sovereign debt. 29% of respondents indicated that they did not hold capital against default risk on sovereign debt rated below AA with 36% holding capital against the risk. There appears to be a split in the approaches used by different firms to model default risk on sovereigns rated lower than AA. We also note that for UK firms, this exposure related to foreign sovereigns (as UK is currently rated at least AA by the three major rating agencies). Therefore, this may reflect some firms general practice to model default risk on foreign sovereign debt. Other respondents indicated that this risk was not applicable and this and this is likely because they held no exposure to either UK and/or foreign sovereign debt. 71

75 RISK CAPITA L 8.9 INSURANCE RISK Approach to insurance risks We asked firms about the approach they use to calculate capital requirements for insurance risks, for each of with-profits, annuity, unit linked and protection business. Graph 8.9.1: What technique best describes your approach to modelling losses arising from life insurance risks for with-profits business? Curve fitting Least Squares Monte Carlo Direct evaluation 18% 5% 77% For with-profits business, 77% (17 out of 22) of respondents calculate insurance risk stresses using direct evaluation. One large respondent uses a Least Squares Monte Carlo approach for with-profits, but curve fitting elsewhere. The remainder of respondents use a curve fitting technique. In general, achieving a good fit for with-profits business is challenging given the complex interactions that are typically modelled. These include features such as the asymmetric nature of the liabilities, bonus smoothing mechanisms, use of dynamic hedging algorithms and management actions. For other products the results were very similar so we have not presented them in graphical form. The proportions using direct evaluation are 75% for annuity business, 72% for unit linked business and 76% for protection business, with the remaining proportions all using curve fitting. Only two firms use different approaches for different classes of business. Overall, the firms using curve fitting and Least Squares Monte Carlo tend to be those who have adopted a simulation based approach to aggregation. 72

76 RISK CAPITA L 8.1 MORTALITY AND LONGEVITY RISK In this section, we consider insurance risk capital for mortality and longevity risks. When compared with market risks, the approaches that firms use for modelling insurance risks tend to have more reliance on expert judgement, in part due to the relatively limited amount of data available and the resulting challenges in calibrating stresses. It is also the case that there has been relatively little change in the Solvency II requirements for life insurance risks to drive significant developments in the industry. Nevertheless, within insurance risks, mortality and in particular longevity certainly stands out as the most material risk for many firms and the one which has received most attention from the industry. For example, we see that the CMI model has become the default choice for UK insurers to model mortality improvements. We have also asked some specific questions on annuities due to the regulatory and political focus they have received. Recent years have seen strong sales and the rise of the enhanced annuity market, but firms are still thinking through the impact of the budget reform on new business volumes and secondary impacts on expenses, for example. 73

77 RISK CAPITA L Stochastic modelling We asked firms whether they used a stochastic time series model of future mortality and, if not, whether they planned to do so in the near future. Stochastic modelling is generally less widely used for insurance risks than for market risks, although there are several popular stochastic models for mortality/longevity. Last year s survey demonstrated that very few firms use stochastic modelling for other insurance risks, so we did not ask about those. Graph 8.1.1a: Do you use a stochastic time series model of future mortality for withprofits products? Yes No, but plan to within 5 year No 4% 4% 92% Graph 8.1.1b: Do you use a stochastic time series model of future mortality for annuity products? 17% 7% 76% 74

78 RISK CAPITA L The results demonstrate that the vast majority of respondents do not use stochastic modelling for mortality and have no plans to do so in the near future. This situation has not changed since last year. In particular, for with-profits business, one firm uses stochastic modelling (and does so for all product classes). In addition one firm (that already uses stochastic modelling for annuity products) plans to do so for with-profits business as well in the next five years. The remaining 92% (25 respondents) do not use or plan to use stochastic modelling. For annuity business, 5 out of 29 respondents (17%) use stochastic modelling and a further two plan to do so within one year. This reflects the general status of longevity trend risk as one of the most material and uncertain insurance risks for firms with annuity portfolios. Graph 8.1.1c: Do you use a stochastic time series model of future mortality for unit linked products? Ye s No 3% 97% 75

79 RISK CAPITA L Graph 8.1.1d: Do you use a stochastic time series model of future mortality for protection products? Ye s No 7% 93% In terms of approaches across other product classes, one firm uses stochastic modelling for all product classes (as mentioned above) and one uses stochastic modelling for annuity and protection business only. No further firms use or plan to use stochastic modelling for unit linked or protection business. All of the firms who indicated that they do use stochastic modelling use different models. The models used are as follows: Lee-Carter (including extensions) Cairns-Blake-Dowd (including extensions) Age-Period-Cohort Cause of death Other in-house model It is worth noting that all of the firms using or planning to use stochastic modelling for mortality risk are relatively large in size. All except one are (Partial) Internal Model firms. This demonstrates that significant scale and expertise is required due to the complexity involved and the lack of an industry standard approach. 76

80 RISK CAPITA L Mortality stress Mortality stresses applied We asked firms about the types of mortality stresses they apply to different classes of business. Across product types, firms most commonly perform a range of individual stresses or perform a combined stress alongside a separate catastrophe stress. Graph 8.1.2a: Which of the following 1-in-2 mortality / longevity stresses do you apply to with-profits business? (1) Mis-estimation of best estimate assumption 2 (2) Trend 18 (3) Volatility 16 (4) Catastrophe 14 Combined stress of (1) to (3) Combined stress of (1) to (4) Number of firms out of 21 respondents perform a standalone level mortality stress with a further 5 performing this stress as part of a combination. As only 3 large insurers perform a standalone volatility stress; it seems that many firms either do not regard this as a material risk or else consider that it is included within their mis-estimation and trend stresses. The Standard Formula includes only a single level stress, but several respondents indicated they assumed this covered both mis-estimation and trend risks. 77

81 RISK CAPITA L Graph 8.1.2b: Which of the following 1-in-2 mortality / longevity stresses do you apply to annuity business? (1) Mis-estimation of best estimate assumption 22 (2) Trend 2 (3) Volatility 18 (4) Catastrophe 16 Combined stress of (1) to (3) Combined stress of (1) to (4) Number of firms For annuity business, almost all respondents apply separate level and trend stresses, with very few performing any combination stresses. The PRA has previously stated that they expect longevity stresses to be broken down into at least level and trend components. Interestingly, four firms apply a standalone longevity catastrophe stress, the impact of which would be expected to emerge over time rather than occur in a single year. 78

82 RISK CAPITA L Graph 8.1.2c: Which of the following 1-in-2 mortality / longevity stresses do you apply to unit linked business? (1) Mis-estimation of best estimate assumption 2 (2) Trend 18 (3) Volatility 16 (4) Catastrophe 14 Combined stress of (1) to (3) Combined stress of (1) to (4) Number of firms Graph 8.1.2d: Which of the following 1-in-2 mortality / longevity stresses do you apply to protection business? (1) Mis-estimation of best estimate assumption 24 (2) Trend 22 (3) Volatility (4) Catastrophe Combined stress of (1) to (3) Combined stress of (1) to (4) Number of firms For unit linked and protection business, the vast majority of firms indicated that they perform both a standalone catastrophe stress and either a standalone misestimation or combination stress. A substantial number still perform trend and volatility stresses, either standalone or in combination. 79

83 RISK CAPITA L Level of mortality stresses applied The tables below summarise what levels of stress are applied to mortality for each class of business. Note that not all firms who indicated the type of stress applied also provided the level of the stress applied. Several responses have been removed where stresses were provided in a format that is not comparable to other firms. Table 8.1.2e: Levels of 1-in-2 mortality / longevity stresses applied for with-profits business Minimum % Median % Maximum % (1) Mis-estimation 9% 17% 25% (2) Trend (p.a.) 1.% 1.25% 1.5% (3) Volatility (4) Catastrophe.15%.3%.31% Combined stress of (1) to (3) 18% 2% 2% Combined stress of (1) to (4) Table 8.1.2f: Levels of 1-in-2 mortality / longevity stresses applied for annuity business Minimum % Median % Maximum % (1) Mis-estimation 5% 13% 4% (2) Trend (p.a.).5% 1.25% 1.75% (3) Volatility.4%.45%.5% (4) Catastrophe.3%.3%.3% Combined stress of (1) to (3) 6% 18% 22% Combined stress of (1) to (4) Table 8.1.2g: Levels of 1-in-2 mortality / longevity stresses applied for unit linked business Minimum % Median % Maximum % (1) Mis-estimation 8% 15% 25% (2) Trend (p.a.) 1.% 1.5% 2.% (3) Volatility (*) (4) Catastrophe.15%.24%.35% Combined stress of (1) to (3) - 2% 42% Combined stress of (1) to (4) Table 8.1.2h : Levels of 1-in-2 mortality / longevity stresses applied for protection business Minimum % Median % Maximum % (1) Mis-estimation 8% 18% 25% (2) Trend (p.a.) 1.% 1.45% 2.% (3) Volatility 18% n/a ( ) 27% (4) Catastrophe.11%.21%.35% Combined stress of (1) to (3) 1% 2% 42% Combined stress of (1) to (4) (*) (*): not shown as only one respondent ( ): not applicable as only two respondents 8

84 RISK CAPITA L Stress levels vary quite widely for all types of business, due to the diverse range of products and policyholder profiles across respondents. Mis-estimation stresses are particularly variable. For mortality-type business, stresses range from 8% to 25%, with the most frequent stress levels being 15% (the Solvency II Standard Formula calibration) and 2%. Some of these assumptions may also be intended to cover other stresses, such as trend and volatility. Annuity mis-estimation stresses vary even more widely, and this may be because firms have greater scope to apportion longevity risk between level and trend components. Trend stresses vary between 1% and 2% for mortality business and between.5% and 1.75% for annuity business. This risk is most important for annuity business, as addressed in the next section. The stress calibration will vary depending on the base assumption and how longevity risk is calibrated between components. Catastrophe stresses vary quite widely, between 11 and 35 per mille, especially for protection business, where this risk is most material. The Standard Formula calibration of 15 per mille is widely used but some firms calibrate a more specific scenariobased stress. Very few firms provided a calibration for a separate volatility stress. The combined stresses calibrations vary widely, which is most likely due to differences in what risks these stresses are intended to capture. Annuities For firms with significant annuity portfolios, longevity stresses tend to be among the most material when considering risk capital. These risks therefore often receive detailed attention and are subject to relatively sophisticated risk modelling. It is also interesting to note the role of industry research in the development of longevity modelling, with the CMI model now the standard for determining mortality improvements. Base annuitant mortality table assumptions Graph 8.1.3a: Base annuitant mortality table used in ICA calculations Pensioners (PN, PC) Retirement annuitants (R) Personal pensioner (PP) Immediate annuitants (I) Assurances (A) Widows (W) Other 4% 4% 4% 18% 4% 9% 57% Note: brackets indicate first letter(s) of table names 81 We asked firms which mortality tables they use as part of their best estimate assumptions for annuity business. The majority of firms use base tables for pensioner mortality, although a wide variety of tables are used.

85 RISK CAPITA L Table 8.1.3b: Base annuitant table multiplier used in ICA calculations Minimum % Median % Maximum % Base - Male 7% 98% 127% Base - Female 7% 1% 127% Stress - Male 64% 84% 126% Stress - Female 6% 88% 11% Respondents indicated a very wide range of table multipliers for base and stress assumptions, with the median base assumptions being 98% and 1% of the base table for males and females respectively. There was relatively little difference between males and females. A few firms indicated that their assumption varies by age. Annuitant longevity stress Graph 8.1.3c: Under your longevity level stress, what is the percentage uplift in expectation of life for your most material annuity product? Number of firms Under 5% 5% upto 1% 1% upto 15% 15% upto 2% 25% upto 25% 25% and over Over half of respondents indicated that the longevity level stress resulted in an uplift in expectation of life for their most material annuity product of between 5% and 15%, although the answers ranged between 1% and 3%. Expectation of life provides an attractive comparison metric as it smoothes out differences in basis and stress approach to relate the impact of the stress to something more fundamental but the following factors should be noted: Several firms indicated that they based this calculation on a male annuitant aged 65; however, a range of different approaches were used. Differences in product mix influence the answers. For example, one of the highest values relates to a firm whose most material annuity product is deferred annuities, which are more affected by the longevity stress than immediate annuities due to the longer time frame. The underlying best estimate mortality basis also influences the results. 82

86 RISK CAPITA L Graph 8.1.3d: Under your 1-in-2 stress scenario, over what time period do you model longevity risk? One year, stress applied at time zero One year, stress applied from end of year 1 Stress applied through to run-off 8% 17% Other 4% 71% The majority of firms apply longevity stresses for ICA using a run-off approach. This has been the standard approach under ICA, but as we move towards Solvency II, the industry has been considering whether this approach is still appropriate. One of the main arguments in support of a one-year approach is that the Solvency II Risk Margin covers the risks from time 1 onwards. This means that the run-off approach potentially results in some double counting of the capital requirement. Generally this issue can be addressed either by moving to a one year stress, or by using a run-off approach with a lower stress calibration. We can understand both approaches and we believe that there are challenges to overcome with both of them. We note that there is currently debate in the industry regarding the issue. 83

87 RISK CAPITA L Annuitant mortality improvements Graph 8.1.3e: Mortality improvement model used in ICA calculations Other Interim Cohurt CMI 211 CMI % 21% CMI 213 8% 25% 25% The majority of firms use a version of the CMI model for annuity business mortality improvements, which demonstrates that it has become the industry standard approach. A range of versions is currently in use, which is likely due to the time it takes firms to get comfortable with the results of the latest models. Two firms use the Interim Cohort projections. The responses are broadly consistent with those for Solvency I, except that several of the firms that use more advanced models for longevity risk capital just use the CMI Model for Solvency I. Of the firms in the other category, two use an in-house model while the rest use simpler assumptions. Note that two of the firms that use the CMI model for best estimate assumptions use a different, more advanced, approach for the stressed scenario. 84

88 RISK CAPITA L Graph 8.1.3f: Base scenario - If you use the CMI model, do you use the advanced or core version of the model? Core Advanced 27% 73% Of those firms that use the CMI model, the majority (73%) use only the core parameters. The four respondents that use the advanced functionality of the model are all large firms, indicating that advanced customisation of the CMI model is not yet standard practice. Graph 8.1.3g: What is your long term improvement factor (CMI model) or underpin (cohort model)? Base - Male Bass - Female Stress - Male Stress - Female Number of firms under 1% 3 1% up to 1.25% %up to 1.5% 1.5% up to 1.75% % up to 2% 1 1 2% and over 85 The majority of respondents indicated that their base assumption lies between 1% and 2%. A much smaller number of firms provided their stress assumption, as many allow for the trend stress in different ways. Of those that provided a stressed long term improvement factor, these were between.5% and 1.75% greater than the base assumptions. About half of firms use a higher base assumption for males than females, with the difference typically being.25% or.5%.

89 RISK CAPITA L Enhanced annuities Graph 8.1.3h: Do the 1-in-2 longevity stresses you apply to enhanced annuities differ from the stresses you apply to standard annuities? Stresses are stronger for enhanced annuities No difference 2% 8% Of those respondents who have enhanced annuity liabilities, the majority (8%) apply the same stresses to enhanced annuities as to standard annuities. Only 2 out of 1 respondents apply different stresses to enhanced annuities and these are all stronger stresses. Stronger stresses can arise for enhanced annuities simply due to the shorter duration of the liabilities where the CMI model is used to derive the stress. The shorter duration means that greater weight is placed on the current higher rates of mortality improvement through the convergence assumption. However, we are also aware of firms considering separate calibrations for stressing enhanced and standard annuities. 86

90 RISK CAPITA L 8.11 MORBIDITY RISK Morbidity is generally a less material stress for firms, so we asked a smaller number of questions. Nevertheless, we did observe that a number of firms model trend and volatility stresses and that these firms tended to be either larger or more specialised health product writers. On the other hand, several firms informed us that they don t apply a morbidity stress as they have immaterial volumes of this business. Morbidity Stress Graph a: If you apply a morbidity stress, which of the following do you apply to your incidence and recovery rates for critical illness business? Indvidual CI incidence Group CI Number of firms (1) Mis-estimation of best assumption (2) Trend (3) Volatility (4) Catastrophe Combined stress of (1) to (3) Combined stress of (1) to (4) For individual critical illness business, all 18 firms perform a mis-estimation stress, either on a standalone basis or as part of a combined stress. Seven firms perform a catastrophe stress, indicating the relative importance of this stress for health business, while it tends to be only the firms more focussed on health and protection business that perform individual trend and volatility stresses. Only seven firms responded in respect of group critical illness business, but there was a similar emphasis on level and catastrophe stresses. 87

91 RISK CAPITA L Incidence and recovery rates Graph b: If you apply a morbidity stress, which of the following do you apply to your incidence and recovery rates for income protection business? Indvidual IP incidence 1 Indvidual IP recovery Group IP incidence 8 Group IP recovery Number of firms (1) Mis-estimation of best assumption (2) Trend (3) Volatility 5 (4) Catastrophe Combined stress of (1) to (3) Combined stress of (1) to (4) For individual income protection business, the pattern is very similar to individual critical illness business, with all 14 firms performing a mis-estimation stress, either on a standalone basis or as part of a combined stress. In general, it is the same more specialist firms that perform a range of individual stresses. Most firms apply the same stresses to incidence and recovery, although three firms omit the recovery catastrophe stress and one the recovery volatility. For group income protection business, the responses were similar, showing that firms perform a mis-estimation stress, often with a catastrophe stress. Again, some firms omit the recovery catastrophe stress. 88

92 RISK CAPITAL Morbidity Tables Graph a: Critical illness: In your BEL calculation, what is the primary morbidity table that you use? CMI Population table 14% Reinsurance rates Own tables 29% 38% 19% Graph b: Income protection: In your BEL calculation, what is the primary morbidity table that you use? 18% 17% 53% 12% 89

93 RISK CAPITA L Tables used for critical illness business are typically informed by data or assumptions from reinsurers (38% of respondents), which indicates that many direct writers are reliant on reinsurers technical support. Firms that use tables from the CMI or internally developed tables tend to be the more specialised protection and health writers. The CMI tables are more widely used for income protection than critical illness, with 53% of respondents indicating they use these. Only the more specialised providers tend to write material volumes of income protection business, which explains the lower dependence on external support. We asked firms to provide the morbidity stress levels that they apply for each type of business. Since only a small number of firms provided a response for each combination of stress and type of business, we have insufficient data to present credible analysis. Furthermore, many of the stress levels reported vary widely due to fundamentally different approaches in terms of combining different risks. The only stress for which there is significant consensus is catastrophe stress, where most firms use a stress of around 3 per mille. 9

94 RISK CAPITA L 8.12 PERSISTENCY RISK In this section we consider life insurance risk capital in relation to persistency risk. Persistency risk is the risk that future lapse rates differ from the expected levels that are allowed for in calculating the best estimate liabilities. Persistency risk is generally more firm and product specific than mortality and longevity risks therefore there are fewer sophisticated industry tools available to model persistency risk than there are for other insurance risks (for instance, the CMI model for mortality risk). The result is that persistency risk methodology is often less detailed than for other risks and there is more reliance on internal expert judgement. Further, persistency risk calibration is likely to rely heavily on a combination of internal data and expert judgement due to a scarcity of credible external historical data. Dynamic policyholder lapse behaviour Graph : Do you capture dynamic policyholder lapse behaviour in your best estimate liability model? Yes No, but plan to within 5 year No 23% 4% 73% We asked respondents if they captured policyholder lapse behaviour in their best estimate liability model. Out of the 3 firms that responded to this question, the majority, 22, stated that they do not capture dynamic policyholder lapse behaviour in their best estimate liability model. Of the large firms that responded, around a third commented that they capture, or plan to capture, dynamic policyholder behaviour. No small or medium firms capture, or plan to capture, dynamic policy lapse behaviour. 91

95 RISK CAPITA L Approach to modelling persistency risk Statistical distribution used to model persistency risk Graph a: What statistical distribution do you use for modelling persistency risk? Normal 1% Lognormal Other 9% 8% 33% 38% 31% 7% Percentage of firms (%) 6% 5% 4% 17% 14% 19% 3% 2% 5% 48% 5% 1% % With-profits Unit linked Protection We asked firms which statistical distribution they use to model persistency risk. Across all product groups, the most common statistical distribution used was the normal distribution, followed by the lognormal distribution. The use of relatively simple distributions is generally seen to be justified by the scarcity of data available to calibrate persistency risk. Moreover, one firm who responded other does use more advanced distributions. Approximately half of the respondents for each of the product groups commented that they do not use a statistical distribution. All of these respondents use a stress based rather than simulation based aggregation approach. Alternative approaches used included using the Solvency II Standard Formula methodology, using a fixed persistency stress, or using expert judgement and industry benchmarking. 92

96 RISK CAPITA L Approach to persistency stress tests Graph b: Which approach do you use for your persistency stress tests? Different impact at different durations 1% Level impact at all durations 9% 21% 18% 22% 8% 7% Percentage of firms (%) 6% 5% 4% 79% 82% 78% 3% 2% 1% % With-profits Unit linked Protection The majority of respondents use a level impact at all durations for their persistency stress test, across all product groups. These firms have likely already captured duration effects in their best estimate assumptions and so retain some duration differences when a level stress is applied. Those firms that used a different impact at different durations were predominantly large firms. 93

97 RISK CAPITA L Graph c: At what level of granularity do you apply your persistency stress tests? Surrenders and PUPs stressed differently 14 Guarantees in and out of the money stressed differently 12 Different impact for different products Different changes in different scenarios 1 Other Number of firms With-profits Unit linked Protection The level of granularity that is most frequently used by respondents in their persistency stress testing is to apply a different impact on persistency rates for different products within a defined product group. This is the same as last year. Only two respondents stated that they stress guarantees that are in and out of the money differently; one firm does so for unit linked business, and one for with-profits business. Whether guarantees are in or out of the money could impact policyholders lapse behaviours, however this could be better captured through modelling dynamic policyholder behaviour than through differing stresses. Two firms indicated that they apply a level impact across all products within a product group. 94

98 RISK CAPITA L The most onerous direction of the lapse stress Graph a: Which direction of lapse stress is the most onerous? Lapses/surrenders up Lapses/surrenders down % 7 Percentage of firms (%) % 1% 88% % With-profits Unit linked Protection We asked respondents which is the most onerous direction of the lapse stress. This will depend on the level of guarantees and the expected pattern of future cash flows. Of the 19 respondents who answered this question for with-profits products, all but one stated that a lapse / surrender down stress would be more onerous than an up stress. Lower than expected lapses will increase guarantee costs. Withprofits products tend to have more guarantees, and in the current low interest rate environment many of these guarantees will be biting. All 22 respondents who answered this question for unit linked products stated that a lapse / surrender up stress would be more onerous than a down stress. Out of the 26 respondents who answered this question for protection business, 23 stated that a lapse / surrender up stress would be more onerous than a down stress. The insurers that stated that an up stress is more onerous generally sell more shorter-term products (e.g. term assurance), whereas the respondents for whom a lapse down stress is more onerous hold more long term business (e.g. whole of life). 95

99 RISK CAPITA L Graph b: At what level do you determine which direction is the most onerous policyholder behaviour stress? Individual policy level Modelling class level Product group level % 5% 14% 8% 12% Fund level Company level Other Percentage of firms (%) % 55% 5% 3 21% 2 14% 12% 1 16% 9% 15% 5% 5% 4% With-profits Unit linked Protection The most common level at which respondents determined the most onerous policyholder behaviour stress, across with-profits, unit linked and protection business, was at product group level. This is the same as in previous years. This level typically offers firms the best balance between accuracy and complexity. 96

100 RISK CAPITA L Levels of lapse stresses applied The table below summarises the level of lapse stresses applied by firms. Table a: With-profits 1-in-2 lapse increase and lapse decrease stress as a percentage change Lapse increase Lapse decrease Duration Minimum Median Maximum Minimum Median Maximum Year 1 33% 5% 87% 34% 5% 85% Years % 5% 87% 21% 5% 6% Table b: Unit linked 1-in-2 lapse increase and lapse decrease stress as a percentage change Lapse increase Lapse decrease Duration Minimum Median Maximum Minimum Median Maximum Year 1 27% 5% 7% 34% 5% 5% Years % 5% 7% 34% 5% 5% Table c: Protection 1-in-2 lapse increase and lapse decrease stress as a percentage change Lapse increase Lapse decrease Duration Minimum Median Maximum Minimum Median Maximum Year 1 19% 5% 7% 15% 5% 85% Years % 5% 7% 15% 5% 85% Three firms provide a stress that varies by duration. Two specify a more onerous lapse stress for year 1, then a level stress for years 2-1, and the other specifies a more complex pattern for protection business. The median response for each type of business is the same as the lapse stress specified in the Solvency II Standard Formula, i.e. increase or decrease in lapse rates of 5%. The large variation in levels of lapse stress applied demonstrates the firm-specific nature of persistency risk and hence the heavy reliance on internal data and expert judgement. 97

101 RISK CAPITA L Impact of budget announcement on persistency risk Graph : Do you intend to change your base lapse assumption and 1-in-2 lapse stress in respect of pension business in the light of the recent budget announcement that retirees would no longer be forced to buy an annuity? Yes No % 7 Percentage of firms % 1% Base Stress We asked if firms intended to change their base lapse assumptions and 1-in 2 lapse stress in respect of pension business in light of the recent budget announcement that retirees would no longer be forced to buy an annuity. These lapse assumptions would apply at the point of retirement. There were 25 responses to this question. Only one respondent stated that they would change their base lapse assumptions and no respondents stated that they intended to change their 1-in-2 lapse stresses. The respondent that intended to change their base lapse assumption explained that they intended to change the GAO take up assumptions. Several insurers who responded No noted that they did not intend to make any changes now but would be monitoring experience in the wake of the budget announcement in order to review the appropriateness of their base lapse assumptions and 1-in-2 lapse stresses going forward. One firm noted that they expect the budget announcement to increase lapses. As it was a lapse decrease that was more onerous for this firm, not making a change to assumptions was a more prudent approach. These results reflect the high level of uncertainty around the impact on policyholder behaviour and suggest the industry is taking a wait and see approach. 98

102 RISK CAPITA L 8.13 MASS LAPSE RISK In this section we consider life insurance risk capital in relation to mass lapse risk. Mass lapse risk is the risk of a one-off change in lapse experience over the course of one year. There is a scarcity of available data for mass lapse risk, therefore the approach to modelling and calibration can rely heavily on expert judgement. Approach to modelling mass lapse risk Accounting for mass lapse in the risk capital calculation Graph a: Do you account for mass lapse in your risk capital calculation? Ye s No 21% 79% We asked respondents whether they currently have a mass lapse stress in their risk capital calculation. Around four-fifths, 23 out of 29, of the respondents stated that they have a mass lapse stress. This has increased from last year when approximately two-thirds of respondents stated they had a mass lapse stress. Of large insurers who responded, 82% have this stress within their risk capital compared with 75% of small and medium sized respondents. 99

103 RISK CAPITA L Expense assumptions in mass lapse stress Graph b: What assumption do you make about expenses in your mass lapse stress? Assume that expenses vary in line with policy numbers Assume that some expenses are overheads which stay fixed and do not vary in line with policy numbers Assume that some expenses are overheads which run-off over time but not directly in line with policy numbers 17% 58% 25% We asked firms about their assumptions regarding expenses in their mass lapse stress. 14 out of 24 firms assume that expenses vary in line with policy numbers. This response was particularly common amongst large firms with 11 out of 15 large firms selecting this response. A large proportion of these firms hold a significant amount of unit linked business. For small and medium firms, the split between the three approaches was approximately equal. Several firms commented that the risk of increased per policy expenses in a mass lapse would be incorporated via the expense risk and correlation assumptions. 1

104 RISK CAPITA L Statistical distribution used to model mass lapse risk Graph c: Which statistical distribution do you use to model mass lapse risk? Normal 1% Student's t-distribution Lognormal 9% 8% 44% 38% 32% Half exponential Other Percentage of firms 7% 6% 5% 4% 11% 6% 13% 6% 11% 5% 3% 53% 2% 44% 38% 1% % With-profits Unit linked Protection The most common statistical distribution used to model mass lapse is the normal distribution, as with base lapse risk. However, a wider range of statistical models are used to model mass lapse risk. This could be because there is less data available and therefore more expert judgement required. One large firm with predominately unit linked business uses the student s t-distribution for their unit linked business. One large firm used the half exponential distribution for all product groups. This distribution is heavily skewed to the tail, and therefore suited to modelling catastrophe like events. Note that one firm that responded other uses an exponential generalised beta distribution. Approximately half of respondents do not use a statistical distribution to model mass lapse risk. All of these respondents use a stress based rather than simulation based aggregation approach. Alternative methods included Solvency II Standard Formula calibration, empirical distribution, industry benchmarking and expert judgement. 11

105 RISK CAPITA L Levels of mass lapse stresses applied Graph : 1-in-2 mass lapse stress by product as a percentage change above best estimate lapses With-profits Unit linked Protection Number of firms Less than 2% Between 2% and 3% Between 3% and 4% Between 4% and 5% 1 1 More than 5% We asked respondents to specify their 1-in-2 mass lapse stress by product group as a percentage point additive change in lapse rate applied either above best estimate lapses or above stressed best estimate lapses. Several respondents commented that they used a similar approach to the Solvency II Standard Formula methodology, i.e. an instantaneous change due to the discontinuance of a set percentage of insurance policies with negative reserves. For some firms, the mass lapse stress varies by product within each product group. 12

106 RISK CAPITA L Method of aggregation of lapse stresses Graph : What method do you use to aggregate the lapse stresses? Add each component together Allow for diversification between each component Take the maximum of all components (i.e. the current Standard Formula approach) Other 11% 4% 31% 54% We asked respondents about the method they used to aggregate lapse stresses. There is an increasing preference to take the maximum of all components (i.e. the current Standard Formula approach). Approximately half of the respondents chose this option, up from 44% in 213 and 17% in 212. There is a mixture of small, medium and large firms that adopt this approach. Just under one third of firms said they allow for diversification between each component. This is a small increase on last year. Only one respondent stated they simply add each component together, therefore assuming no diversification. This is down from seven respondents last year, and is more similar to 212 when no respondents indicated that they would adopt this approach. Several firms described other approaches which involved combining the lapse stresses in various ways, such as considering a scenario with a mixture of up and down lapse stresses depending on which is the most onerous at a product level. 13

107 RISK CAPITA L 8.14 EXPENSE RISK In this section we consider life insurance risk capital in relation to expense and expense inflation risk. Expense risk is the risk of loss, or an adverse change in the value of liabilities, caused by the fact that the timing and/or the amount of expenses incurred differs from those expected. Firms tend to use historical internal data, expert judgement and industry benchmarking when modelling expense risk. Often published inflation data, for example the Retail Price Index or published earnings indices, is used to model expense inflation risk. Approach to modelling expense risk Statistical distribution used to model expense and expense inflation risk We asked respondents which statistical distribution they use to model expense and expense inflation risk. Graph a: Which statistical distribution do you use for modelling expense risk? Normal Lognormal Other 33% 54% 13% Over half of respondents do not use a statistical distribution to model expense risk. Other methodologies used included a Solvency II Standard Formula approach, industry benchmarking and expert judgement. For those firms that do use a statistical distribution to model expense risk, the most commonly used distribution is the normal distribution, followed by the lognormal distribution. One firm who responded other uses an exponential generalised beta distribution. 14

108 RISK CAPITA L Graph b: Which statistical distribution do you use for modelling expense inflation risk? Normal Student's t-distribution Lognormal Other 29% 59% 4% 8% The majority of firms take the same approach to model expense inflation risk as they do expense risk. Expense Stress Expenses subject to the 1-in-2 expense stress Graph a: Which of your expenses are subject to the 1-in-2 expense stress? All expenses All expenses except investment expenses 4% Only internal expenses (i.e. not those governed by outsourcer arrangements) Other 25% 53% 18% We asked firms which of their expenses are subject to the 1-in-2 expense stress. The majority of respondents subject all expenses to a 1-in-2 expense stress. A greater proportion of large firms consider all expenses whereas the approach taken by small and medium firms is more mixed. 15

109 RISK CAPITAL Level of expense and expense inflation stresses applied Graph b: 1-in-2 expense and expense inflation stresses as a percentage change 7% 6% 5% Percentage change 4% 3% 29% 2% 15% 1% % Expenses Expense inflation Note that the graph shows as a box and whisker plot the distribution of expense and expense inflation stresses. The box represents the inter-quartile range, the whiskers represent the minimum and maximum survey responses, and the dot represents the median or 5th percentile. We asked firms to specify their 1-in-2 expense and expense inflation stresses, as a percentage change (e.g. if inflation assumption goes from 4% in base to 5% in inflation increase stress, the answer should be 25%). Note that we have removed those firms who provided an additive inflation stress. There is a larger variation in the stresses used by firms for expense inflation. It is often a mixture of expert judgement and published inflation data, for example the Retail Price Index or Average Earnings Index, which is used to model expense inflation risk. The larger variation in expense risk stresses could be due to differences in the underlying inflation data. 16

110 RISK CAPITA L 8.15 LIQUIDITY RISK Liquidity capital requirements The number of respondents holding capital for liquidity risk has fallen from 21% in 212, to 14% in 213, and finally to 1% in 214. Two large firms have stopped holding capital for liquidity risk since 213, whilst one large firm has recently started to hold capital for this risk. We note from the 212 survey that 38% of respondents were considering liquidity risk as part of their ORSA, and from our experience. Due to increased interest from the PRA in liquidity and funding, we have observed in the market that this number has risen since then. We anticipate that under Solvency II the majority of firms will not hold Pillar 1 capital in respect of liquidity risk but will consider this risk as part of the Pillar 2 requirement. Graph : Do you hold capital for liquidity risk? Ye s No 1% 9% Liquidity risk strategy / appetite We asked firms whether they have a clearly defined risk strategy / appetite. As predicted in last year s survey, the number of firms stating they had a clearly defined liquidity risk strategy and risk appetite has increased (from 6% to 9%). As mentioned above, the PRA has shown an increased interest in liquidity and funding which, combined with ORSA development ahead of Solvency II implementation, is likely to be a key driver of this increase. 17

111 RISK CAPITA L Degree of mass lapse required to give rise to a liquidity issue Graph : What is the degree of mass lapse that you consider would be required to give rise to a liquidity issue? Beyond a 1-in-2 year event Not considered in relation to a 1-in-2 year event 17% Not considered 52% 31% We asked firms the degree of mass lapse (i.e. beyond or within a 1-in-2 year event) that they consider would be required to give rise to a liquidity issue (i.e. liquidity management actions would need to be taken). We also provided respondents with the option to state that they had not considered this (either not considered at all, or not considered in relation to a 1-in-2 year event). We observe that the number of respondents who do not consider the impact of a mass lapse on liquidity has decreased, with only 17% of firms currently not considering this risk, compared to 51% in the 213 survey. We anticipate the reason for this as increased interest from the PRA on reverse stress testing particularly the level of mass lapse required to cause a liquidity issue. For those firms who do consider this risk in relation to a 1-in-2 year event, all of these state they perceive the level of mass lapse required to cause a liquidity issue to be beyond a 1-in-2 year event, which now represents 52% of respondents compared to 23% in 213. There has been an increase in the number of firms considering this risk, but not in relation to a 1-in-2 year event, from 17% to 31%. 18

112 RISK CAPITA L 8.16 OPERATIONAL RISK Operational risk is defined under Solvency II as the risk of loss arising from inadequate or failed internal processes, personnel and systems, or from external events. Operational risk should include legal risks, and exclude risks arising from strategic decisions, as well as reputation risks. In the UK there is already a requirement to quantify operational risk capital under ICA and there has been further significant investment in modelling techniques, by Internal Model firms in particular, following the advent of Solvency II. However, the quantification of operational risk remains subjective and expert judgement driven compared to other key risk types, due to the relative lack of internal and external data. Treatment of operational risk has often been driven by a firm s choice between Solvency II Internal Model and Standard Formula. The Standard Formula specifies a factor based approach which is relatively simplistic compared to many of the approaches that firms have adopted for ICA, as it is not risk sensitive. Use of an internal loss capture database Graph : Does your firm use an internal loss capture database to record operational risk loss data? Ye s No 19% 81% Most respondents (25 out of 31) use an internal loss database to record operational losses. There is no obvious relationship to the size of respondents. It is now generally considered standard practice to record internal operational losses and having such a database will be a governance requirement under Solvency II. However, the implementation of internal loss databases is a recent industry development, with very few firms having even 5 years of data and correspondingly few tail events with which to parameterise assumptions. Therefore, there is an increasing tendency for respondents to seek other sources of loss data, such as external loss databases, to complement the internal data. 19

113 RISK CAPITA L Approach to modelling operational risk loss Graph : What technique best describes your approach to modelling operational risk loss? Scorecard approach Deterministic scenario Stochastic modelling with expert judgement 2% 7% 17% Stochastic modelling with loss data Factor based (Standard Formula) 3% Other 13% 4% We asked firms which technique best described their approach to modelling operational risk for ICA. Two fifths of respondents indicated that they use stochastic modelling for operational risk and only one small firm uses the Standard Formula factor based approach. This can be compared with last year, when approximately one third of respondents indicated that they used stochastic modelling and five firms responded that they used a Standard Formula factor based approach. Broadly speaking, the results demonstrate that standard practice has emerged amongst larger Internal Model firms to use a scenario based stochastic approach for operational risk capital modelling, where relevant internal and external data is augmented by expert judgement to parameterise the model. Answers in the other category mostly use some form of expert judgement. One of these firms reported the use of Bayesian networks. The range of responses reflects the challenges of effective operational risk capital modelling and the different stages firms are at in their development programmes. 11

114 RISK CAPITA L Sources of operational risk loss data Graph : What is the primary source of your operational risk loss data? No source, risk modelled on plausible operational loss scenarios Some actual internal operational risk loss data and scenarios Combination of internal and external loss data 4% 7% 25% The ABI database / OpRisk database Other 21% 43% Most firms have only started to capture operational loss data relatively recently with the result that calibrating loss distributions with any degree of confidence can be challenging. Two thirds of firms use a mixture of operational risk loss data and the most common practice among large Internal Model firms is to use a combination of internal and external data sources. A typical approach is to hold scenario workshops, with detailed briefing notes to workshop participants containing the purpose of the workshop, details of relevant internal and external loss data, risk and control self assessments and other relevant data. 111

115 RISK CAPITA L Comparison of Standard Formula operational risk capital with ICA operational risk capital Graph : Does Standard Formula or ICA produce the highest operational risk capital (post diversification)? Standard Formula ICA Little material difference 1% 14% 76% In general, the ICA approach produces higher operational risk capital requirements than the Standard Formula and this year s result shows 76% of firms hold higher capital under ICA than Standard Formula. Interestingly, this is post-diversification, demonstrating that the Standard Formula parameterisation is very different from firms own view of their operational risk capital requirements. In particular, the drivers of the operational risk capital charge under the Standard Formula are premiums, technical provisions and gross expenses in respect of unit linked business. Firms own ICA calibrations can be based on a much wider range of risk drivers and incorporate scenario approaches. 112

116 RISK CAPITA L 8.17 AGGREGATION Approach to risk aggregation We asked firms what aggregation approach they use under the current ICA regime. Approaches to aggregation include using a correlation matrix or using copula techniques: The correlation matrix approach assumes linear loss functions and dependency. Some components may be calculated based on simulation (for example market risk) but most components are stress based and the ultimate aggregation is via a correlation matrix. Copula techniques imply use of a simulation based approach across the full taxonomy of risks. Copulas can allow for non-linear loss functions and non-linear tail dependency, depending on what type of copula is chosen. For example, Gaussian copulas do not allow for tail dependency, whereas this is possible under the Student s t copula. As predicted in last year s survey, the number of firms using the correlation matrix approach has declined as Internal Model firms move to use their Solvency II methodologies under ICA+. We see that 68% of firms are using a correlation matrix approach compared to 82% last year. We note that the proportion of firms using copula techniques has increased from 12% to 26% this year, which is indicative of firms using their Solvency II methods under ICA+. Two firms out of 31 respondents are using an advanced copula such as a Student s t copula, with six further firms using a Gaussian copula. All firms that currently state they use copula methods are large. Graph : What is your approach to risk aggregation Correlation matrix Gaussian copula Advanced copula (e.g. student's t) Other 7% 6% 19% 68% 113

117 RISK CAPITA L Method of setting correlations between risk categories As covered in the previous question, most firms have indicated that they used a correlation matrix or copula technique for aggregation. We asked these firms how they set the correlations between different risk categories (note that firms were able to select multiple categories in their response). Consistent with the responses in last year s survey, general reasoning / expert judgement is the most widely selected response to this question (28 firms). Responses indicated that historical data (18 firms) and benchmarking / external advice (18 firms) were also widely used. Nine firms use internal experience data. As may be expected, 8 of these are large firms, as these firms typically have access to the volume of data required for sufficient statistical credibility. No firms stated that they used other sources such as rating agencies or reinsurers. Graph : If using a correlation matrix or copula approach, how have you set the correlations between different risk categories (e.g. market to insurance risks)? Using general reasoning and/or expert judgement 3 Using data/benchmarking/advice provided by external consultants 25 Using historical insurance industry data and financial market data Using Solvency II standard formula 2 Using firm's internal experience data Using regulatory guidance Number of firms Using data provided by other industry bodies

118 RISK CAPITA L Correlation assumptions We asked firms to populate a template correlation matrix in respect of their 213 ICA calculation. Given that firms use a variety of risk drivers in aggregation, and also that features of some of these risk drivers are very specific to firms, caution should be exercised when using these results. As discussed earlier in this report, the approach to setting correlations varies across firms. However, in general, the following statements hold: Given the availability of data, market-to-market correlations can be expected to rely less on expert judgement; Conversely, market to non-market correlations and non-market to non-market correlations can be expected to rely more heavily on expert judgement. Consequently, market-to-market correlation values are often set at a more granular level of rounding than for market to non-market correlations and non-market to nonmarket correlations (which are often set in 25% steps given expert judgement). Note that graphs a, b & c show as a box and whisker plot the distribution of correlation pairs. The box represents the inter-quartile range, the whiskers represent the minimum and maximum survey responses, and the dot represents the median or 5th percentile. It should be noted that the number of respondents for each correlation pairing will vary, as different firms model correlations at different granularities. The correlations shown here are correlations of the directions of the respective risk drivers for example, a negative correlation between equities and credit spreads would mean that equity markets fall in conjunction with a rise in credit spreads. 115

119 RISK CAPITAL Market-to-Market correlations Market to market correlations are generally set by reference to historical data, and consequently are set at a more granular level than those involving non-market risk drivers. For many correlations, such as equity versus credit spreads, a large amount of external data exists, and the firms exposures to these risks are similar in nature. Therefore, we observe a narrow inter-quartile range for these risk pairings. For correlations involving the interest rate level (PC1), a fall in interest rates is generally the more onerous stress, and hence we observe many firms correlating falling interesting rates with other adverse market stresses (such as falling equity and property values, and rising credit spreads), which is generally what recent historical data has suggested. However, we are aware that the interest rate up stress is more onerous for certain firms, and hence the rise in interest rate is correlated with other adverse market stresses, as this is more prudent in terms of the capital requirement. As a result, we observe a range of correlations (over positive and negative values) of the interest rate risk driver with other market risks. In general, where these are comparable, we observe that the median market correlations are generally less onerous than the Standard Formula equivalents. Graph a: Market to Market Correlations Interest rate level/pc1 v Equity.29 Interest rate level/pc1 v Equity volatility Interest rate level/pc1 v Property Interest rate level/pc1 v Credit spread Interest rate level/pc1 v Credit default -.25 Equity v Equity volatility Equity v Property Equity v Credit spread Equity v Credit default -.5 Equity volatility v Property Equity volatility v Credit spread Equity volatility v Credit default Property v Credit spread Property v Credit default Credit spread v Credit default Correlation 116

120 RISK CAPITAL Non-market-to-Non-market correlations Due to the reduced availability of data (when compared with market to market correlations) non-market to non-market correlations tend to rely to a greater extent on expert judgment. Generally, we observe that the market medians of correlations involving demographic risks, such as mortality and longevity, are close to zero. For persistency and mass lapse correlations, we observe a wide variation of assumed correlations. For example, persistency versus expenses varies between and.7. We noted earlier that persistency assumptions are very product-specific, and we expect these to vary widely between firms. We also observe similarly wide variation in the correlations between persistency and mass lapse, and also between mass lapse and operational risk. For operational risk, the median correlations are generally positive, which is the more onerous direction. The median of the longevity level versus mortality correlation is, which is more onerous than the Standard Formula correlation of It may be the case that firms view the individuals who buy protection products as generally distinct from annuitants, and hence assume a correlation of for prudence. Graph b: Non-market to Non-market Correlations Longevity level v Longevity trend Longevity level v Mortality.7. Longevity level v Persistency Longevity level v Mass Lapse Longevity level v Expenses Longevity level v Operational Longevity trend v Mortality Longevity trend v Persistency Longevity trend v Mass Lapse Longevity trend v Expenses Longevity trend v Operational Mortality v Persistency Mortality v Mass Lapse Mortality v Expenses Mortality v Operational Persistency v Mass Lapse Persistency v Expenses Persistency v Operational Mass Lapse v Expenses Mass Lapse v Operational Expenses v Operational Correlation

121 RISK CAPITAL Market to Non-market correlations For the risk drivers involving demographic risks, we observe a near-zero correlation assumed with market risks, and as a result these correlations have not been included in Graph c. The Standard Formula suggests that market risks versus life risks should be correlated at.25, we observe that a range of firms have adopted a lower correlation than this. We observe that both persistency and mass lapse risk drivers are generally negatively correlated with improving market conditions, i.e. lapse rates will decrease when property and equity markets are rising. However, we do observe a wide variation of assumptions for persistency correlations with the market, and these are likely to be specific to firms, such as the types of products sold, and furthermore, the exact features of these contracts. For example, adverse market conditions may increase the moneyness of guarantees under with-profits business, thus making them more attractive and reducing lapse rates. However, for unit-linked business, falling equity and property values would lower investment returns, and may be assumed to lead to an increase in lapse rates. The expense risk drivers are generally correlated in a manner that implies a rise in per policy expenses under adverse market movements. Expenses could be indirectly affected by stresses that reduce the number of in-force policies (through higher lapses), although this depends on the type of product, and wouldn t be relevant to, for example, annuity business. Furthermore, adverse market movements would tend to increase investment expenses as a proportion of the fund. The risk driver of operational losses is positively correlated with the most onerous direction of the market risk drivers for all firms. This is in line with the expectation that operational losses will increase in times of market turmoil. Graph c: Market to Non-market correlations Interest rate level/pc1 v Persistency -.5 Interest rate level/pc1 v Mass lapse -.5 Interest rate level/pc1 v Expenses. Interest rate level/pc1 v Operational -.2 Property v Persistency -.16 Property v Mass lapse Property v Expenses Property v Operational Equity v Persistency Equity v Mass lapse -.25 Equity v Expenses -.22 Equity v Operational Credit spread v Persistency Credit spread v Mass lapse Credit spread v Expenses.12 Credit spread v Operational.25 Credit default v Persistency.15 Credit default v Mass lapse.25 Credit default v Expenses Correlation 118

122 RISK CAPITA L Software used to perform aggregation We asked firms what software they used to perform their aggregation. Observing similar responses to last year, out of 31 responses, the majority (22 firms) indicated that they used an internally or externally developed Excel-based tool. We note that those firms using Excel based tools tend to use a stress-based / expert judgment approaches to aggregation. Only a small proportion of those firms using Excel based tools use simulation-based approaches, as these firms tend to prefer offthe-shelf solutions, presumably to increase the efficiency of the calculation. RiskAgility is used by 5 respondents for capital aggregation (2 of which have started using RiskAgility since last year). All respondents using RiskAgility use copula approaches in capital aggregation. Two firms also responded that they used IBM Algorithmics software. Graph : What software are you using to perform your aggregation? Algorithmics - IBM 25 RiskAgility - Towers Watson Internally developed Excel-based tool 2 Externally developed Excel-based tool Other Number of firms

123 RISK CAPITA L Approach used to allocate capital diversification benefit Graph : What approach is used to allocate the capital diversification benefit to lower levels of granularity? Pro-rata Discrete marginal approach Continuous marginal approach (Euler method) 21% Variance-covariance method 34% No allocation of diversification benefit 7% 3% 35% We asked firms what approach is used to allocate the capital diversification benefit to lower levels of granularity, if this is done. This year, more firms are allocating their diversification benefits, with 66% of respondents stating they do this, which is an increase from 57% in the 213 survey. Consistent with last year s results, the two most widely used approaches are the pro-rata method (which allocates capital according to some basis, such as reserves or premium income) and the continuous marginal approach (also known as the Euler method, whereby the diversification benefit is allocated with reference to the partial derivative of the capital requirement). However, there is currently no unique and commonly accepted practice of what should be considered as a fair allocation. The percentage of firms using the pro-rata method has remained stable since last year (21% compared to 23% previously), whereas we observe an increase from 23% to 35% in the proportion of firms using the continuous marginal approach (Euler method). Of the 35% (1 firms) that use the Euler method, 8 of these are large firms as this method is generally more computationally intensive and is usually adopted by firms with more sophisticated approaches for their ICA. The pro-rata and discrete marginal approaches were used more widely by small and medium firms, where stress-based approaches to the capital calculation are more common. 12

124 RISK CAPITA L Level of business that capital is allocated at Firms allocate risk capital to more granular levels for a number of different purposes, such as: Supporting effective capital optimisation across the organisation Supporting performance evaluation Supporting key decision processes such as pricing and business plan projections Regulatory compliance and financial reporting Allocating to legal entity level is the most widely used approach (13 out of 27 firms), with equal splits between product level (6 firms) and fund level (6 firms). There is no meaningful relationship between the size of the firm and the granularity at which it allocated risk capital. This is possibly because smaller firms, despite generally taking less complex approaches to allocation, have a lower number of products and funds, thus making it relatively more manageable to allocate capital to this level. Graph : At what level of business do you allocate risk capital? Legal entity Fund level 8% Product level Other 22% 48% 22% 121

125 RISK CAPITA L Method of allowing for non-linearity We asked firms about their approach to allowing for non-linearity. Firms may allow for this via a variety of methods, outlined below: Under the Killer Scenario approach, all individual risks at a 99.5th percentile are calculated and then a combined scenario of all risks is calculated. The non-linearity adjustment is the capital requirement under the combined scenario divided by the sum of the capital requirements under the individual risk scenarios. The Risk Geography approach considers combined scenarios in the proximity of the expected insolvency areas and examines the relative capital requirements of these scenarios, determining a range for the capital requirements. Firms who use proxy models for modelling risks can allow for non-linearity by including cross-terms reflecting the interaction of risk drivers. Graph : Which method that best describes your approach to allow for nonlinearity? Killer scenario approach (e.g. medium bang, big bang, etc) Inclusion of cross-terms in proxy models 17% Risk geographies No allowance Other 7% 3% 52% 21% We observe similar results to last year s survey, with approximately half of firms (15 out of 29) stating they use a Killer Scenario (e.g. big bang or medium bang) approach to allowing for non-linearity. Six firms allow for non-linearity through the inclusion of cross-terms in the proxy models, while other approaches include producing scenarios to check the appropriateness of results derived from the correlation matrix approach, simultaneous modelling, as well as using Monte Carlo simulation to derive the adjustment. 122

126 RISK CAPITAL Magnitude of non-linearity adjustment We asked firms to state the magnitude of the adjustment for non-linearity, expressed as a percentage of the unadjusted capital. The mean of this adjustment is 3.8%, with a median of 2.%. We note that the majority of responses for the magnitude of this adjustment are below 1%, although one firm has an adjustment of 19%. In terms of the direction, we would generally expect the allowance for non-linearity to be positive, as we are aware of firms that floor the adjustment at zero if it became negative, which is a prudent approach. However, some firms may use a negative adjustment, if justified appropriately. Graph : What is the magnitude of the adjustment for non-linearity, expressed as a percentage of diversified capital requirements? 2% 18% 16% Adjustment for non-linearity 14% 12% 1% 8% 6% 4% 2% 2% % Note that the graphs show as a box and whisker plot the distribution of the nonlinearity adjustment. The box represents the inter-quartile range, the whiskers represent the minimum and maximum survey responses, and the dot represents the median or 5th percentile 123

127 RISK CAPITA L 8.18 CAPITA L FUNGIBILITY Approach to accounting for capital fungibility We asked firms to state their approach to account for capital fungibility (i.e. the extent to which capital can be transferred between different legal entities of the group). A similar percentage (43%) to last year s survey stated that there are no fungibility issues in their business, and consequently do not require consideration. Two firms account for fungibility restrictions explicitly in modelling ( bottom-up approach ), both of which are large firms. We observe improved market practice in this area; the number of firms not considering fungibility having decreased from last year (29% to 13%). There has been a corresponding increase in the number of firms allowing for fungibility via an out-of-model adjustment ( top-down approach ), which has risen from 14% to 34%. Graph : Which best describes your approach to accounting for capital fungibility? Top-down approach (i.e. don t explicitly account for fungibility in actuarial models and make an adjustment at the end) Bottom-up (i.e. account for fungibility restrictions in modelling) Not applicable as fungibility is not relevant to the business 13% 3% 34% Not applicable, as this risk has not been considered Other 43% 7% 124

128 9 Modelling 9.1 MODELLING PLATFORMS We asked firms what modelling platforms they use for primary modelling purposes (in particular, valuation purposes) and also for pricing purposes. Respondents were able to select multiple modelling platforms. For primary modelling purposes, 44% of firms use Prophet only, a further 25% use Moses only, with the remaining 31% using other or multiple platforms. In line with our expectations, this is broadly consistent with last year s survey, where the equivalent proportions were 46%, 2% and 34%. We see very similar proportions in relation to the modelling platform used for pricing; however we note a marginal increase in the use of Prophet and Moses when compared to last year s survey. The other modelling platforms that firms included in their responses were Mo.Net, MG-ALFA, AXIS, and Excel. Graph 9.1.1: What modelling platform are you using for primary purposes and pricing purposes? Single Platform - Prophet 1% Single Platform - Moses Single Platform - Other 9% 8% 44% 38% Using Multiple Platforms 7% Percentage of firms 6% 5% 4% 3% 25% 16% 19% 22% 2% 1% 16% 22% % Primary modelling Pricing 125

129 MODELLING 9.2 ECONOMIC SCENARIO GENERATO RS Graph 9.2.1: Which company provides your Risk Neutral ESG? Moody's (Barrie and Hibbert) Towers Watson 14% Internal ESG 1% 76% Continuing the results seen in recent years, Barrie & Hibbert remains the most widely used ESG provider. In particular, 76% of respondents who use a Risk Neutral ESG indicated that this was provided by Barrie & Hibbert. We observe that the use of ESG providers has remained stable over the year, but note that two firms have swapped provider, both moving to the Towers Watson ESG, from an internal ESG and the Deloitte ESG respectively. Out of the three respondents who state they use an internal ESG, two of these are large firms, and one is a small firm. 126

130 MODELLING 9.3 PROJECTING THE BALANCE SHEET Number of years considered for balance sheet projection and business planning We asked firms how far they project a number of key balance sheet metrics into the future, and also how long a period is considered for business planning purposes. We find that for all balance sheet metrics and business planning, the most frequently used period is 5 years. For the ORSA and economic capital metrics, we note the majority of firms (88% and 83% respectively) project these for 3 to 5 years. This is consistent with what we observe for the business planning period, with 96% of firms planning for this length of time. This is not surprising, as we would expect firms economic capital and ORSA metrics to be an integral part of the business planning exercise. The projection period of the ICA/ICA+ metric is more varied, with only 61% projecting this between 3 and 5 years. We also observe 4 firms projecting this metric beyond 1 years, although we note that none of these firms projects an economic capital metric. Graph 9.3.1: How many years do you project your balance sheet? 6% 52% 56% 56% 5% 41% Percentage of firms 4% 3% 28% 35% 2% 17% 17% 17% 17% 1% 11% 4% 4% 4% 6% 9% 8% 11% 8% % % 1 % % 5 >5 % % Years ICA/ICA+ ORSA EC Business planning 127

131 MODELLING Methodology for projecting future capital requirement The methodology used for projecting capital requirements has changed over the past few years, reflecting the fact that firms are continuing to develop their approach in advance of the implementation of Solvency II. As in previous years, a risk driver based approach is the most common approach for projecting capital requirements; 56% of respondents have used this approach compared to 5% in last year s survey. However within this group there has been an increase in respondents selecting drivers at the risk module level, with a corresponding decrease in firms selecting drivers at the risk module and block of business level. The number of respondents selecting a single risk driver remains largely unchanged. We also observe the number of firms using a combination of modelling and risk drivers to project future capital requirements increasing from 14% in 213 to 25% in 214. The majority of firms (5 out of 8) using this technique are large firms, but we also observe this approach being taken by 2 medium and 1 small firm. Graph 9.3.2: How does your company project its future capital requirement? Other 3% Whole capital measure is projected using a single risk driver (e.g. assumed to run-off in line with BEL) 9% A risk driver approach where separate risk drivers are selected for each risk module 28% A risk driver approach where separate risk drivers are selected for each risk module and each block of business 19% A combination of modelling and risk drivers is used for the different capital requirements for each risk 25% Actuarial model is able to perform stresses at future dates for each risk and capital is then aggregated outside the model 9% % 5% 1% 15% 2% 25% 3% Percentage of firms 128

132 MODELLING 9.4 PROJECTING NEW BUSINESS Type of projection method used for new business 52% of respondents stated that they now run representative new business model points through the actuarial model, which is an increase of 1% from last year s survey, which is offset by a similar decrease in firms scaling current business mix to arrive at new business. As the former approach is arguably more sophisticated, this might indicate an improvement in market practice in this area. However, we observe that there has been an increase in the number of firms using the approximate method from 16% to 26%. We note that these movements may have been driven by changes in approach but may also be the result of variation in participants in the survey or in how this survey question has been interpreted. Graph 9.4.1: How does your projection method build in new business? Representative new business model points are run through the actuarial model New business is assumed to follow the same mix as current business so results are scaled 4% New business is added in approximately 26% Other 52% 18% 129

133 MODELLING How much new business do you allow for in your projections? Graph 9.4.2: How much new business do you allow for in your projections? No allowance for new business 1 year 14% 2 years 3 years More than 3 years 11% 47% 7% 21% Last year 69% of respondents indicated that they allowed for more than 1 year s new business in their projections; this has increased to 75% in this year s survey and within this group the majority of respondents allow for more than 3 years of new business. This is consistent with the responses discussed above relating to the period for business planning. With the exception of one medium sized firm, the firms that do not make any allowance for new business are either closed to new business or are small. 13

134 1 Solvency II 1.1 TRANSITION FROM ICA TO SOLVENCY II Solvency II elements incorporated in this year s ICA methodology Graph 1.1.1: Which of the following Solvency II elements do you expect to incorporate in this year s ICA methodology (based on YE13)? Solvency II risk free rate 2 Contract boundaries Liquidity premium / matching adjustment Capital eligibility rules 15 Volatility adjustment Risk margin None Number of firms We asked respondents which elements of their Solvency II methodology they were planning to incorporate into their ICA methodology for 214. Of the 32 respondents, 15 indicated that they are making adjustments to their ICA methodology to align it with Solvency II. The majority (11 out of 17) of the firms that have not sought to align the methodologies are Standard Formula firms. A number of respondents commented that whilst they have incorporated a liquidity premium into their ICA methodology, this liquidity premium is in line with their internal methodology and not the Solvency II Matching Adjustment rules. Those respondents incorporating Solvency II contract boundaries into their ICA methodology are primarily unit linked and reinsurance providers. 131

135 SOLVENCY II Basis for YE 213 ICA capital requirement calculation Graph 1.1.2: Which of the following are you aligning your ICA calculation to? 27% 43% % 17% Number of firms Historic ICA approach Solvency II or Economic Capital A partial move to Solvency II or Economic Capital Other Solvency II standard formula Solvency II internal model Economic capital model (if different) We asked respondents what their year-end 213 ICA calculation would be based on with the intention of finding out whether firms were looking to switch to their economic capital or Solvency II bases for the purpose of their ICA. Only 27% of the 3 respondents to this question are looking to maintain their historic ICA approach, with 43% indicating that they would base their ICA calculation on Solvency II or economic capital, with a further 17% partially moving in that direction. As Graph shows, only one of the 43% is moving to an economic capital basis, with the rest moving towards Solvency II. Overall, this indicates a high degree of alignment between ICA and Solvency II approaches, much more so than last year where 47% of respondents maintained their historic ICA approach. This is not surprising given the greater degree of certainty around the implementation date of Solvency II and the details of the Solvency II calculation 132

136 SOLVENCY II Respondents time-frame for aligning ICA with Solvency II Internal Model Graph 1.1.3: When do you intend to align your ICA with your Solvency II Internal Model? Already done For YE13 reporting 2 For YE14 reporting Never Other We asked respondents about their time-frame for aligning ICA with the Solvency II Internal model. There were 18 responses to this question. More than half the firms that intend to align their ICA and Solvency II Internal Model have already aligned them. Two firms selected other as their answer, for one of these firms this was because the ICA capital requirement is aligned but the balance sheet is not. The other firm is going to make this transition post 1st January 216. Four firms stated that they did not intend to align their ICA and Solvency II Internal Model. 133

137 SOLVENCY II Use of existing ICA documentation for Solvency II Internal Model Application purposes Graph 1.1.4: How much use will you make of existing ICA documentation for Solvency II Internal Model Application purposes? Significant use Some use 2 No use 6 8 We asked the respondents how much use they will make of existing ICA documentation for the Internal Model Application Process there were 16 responses to this question, i.e. those with an Internal Model. Nearly all respondents have used ICA documentation to some degree (14 out of 16) and nearly half of those make significant use of ICA documentation in their IMAP. There doesn t appear to be any significant difference in the degree to which ICA documentation is used between the firms which have a full Internal Model and those which just have a Partial Internal Model. 134

138 SOLVENCY II Expected level of changes in resources involved in the ICA when moved to Solvency II Graph 1.1.5: How much do you expect the level of resource involved in the current ICA process to change when you move to Solvency II? Significant increase Moderate increase No change 3 5 Decrease 6 16 The respondents were questioned about their expected level of resource involved in producing the ICA when they move to Solvency II. There were 3 responses to this question. The majority expect to see an increase in resource required for ICA, most of these firms expect this to be a moderate increase rather than a significant one. 3 firms expect to see a decrease in resource required. There appears to be no clear link between the expected resource requirements and the size of the firm or whether they are using an Internal Model. 135

139 SOLVENCY II Time taken in ICA to produce final report from base balance sheet Graph 1.1.6: How long does your ICA take to produce, from production of base balance sheet to finalisation of report (pre Board sign off)? Less than 1 month months months months Greater than 6 months Number of firms A large number of firms (15 out of 31) indicated that they take between 1 and 2 months to produce their ICA, which is a significant improvement on recent years. There is no obvious trend between type or size of firm and the time taken to produce the ICA. In particular, the 9 who take between 3 and 6 months include small Standard Formula firms and large Internal Model firms and include those with very simple business and those with complex business. Some firms commented that the process to produce the ICA started before the production of the base balance sheet i.e. including assumption setting, sign off etc and so in reality the timescales presented here are smaller than the total production time. However, we designed this question with ease of comparison in mind. 136

140 SOLVENCY II Time since PRA/FSA reviewed ICA Graph 1.1.7: When was the last time the PRA / FSA reviewed your ICA? Currently reviewing 16 In the last 6 months 14 In the last year 12 In the last 2 years More than 2 years Never Number of firms We asked the respondents when was the last time the PRA or the FSA reviewed their ICA. There were 3 responses to this question. 9 of the 3 firms are currently having their ICA reviewed by the PRA and a further ten have had their ICA reviewed in past year. A large majority of the Full and Partial Internal Model firms have had their ICA reviewed in at least the last year (12 out of 14). This shows that there is currently a high level of scrutiny of firms ICAs ahead of the Internal Model Approval Process. Only one firm has never had their ICA reviewed and they are a small Standard Formula firm. A further 7 out of 16 Standard Formula firms haven t had their ICA reviewed in the last 2 years. We have seen an increase in the number of ICA reviews over the last 18 months as the PRA prepares for Solvency II. 137

141 SOLVENCY II Distribution of biggest driver in ICA Graph 1.1.8: What was the biggest driver of movement in your ICA over this period? Market risk 16 Insurance risk 14 Credit risk Operational risk 12 Liquidity risk Group risk Diversification benefits Number of firms Model recalibration Modelling improvements 4 Other The main driver of change in the ICA over the last year appears to be market risk, with 15 of the 3 showing that this was the main driver. Other than 4 firms selecting insurance risk, the other main drivers were related to modelling and model calibrations all but two of these firms are Standard Formula firms. Several firms listed other reasons for movement in ICA over this period and they included ICG add-ons, moving to an IFRS basis within the ICA, and new business / exposure impacts. 138

142 SOLVENCY II 1.2 SOLVENCY II APPROAC H Approach used to calculate the Solvency II Solvency Capital Requirement Graph 1.2.1: With which method is your firm planning to calculate the Solvency II Solvency Capital Requirement? Full internal model Partial internal model 19% Standard Formula 53% 28% Over the last two years, firms have found the Internal Model Approval Process particularly challenging and while some firms have been requested by the regulator to re-enter the approval process, a number have dropped out or reduced the scope of their Internal Model to a Partial Internal Model. This is reflected in the results of our survey for this year with 19% of respondents indicating that they are taking a full Internal Model approach in comparison to 34% of respondents in 213 (37% and 43% in 212 and 211 respectively). This result reflects the recognition within the UK industry that there are significant barriers to achieving Internal Model approval and this is discouraging firms from remaining in IMAP. Where a Standard Formula approach is adopted instead, the onus remains on firms to justify the appropriateness of the Standard Formula and this is an area that a number of firms are currently working on. 139

143 SOLVENCY II Extent Internal Model currently used when making business decisions Graph 1.2.2: If you answered, Internal Model or Partial Internal Model, have you submitted any pre-application documentation for the approval of your (Partial) Internal Model? Ye s No 5 1 In order to ensure that the PRA does not have an excessive amount of documentation to read when an approval is submitted, the documentation has been reviewed in stages or themes. Therefore, some firms will have already submitted documentation and received feedback. Of the 9 Partial Internal Model firms in our survey, 7 have now submitted pre-application documentation to the PRA. Of the 6 full internal firms only 3 have submitted documentation to the PRA. All firms who have submitted documentation have received feedback mainly to a limited extent although one firm responded that they have received feedback on everything they have submitted to date. 14

144 SOLVENCY II Extent Internal Model currently used when making business decisions Graph 1.2.3: To what extent do you currently use your Internal Model when making business decisions? Fully 1 To a large extent To some extent 8 Very little Not at all Number of firms All full or Partial Internal Model firms indicated that they use their model in making business decisions. The majority (9 out of 15) indicated that they use their model in some business decisions with 4 indicating they use their model to a large extent. The 2 firms that indicated that they use their model in very few decisions are full Internal Model firms. While there has been a slight increase in the use the Internal Model for making business decisions, this indicates that firms still have a significant amount of work to do to fully embed Solvency II which is crucial to satisfying the requirements of the Use Test. This result also suggests that the decision to remain in IMAP is primarily driven by the inappropriateness of the Standard Formula rather than firms having existing and available economic capital/internal Model data which they use to steer their business. 141

145 SOLVENCY II Extent to which respondents use their Internal Model for making business decisions Graph 1.2.4: Have you rolled your Solvency II programmes into Business As Usual? Yes fully integrated 35 Partially integrated Not at all integrated N/A Number of firms Reserving Pricing Capital calulations Risk management Comparing to the results of our 213 survey the level of integration of Solvency II calculations in BAU processes remains the same. As expected, risk management and capital calculations are fairly well integrated into the BAU processes with pricing and reserving much less well integrated. A number of respondents indicated that integrating Solvency II into reserving and pricing was simply not applicable; all of these firms are adopting a Standard Formula approach. 142

146 SOLVENCY II Time taken to produce Solvency II numbers Graph 1.2.5: How long does it currently take you to produce Solvency II numbers, from production of base balance sheet to finalisation of report (pre Board sign off)? Less than 1 month months months months Greater than 6 months Number of firms The time taken to produce the Solvency II balance sheet is very similar to the production timescales for producing the ICA with the majority of firms taking between 1 and 2 months to complete the work. A large proportion of firms indicated that they take between 2 and 3 months to produce the Solvency II balance sheet potentially recognising the greater amount of work required for Solvency II submissions. The question did not specify whether this production timescale included production of all of the Solvency II documentation i.e. SFCR, RSR and QRTs. However, it is clear that some firms will require significant transformations in their process to be able to meet the ultimate Solvency II reporting requirements of 14 weeks for annual and 5 weeks for quarterly submissions. 143

147 SOLVENCY II Intention to apply for a Matching Adjustment Graph 1.2.6: Do you intend to apply for a Matching Adjustment? Ye s No 53% 47% Graph shows the proportion of all respondents (both annuity providers and non-annuity providers) that indicated they intend to apply for the Matching Adjustment. Of the 11 firms in the survey with over 1bn of annuity liabilities, 1 are planning to apply a Matching Adjustment, as expected. A further 11 have less than 1bn (but greater than zero) of annuity liabilities. Of these, 2 intend to apply for a Matching Adjustment. It is likely that a number of these do not see the capital benefit given relatively small annuity portfolios. Additionally, some of these reinsure their annuity books or back their annuity portfolio with significant proportions of non traditional fixed interest assets. There were also reinsurers in the list of those intending to apply. 144

148 SOLVENCY II Applicability of the Matching Adjustment to assets Graph 1.2.7: What is your working assumption with regard to the applicability of the Matching Adjustment for the following assets? Matching adjustment is assumed to be fully applicable 1% 6% Matching adjustment is assumed not to apply 9% Intend to restructure in order for the matching adjustment to be applicable Undecided / unclear at present Percentage of firms 8% 7% 6% 5% 4% 29% 6% 55% 5% 1% 45% 18% 3% 59% 2% 45% 4% 36% 1% % Callable bonds Commercial mortgages without make whole clauses Commercial mortgages with make whole clauses Equity release mortgage assets There is still significant uncertainty over whether certain assets (i.e. non vanilla fixed interest assets) that are used to back annuities will be admissible for Matching Adjustment purposes. We asked respondents whether they thought the Matching Adjustment would be applicable for a number of these asset types. Respondents included those who were not intending to apply for a Matching Adjustment but who had formed a view on the asset eligibility. Therefore, 21 of the 32 firms included in the survey responded. In the case of callable bonds a significant portion of respondents (59%) indicated that they are undecided or it is not clear whether the Matching Adjustment will apply and 29% indicated that they do not think it will apply. One firm has assumed a degree of restructuring will need to be performed to get them to apply and another indicated that they assume it will apply. The latter appears an optimistic view, particularly in the absence of suitable make whole or Spens type clauses. In the case of commercial mortgages, it was generally felt that unless suitable make whole clauses are in place, these assets would not qualify for a Matching Adjustment although some firms are still undecided/unclear. This is in line with our expectation as it seems this area of the asset eligibility rules is now more certain. In the case of equity release mortgages, 5 out of 11 respondents indicated that they do not think that they will achieve a Matching Adjustment in respect of these assets. 2 firms indicated that they are considering restructuring their portfolio in order to achieve Matching Adjustment and we are aware of a number of other firms in the market intending to take this approach as more firms are accepting that these assets will not be eligible in their existing form. 145

149 SOLVENCY II Overall, there is still uncertainty in the industry as to which assets the Matching Adjustment will apply. There are papers written by industry bodies to try to help drive a more consistent approach given that the rules are open to significant interpretation. The PRA has also released letters to help provide clarity and the trial Matching Adjustment submission may also assist with this. It will be difficult for the PRA to give a definitive yes or no as they will need to ensure they demonstrate a consistent approach across Europe. However, the emphasis on fixity of cash flows has helped to rule out some asset classes, such as equity release, without some form of restructure. Use of the Volatility Adjustment Graph 1.2.8a: Do you intend to use the Volatility Adjustment for all business where a Matching Adjustment does not apply? Ye s No 47% 53% 146

150 SOLVENCY II Graph 1.2.8b: Firms applying for the Matching Adjustment and intending to use the Volatility Adjustment Applying for MA and intend to apply VA to all other businesss 1 Not applying for MA but intend to apply VA Applying for MA and do not intend to apply VA to other business 8 Do not intend to apply either MA or VA Number of firms We asked respondents whether they intended to use the Volatility Adjustment for all business where a Matching Adjustment does not apply. Just over half of the respondents indicated that they do intend to use a Volatility Adjustment where a Matching Adjustment does not apply. Graph 1.2.8b shows the result of analysing the responses to this question alongside the firms responses to whether they intend to apply a Matching Adjustment. The 8 firms that indicated that they intend to apply for Matching Adjustment as well as using the Volatility Adjustment on other business are predominantly providers with a high concentration of annuity business as well as a broad mix of other types of business. Of those firms that do not intend to use the Volatility Adjustment, 7 are applying to use the Matching Adjustment and in some cases this would cover all or at least the majority of their business. A number of respondents indicated that they are still investigating the use of the Volatility Adjustment and some indicated that they do not think that the Volatility Adjustment would apply to unit linked business although this is not something that is specifically noted in the rules. We were surprised to see that such a large number of firms are not intending to use the Volatility Adjustment, although this could be to do with simplicity of approach given that the result without Volatility Adjustment also needs to be disclosed. Given the HM Treasury consultation which, at the time of writing, has just been released, it is not clear whether firms will need to apply to use the Volatility Adjustment. If this is the case, then this is likely to reduce the attractiveness of the adjustment. 147

151 SOLVENCY II Use of transitional provisions Graph 1.2.9: Do you intend to apply to use transitional provisions? Yes - technical provisions 14 Yes - risk free rate Yes - other 12 No 1 Not yet decided Number of firms Of the 7 firms intending to apply for transitional provisions, 6 of these are large firms and the remainder is a niche firm who is impacted by Matching Adjustment rules. All of those that intend to apply for transitional provisions are intending to use the technical provision transitional, which is the broadest transitional provision. A large number of respondents (13 out of 32) indicated that they have not yet decided whether they will apply for transitional provisions. This reflects the uncertainty that still exists around Pillar 1 i.e. whether the Matching Adjustment will be granted. One respondent commented that they are only considering transitional provisions as a contingency plan. We are also aware of a number of other firms that are considering transitional provisions as a contingency plan depending on the ultimate position on other aspects of Pillar 1 that are currently uncertain. However, firms will need to make these decisions quickly as they will need to apply for approval and will also have had to give an indicative view of whether they intend to apply for this approval to the PRA earlier this year. 148

152 SOLVENCY II Highest priority part of the Solvency II programme Graph 1.2.1: Which part of the Solvency II programme is your current highest priority? Standard formula 8 Internal Model and Partial Internal Model 6 Number of firms Pillar 1 methodology 1 Pillar 1 valuation ORSA 1 1 Embedding and Use Test 1 1 Analysing requirements of the QRT s Speed of reporting IMAP We asked respondents which part of their Solvency II program was currently their highest priority. We have not asked this question in previous years but we know from working with a number of clients on Solvency II over the years, that for large firms, the focus has shifted from Pillar 1 methodology and ORSA to Pillar 3 over the last year or so. Some smaller firms however have just started to consider the Pillar 1 methodology and valuation and the ORSA. The results largely back up this general observation. For the large Internal Model or Partial Internal Model firms, their main focus is IMAP (8 out of 15), with the Pillar 3 requirements, including the speed of reporting (4 out of 15), being the next biggest concern. For the Standard Formula firms, the main concerns are around the ORSA (6 out of 17), the Pillar 1 valuation (5 out of 17) and the analysis of Pillar 3 requirements (4 out of 17). The results here are close, which may imply that the Standard Formula firms have only recently started to fully engage with Solvency II and are having to pick up their projects across all three pillars. Only 2 respondents indicated that Pillar 1 methodology was their highest priority reflecting that firms have now, in general, reached a point where their intended methodology is well defined and stable or that any uncertainties are dependent on external factors such as emerging regulation or regulatory approval. Interestingly, these two are Internal Model firms it may be that the focus is on the documentation and validation of the methodology rather than the methodology itself. 149

153 SOLVENCY II Valuation and stress of pension scheme deficit Graph : How do you value/stress your pension scheme deficit for the purpose of the Solvency II balance sheet and SCR? IAS19 basis Funding basis Best estimate basis 1 1 Other Not currently stressed 5 Solvency balance sheet SCR We asked survey participants what methodology they were planning to use to value their defined benefit pension scheme under both the base Solvency II balance sheet and in the SCR. The results above are shown only for the 18 firms to whom this question is applicable. The technical specifications for the various assessment exercises (including those for the preparatory phase, released on 3 April 214) suggest that the IFRS (i.e. IAS19) valuation is consistent with the Solvency II requirements for the base balance sheet. This is the method adopted by most firms (62%). However, the funding basis is used by 5 firms, which has typically been a requirement of the current ICA regime. One firm uses a funding basis in the base case and then stresses the IAS19 position and a further 2 firms use different (unspecified) methods in their stress position compared to the base. In addition 2 do not currently stress the base position both of these use IAS19 valuation in the base case. 15

154 SOLVENCY II Risks included when stressing pension scheme deficit Graph : Which risks do you include when stressing your pension scheme deficit on the Solvency II balance sheet? Number of firms Property risk Interest rate risk Equity risk Credit risk Currency risk Inflation risk Mortality risk Longevity risk Morbidity risk Expense risk Other Not applicable The Level 3 guidelines and the preparatory phase technical specification have suggested that only market risks need to be stressed when calculating solvency risk capital on the pension scheme under the Standard Formula. However, it isn t clear whether this will also be the case for an Internal Model calculation we suspect it will not be the case given the PRA s general requirement to stress non-market risks under the ICA regime. From Graph , it is clear that the market risks are the most prevalent stress, but non market risks are also stressed by a large number of firms. Additionally, we asked firms whether they allowed for pension scheme risk in their Solvency II Risk Margin calculation. The Solvency II Risk Margin calculation assumes a transfer of the insurance obligations of a firm to another (reference) firm. As the pension scheme is not a part of the insurance obligations, we were surprised that 2% of firms include pension scheme in their Solvency II Risk Margin calculation. 151

155 SOLVENCY II Overall Solvency II effect Graph : Please indicate the overall Solvency II effect for your company by comparing BEL + ICA to Solvency II BEL + RM + SCR 11 4 Broadley neutral Beneficial 11 4 Adverse, material Adverse, not material We asked respondents what the overall impact of Solvency II was in comparison to ICA. i.e. comparing BEL + ICA to BEL + RM + SCR. Half of the respondents indicated that they are worse off under Solvency II, and of those 15, 11 indicated that they are materially worse off. It is not a surprise that 8 of those 11 have over 1 billion of annuity liabilities. The majority of the partial or full Internal Model firms (9 out of 14) are worse off under Solvency II than ICA. Given the alignment of ICA and SCR that has been observed, this reflects the impact of the additional Risk Margin under Solvency II that is absent from the ICA regime. 1 of the 16 Standard Formula firms either indicated that they are better off under Solvency II or that the impact is broadly neutral. This indicates that for these firms there is a benefit from replacing the ICA calibration by the Standard Formula calibration that offsets the inclusion of the Solvency II Risk Margin. 152

156 SOLVENCY II Solvency II Risk Margin as a proportion of SCR for non-profit business Graph : What is the proportion of your Solvency II Risk Margin to your postdiversification SCR for your NP fund? % RM < 15% 5 15% RM < 2% 2% RM < 25% 4 25% RM < 3% 3% RM < 35% 35% RM < 4% 4% RM < 45% Mean number of firms % RM < 5% RM 5% We asked respondents what their Solvency II Risk Margin as a proportion of SCR is for non-profit business. As Graph shows we received a wide range of responses that are fairly evenly distributed between % and over 5%. As the cost of capital is defined as 6%, the variability must be because of differences in the mix of business, the run-off methodology used, as well as differences in the run-off profile driven by the demographics of the book. We aware that a number of firms are using different techniques, the most common being a risk driver approach which can be either fixed or time dependent. Other approaches taken include using proxies such as BEL or assets as well as a full projection of the base and stressed balance sheet. 153

157 SOLVENCY II Management actions resulting in a zero SCR for with-profits business Graph : Does your ability to apply future management actions mean that you have a zero SCR for with-profits business (excluding operational risk) Ye s No 11% 89% We asked respondents whether the application of management actions means that they have a zero SCR for with-profits business other than operational risk. Under Solvency II, the adjustment for the loss absorbency of technical provisions allows a firm to reduce the SCR by the amount of loss absorbency i.e. the discretionary element of the BEL that can be removed if a stress scenario occurred. If the discretionary element of the BEL is greater than the Basic SCR then firms can report a zero SCR other than the SCR for operational risk which is added to the Basic SCR after the adjustment for loss absorbency of technical provisions is applied. Under the current regulatory regime, closed with-profits funds typically show zero RCM in PRA Form 19 where they have a management action to remove discretionary liabilities in the stress scenario, and this discretionary element is greater than the RCM. Only 2 firms indicated that they expect they will have a zero SCR once management actions are allowed for under Solvency II. Of the 18 firms that indicated this question was applicable, 6 firms currently show a zero RCM for all of their funds and a further 3 firms show a zero RCM for the majority of their funds. This indicates that either funds are not applying the same management actions for Solvency II or that the SCR under Solvency II is significantly more onerous than the current RCM. For those firms that are in run-off and that are distributing their estate this will have significant implications; in particular, the distribution of their estate in a fair manner, whilst remaining solvent throughout the run-off. 154

158 SOLVENCY II Differences in treatment between economic capital and Solvency II calculations Graph : Which of the following areas do you treat differently when performing Economic Capital versus Solvency II calculations? Number of firms Risk free rate Contract boundaries Risk margin cost of capital charge Treatment of subordinated debt Treatment of pension Treatment of asset managers US equivalence DTA allowance scheme risk Credit risk adjustment Matching adjustment Volatility adjustment More management actions in SII More management actions in economic capital Other We asked respondents which areas they treat differently between their internal economic capital calculations and Solvency II calculations. As expected, the key areas where the treatment differs is the discount rate (including Matching Adjustment / liquidity premium allowance, Volatility Adjustment, credit risk adjustment), contract boundaries, the cost of capital charge and the treatment of pension scheme risk. These are the known areas of difference between current ICA and the proposed Solvency II requirements; therefore these are generally in line with our expectations. It is interesting to note that 2 firms adopt more management actions in their economic capital calculation (potentially due to there being less rigorous requirements around justifying the actions) and one firm adopts more management actions in Solvency II. 155

159 SOLVENCY II 1.3 PROFIT AND LOSS ATTRIBUTION Items covered by Profit and Loss Attribution Graph 1.3.1: Which items will your Profit and Loss Attribution cover? Best estimate liability 25 Risk Margin SCR 2 MCR Assets Own funds only Number of firms The Profit & Loss Attribution allows a firm to define what profit and loss is and could, for example, either include or exclude capital requirements. We asked about the granularity of the analysis performed. It is clear that there are various different approaches, with 6 firms only considering the change in Own Funds (i.e. not drilling into any more detail) and others considering the change in best estimate liabilities, Solvency II Risk Margin and assets. There are 12 firms who include the SCR in their Profit & Loss Attribution. Although a change in SCR is likely to be required by Boards and/or senior management to understand the reasons for the movements, it is not a Solvency II requirement. Of those 12 firms, 7 are large internal (or partial internal) model firms and may have been performing such analyses on an ICA basis. 156

160 SOLVENCY II Methodology used to quantify the movements Graph 1.3.2: If you analyse the change in Risk Margin or SCR, how do you quantify the movements within your Profit and Loss Attribution? Ratio in line with BEL movements Perform full model runs Use sensitivity runs and ratio these to observed events Other 6 Change in Solvency II Risk Margin Change in SCR 1 4 We asked respondents what methodology they used to analyse the Solvency II Risk Margin and the SCR. We have presented the results of those who indicated that they perform this analysis based on the previous question. For Solvency II Risk Margin purposes, the most widely used method appears to be to approximate by using a ratio in line with the movements in best estimate liability. However, 3 firms do perform full model runs to analyse the change these are either Standard Formula or small Partial Internal Model firms. The other responses largely indicated that the methodology is still to be developed. For the change in SCR, the most widely used method was to use full model runs. These are the same 3 firms mentioned above, but also one of the large Internal Model firms. This was unexpected, given how onerous this calculation is likely to be. Similarly, the other responses indicated that the methodology is still to be developed. 157

161 11 Financial Reporting Over the last few years we have seen a number of developments in the financial reporting arena, most notably a move away from the more traditional embedded value reporting with increased focus on Solvency II, economic capital and IFRS 4 phase II. In this section we report our findings from the questions in key areas including analysis of change, experience analysis and solvency monitoring, IFRS and embedded value reporting (both now and in the future) ANALYSIS OF CHANGE Reporting metrics We asked firms on which reporting metrics they currently perform or intend to perform an analysis of change. Of 32 respondents, most of these perform an analysis of change on Solvency I Pillar 1, with approximately two-thirds also performing an analysis of change on ICA/ICA+, embedded value and Solvency II. Graph : On which reporting metrics do you currently or do you intend to (in 214) perform an analysis of change? Solvency I, Peak 1 3 Solvency I, Peak 2 ICA / ICA+ 25 IFRS 2 Embedded value Solvency II, Pillar 1 Solvency II, Pillar 2 Economic capital Number of firms

162 FINANCIAL REPORTING Granularity of analysis of change To gain an understanding of the level of granularity of analysis of change, we asked respondents at which level they perform the analysis of change under the metrics included in the previous question. Graph : What level of granularity do you, or do you intend to, perform your analysis of change at? Solvency I Peak I Group level 16 Entity level 14 Fund Major product category 12 Other Number of firms Solvency I Peak II Group level 16 Entity level Fund Major product category Other Number of firms

163 FINANCIAL REPORTING Graph (continued): What level of granularity do you, or do you intend to, perform your analysis of change at? ICA/ICA+ Group level 16 Entity level 14 Fund Major product category 12 Other Number of firms IFRS Group level 16 Entity level Fund Major product category Other Number of firms

164 FINANCIAL REPORTING Graph (continued): What level of granularity do you, or do you intend to, perform your analysis of change at? Embedded Value Group level 16 Entity level 14 Fund Major product category 12 Other Number of firms Solvency II Pillar I Group level 16 Entity level 14 Fund Major product category 12 Other Number of firms

165 FINANCIAL REPORTING Graph (continued): What level of granularity do you, or do you intend to, perform your analysis of change at? Solvency II Pillar II Group level 16 Entity level Fund Major product category Other Number of firms Economic Capital Group level 16 Entity level Fund Major product category Other Number of firms Across the different reporting metrics, by far the most widely used level of granularity is at the major product category. This is a change from last year when majority of respondents were performing the analysis of change at entity or fund level. This change may be driven by the Solvency II profit and loss requirements. 162

166 FINANCIAL REPORTING 11.2 THE MOST IMPORTANT METRIC TO THE BOARD The responses to this question indicate that the Board currently gives most weight to IFRS results closely followed by ICA/ICA+ results. It is interesting to note that embedded value was the joint most widely used measure in last year s survey; however it appears to have had a significant decline in importance to the Board over the last year. Unsurprisingly, Solvency II and economic capital stand out as being the dominant metrics in the future. Graph : What metric is currently given the most weight by Board and senior management, and which do you expect to be in the future? Number of firms Solvency I, peak 1 2 Solvency I, peak ICA / ICA+ IFRS Embedded value Solvency II, Pillar Solvency II, Pillar 2 Economic capital Other Currently In the future 163

167 FINANCIAL REPORTING 11.3 EXPERIENCE ANALYSIS As expected, most firms conduct their experience analysis investigations annually. Where more frequent investigations were carried out, this was typically for withdrawals or expenses rather than mortality. This was also as we expected, as lapses and expenses are usually more volatile than mortality claims. Smaller firms were less likely to carry out more frequent than annual investigations, most likely due to resource constraints. Graph : How often are experience analyses performed? Number of firms More frequently than annually Annually At least every two years At least every three years Never Annuitant mortality Non-annuitant mortality Expenses Withdrawals 164

168 FINANCIAL REPORTING 11.4 SOLVENCY MONITORING This part of the survey focuses on how respondents monitor each of their solvency capital measures both in terms of frequency and approximation of modelling. Frequency of Solvency I monitoring The PRA expects life insurers to continuously monitor and meet their capital resource requirements. Under INSPRU, insurers must also continually monitor their solvency. Graph a: How frequently are you monitoring your Solvency I Pillar 1 capital measure? 2 15 Number of firms Daily Weekly Monthly Quarterly Half-yearly Annually Via approximation Via full quantification 165

169 FINANCIAL REPORTING Graph b: How frequently are you monitoring your Solvency I Pillar 2 capital measure? 2 15 Number of firms Daily Weekly Monthly Quarterly Half-yearly Annually Via approximation Via full quantification On a Pillar 1 basis most respondents carry out full quantification on either a halfyearly or a quarterly basis, with approximate methods being used to monitor solvency more frequently than this. A similar pattern is observed on the Pillar 2 basis (i.e. ICA), although it is noted that fewer respondents carry out full quantifications on a quarterly basis. It is expected that the approximations will include a subset of risks being monitored more frequently than on a full calculation. These include for example equity, interest rates and credit risks. The derivation of the approximation is likely to be related to movements in particular market indices and other market movements. 166

170 FINANCIAL REPORTING Frequency of Solvency II and ORSA monitoring We asked respondents to comment on their practice in relation to full and approximate regular solvency monitoring of all capital metrics. Graph a: How frequently are you monitoring your Solvency II capital measure? 2 15 Number of firms Daily Weekly Monthly Quarterly Half-yearly Annually 4 Via approximation Via full quantification 167

171 FINANCIAL REPORTING Graph b: How frequently are you monitoring your ORSA capital measure? 2 15 Number of firms Daily Weekly Monthly Quarterly Half-yearly Annually Via approximation Via full quantification Just over half of the respondents are monitoring Solvency II Pillar 1 via full quantification on an annual basis, with the rest of the respondents performing halfyearly or quarterly calculations. A greater proportion of respondents are monitoring the ORSA via full quantification on an annual basis, i.e. there is less monitoring of the ORSA carried out half-yearly or quarterly. There is limited monitoring of either Solvency II Pillars via approximate methods, which indicates that respondents are still developing these processes. 168

172 FINANCIAL REPORTING Frequency of Economic Capital monitoring Graph : How frequently are you monitoring your Economic Capital? 2 15 Number of firms Daily Weekly Monthly Quarterly Half-yearly Annually Via approximation Via full quantification Our survey has shown that half of the respondents monitor their economic capital (via a full quantification) more frequently than annually. The remaining half performs a full calculation on an annual basis. Monthly monitoring is typically performed via approximation 169

173 FINANCIAL REPORTING 11.5 IFRS 4, PHASE II Preparation for changes In this section, respondents were asked about the proposed changes to IFRS reporting and their status of preparedness. Approximately two-thirds of the respondents are aware of the changes in IFRS 4 phase II but few have started to understand the balance sheet impacts, the system implications or the programme management requirements to implement this. Those who have started to understand the implications tend to be either large firms or subsidiaries of large firms. Graph : To what extent are you prepared for the changes in IFRS accounting for insurance contracts (IFRS 4) Aware of changes, but not started to consider the implications 2 Started to think through the numerical impacts on the balance sheet, emergence of profit etc Started to think through the system implications 15 Started to think through the programme management implications Haven't given this any consideration Number of firms

174 FINANCIAL REPORTING Areas causing most concern In June 213, the IASB re-exposed the draft standard for IFRS 4 phase II and specifically asked for feedback on following six areas, the use of other comprehensive income, unlocking of the contractual service margin (CSM), mirroring, presentation, transitional arrangements; and, the complexity and clarity of drafting. We asked respondents which of the above areas were the cause of most concern and the 3 areas causing most concern were presentation, complexity and clarity of drafting, and unlocking of the CSM. Graph : Rank the following topics in order of greatest concern under IFRS 4 phase 2 OCI High cause for concern 1 Moderate cause for concern Low cause for concern 8 No cause for concern Number of firms

175 FINANCIAL REPORTING Graph (continued): Rank the following topics in order of greatest concern under IFRS 4 phase 2 Unlocking of CSM High cause for concern 1 Moderate cause for concern Low cause for concern 8 No cause for concern Number of firms Mirroring High cause for concern 1 Moderate cause for concern Low cause for concern 8 No cause for concern Number of firms

176 FINANCIAL REPORTING Graph (continued): Rank the following topics in order of greatest concern under IFRS 4 phase 2 Presentation High cause for concern 1 Moderate cause for concern Low cause for concern 8 No cause for concern Number of firms Transitional arrangements High cause for concern 1 Moderate cause for concern Low cause for concern 8 No cause for concern Number of firms

177 FINANCIAL REPORTING Graph (continued): Rank the following topics in order of greatest concern under IFRS 4 phase 2 Complexity and clarity of drafting High cause for concern 1 Moderate cause for concern Low cause for concern 8 No cause for concern Number of firms The CSM is a new component to the balance sheet and there isn t an equivalent component under existing balance sheets, so it is not surprising that this is a cause for concern. There are also a number of new presentational requirements which will lead to process, data and system changes; hence it is understandable that this features heavily in the results. The re-exposure draft issued in 213 is generally considered to be one of the most complex set of accounting requirements to be issued by the IASB, hence the IASB have said that they will allow approximately 3 years from issuing the final standard to the implementation date. The concerns over the complexity are reflected in the results above. 174

178 FINANCIAL REPORTING 11.6 CURRENT EMBEDDED VALUE REPORTING Out of the 32 respondents to the survey, 63% (2) currently compute an embedded value, with the majority of these firms disclosing results externally. 37% do not currently compute an embedded value, of which over half are mutual firms. These results are largely unchanged from last year. Graph : Do you currently compute an embedded value? Yes disclosed externally Yes for internal use only No 37% 47% 16% The type of embedded value balance sheet produced is shown below. This shows that there is a variety of methods being used, with no single dominant approach being adopted in the market. Compared to last year, there are now more firms adopting full Market Consistent Embedded Values or Market Consistent European Embedded Values. Graph : What type of embedded value do you currently use? Market Consistent Embedded Value (full in line with CFO Forum Principles) Market Consistent European Embedded Value 2% Traditional European Embedded Value Traditional Embedded Value 35% 25% 2% 175

179 FINANCIAL REPORTING 11.7 EMBEDDED VALUE REPORTING IN THE FUTURE With the increased focus on Solvency II and the tight reporting timelines, firms are considering how they will streamline their processes to ensure they can produce all the required metrics in the required timeframe under the new regime. In particular, firms have been considering the value of embedded value reporting once Solvency II is in place. Solvency II is intended to be market consistent and be based on best estimates, and therefore there may be a reduced benefit for insurers in continuing to perform and report on an embedded value basis. Conversely, embedded value reporting post- Solvency II could still be useful for similar reasons as currently, i.e. to give insight into any prudence embedded in the regulatory reporting figures (for example the application of contract boundaries required for Solvency II) and to show a market consistent value of the business to shareholders. From the survey responses and our discussions with firms, it is clear that many respondents are deferring decisions on embedded value until the final Solvency II rules are published, and the CFO Forum provides guidance. We asked respondents whether they expect to compute an embedded value upon implementation of Solvency II and the results are shown in Graph Graph : Do you expect to compute an embedded value upon implementation of Solvency II? Yes for first 3 years only Yes beyond first 3 years 5% No Not yet decided 2% 55% 2% The number of firms expecting to move away from embedded value once Solvency II is implemented continues to increase. Out of 2 respondents who currently produce embedded value results, a quarter of firms intend to continue reporting embedded value, just under a quarter of firms intend to discontinue reporting embedded value, and just over half of firms are undecided. 176

180 FINANCIAL REPORTING 11.8 KEY DIFFERENCES BETWEEN EMBEDDED VALUE REPORTING AND SOLVENCY II IN THE FUTURE Differences between PVFP and Solvency II BEL For respondents that currently produce an embedded value balance sheet, we asked about the differences in assumptions that will exist between those used to calculate PVFP and Solvency II BEL. Graph : What differences in assumptions will there be between those used to calculate your embedded value PVFP and your Solvency II BEL? Embedded value will use a different risk free rate curve to Solvency II Embedded value will use different liquidity premium assumptions to Solvency II Embedded value will use different contract boundary definitions to Solvency II Embedded value will use different investment returns assumptions to Solvency II Other differences Number of firms The contract boundaries assumption continues to be the assumption change most commonly made, followed closely by economic assumptions. Generally we are seeing a lower number of respondents making a change between the Solvency II and embedded value metrics; however this may simply reflect the fact that fewer respondents intend to produce an embedded value balance sheet after the implementation of Solvency II. 177

181 FINANCIAL REPORTING Differences between CNHR and Solvency II Risk Margin We also asked respondents how their embedded value CNHR will differ from the Solvency II Risk Margin, with results shown below. Graph : How will your CNHR differ from the Solvency II Risk Margin Different risks assumed to be non hedgeable 1 Different cost of capital assumption used Different diversification benefits used 8 CNHR not calculated Other Number of firms Out of 2 respondents, 45% of respondents (9) indicated they will use a different cost of capital assumption. 3% of respondents (6) indicated they will use different diversification benefits. All of those using different diversification benefits are also using a different cost of capital assumption. 178

182 12 Tax A balance sheet drawn up under any reporting basis should generally include deferred tax balances. The Solvency II economic balance sheet is no exception. Furthermore, tax has an important role to play in the calculation of the Solvency II SCR, which requires the deferred tax balances which would appear on a hypothetical post-stress balance sheet to be computed. The movement in the deferred tax balances relative to the Solvency II economic balance sheet is referred to as the loss absorbing capacity of deferred taxes ( LADT ) and has the potential to materially mitigate the overall capital requirement under Solvency II. There has been increased regulatory focus on tax during 214, with the PRA releasing Supervisory Statement 2/14 Solvency II: recognition of deferred tax in April and the draft guidance on the Standard Formula LADT, published by EIOPA on 2 June. While there remain areas of uncertainty, it is becoming clear that large tax credits and/or net deferred tax assets ( DTAs ) will be a particular area of scrutiny for supervisors. Accordingly, firms are assessing the sources and quality of the evidence that supports their LADT, particularly where future profits are assumed. One of our principal observations in this section of the survey is that around half of respondents are recognising net DTAs on their stressed balance sheet. That is broadly in line with last year but there is some evidence of developments in methodology for example a shift towards relying on a partial release of the Solvency II Risk Margin (as opposed to a full release) as a source of future profits. The first four questions in this chapter consider the calculation of the LADT and the support of the deferred tax position on the post-stress balance sheet; the next two questions consider changes to the methodology in future periods and how tax is dealt with by the Internal Model; the final question concerns the consistency of tax methodologies across different reporting bases. 179

183 TA X 12.1 DEFERRED TAX ON THE STRESSED BALANCE SHEET Treatment of non-zero loss absorbency of tax The LADT component of the SCR can be thought of as the tax credit which arises as a consequence of the shock event(s) assumed when deriving the SCR. The loss arising from a shock event potentially gives rise to a DTA, which can only be valued to the extent it can be offset against appropriate deferred tax liabilities ( DTLs ) on the economic balance sheet or to the extent that there is evidence of future profits, in line with IAS 12. In cases where a LADT was calculated in respect of the whole business (or a significant part of the business), we were interested to know the extent to which this LADT was recognised. Graph sets out the responses to this question. As can be seen, there was a wide range of practices: Graph : If for a significant part of your business a non-zero loss absorbency of tax (LADT) is calculated, is that LADT: Fully recognised on that basis that it was less than the deferred tax liability (DTL) on unstressed balance sheet Capped at the level of the DTL on the unstressed balance sheet Supported by reference to DTLs and prior year income or profits Supported by reference to future profits (in addition to first four items above) 1 4 Supported by reference to future management actions and/or future tax planning (in addition to the first five items above) Recognised at the full tax rate but not fully tested Other 7 2 Of the 21 respondents who did not select n/a to this question, 43% (nine) of them indicated that their position was supported by a relatively conservative methodology (i.e. with shock losses offset against DTLs and prior year profits and no net DTA being recognised). In 4 of these cases, the LADT was capped at the level of the DTL implying that perhaps there would be scope to recognise a larger LADT if future profits were also anticipated. In contrast, eight of the relevant respondents (38%) also relied on future profits to support the LADT and in one case this included future management actions or tax planning. Taking full credit for the LADT without testing whether it can be supported in full remains a minority practice. Other responses referred to the movement in unrealised capital gains and capped at the future tax payments projected by the valuation model. Overall, the broad pattern of practices was consistent with the 212 and 213 surveys. 18

184 TA X Reasons for zero loss absorbency of tax We were also interested to understand why some firms may not recognise LADT for a significant part of their business. The majority of these respondents were friendly societies or mutual organisations. As can be seen from Graph , for the majority this would seem to be a reasonable result due to unvalued tax losses already existing on the Solvency II base balance sheet. Four respondents were yet to do sufficient work to recognise a benefit. Graph : If for a significant part of your business, the loss absorbency of deferred tax was assumed to be zero, was this because... There were unrecognised deferred tax assets in respect of losses relating to that part of the business even in unstressed conditions Preliminary work indicated that the calculation would be too onerous or the benefit too marginal due to the difficulty of obtaining supporting evidence No detailed work has been undertaken, therefore no benefit of LADT is recognised

185 TA X Assumed sources of future income / profits to value deferred tax assets on the stressed balance sheet We then asked which sources of future income or profit were included in the methodology for valuing DTAs on the stressed balance sheet. Where the LADT cannot be netted with DTLs, some projection of future taxable profits/income is required. Respondents were asked to select all of the options which applied. Graph : Which of the following sources of future income or profits does your tax methodology allow you to anticipate when valuing deferred tax assets on the stressed balance sheet? Other We use business as usual forecasts which will include some or all of the above items but these are not separately identified Post shock management actions (e.g. recapitalisations) 2 1 Income or profits in excess of the risk free rate Post shock tax planning to move income or profits and losses into the same entity Income or profits from non-insurance entities in the group 5 2 Reversal of interest rate shock on bonds which can be held to maturity Equity markets reverting to mean post-shock 2 1 Investment return on excess capital 11 Taxable profits arising on future new business 9 Unwind / release of some or all of the risk margin (or equivalent allowance for variation from the best estimate) Number of firms There were 17 respondents to this question and graph sets out the responses. Note that 17 respondents is considerably more than the number of respondents who stated in the first question that they have recognised an LADT in excess of DTLs; this is not necessarily inconsistent as this question concerned what was allowed in the methodology. Thus these responses indicate that respondents may be considering generalising their methodology so they are not solely dependent on an offset against DTLs and prior year profits in future should circumstances change. The four most common sources of income anticipated by respondents were the release of the Solvency II Risk Margin, allowance for new business profits, investment return on excess capital and, to a lesser extent, income/ profits in excess of the risk free rate. This pattern was consistent with our 213 survey. Most respondents selected at least two sources. Other responses noted a movement in unrealised capital gains and future tax payments projected by their valuation model. Two respondents relied on profits from non-insurance entities in the group. It is expected that Solvency II guidance will not allow this for the Standard Formula. Of the respondents selecting this option, one was using a Partial Internal Model and one an Internal Model. 182

186 TA X Release of Solvency II Risk Margin as a source of future profits As a follow-up question, we asked about the extent to which the release of the Solvency II Risk Margin was assumed as a source of future profit, where applicable. Graph : Under Solvency II, when valuing deferred tax assets on the stressed balance sheet, does your methodology consider that the release of the Risk Margin is a source of future profits? Yes, in full Yes, for closed books/products only 2 Yes, to the extent releases are expected within the normal business 2 Yes, in part, on some other basis No It is interesting to note that there is an even split between respondents who anticipate the release of some (or all) of the Solvency II Risk Margin and those who do not. This pattern is consistent with our 212 and 213 surveys. However, fewer respondents are now using the release of the Solvency II Risk Margin in full (two respondents this year compared to nine in 213). There appears to be a trend towards restricting releases (e.g. to that expected within the normal business planning cycle or which relate to closed products). Probing the data more deeply reveals that 4 respondents have become more conservative on this point since 213 (generally smaller firms moving from full recognition to no recognition) while 5 respondents had become less conservative (generally medium or large firms moving from no recognition to partial recognition). In summary, the current position is that smaller firms generally have not anticipated any Solvency II Risk Margin release but that larger ones have adopted a range of different practices to allow for some release. It will be interesting to see how practice continues to evolve on this point. The PRA s statement 2/14 indicates their scepticism on the point. It may be the case that either firms are not yet prepared to concede the point in full or respondents have yet to reassess their methodology following the release of the PRA supervisory statement. Three respondents indicated that some Solvency II Risk Margin release was anticipated in this question despite not selecting that option in the previous question. In one case, the release was in respect of closed books/products only, in another it was on some other basis. This perhaps indicates that some respondents are being flexible in their application and depart from their general approach in specific circumstances. 183

187 TA X 12.2 FUTURE POSITION AND TAX MODELLING Following the change to the UK life insurance corporation tax regime from 1 January 213, a number of firms have large DTLs on their IFRS/ Solvency II base balance sheet reflecting the transitional adjustment arising. In most circumstances this transitional adjustment is released into current tax over a ten year period on a straight line basis. In a new question for 214, we were interested to know whether the run off of this DTL would materially impact the ability to take credit for the LADT in future. Graph : Will the run-off of IFRS deferred tax liabilities relating to the transition to the post-212 tax regime materially impact your ability to take credit for the loss absorbency of deferred tax? Yes but we have quantified the impact and will explore any relevant mitigating steps The impact on future LADT is expect to be adverse but has not been considered further No, the impact is immaterial No, we have net deferred tax assets in respect of trade profits timing differences Unknown 1 9 Of the 22 respondents who did not answer n/a to this question, 9 (41%), including several large firms, did not expect a material impact. This could be because the transitional DTLs are not material or because there are material other DTLs or sources of future profits. Comparing the result to Graph indicates that most of those who expected a material adverse impact are relying only on DTLs and/or prior year profits to support LADT. We would expect these firms to more actively consider other sources of future profit as the transitional DTLs run off. Conversely, most of those who expected an immaterial impact are already relying on future profits. 184

188 TA X Determination of the deferred tax asset arising from the shock loss We next went on to ask about how firms were modelling tax within their Internal Model. Graph : In terms of calculating the LADT, how is the deferred tax asset arising from the shock loss determined The internal model (or equivalent) calculates a pre tax SCR and the tax adjustment is a discrete final step outside of the model 3 The internal model (or equivalent) works out the loss(es) and the potential DTA but caps the DTA by reference to a tax capacity cap introduced as a user-defined parameter The Internal Model (or equivalent) works out the loss(es) and the capacity to value tax losses is determined dynamically by the Internal Model Other Just over half of the relevant respondents (1 out of 18) indicated that their Internal Model calculated a pre-tax SCR in the first instance, with the tax adjustment being a discrete final step. A further five respondents (28%) stated that their Internal Model valued tax losses by reference to a measure of tax capacity, and for two of them the tax cap was valued dynamically. The pattern is similar to that observed in the previous two years. The smaller insurers who answered this question tended to indicate that the tax adjustment was outside the model or selected other. 185

189 TA X 12.3 TAX METHODOLOGY Finally, we asked if there are any areas where the tax methodology differed (or was expected to differ after taking into account of planned developments) between the various bases used (ICA, ICAS+, Solvency II and economic capital, where relevant). Graph shows the range of differences indicated by respondents. Graph : After taking into account planned developments, are there any areas where your expected tax methodology will vary between the bases used? Other 4 The future profits used to support deferred tax assets are re-computed (for example as a consequence of taking a different view on contract boundaries) 5 Time horizon over which deferred tax assets can be recovered. 7 The extent to which loss absorbency of tax is taken into account in calculating the capital requirement run-off used to define the risk margin (or equivalent) 3 Whether deferred tax assets on the base balance sheet are tested for impairment as a result of a shock 7 Discounting of deferred tax balances Number of firms Half of the respondents (16 out of 32) indicated that they used (or planned to use) differing tax methodologies in at least one of the above areas across different reporting bases. There was a clear bias within this subset in favour of medium-sized / larger firms (13 of the 16 relevant respondents). Where other was selected, respondents had still to consider the position. Interestingly, the responses provide evidence that there is some departure from IAS 12 and/or Solvency II rules as regards the quantification of deferred tax balances on some reporting bases (for example on the discounting of deferred tax balances and writing down DTAs in stressed conditions). Such departures are not necessarily inappropriate. It will be interesting to see if the areas of divergence become more numerous. If, for example, the Regulator insists on prudent interpretations or an onerous level of evidence we could see divergence of practice between regulatory measures and economic capital measures. 186

190 13 Acknowledgements Many people dedicated their time to this survey to make it a success and we would like to take this opportunity to thank each of them. We had responses from 32 different respondents and many more actuaries gave up their time for surveys to be completed. We appreciate the time and care that is put into your responses, the feedback you provided to us and your willingness to discuss your answers with us. A list of the respondents participating in this year s survey can be found in Section 14. CORE TEAM John Jenkins Partner Jane Parker Survey Lead Kim Owen Survey Manager Simon Hogley Survey Manager 187

191 ACKNOWLEDGEMENTS AUTHORS Graeme Tweedy Executive Advisor Richard Dyble Senior Advisor Florin Ginghina Executive Advisor Thomas Evans Senior Advisor Maynard Kuona Executive Advisor Jonathan Martin Senior Advisor Michael Taylor Executive Advisor Kate Fry Senior Advisor Vasu Patel Executive Advisor Ravi Dubey Executive Advisor Simon Tomlinson Executive Advisor WITH ADDITIONAL INPUT FROM... Natasha Naidoo Executive Advisor Peter Stanley Executive Advisor Bob Gore Principal Advisor Danny Hurley Principal Advisor David Honour Principal Advisor Nick Ford Principal Advisor Sandy Trust Principal Advisor Gordon Gray Principal Advisor 188

192 14 List of participants PARTICIPANTS AEGON UK Ageas Protect AXA Wealth Chesnara and Countrywide Assured Engage Mutual Assurance Equitable Life Friends Life Guardian Assurance Hannover Re UK Life Branch Hodge Lifetime HSBC Life (UK) Legal and General Assurance Society Liverpool Victoria Friendly Society Lloyds Banking Group National Deposit Friendly Society NFU Mutual 189