Practical considerations for ORSA modelling - Risk Library

Practical considerations for ORSA modelling May 2014 kpmg.co.uk

01

EXECUTIVE SUMMARY The Forward Looking Assessment of Own Risk (FLAOR) or Own Risk and Solvency Assessment (ORSA) has been one of the more successful areas of Solvency II implementation; both UK and European regulators expected1 insurers to be prepared to implement their ORSA from 1 January 2014. In addition to the significant progress of Solvency II in 20132, the utility of the ORSA has also seen its popularity grow across multiple regions worldwide, with the skills and analysis accepted by most insurers as valuable to their business. There are parallels between the ORSA and the experience of UK firms in implementing the Pillar 2 - Individual Capital Assessment (ICA). The ICA regime required firms to improve their approaches to measuring and quantifying risk-many UK firms are now well-advanced in their ICA development and have obtained recognisable benefits. In conjunction with this, many firms also substantially evolved their modelling capability to support the analysis. We believe companies will be able to use that experience for the ORSA going forward and expect that further impetus will be given to the industry’s modelling capability to capture more accurately how a firm’s risk exposure evolves over time or changes in different environments. This paper outlines three key areas of the ORSA requirements, for which we observe a need for further improvement to actuarial modelling techniques. The ORSA guidance establishes that an ORSA report should contain analysis that includes: a) A forward looking assessment of risk and solvency levels for future years (Balance sheet projection capability); b) An ability to report on the monitoring of solvency on a continuous basis (Solvency Monitoring); c) A sufficient analysis of risk through tools such as Stress, Scenario and Reverse Stress Testing. KPMG’s benchmarking insights highlight the varying degrees of sophistication between market participants in the three key areas identified. Firms may be expected to move to a more mature approach that could require increasing capabilities if their methodology has been identified as insufficient. We believe firms need to address these areas to fully satisfy the ORSA requirements as well as ensuring that the ORSA adds value in regards to business planning and decision making. This paper is focused on small and medium sized insurers and in parallel to expanding on the three requirements above, we present a discussion outlining some of the different actuarial modelling techniques firms could use to deal with the increased expectations for the ORSA. Similar to the ICA, the underlying principles of the ORSA will remain the same for each firm. It will be important for each firm to develop the right approaches that are proportionate to their firm.

“I continue to be encouraged at the pace of change in the insurance industry both in the UK and internationally. The ORSA is part of that change - it is no longer a term just discussed in darkened rooms by actuaries and is a part of most companies’ business processes. However, the challenge now is to develop it so it is efficient and can really be used by the business. In this article we discuss some ways in which this can happen.”

Nick Dexter, Director KPMG

“While there is no doubt the ORSA presents challenges technically, it should be reassuring to businesses to see there are solutions available which can meet their stringent requirements. In this article we aim to highlight ways that businesses can supplement existing technology to ready themselves for the ORSA now and protect themselves against the future.”

Candice Thompson, Senior Director IBM

02

KPMG

IBM |

RECENT DEVELOPMENTS IN THE REGULATORY AND MARKET ENVIRONMENT We have seen significant developments around Solvency II in 20132 which remove some of the uncertainty around the implementation of Solvency II and although the timeline for finalising delegated acts and implementing measures remains tight, there is now a much stronger commitment by all parties to achieve a 1 January 2016 implementation date. Of most importance to ORSA is the European Insurance and Occupational Pensions Authority (EIOPA) issued report “A Final Report on Public Consultation No. 13/009 on the Proposal for Guidelines on Forward Looking Assessment of Own Risks (based on the ORSA principles)1” that was issued in September 2013. The consultation paper was released in March 2013 and provides guidelines designed to guide National Competent Authorities (NCAs) in their supervision of their relevant firms’ ORSA developments. These guidelines covers the requirements that are expected of an FLAOR / ORSA, we highlight three of the key main requirements below that will be covered in more detail throughout this paper: The ORSA guidance1 establishes that an ORSA report should include: a) The FLAOR / ORSA should be a forward looking assessment of risk and solvency levels for future years (Projection capability) b) The FLAOR / ORSA should be able to monitor solvency on a continuous basis (Solvency Monitoring) c) The FLAOR / ORSA should be able to carry out a thorough analysis of risk through tools such as Stress, Scenario or Reverse Stress Testing. Firms are also now expected to comply to the FLAOR / ORSA timetable (regardless of Solvency II’s progress) as set out below: a) By February 2014, NCAs were required to report to EIOPA on progress towards implementation of Forward Looking Assessment of Own Risk (FLAOR) / ORSA. b) In 2014, to submit an ‘in development’ FLAOR / ORSA and this can be based on Solvency I / ICA capital requirements / projections. A working ORSA policy will need to be put in place even if it subsequently changes. c) In 2015, to submit an ‘in development’ FLAOR/ ORSA which must be based on Solvency II technical provisions and SCR. It is expected to contain projections of quantity / quality of Solvency II capital over the business planning period. It must include a comparison of the risk profile to Standard Formula assumptions (non-IMAP firms only).

Internationally, the “ORSA” concept has gained ground with firms and regulators in a number of jurisdictions moving towards ORSA type requirements. In North America, the US national insurance regulator (NAIC) is running a pilot exercise3 prior to formal submissions on 1 January 2015, whilst the Office of the Superintendent of Financial Institutions (OSFI) in Canada has issued a consultation paper to the industry4. Other examples include; regulators in South Africa, Taiwan, Malaysia and Singapore, who are all developing ORSA like guidance broadly in line with Solvency II requirements. While the Solvency II related papers uses the term “FLAOR”, we will however use the term “ORSA” throughout this paper as it is more widely used globally. In the United Kingdom, the Prudential Regulatory Authority (PRA) is asking firms to assess the viability, sustainability and vulnerability of their business models5 – the extent to which a company can demonstrate these characteristics will be considered as part of the determination of the level of supervision going forward. Hence, the PRA will require firms to conduct a Business Model Analysis (BMA) that will incorporate the articulation of a clear strategy, a business model to achieve that strategy and evidence that the business is sustainable and resilient over the business planning cycle. The work that firms are expected to perform as part of the BMA includes a detailed analysis of the business model and quantification of key metrics in the business strategy over the plan period. The PRA requirements are likely to require complementary analysis to that expected under the Solvency II ORSA. On top of the above external regulatory drivers, we see a trend of firms moving from traditional disclosures of a Solvency I balance sheet to more sophisticated Economic Capital (EC) metrics. As part of KPMG’s EC disclosure benchmarking, we note that for year end 2011 there were at least 10 large European firms that disclosed EC metrics6. This number increased to 12 for year end 20126. Such disclosures have increased the pressure on firms to understand, manage and explain the changes in the EC metric to key stakeholders, shareholders and market analysts. Many firms are investing in tools and processes to forecast and analyse the EC results. All of the above suggests that ORSA should be the current key focus for firms regardless of the actual implementation date of Solvency II and, from our recent surveys7, 8, there is still a wide spectrum in the degree of ORSA readiness and complexity of approaches adopted. To assist firms in preparation for the implementation of ORSA, this document will consider further the requirements of these three key objectives and the financial analysis that underpins it.

KPMG

IBM | 03

PROJECTION METHODOLOGY

Through ORSA, firms should be better placed to make risk-oriented decisions and adopt robust business strategies as it also encourages firms to be more forward looking and to carry out business planning exercises for the foreseeable future. Article 45 of the Solvency II directive9 requires that firms’ assessment of their overall solvency is forward-looking, including a medium or longer term perspective as appropriate. (This point was further emphasised in the Consultation Paper CP 13/0091 and in Guideline 9 of the Consultation Paper on the Proposal for Guidelines on Own Risk and Solvency Assessment10.) This effectively means that the typical one year solvency position assessment is no longer enough; solvency position and risks will need to be projected forward and assessed for each future year of the firm’s business planning period. Furthermore, in the projections, firms should take into account the following: a) changes in risk profile b) business mix and volume (including new business) c) allowing for expected changes in external factors d) allowance for management actions

e) projections of the quantity, quality and composition (across tiers) of own funds across the business planning period. However, the notes to the Guidelines makes it clear that it is up to firms to decide and justify the methods, assumptions, parameters, dependencies and levels of confidence used in the projection.

KPMG’s Insights The ‘average’ insurer will align their projection period with their annual business plan. The projections will be performed on both a Pillar 1 and Pillar 2 basis, focusing on the biting peak for regulatory capital. Risk drivers are typically rolled forward at a risk level. Non-profit, Unit-Linked base balance sheets and With-Profit asset shares will be projected using actuarial liability models (including assets). The WP business will include a roll-forward of the Cost of Guarantees. Other asset projections may be incorporated into the liability models or projected externally using a simple earned rate.

04

KPMG

IBM |

KPMG’s Insights Which of the following best describes your method of projecting the capital measure for the ORSA report? While our Technical Practices Survey 2013 showed that over 75 percent of the respondents in the UK market are using the risk driver approach for capital projections, there is still a wide spectrum of sophistication in this area. In particular, six percent of respondents indicated that they used a single risk driver to project their whole capital measure; whereas 11 percent have an actuarial model that is capable of performing stresses at future dates for each risk with capital aggregation done outside such models. 30

29%

25 20 15

14%

10

14%

14% 11%

11% 5

6%

0 Actuarial model is able to perform stresses at future dates for each risk and capital is then aggregated outside the model

A combination of modelling and risk drivers is used for the different capital requirements for each risk

A risk driver approach where separate risk drivers are selected for each risk module and each block of business

A risk driver approach where separate risk drivers are selected for each risk module

Whole capital measure is projected using a single risk driver

Other

Not decided

Source: KPMG’s Technical Practices Survey 2013 Figure 1 – Method of projecting the capital measure

Projection of capital was noted as an area for development. While we understand that developments are ongoing and firms are making progress in this area, KPMG believe that the key next steps should focus on a number of practical issues. These include: a) Process improvements – By building projection processes as a natural extension to solvency monitoring, a more aligned and robust process may be employed. Business driver information and data can be shared effectively across individual teams to promote an industrialised process that will allow better accessibility and transparency. b) Integration / consolidated company projection – We still see different items in the balance sheet being projected separately to others (as per Figure 1). In order to accurately project the capital position, the methodology should be able to capture the structure of firms and decision rules (such as management actions) as accurately as possible and be able to include assets, liabilities and interactions between them in the same projection.

c) Validation / direct calculation of costs of options and guarantees – typically, insurers will project WithProfit asset shares using actuarial liability models but employ a roll-forward approach to evaluate any cost of guarantees. Stochastic-on-deterministic approaches for With-Profit business with recalibrated Economic Scenario Generators are likely to produce a more accurate assessment. d) Various approaches to approximations and challenges for each – As set out above, there is quite a wide spectrum of sophistication in the approximations (e.g. risk drivers, full nested stochastic, proxy modelling) of projection of future periods. Companies should understand the implications of each in regards to complexity and accuracy before adopting a specific approach. e) New business – Given the wide breadth of options in layering in future new business into projections, firms will need to consider the best option in regards to practicality and accuracy.

| KPMG

IBM | 05

Projecting risk and solvency levels for future years (projection capability) Integration / Consolidated company projections

Figure 2 – A typical group structure (model life office) which might need to be modelled in actuarial modelling software

As highlighted in bullet point b) above, many actuarial solutions do not have the functionality required to provide an integrated and consolidated company projection that allows a complete company model to reflect the company’s corporate structure. A simple example of a group structure that might need to be modelled within actuarial modelling software is shown in figure 2. More advanced models can enable decision rules to be applied at any level within an organisation, for instance, at group, company, fund and individual product level. This means that users could realistically model the decisions and management actions that would apply in practice, such as decisions about intra-group money transfers, the buying and selling of assets, taxes (at a companywide level), declared bonuses and even rules around selling a company.

As well as including natively modelled assets, a possible improvement to traditional methods is to integrate the actuarial solution with a separate, more specialised, asset tool in order to optimise accuracy. Since modelling liabilities does not affect the value of assets, only the amount held, the assets valued at an instrument level could be imported from a specialist asset modelling tool to achieve a full valuation Asset Liability Models (ALM). Improved accuracy would provide management and other stakeholders with much greater confidence in the numbers and the ability to make better informed business decisions. This may also free up more capital to help grow the business if there is less need for excessively prudent assumptions in the capital calculations.

06 | KPMG

IBM

Challenges faced in preparing for modelling approximations

Figure 3 – Nested projections to accurately capture values at future time steps

KPMG’s ORSA benchmarking results above suggest that companies are having difficulty ensuring that their model approximation techniques (discussed later) are based upon accurate results. A possible way of improving accuracy is for the actuarial modelling software to carry out calculations based on a hierarchical system (i.e. projections within projections). A graphical representation of a hierarchical projection is set out in figure 3, where different “layers” (green boxes) carry out calculations as required and feed results to the overarching layer on top (e.g. reserves are calculated at the Reserve layer and these calculated reserves value are fed up to the Realistic layer). This provides flexibility, giving the ability to recalculate anything at any time. This is best illustrated when considering point c) from the benchmarking results i.e. how complex guarantee costs can be calculated at all future time steps. For example, in figure 3, the Realistic layer would project cash flows under realistic assumptions through time until a point where the value for the guarantees was needed. It would then trigger the Guarantee_Costs layer which would do a full cash flow projection using up to date information from the Realistic layer to calculate the guarantee costs at that time. The results of this guarantee cost calculation would be fed back into the Realistic layer to continue its projection. Arguably most accurate way for an insurer to allow for any guarantees would be to use a full nested stochastic calculation. Using the layer system in figure 3 it would be straight forward to create complex nested stochastic models if all the separate layers in the model were each able to run stochastically.

Increased model accuracy will invariably come at the expense of run time, especially if a nested stochastic calculation is being considered. However, there have been numerous advances with actuarial modelling software which serve to remedy this. Examples of some features which actuarial software could use to vastly improve the speed of runs and prevent compromises on accuracy are: s Distributed processing means that other computers on a network, a set of servers, a grid or cloud can be used to process models. Additionally the inclusion of multithreading allows the utilisation of all available cores on the server / computers. s Code optimisation would mean that a model, such as an ORSA model, which is results driven, would not have to run code that is not needed to produce the required model output report s Active memory management can enable maximum use of the available memory and avoid unnecessary and time consuming writing of data to and from a disk. s Ability to run on 64 bit platforms to enable increased amounts of memory to be utilised Utilising some or all of these features enables high performance models to be built and run more frequently. The more timely availability of the modelling results would mean that they could be embedded more easily into dayto-day decision making of a business.

| KPMG

IBM | 07

New business Modelling future new business (in other words policies which are not currently in force) is an important requirement for ORSA because part of any business plan will usually be to grow the business. This can be difficult to achieve because the amount and type of new business will depend, amongst other things, on the insurers’ strategy for future growth, prevailing economic conditions as well as the availability of capital to finance it.

An actuarial model for new business should allow the option of initial values for new records to be either read directly from external files or created within the model itself. The option to recalculate in the model offers the advantage of adopting a more dynamic approach. For example, the type and amount of new business may directly correspond to an insurer’s expectations under the modelled economic situation and capital position.

08

KPMG

IBM |

SOLVENCY MONITORING

To be of most use to the business, the ORSA should be reassessed when conditions change or the business strategy is to be emended or developed. – indeed Article 45 (1.b) of the Solvency II Directive Level 1 text9 states that the Own Risk and Solvency Assessment shall be ‘assessed on a continuous basis for the compliance with the capital requirements’. Further guidance can be found in the additional notes to Guideline 10 of the Consultation Paper on the Proposal for Guidelines on Own Risk and Solvency Assessment10 which makes it clear that the continuous basis for compliance does not constitute an obligation to recalculate the FULL regulatory capital requirement all the time. However, firms must demonstrate their ability to assess their solvency to an extent that is commensurate to the nature, scale and complexity of their business. Effectively, a life insurance company will need to monitor their solvency levels more frequently than their usual statutory reporting (annually) or internal reporting requirements where a full model rerun and data updates are expected.

The PRA has also notified Internal Model firms of its intention to implement a series of Early Warning Indicators11 (EWIs). The intention of these is to act as a tool for ensuring the on-going appropriateness of firms’ Internal Models. The PRA are still in the process of developing and refining these tools at the time of writing this document. It is however expected that firms should be able to discuss their performance against these indicators with the PRA at some point in the near future and therefore continuous solvency monitoring should be high on firms’ agendas. Estimation and approximations are allowed where appropriate. The choice of allowable estimations and the frequency of the calculations will be determined by each individual firm but such decisions should be appropriately documented and justified.

KPMG

IBM | 09

KPMG’s Insights How frequently will you be monitoring your ORSA capital measure?

25 22

20

15

11 7

7

5

7 5

5

Source: KPMG’s Technical Practices Survey 2013

Annually via full qualification

Half-yearly via approximation

2 Quarterly via full qualification

2 Quarterly via approximation

1

Monthly via approximation

2

Weekly via full qualification

0

Daily via approximation

5

Annually via approximation

11

Half-yearly via full qualification

10

We see a clear trend towards more frequent monitoring. Sixty-nine percent of respondents intend to monitor their ORSA via an approximate method more frequently than half yearly, including 38 percent who suggest that they would monitor monthly. Sixty-seven percent of respondents indicated that they would monitor their ORSA annually at a minimum through full quantifications and 20 percent indicated they would do a full quantification quarterly. Figure 4 – Frequency of ORSA monitoring

Full solvency evaluations will require firms to produce calculations based on a nested stochastic approach, which, due to the number of runs necessary, is impractical for continuous solvency monitoring purposes. Rather than carrying out this detailed approach, a number of firms are now considering proxy modelling techniques to carry out acceptable estimates with more appropriate run times. KPMG’s Economic Capital benchmarking7 highlights the wide variety of techniques used; popular methods include replicating portfolios, curve fitting and Least Square Monte Carlo. Firms should consider the estimations allowed within the calculations; practical considerations of the accuracy, speed and resource requirement of calculations will need to be made. IBM, in the next section, sets out some examples of proxy modelling which could be used, as well as methods of validation and likely issues and discrepancies between a full model run and proxy models.

KPMG’s Insights Based on our knowledge of the insurance market in the UK, we believe that currently firms tend to already carry out monthly Solvency Monitoring on both Pillar 1 and 2 (typically with a focus on market risk only). Pillar 1 figures will usually be based on model runs for sole entities, but a roll-forward will be used for Group consolidated figures, and Pillar 2 will usually exclusively use a roll-forward basis. Manual reserves are unlikely to be updated and new business and claims are often not allowed for.

10

KPMG

IBM |

Improving performance using approximation methods within an actuarial modelling software: The most common proxy methods that are being utilised in the market at the moment are replicating portfolios, curve fitting and Least Square Monte Carlo (LSMC). Traditionally, these proxy methods are calculated using specialised tools outside the actuarial modelling software. This provides more flexibility and complexity to the proxy modelling techniques since they are not restricted by the limitations of actuarial models. However, a fully integrated approach to proxy modelling offers advantages and developments have been ongoing within actuarial modelling software such that some now have the ability to perform some of the proxy techniques. This can be achieved by utilising the same technology usually only found in the specialised “Economic Capital” tools. The advantages of such an integrated approach are set out below: s Reducing Operational Risk By performing modelling and fitting in one integrated system, the need for any intermediate stage is removed since there is no need to adapt the results of a model to work with a separate fitting tool. This helps to reduce the Operational Risk associated with the process, and saves time by reducing the number of actions undertaken.

s Improving back testing and validation Having inbuilt proxy methods along with the ALM models used for fitting in a single integrated system allows the user to perform back testing and to validate their proxies regularly, easily, ensuring they remain fit for purpose. An integrated model would ensure that the basis, assumption and approaches (e.g. approach for applying stress) are always consistent for the purpose of comparing results and validation. s Realising business benefits by applying the curve to address practical tasks Once generated, the curve can be applied to different functions within the actuarial software to derive a range of business benefits. For example, applying the curve with nested stochastic functions would enable capital to be monitored on a frequent basis since capital requirements can be calculated quickly and efficiently.

| KPMG

s Improving transparency and understanding to assist with management decision making It is important that once results have been calculated they are displayed in a way that is useful to the user. For Solvency II monitoring purposes, for example, it will be vital to be able access the results when required, and for those results to be as transparent as possible. Some actuarial software is compatible with dashboard software that is customisable and allows the user to see at a glance exactly the results they want. If the fitting tool is integrated then it will also benefit from the use of such dashboards.

IBM

11

s Preparing for the future An additional advantage of having both the fitting functions and underlying models within the same modelling system is that as additional or newer hardware becomes available, the proxy models can gradually be replaced with the underlying nested stochastic calculations upon which they are based. This prevents the business being left with a defunct system that they have spent time and money developing.

Figure 5 – An example of a results dashboard to aid management decision making

12

KPMG

IBM |

STRESS, SCENARIO AND REVERSE STRESS TESTING Understanding the risks, the impact and possible mitigations of such risks is essential to any decision making and business planning process of a firm. Article 45(2) of the Solvency II Directive Level 1 text9 states that firms should have in place processes which are “proportionate to the nature, scale and complexity of the risks inherent in its business and which enable it to properly identify and assess the risks it faces in the short and long term to which it could be exposed”. Stress, scenario and reverse stress testing are processes which firms could employ to assess such risks the exact and impact of the risk as well as effectiveness of possible mitigation.

On top of the usual issues that firms already encounter as part of the stress testing process, firms also face additional complexity due to the need for projections of future periods under stress conditions and also the need for more complex and detailed stresses under Solvency II. We note that firms face the following issues amongst others:

The Guideline 9 of the November 2011 Consultation Paper10 (Forward-looking perspective Article 45) mentions that firms, as part of the business and capital planning process, need to carry out stress, scenario and reverse stress test to feed into the overall ORSA process. It makes it clear that the scope and frequency of such exercises should be decided by firms taking into account of the nature, scale and complexity of the firm’s business and risk profile. Guideline 8 from the same paper, stress test and scenario analyses are explicitly mentioned as being required to provide an adequate basis for the assessment of solvency needs. Hence, solvency monitoring and projections will need to be considered in the stressed environment for firms to be able to properly assess the extent of the risks and the possible mitigations.

b) Validation of proxy models under stress – firms may not be able to validate loss functions to the desired granularity under stress conditions or may struggle to ensure tolerance levels are not breached.

Stress, scenario and reverse stress testing are of course familiar to UK firms, having already implemented such techniques in the calculation of Pillar II ICA results or as part of overall existing risk management process.

a) Volume of runs – the number of runs under Solvency II will increase significantly as more stress, scenario and reverse stress testing is needed. Furthermore, extra runs will be required for calibrations including those under stress conditions.

c) Assessment of stress and scenario testing over time – Impacts should be projected over the business planning period, showing recovery paths where appropriate d) Calibration – the number of scenarios used per function, the method of selection and number of risk drivers per function will be established under typical modelling scenarios. Firms may have to reassess these requirements when calibrating results under stressed conditions e) Management action – Management actions are often not well integrated in the models and may be allowed for through manual adjustments. This would not be practical under stresses, scenario and reverse stress testing as the number of results that need management action adjustments will increase significantly.

| KPMG

IBM | 13

Addressing the challenge of stress, scenario and reverse stress testing with actuarial modelling software Reducing costs and risks with improved run scheduling

Actuarial modelling software should be able to automate stress testing by being able to re-use the same model(s) a number of times with varying assumptions under each run. This will reduce both the effort (hence cost) and risk associated with manually scheduling multiple models with different sets of assumptions. The most advanced actuarial modelling software allows assumptions to be maintained separately (as opposed to being held within the model itself), for example, in a database or external spreadsheets and then read from that source at run time. Separating the maintenance of assumptions from the model means that the model itself can be locked down and the assumptions changed independently as and when required. Overall, this helps to reduce operational risk.

Compliance and Governance

Batch or run scheduling functionality allows for a set of models to be grouped so that they can be run together (potentially on a grid) or in a customised sequence (for instance, if one model relies on the results of another, they can run in the correct order) which can be run without any manual intervention. By combining that with the ability to separately maintain assumptions (mentioned above) means the same model can be included in a batch multiple times and use different assumptions for each run. This allows the complete set of results required as part of a stress test to be achieved in one go, without manual intervention.

A proper governance framework needs be put into place around the actuarial models which means access to models should be managed through fully customisable user profiles and defined roles/ permissions. Any aspect of usage within models should be capable of being enabled or restricted in order to improve flexibility for those administering them. Any governance system set-up should also ensure all activity is fully traceable so the designated system administrator can monitor activity carried out within the models.

As well as speeding up processes, it is also important to ensure that the correct models, data and assumptions are used in the stress and scenario runs. Once complete, the models or batches described in the run scheduling section should be locked down so that the production of numbers is secure and can be automated. The approved locked down model should then be transferred from the development environment to a separate production environment, which could be an area such as a secure web server, to prepare, manage and schedule runs. If a company has an existing scheduling tool then it is important that it is straight forward that any production models can be accessed by it.

Figure 6 shows an example of a secure environment using a set of external tools with the arrows indicating the information flow. The diagram illustrates a framework environment to enable all assumptions and data used for running models and results generated to be fully secure, locked down and auditable.

14 | KPMG

IBM

Figure 6 – A typical production environment and process flow

In this example:

Calibrating the model

s Data and assumption sets are uploaded and securely stored in the data repository, giving a version controlled, static set of secure data.

As discussed by KPMG’s analysis of the issues faced by insurers in this area, model calibration and validation is a key part of any stress or scenario testing process. A modular structure in the actuarial modelling software helps users calibrate and test model results more efficiently because it allows code to be shared across models. The ability to test individual code segments with known inputs, in isolation to the rest of the model, enables more effective calibration and validation. For example, if a user were looking at the guarantee costs in the model in figure 3 they would be able to do so without having to run the full model. Instead the calibration and testing could be performed on a defined sub segment using known inputs. By limiting the calibration and testing to a defined part of the full model, it is easier to assess in isolation whether the results are as expected.

s Models, developed in the development environment can be securely uploaded to an enterprise system where they can be accessed by a wider set of users and executed with the correct assumptions, s Once models have been executed, results are brought back into the data repository, providing a permanent end to end link between data, models and results. s The entire process is managed and controlled by a workflow manager. This type of process control can reduce the operational risk associated with manual processes and can also help to streamline those processes by making companies think about the optimal way to achieve the end goal. By customising the workflow process and prompting users as and when they have outstanding actions, insurers build in an appropriate level of checking and review. This can help reduce the risk of incorrect results being produced by ensuring only the actions required by the process at that stage are carried out.

Management Actions A model should accurately capture and model management actions at all levels as closely as possible to how they operate in an insurer in practice.

| KPMG

It is important that these modelled management actions can be vigorously tested, in an efficient and controlled manner, under extreme scenarios to confirm that they perform as the user intended. The same principle of modularity can be applied to calibrating and testing the impact of stresses on the coded management actions and since the code is shared, any changes made should instantly update the full model. This helps to reduce the overall workload involved in calibrating and testing the model. Having the same code throughout also has added advantages: s It helps reduce the risks normally associated with having to repeat code (even though it performs similar calculations) several times. s It helps maintain consistency across the model suite since only one set of code is required, whilst allowing detailed calibration and testing to be carried out in a more effective manner. The ability to run the same model with different assumptions assists calibrating and testing results by ensuring that it is easy to make changes to those assumptions. By removing the need to change code, and hence risk the possibility of accidently corrupting models, it also reduces the time to set up the calibration runs.

IBM | 15

An advanced model should be capable of using scenario files from a wide range of Economic Scenario Generators for stochastic runs so insurers are not limited to the source of these files. This flexibility means that the same model can be run a number of times with a different set of scenarios used each time. This enables the automation described earlier to reduce both the time and cost involved in setting up calibration processes.

CONCLUSION

Although the Solvency II delays have caused many insurers to consider “downing tools”, the external drivers of the PRA’s Business Model Analysis and further EIOPA guidance will ensure that the ORSA developments that are being made will continue to be high on the agenda of insurers. Firms are starting to realise the additional benefit of implementing a “good ORSA” that effectively captures the risk and complexities of the firm. These benefits are as well as from a business planning, strategy and decision making perspective. KPMG have observed a number of areas where firms need to improve the quality of their analysis, in the key areas covered by this report, to meet the requirements the exaggeration the ORSA guidelines.

However, there remains a place for a variety of actuarial techniques to be used in the market are varied in terms of efficiency and complexity. It will be important for each firm to develop the right approach that are proportionate to their firm.

KPMG

IBM | 17

HOW WE CAN HELP?

KPMG is a global network of professional firms providing Audit, Tax, and Advisory services. We operate in 156 countries and have 152,000 professionals working in member firms around the world. Our UK Actuarial and Insurance Risk Practice alone consists of about 150 consultants across the Insurance Life, General and Risk teams. Our dedicated insurance team provides a full range of product approach including transactions support, post-merger integration, capital management, modelling, Solvency II, IFRS, finance transformation, financial reporting, restructuring, RDR, reserving and audit support.

Areas where we have assisted clients include the following:

Over recent past years, we have supported a wide range of insurance companies with their Solvency II and Economic Capital implementation including ORSA. Hence, KPMG has extensive experience in the development of ORSA and is well placed to assist firms in meeting their ORSA needs.

s Provision of technical support to the balance sheet forecasting process and Risk function.

s Design and build of an robust ORSA framework including a policy and report. s Development of a ORSA process to interact with the BAU activity of a firm. s Undertaking gap analysis of key ORSA inputs and completion of dry run exercises. s Development of appropriate capital and risk frameworks

s Complete scenario and RST analyses. s Provision of advice on contingency planning, possible management actions and their quantification.

18 | KPMG

IBM

Whilst implementing an ORSA is undoubtedly a complex task, this article seeks to reassure companies that there are effective solutions available and many of the problems being encountered can be overcome with the right tools and the correct application of those tools. Indeed, the ORSA need not only be seen as regulatory compliance, but as the heart of an integrated ERM strategy and investing in and upgrading systems to adopt an ORSA approach early can provide both a future proof solution and potential competitive advantage. IBM has solutions available to address the problems outlined in this article:

IBM Algo Financial Modeler: IBM Algo Financial Modeler is a leading software solution for financial and actuarial analysis, designed in response to the increasing focus on risk and capital management and the demand for more realistic modelling in financial analysis. It has been specifically designed to: s Make actuarial models easy to use, understand and review. s Simplify the building and understanding of large and complex company models. s Make the implementation process straightforward and effective. s Be fast, especially for ALM and stochastic modelling s Remove any workarounds currently needed to handle complexity. s Maximize flexibility and transparency to easily handle any likely future requirements. s Utilise proxy methods including curve fitting and LSMC

IBM Algo Financial Modeler Enterprise: This powerful, server based platform provides total separation of the model development and production environments, providing a completely automated, end-toend process that is increasingly important in being able to demonstrate that a robust, secure, and stable system operating process is in place to auditors and regulators.

Add on packs - ORSA and Proxy modelling in IBM Algo Financial Modeler: This add on pack offers insurers a new approach to improving the efficiency of full and partial internal models for projecting solvency capital requirements (SCRs). It also allows businesses to support their own ORSA calculations without investing heavily in new hardware.

By providing a flexible framework for users, insurance firms can implement either a fully nested stochastic model or a fitting method of their choice. The sophisticated design of this model enable insurers to use proxy modelling in IBM Algo Financial Modeler with existing actuarial models providing maximum benefit for minimum effort.

Native Asset Modelling IBM have an extensive asset modelling product coverage list with over 400 different pricing models to accurately capture the cash flows and payouts of an individual security. Models are the most commonly used valuation methods. Terms & conditions and associated risk factors are specified for each product type and loaded into the transaction database. A standard function for valuation and simulation of the trade, allowing settlement (cash and physical) and through-time valuation to be realized; sensitivities, cash flows, and other attributes are also calculated. In many cases, model calibration algorithms are also provided, ensuring consistency and facilitating validation. The flexibility within the IBM framework means that clients can model a wide variety of products that exist in the market today.

IBM Cognos Business Intelligence IBM Cognos Business Intelligence provides reports, analysis, dashboards and scorecards to help support the way people think and work when they are trying to understand business performance. You can freely explore information, analyse key facts and quickly collaborate to align decisions with key stakeholders. Capabilities include; s Reports - From professional report authors to business users to people in the field, IBM business intelligence software includes capabilities for authoring, viewing and modifying reports and interactive visualizations—online or off, in Microsoft Office applications or in-process applications, in the office or on the go. s Dashboards help users access, interact and personalize content in a way that supports how they make decisions. Historical information alongside current data, data in motion and predictive analytics help you quickly move from insight to decision—all in one dashboard. s Analysis capabilities provide access to information from multiple angles and perspectives so you can view and analyse it to make informed decisions. Explore your data with trend analysis, “what-if” scenarios, drill-down and more.

SOURCES 1.

CP-13/009 - Consultation Paper on the Proposal for Guidelines on Forward Looking assessment of the undertaking’s own risks (based on the ORSA principles), including Impact Assessment https://eiopa.europa.eu/en/consultations/ consultation-papers/2013-closedconsultations/march-2013/guidelines-onpreparing-for-solvency-ii/index.html?no_ cache=1&cid=5792&did=30493&sechash=849321be

2.

NAIC OWN RISK AND SOLVENCY ASSESSMENT (ORSA) GUIDANCE MANUAL – As of March 2013 http://www.naic.org/documents/committees_e_orsa_ wg_related_docs_guidance_manual_2013.pdf

4.

Draft Guideline E-19 – Own Risk and Solvency Assessment http://www.osfi-bsif.gc.ca/app/DocRepository/1/eng/ guidelines/sound/guidelines/e19_dft_e.pdf

5.

The Bank of England, Prudential Regulation Authority – The PRA’s approach to banking supervision http://www.bankofengland.co.uk/publications/ Documents/other/pra/bankingappr1210.pdf

KPMG Economic Capital Disclosure Update – May 2013 (contact [email protected] for more information)

7.

KPMG Economic Capital Modelling in the Insurance Industry: A solid foundation for future advantage? http://www.kpmg.com/Global/en/IssuesAndInsights/ ArticlesPublications/economic-capital/Documents/ economic-capital-modeling-july-2012.pdf

Illuminate – KPMG’s Insurance regulatory newsletter Issue 01. http://kpmgmail.co.uk/rv/ ff0014132f4151f2b3699907fe409ba57821c680/ p=7712430

3.

6.

8.

KPMG Life Insurance Technical Practices Survey 2013 http://www.kpmg.co.uk/email/09Sep13/286049/ Technical_Practices_Survey_2013/index.html

9.

DIRECTIVE 2009/138/EC OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL of 25 November 2009 on the taking-up and pursuit of the business of Insurance and Reinsurance (Solvency II) http://eur-lex.europa.eu/LexUriServ/LexUriServ. do?uri=OJ:L:2009:335:0001:0155:en:PDF

10. Consultation Paper on the Proposal for Guidelines on Own Risk and Solvency Assessment https://eiopa.europa.eu/consultations/ consultation-papers/2011-closed-consultations/ november-2011/solvency-ii-consultation-paperon-the-proposal-for-guidelines-on-own-riskand-solvency-assessment/index.html?no_ cache=1&cid=4624&did=18309&sechash=0f79b42b 11. Monitoring levels of capital and early warning indicators http://www.bankofengland. co.uk/pra/Documents/solvency2/ introductionofearlywarningindicators23May2013.pd

| NOTES

Contact us For more information please contact:

Your KPMG Contacts:

Your IBM Contacts:

Nick Dexter Director T: + 44 (0) 207 311 5443 E: [email protected]

Candice Thompson Senior Director T: +44 (0) 1737 232 512 E: [email protected]

James Isden Principal Advisor T: + 44 (0) 776 099 0765 E: [email protected]

Dan Wainwright Principal Consultant T: +44 (0) 7584 473 189 E: [email protected]

Zaid Hoosain Principal Advisor T: +44 (0)7771 586 456 E: [email protected]

Chris Bailey Senior Actuarial Consultant T: +44 (0) 7867 452 44 E: [email protected]

www.kpmg.co.uk The information contained herein is of a general nature and is not intended to address the circumstances of any particular individual or entity. Although we endeavour to provide accurate and timely information, there can be no guarantee that such information is accurate as of the date it is received or that it will continue to be accurate in the future. No one should act on such information without appropriate professional advice after a thorough examination of the particular situation. © 2014 KPMG LLP, a UK limited liability partnership, is a subsidiary of KPMG Europe LLP and a member firm of the KPMG network of independent member firms affiliated with KPMG International Cooperative, a Swiss entity. All rights reserved. The KPMG name, logo and “cutting through complexity” are registered trademarks or trademarks of KPMG International. Oliver Marketing for KPMG I OM013346A I May 2014 |

IBM, the IBM logo, ibm.com, Algo, Algo Financial Modeler, and Algorithmics are trademarks of International Business Machines Corp., registered in many jurisdictions worldwide. Other product and service names might be trademarks of IBM or other companies. A current list of IBM trademarks is available on the Web at “Copyright and trademark information” at www.ibm.com/legal/copytrade.shtml.