Speech by Adair Turner, FSA Executive Chairman at the Swedish Central Bank
Thank you very much for the invitation to speak here today. I think it is fair to say that the question I have been asked to address is both an ambitious and important one.
It is a commonplace that those who do not learn from their history are destined to repeat it and, given the extent, severity and duration of the crisis we are currently experiencing in financial markets, we must surely be absolutely determined not to make the same mistakes again. And that’s particularly the case with insurance, given the essential services which the sector provides to the UK economy. Risk transfer and pooling which are at the heart of the insurance business model are vital for much economic activity and, similarly, in their role as financial intermediaries and long-term investors, insurers are crucial for economic growth and development.
It is in that context that we need to consider as insurance supervisors what we can learn from the banking crisis.
We have, of course, already made a number of changes to our supervisory approach since the financial crisis, and shortly we will be operating under significantly different regulatory architecture. The Bank of England will assume responsibility for the micro-prudential supervision of both banks and insurance companies along with new responsibilities for macro-prudential supervision, and alongside this will be the creation of a new, separate Financial Conduct Authority. Whilst these reforms have primarily been driven by bank failures, they will nevertheless advance our approach to supervising insurers.
There is a temptation then, to simply reflect upon these developments. However, I have discussed the outline of our new supervisory approach and philosophy previously and I don’t want to take up time by repeating what has been already said in our various Prudential Regulation Authority (PRA) approach documents. Instead, I would like to discuss the practical challenges that the financial crisis has raised for prudential regulators of insurance companies. And perhaps in doing so, I will be able to capture some of the implications for the PRA’s supervisory approach.
But before we can consider the lessons for insurance supervisors, I think we need to have some idea of the main lessons which emerged from the crisis. I would suggest three overarching themes:
- there is an imperative to ensure that the regulatory capital regime appropriately captures the risks assumed;
- that no single firm or group should either be too big to fail or whose failure should be allowed to seriously disrupt the provision of core financial services to the real economy; and finally
- there is a need for regulators to be alert to changes in the activity of non-bank firms (often – but not always – which sit outside of the regulatory perimeter), which result in them assuming risk which economically looks like banking.
What I would like to do for the rest of my time with you this morning is to examine the extent to which insurance supervisors have reflected on these themes. I would like to discuss how our collective response to some contemporary key issues are conditioned by how we respond to some of these imperatives.
Let me begin with the regulatory capital framework. It is now readily apparent that the pre-crisis regulatory framework for banks was deficient principally in its under-estimation of the amount of capital which banks needed to cover the market risk on their so-called trading books, as well as their traditional core banking books. It is instructive to note that for most of the pre-crisis period (and in particular during the build-up of unsustainable leverage) two conceptually different approaches to assessing bank capital were deployed by supervisors. They could broadly be described as straight forwardly formulaic (Basel 1) and internally developed models. But both approaches failed to reflect adequately the build-up of risk in the decade prior to 2007 – they failed for different reasons but equally comprehensively. It may be instructive to review what went wrong with both approaches.
What went wrong with banking models?
First the world quickly became more complex than could be captured by either approach. The formulaic approach of Basel 1 was exposed fundamentally because the risk factors used were only valid when banks carried on the activities they had engaged in when the approach was developed. When this changed – as it so clearly did in the late 1990s and in the first part of the new century – the formulae didn't and, as a consequence, risk was under-estimated both in the regulatory framework and also in the assessment at the front end of the business.
Many internal models were exposed because they rested on assumptions which turned out not to hold when bad times came. The development of internal models enabled firms to adopt a more positive outlook which was conditioned by the review period from which much of the data that was used to populate the models were drawn. Assumptions were used to parameterise models with insufficient rigour and independence from the front end of the business and management attention was too often focused on those parameters considered too conservative at the expense of those that were insufficiently prudent.
These problems were compounded by what we can now see were flawed technical assumptions underpinning the probability of extreme ‘tail’ events in the construction of models. Crucial in this regard was the assumption that one can accurately estimate the probability of extreme events from data derived in large part from normal times. This led to insufficiently robust assumptions, in particular concerning the distribution of losses, and modelled losses were often based on insufficient data.
A third type of failure related to the feedback loops created by the widespread use of models. This is because the assessment of the risk of various asset classes for regulatory purposes itself drives investment decisions and hence asset returns, which then serve as inputs into risk models. This gives rise to the potential for destabilising feedback loops as under-estimation of risk helps drives asset bubbles during good times leading to an inevitable adjustment and downward spiral in bad times.
And finally the widespread use of models rested on the incorrect assumption that if each firm appeared adequately capitalised on risk-based measure, then the whole system would be safe: we can now see that this was not the case.
Lessons for Solvency 2
Given that we are in the process of implementing Solvency 2, which makes extensive provision for the deployment and use of internal models as well as promoting a standard formula approach, it's worth reflecting on what we can learn from this experience.
The first seems to me to be the need to remember that, however good we think our solvency regime is, we need to consider whether the business written by insurers is becoming too complex for either approach to capture appropriately or whether the risk captured is itself changing. It may well be that internal models are better placed to achieve the comprehensive risk capture that supervisors want for certain types of activity but that, if that is the route that we choose to go down, we need to be sure that the data underpinning the model are robust and the assumptions appropriately conservative. We can be helped in this regard by the much greater use of imaginative tests of resilience to deeply stressed scenarios to identify potential vulnerabilities in a timely manner. Hence our current emphasis on stress and reverse stress testing.
However, technical advances in modeling technology alone will not be sufficient. There remain some fundamental uncertainties, particularly the paucity of relevant historical data for the calibration of tail dependencies between risks. It is important to strike a balance between simplicity and reliable capture of tail dependency. It is also critical that the limitations of the approach are recognised and conservatism is built into the calibrations to address them. Arguably, these are more complex in insurance since correlations in the tail are likely to be asymmetric in nature: we could expect a very large catastrophe or mortality event to have major market impacts but a wave of corporate defaults will never cause a tsunami.
There are also notable differences between modeling insurance risk and modeling credit and market risks. Importantly, the widespread adoption of quantitative techniques that seek to understand an exogenous risk such as natural catastrophe events or longevity will not change the nature of the risk itself.
I think the main lesson from the crisis is that we must not blindly accept the outputs of these models. This has been shown to reduce the quality of prudential supervision. But that in my view does not mean we should reject the use, or discourage the development, of risk modeling per se. The PRA will want to maintain a risk-sensitive approach to prudential regulation and relying on a solely mechanistic approach to setting capital requirements can be inadequate for certain types of activity.
Importance of non-modeled cross checks
The recognition of both the value and limitations of risk-based internally developed capital models underpins our work to develop more simple metrics to help monitor the appropriateness of internal models. These metrics, which I have termed ‘Early Warning Indicators’, will be independent from the firm’s modelled Solvency Capital requirement (SCR), to help us ensure that the level of solvency standards calculated by firms does not deteriorate over time.
Our preliminary analysis confirms that the ratio between the Solvency Capital requirement and the Minimum Capital Requirement (MCR) could be used as a ‘backstop’ for the SCR calculating using internal models. We have proposed using the modeled Solvency Capital Requirement, which falls outside these predetermined ranges as a notification event which will trigger an immediate supervisory review of the appropriateness of the model. This could result in the revision of the parameters and the imposition of a temporary add-on until this work is complete. The aim of such action would be to maintain confidence that the Solvency Capital Requirement meets the Solvency II calibration requirement.
Imperative for judgement-led forward-looking supervision
Ultimately, supervisors’ use and response to the outcomes of these non-modeled Early Warning Indicators will rely upon judgement. And it is such forward-looking judgement that is at the very heart of the PRA’s philosophy of supervision. I said earlier that I did not want simply to run through our approach document. But it perhaps bears repeating that one key element of the PRA’s supervisory approach will be to give supervisors the tools and skills they need to exercise effective supervisory judgement. Early Warning Indicators will be one such tool.
II Too big to fail
I want to move on now to consider what lessons we can learn from the consequences of firm failure in the crisis. In a competitive market place, we must recognise that insurance companies and banks are in a risk-taking business and, as such, they can and will fail. This is why we will not be operating a zero failure regime. But the banking crisis has shown us that, in such a regime, we must consider not only our prudential framework for assessing the risks that firms as a going concern take but also our framework for dealing with them when they fail. The crisis exposed serious gaps in our resolution framework for banks, gaps which are in the process of being addressed. But I would like to offer a few thoughts on the extent to which gaps might also exist in insurance and how those gaps are being considered both domestically and internationally.
It should be an obvious truism that we must have a system that that is capable with dealing with firm failure in a way that is consistent with our objectives and, importantly, does not impair the provision of core financial services to the real economy. Before coming on to the read across for insurance, I think it is worth reflecting upon the services that insurers offer in this regard.
Fundamentally, insurance facilitates the transfer and pooling of risk and for this reason insurance plays a unique role in society. It fulfils social functions such as retirement provision, income protection and funding for long-term and health care. It allows day-to-day business activities to be carried out smoothly through certain types of cover, such as public liability, employers’ liability, trade credit and transport liability. If insurance provision was to be disrupted on a significant scale then policyholders could lose critical covers or the payment of vital income and economic activity could be disrupted. It is probably fair to say that the failure to maintain continuity for such key services is unlikely to be acceptable from a societal perspective.
However, this is not to say that an insurer failure would be unacceptable. How do we therefore reconcile the need to maintain the provision of insurance and the allowance for insurer failure? To address this challenge, we need to have confidence that no insurer is too big, too complex or too interconnected to fail. Insurers should be able to exit the market in an orderly fashion which provides continuity of access to critical services. How do we get there?
There is not currently a resolution regime for insurers in the UK as there is for banks. That does not necessarily mean that we need one but it should prompt us to consider the question. Generally, failing insurance firms go into ‘run-off’, a process in which a firm’s permission to conduct new business is withdrawn and existing liabilities mature or ‘run off’ over time. The run-off may be combined with a Scheme of Arrangement, which allows the firm to restructure its liabilities. In addition, formal insolvency mechanisms are available in the form of a modified administration or liquidation.
This framework has generally proven adequate for dealing with general insurance companies. However, we have no experience of very large failures, in particular of life insurance groups, and must not be complacent. If there is one thing that the financial crisis has taught us it is that we should be prepared to face the unthinkable: before 2007 after all we had not experienced a run on a British bank for over a hundred years and never one which had taken place on television and the internet with all the fairly instantaneous feedback loops that created. Is it right to wait for the next failure whilst it is uncertain that our current available tools would be able to cope with the most extreme scenarios?
What are we concerned about? Generally, we believe that continuity of cover could be at risk in insolvency. In a run-off scenario, there is a trans-generational risk for policyholders whose contracts mature later which may be particularly acute in with profits offices. Schemes are complex and court led processes and Part VII transfers are dependent on the existence of a willing third-party purchaser. Finally, the value of the firm is significantly destroyed whilst it gradually descends towards insolvency, leaving less for policyholders and that may mean that the overall net cost could not be absorbed by the Financial Services Compensation Scheme (FSCS).
We are therefore examining what changes may be needed for us to have the confidence that no insurer is too big, too complex, too interconnected to fail and that continuity of critical contracts is maintained. Domestically, we are exploring ways to improve the FSCS’s operations and our insolvency arrangements that would allow the best chance of continuity for the most critical policies. The PRA, working closely with the FCA, will be responsible for setting the rules of the FSCS. The design of these including, for example its funding arrangements, its capacity to support insurer resolution and the arrangements for publicising the extent of cover under the scheme, falls within the scope of this work. Improving the operational capability of the FSCS does not constitute a UK insurance resolution regime but it may well improve matters for policyholders. Beyond changes to the FSCS, we will need to discuss additional options with relevant authorities, including the Treasury.
At international level, the Financial Stability Board (FSB) has set out minimum standards that should apply to any financial institution that could be systemically significant or critical if it fails (the ‘Key Attributes’), which include stabilisation type powers. The International Association of Insurance Supervisors (IAIS) has indicated that they will be releasing the first list of Globally Systemic Important Insurers (G-SIIs) in the summer of this year and that they will over the course of the next 18 to 24 months work on the identification of Domestically Systemic Insurers (D-SIIs). The challenge here is to recognise the specificities of insurance compared to other financial sectors. This is why Paul Tucker, as Chair of the FSB’s group on resolving large and complex financial firms, is setting up a workshop on insurance resolution. Based on this information, we will need to assess whether a special resolution regime could be needed for insurers and, if so, what characteristics it would need.
Emerging risks in insurance markets – systemically important insurers
I want to turn now to the final area in which insurance supervisors may need to learn lessons – whether insurers (or any parts of their groups) undertake activity which could result in them undertaking banking type maturity transformation. That subject is of course central to the debate about whether they could be characterised as systemically important. The concept of Globally Systemic Important Insurers has, perhaps unsurprisingly, caused some degree of consternation. The market has strongly argued that traditional insurance business does not give rise to systemic risk and that the size of the balance sheet is no indication of the threat posed by an insurer to the real economy. There are merits to both of these arguments. Insurance companies have fundamentally different balance sheets to banks, do not form part of the payment system, are not as interconnected and fail at different speeds to banks.
Nevertheless, we must be alive to the potential for new threats to the financial system to emerge from insurance markets as they develop. I think in this context we can learn some important lessons from our banking colleagues. We should ask ourselves three questions in this regard: whether insurers assume leverage to take on risk, whether they engage in maturity or other forms of asset transformation or whether they assume credit risk in significant scale. If the answer to any one of these questions (let alone in combination) is positive then we need to consider carefully whether they are systemically significant.
Taking each of the questions in turn; leverage first. We know that insurers engage in stock lending activities where generally low risk government and quasi-government securities are repo’d for cash which can then be invested in higher risk paper. Another way in which insurance groups can contribute to leverage is by enabling non-insurance parts of the group to borrow (either implicitly or explicitly) on the strength of the insurance activities. There seems little doubt that AIG Financial Products would not have been able to borrow as much as it did without the implicit strength of support that was judged would be extended to the rest of the AIG Group.
What about maturity transformation? I accept that this is an area where insurers are genuinely different from banks although I note that many life insurers use their balance sheets to sell long term (ie multi year) investment products but with a promise that investors can cash in the products early, albeit generally with early termination clauses in some cases. They can also – through, say, variable annuities – take short-term derivative promises from the financial markets and transform them into long-term quasi exotic derivative promises to policyholders. This doesn’t make them ‘banks’ per se but it is another area we might want to consider when assessing the changing nature of insures’ risk profile.
Assumption of significant credit risk? Prior to 2007 each of the three largest mono line credit insurers had individually wrapped almost $1tn of corporate paper of various quality and transformed them into Triple A rated paper. The total global market for all financial assets currently stands at around $76tn. And to take an example from more ‘traditional’ insurance, most life firms fund to a significant extent their annuity liabilities through corporate exposure that is subordinate to the debt provided by the banking sector.
None of these are knock-out arguments but I think they should give us pause for thought, not just to consider how such activities should be captured for solvency purposes but also how they should be supervised. This is especially the case if we are not able to supervise the whole group, only the insurance entity. These matters are of course a focus of concern for both the IAIS and FSB. The view that an insurance group without non-traditional, non-insurance activities cannot be systemic is widely held. It may even turn out to be true, but frankly the lessons which I have discussed suggest we should not rush to accept it without much greater thought.
My intention in revisiting this topic is to show how important it is to challenge the conventional wisdom and to adopt a forward looking approach in terms of assessing how new and innovative insurance business models may provide threats, not only to policyholders, but also to the economy more generally.
I said at the outset of my talk that I wished to avoid simply repeating the PRA’s approach document. Instead I have examined the implications of three important challenges facing insurance supervisors at the moment namely: how we ensure that the solvency regime (whether we use Standard Formula or internally developed models) adequately captures the risks being assumed by insurance companies, what resolution in a non-zero failure world might look like and the potential for the emergence of new risks in insurance markets that could give rise to risks to the wider financial sector or society more generally.
In essence, we cannot be said to have learnt the lesson of the crisis unless we apply effective and critical judgement when considering the outputs of the capital framework we use regardless of whether they are generated by standardised formulae or internal models.
We have to build a robust resolution framework such that it sits comfortably with a ‘non zero-failure’ regime, otherwise we cannot have properly reflected upon failings in the banking sector.
And without adopting a forward-looking approach that considers not only the risks in the current operating environment but also how they might be shaped by possible developments to the economic environment, we will again suffer from the lack of foresight which encouraged such overconfidence in the middle part of the last decade.
I am sure we are all agreed about the need to avoid that outcome. So, much still undecided and much still for us all to do.