Speech Supervision of Market Risk: The State of Play
Assistant Governor (Financial Institutions)
Talk to the AIC Integrated Risk Management Conference
In just over 18 months from now, it is expected that a wide range of countries will formally implement amendments to capital requirements that will, for the first time, impose a capital charge on banks' market-related exposures.
In my remarks today, I want to cover a number of matters that will be important for the implementation of the market risk capital requirements in Australia from end 1997. First, I will note where we are up to at present and how I see the implementation process evolving over the months ahead. Second, I will say something about the issues that still need to be decided before new guidelines can be introduced. Finally, I will offer a few observations about risk management practices in Australian banks.
The Implementation Process
As you will recall, the Basle Committee on Banking Supervision released proposals for a capital charge to cover banks' market risks back in April 1993. These proposals set out a standard model for measuring risks arising from changes in prices of debt instruments, foreign exchange and equities. Following an extensive comment and discussion period, revised proposals were released in April 1995. The main development in these revised documents was the recognition of banks' own internal models for the calculation of market risk capital charges. This was subject to these models meeting a range of quantitative and qualitative criteria. The risk classes concerned were also expanded to include risks arising from changes in commodity prices, a recognition of the widening range of market activities carried out by banks.
A further comment period followed, and in January this year final amendments to the Basle Capital Accord were released, for implementation at the end of 1997. The Reserve Bank distributed these documents to banks, and made clear to them that we intended to introduce Basle market risk requirements in Australia on the same timetable to be used internationally. Incidentally, we also indicated that if a bank wished to move to the new standards ahead of their application to all banks, we would not object. We noted, however, that it would be necessary for any such bank to report on both the current and new basis until the market risk requirements became binding on all, so that a basis for comparison could be maintained.
The next step is for the Reserve Bank to turn the Basle documents into a draft Prudential Statement. The Basle proposals leave a number of issues to ‘national discretion’. In other words, they are to be decided by supervisors in each country in the light of local conditions. As well, there are some aspects of the guidelines which we are considering adjusting. I will say more about these in a moment.
We hope to be able to release a complete draft Prudential Statement fairly soon. To repeat, this draft Statement will mostly follow the already released Basle documents, but will also contain our proposals for issues left to national discretion and some other adjustments. We will allow a period for comment on the material in the draft Statement not covered by the Basle document. It is our intention, however, to have a final Prudential Statement in place in good time for full implementation, as I have already said, at end 1997.
As we proceed down this path, banks will need to be thinking about how they wish to be treated under the market risk regime. Obviously, there is an interconnection between what banks will choose to do and our final decisions on the detailed framework. But most of the main features are already known, and in general terms banks are thinking now about the broad issues – for example, whether they see themselves as standard model users or as candidates for the use of internal models.
Standard model users will typically be those with less extensive levels of market-related business. As a starting point, they will need to be able to organise their data on positions in the format specified in the standard model (for example, interest rate positions will need to be disaggregated into a series of time buckets). Even within the standard model, there are choices to be made about the selection of simple rules. For example, where options form part of a banks' trading strategy, the bank will need to consider which of the alternative approaches to the treatment of options they wish to apply.
Where the internal model approach is adopted, the process will be somewhat more complex. By end 1997, banks need to be in a position to satisfy all components of the qualitative and quantitative standards set out in the January 1996 Basle guidelines. The Reserve Bank will not be amending any of those requirements. We will be holding discussions with banks, and using our visit program, to satisfy ourselves that the risk management practices in place within banks meet the requisite standards, and that models perform with acceptable accuracy. If they do not, then model user status will not be approved.
The Basle Proposals – Issues for National Discretion
Let me turn now to some of those matters which are left to national discretion in the Basle guidelines.
One of the decisions we have already made relates to the treatment of Tier 3 capital. As you may be aware, we have decided against introducing this aspect of the Basle arrangements into the Australian market risk framework. You will remember that Tier 3 capital is short-term (maturity of 2 years or less) subordinated debt, with a lock-in clause that would prevent repayment if that would cause minimum capital requirements to be breached.
The Reserve Bank has a history of taking a relatively strong stand on the definition of capital. As many of you will know, the current treatment of subordinated debt, outlined in the capital adequacy guidelines, is tighter than that adopted by most other bank supervisors (we require a seven-year minimum term compared with the five-year term adopted by most others). Banks have appreciated, especially in recent years, that the only capital that can be relied upon in the last resort is the permanent capital contribution made by the shareholders. We had no wish to make changes that could lessen the emphasis on this ‘highest quality’ capital. We know that the capital position of the Australian banking system is strong and we believe that the increase in the minimum capital charge for the system in relation to market risk will be small in comparison with the level of excess capital currently held in banks. We decided, therefore, not to accept this aspect of the proposals for the time being at least. We may reassess the applicability of Tier 3 capital once the market risk arrangements are actually in place and operating. We will watch for, and take into account, any signs that inability to access Tier 3 is putting Australian banks at a competitive disadvantage.
A second area with which you may be familiar is our decision to proceed with that part of the guidelines dealing with so-called qualifying securities. Amongst other things, this provides for concessional risk weights to be applied to debt issues of certain high quality corporates – the specified risk weightings are significantly lower than applied to loans to the same counterparty. We will need to set out the criteria for securities to ‘qualify’. Basically, those securities rated investment grade by at least two approved credit ratings agencies will be regarded as qualifying for the concessional risk weights. We will publish a list of rating agencies acceptable for this purpose. Securities issued by Australian banks will automatically qualify, while specific risk weights on traded Government securities will become zero in the new regime.
One very significant issue we've had to deal with concerns the definition of the trading book. While the market risk charge for foreign exchange and commodities is to be levied on both traded and non-traded positions (with some limited exceptions), in the case of interest rate and equity exposures the market risk charge is to apply to traded portfolios only. The Basle guidelines define the trading book as comprising those positions intentionally held for short-term resale and positions arising from market making. In our view positions may be regarded as being held with a trading intent if:
- they are regularly marked to market as part of the internal management processes; and
- the position takers have autonomy in entering into transactions within predetermined limits.
Since the definition of a trading book is strongly intertwined with the broader risk management practices of each bank, we will be looking for each bank to lodge a policy statement with the Reserve Bank concerning the activities that the bank regards as constituting part of the trading book and detailing the arrangements in place to prevent inappropriate switching of transactions between the trading and non-trading books.
The approach we have adopted for the definition of the trading book has direct implication for the treatment of PAR assets in the (general) market risk framework. The Basle guidelines made no reference to the treatment of highly liquid assets held as part of a compulsory supervisory or regulatory arrangement, such as PAR assets. Given their presence in the Australian system, one approach would have been to include all PAR assets within the market-risk framework because they represent readily traded securities. We observed that some banks clearly trade PAR assets actively, mark the securities to market, and in all significant respects, treat them as trading stock. That being the case, we saw no reason to exclude them from the market-risk requirements. But in other banks, the nature of dealing in PAR assets is often very limited in scope, and falls well short of what would be regarded as trading in any sense of the word. Some banks do not trade their PAR assets at all, effectively holding them to maturity and treating them purely as investments for accounting purposes.
We formed the view that applying a single treatment to the PAR assets of all banks, though administratively simple, was not justifiable as part of the introduction of the market risk requirements. Instead, we consider that banks should use the trading book criteria discussed above to determine whether their PAR portfolio falls within or outside the trading book, and therefore the ambit of the market risk requirements.
Some of the matters we have had to consider reside at the more technical end of the spectrum. You'll recall that the Basle proposals specify two broad approaches which banks can adopt for the measurement of market risk. The first is the standard model, an approach which sets out some relatively simple rules for the measurement of risk on interest rate, foreign exchange, equities and commodities positions. Also specified separately under the standard model are a number of alternative ways of measuring risks associated with banks' option trading portfolios. One of these alternatives is a scenario-based approach which requires options positions to be measured against a series of price and volatility movements.
The second broad approach set out in the Basle guidelines involves the use by banks of their internal risk-measurement models. In essence, Basle has said that subject to meeting a number of quantitative and qualitative tests, banks may use internal value-at-risk models to measure their market exposures. The qualitative tests are basically there to ensure that risk-management processes and practices are of an appropriately high standard. The quantitative tests set the main parameters to be used in the value-at-risk models (a 99 per cent confidence level, 10-day holding period and so on). One of the controversial requirements of the internal model option is that the output of such models – that is the aggregate risk estimate – be subject to so-called ‘multiplication’ and ‘plus’ factors. The objective is to scale up the measure of risk and the resulting capital charge to provide greater overall conservatism to the capital requirement and to cover possible flaws in model estimation.
One issue to which we have given considerable thought is the question of whether greater flexibility in model types should be allowed under the market risk guidelines. One particular issue we are considering is whether to allow greater use of scenario-based approaches, such as those specified in the standard model treatment of options, for the measurement of risk in trading books more broadly.
The practical issue in adopting scenario approaches for supervisory purposes, is how to specify the size of the price shock to be used as the basis for a capital charge. One approach would be to use the price shifts specified or implicit in the standard model.
What this would deliver is a hybrid approach which sat between the standard and the internal model approaches. An alternative would be to specify price shifts for use in the scenario-based approach which corresponded to a 99 per cent confidence level. This would involve regular re-estimation of price volatilities but the result would be an approach not far removed from a value-at-risk model.
Another alternative we have been contemplating is quite different in nature. It would permit banks to generate a capital charge based on their internal, board-approved limit system. This is an approach which might appeal to banks doing only a small amount of market-related business. It would allow them to hold a relatively constant level of capital over time, rather than developing a standard or internal model methodology to measure changing, albeit small, market-risk exposure. The price would be that the resulting capital charge would almost certainly be higher on average than it would be under the permissible alternatives. To make such an approach acceptable, there would need to be strict processes and procedures in place to ensure board-approved limits were not breached. We would need also to ensure that the limits were not set unrealistically low in order to reduce the capital charge.
Before moving on, I would like to explain something of our motivations in contemplating these changes.
We think that in accepting the use of bank value-at-risk models, supervisors should leave open the possibility that alternative approaches to measuring risk might be developed and implemented. There is rarely a single, best way of doing things, particularly in such a dynamic area as risk management. We do not wish to constrain banks exploring new ways of measuring their risks and, providing they are acceptable, using such approaches for the measurement of regulatory capital charges. Nor do we want to encourage the development of dual risk-management systems – one for internal use in a bank and one for regulatory purposes only. Underlying our thoughts on this matter is the fact that we intend applying the market-risk guidelines to all Australian banks, not just the ‘internationally active banks’ which will be the focus of attention in some other countries. In these circumstances, we feel there is a case to be more flexible than provided for in the international guidelines.
Of course, in providing possible alternatives, we also need to be satisfied that whatever approach banks use to generate a regulatory capital charge, there is a broad consistency of treatment across the banking system. To assist in reaching views about consistency, we have in mind conducting some exercises in the near future to determine how the various alternative methods perform using some benchmark portfolios. These are issues we will be discussing with banks in the period ahead.
Risk Management Practices in the Australian Banking System
Before concluding, I thought it might be useful to offer a quick overview of how we see risk-management practices and standards within the Australian banking system. As most of you will be aware, we have been conducting a program of visits to Australian banks over the past year or so. This program supplements the visit program already in place to analyse credit-related risk within banks. Largely as a result of the visit program, we've been able to get some insights into risk-management practices of banks that were previously difficult if not impossible to obtain. Let me chance a few generalisations.
In aggregate terms, the level of pure market risk is not high in the Australian banking system. Financial instruments used and traded in banks remain predominantly at the vanilla end of the spectrum and open positions are constrained by reasonably conservative limits imposed by banks' boards. This is the basis of our view that the introduction of the Basle capital framework will not generate the need for much new capital. I think it would be fair to say also that a good deal of thinking is being given to risk management issues within banks and some of our banks at least seem to be as advanced in this as any in the world.
For those banks active in trading, our perception is that risk measurement methodologies are relatively advanced by international standards. For those banks where trading is not such a focus, approaches to risk management seem to be broadly appropriate to the level and sophistication of the business being conducted. While these conclusions hold generally for each of the banks, there is nevertheless a lot of variation between the different banks. We have certainly seen instances of specific shortcomings that we have taken up with bank management. While the standards are generally good there is always room for improvement.
This generally favourable assessment provides no grounds for complacency since, as we all know, good risk management methodologies do not preclude problems from emerging due to operational shortcomings. One of the main observations from our visits is that thinking about risk issues usually runs well ahead of implementation. Sometimes banks face some pretty basic issues in relation to risk management, for example, getting position or risk data together and aggregated, sometimes globally or ensuring that new risk measurement approaches are being applied on a consistent basis across portfolios. It is also apparent that many risk management systems are the product of decisions taken, sometimes, over a long period of time. Very often, the catalyst to system development has been the decision to enter new markets or to introduce new instruments. Resulting risk management structures can end up disjointed or less integrated than they ideally would be. This can make it more difficult for management to have a clear and consistent picture of the overall risks the bank faces. As well, the more linkages that need to be made between diverse systems the greater the potential for mistakes of various sorts to occur. This is well recognised by bank management and many Australian banks are working towards more modern and integrated systems, often involving heavy investment in new technology.
Another area where work needs to be done is in relation to stress testing. Our feeling is that few banks to date have the capability to fully meet Basle standards in this area.
As we all know, value-at-risk models deliver a statistically determined, and usually aggregate, estimate of risk. We also know that the price or volatility movements implicit in risk measured at the 99 per cent confidence level, are movements we can expect to see occur two or three times a year. But these are not the market shifts that will ‘break the bank’. In fact, experience tells us that large price movements occur more frequently than implied by the normal distribution which is the foundation of most value-at-risk models. Sometimes, market price movements can be many multiples of what probability theory would suggest. For their own internal risk management purposes, banks need to consider the impact of large price changes on their portfolios. Stress testing permits such an analysis.
Banks should also be thinking about the impact of particular events on their estimates of portfolio risk. What those particular scenarios might be cannot be pre-specified easily. Usually, they will be peculiar to the characteristics of individual portfolios. The point is clear nonetheless. Banks need to turn their mind to what we might call ‘unexpected’ events as well as ‘expected’ events that can generate losses and make use of that information in the risk management process. Capital, after all, is there to ensure a bank lives to fight another day after the unexpected has become reality.
Risk management methodology and practice is a rapidly growing focus of interest within the banking system. Supervisors are also vitally interested in exploring ways to more accurately measure risk exposures and set capital charges. In this sense, both banks and banking supervisors, here and internationally, have been moving in the same direction, with very similar objectives. The market risk guidelines represent an important departure in supervisory practice in recognising the more rigorous risk management methodology contained in banks' internal models for the purpose of setting capital standards.
I have little doubt that the experience we've gained from developing and implementing the market-risk standards will spill over, in time, to a focus on more ‘scientific’ approaches to the supervisory treatment of other forms of bank risk – the most obvious candidates being credit risk, interest-rate risk on the balance sheet and perhaps even the very difficult area of operational risk. These methodologies logically feed into a closer analysis and recognition of the economic measurement of capital. Banks are already moving down these paths and we are expanding our own expertise and research and analytical capabilities to address these issues.
Of course, we also recognise that risk management is about more than sophisticated statistical techniques. It is also about sound management judgment and practices, including applying policies effectively in practice. Failure to observe basic requirements like separation of back and front offices can have the most serious consequences. From our perspective, the process of expanding the capital adequacy framework to encompass market risk is as much about developing a sharpened risk consciousness within banks as it is about the generation of a capital charge for market risk.