Value at risk models have been developed by securities firms as an aid to improved internal risk management. Furthermore, only a very few firms yet have the ability to measure their risk on a consistent basis across their whole trading portfolio using these models. However as the techniques involved in these models become better developed there will be increasing demands on the securities regulators to recognise the output of these models for the purposes of calculating capital requirements for market risk. This development carries significant implications, which require careful consideration.
3.1 Changing the rationale for using models
There is an inevitable risk that, in moving to recognise models for capital calculations, the rationale for using models within firms will change. At present, those firms that have developed value at risk models have done so of their own volition for the purpose of managing their market risk. It is therefore argued by some firms that, in recognition of their superior risk management capabilities, these firms should benefit from a more favourable models based regime for capital calculation. Moreover, it is argued that other firms will be encouraged to develop similar risk management practices by setting more favourable regulatory capital standards based on models. However, if the primary incentive to use models is no longer the internal objective of better risk management, but becomes instead the prospect of lower capital requirements, regulated firms could be incentivised to use models in an inappropriate manner. Regulators could then come under pressure to approve the use of models with the primary objective of enabling firms to operate with a lower capital requirement. There could also be an incentive for firms to seek out models which delivered a lower capital requirement rather than improved risk management. This underlines the burden of responsibility which a move to a models based capital requirement will place on the regulators to be satisfied that firms are indeed using models for the primary objective of good internal risk management and only secondarily in order to benefit from a more favourable capital standard.
3.2 The purpose of regulatory capital
A further consideration is the difference between the circumstances which models address and those which concern regulators. Value at risk is a statistical measure based on historical observations of the likelihood of a firm's profit or loss falling outside a predetermined range. It therefore addresses what is likely to happen in normal circumstances. It does not seek to measure what might happen in face of an unusual event or when the unexpected happens. Yet it is precisely these very occasional but significant events (sometimes termed outliers) which regulators are most concerned about, and where capital is most required to protect the firm and the wider financial system. Consequently there will always have to be a cushion of capital for regulatory purposes over and above that implied by the model, but the amount of that cushion will be the subject of regulatory judgement. The Basle Committee has decided to establish a cushion of this type by requiring a multiplication factor of 3 to be applied to the value at risk calculation. As yet there is not enough regulatory experience of value at risk models to form a clear view on whether this approach will provide a sufficient cushion to deal with large adverse market movements, or whether an alternative approach, perhaps based on stress testing, would prove more appropriate.
3.3 Internal models or a regulatory model?
The case for recognising firms own value at risk models for the purpose of capital calculation is based on two main justifications. First, since there is no definitive model of market risk, it is argued that it is preferable to have a range of estimates rather than for the regulators to attempt to prescribe one model which firms should use. Secondly it is argued that firms which already have well developed value at risk models should not be required to undertake separate calculations of regulatory capital requirements.
However, for a number of reasons regulators will find that they will have to become involved hi specifying a number of the key parameters which should apply to the models which firms will be allowed to use for the purposes of calculating regulatory capital. At present different firms use a range of different techniques and assumptions in calculating value at risk for their own risk management purposes. Variables which may differ between firms include:
- the observation period over which model relationships are calculated
- the confidence level which is applied
- the minimum holding period which is assumed for the portfolio
- the acceptability of correlations within product types - e.g. between different equities and different equity markets
- the acceptability of correlations between product types - e.g. between interest rates and equity markets
In the absence of agreed conventions about these factors it is possible for different firms' models to produce significantly different value at risk numbers for the same portfolio. Clearly this would not be acceptable in the context of regulatory capital where a significant measure of equality of treatment is required as between different regulated firms. Consequently it is necessary for the regulators to require a reasonable measure of commonalty between the models which firms use for calculating regulatory capital, and to have a clear view about the acceptable degree of dispersion between the results of different firms'' models. Indeed, much of the work which the Basle Committee has undertaken recently has been associated with specifying certain of the key parameters such as the confidence level, the observation period, the holding period and the acceptable cross correlations with which they propose banks' models must comply in order to be recognised for the purposes of capital calculation.
It is therefore unavoidable that a move towards the recognition of value at risk will require regulators to set standards for the acceptable parameters. This will mean that firms may still find it necessary to undertake two separate calculations of value at risk - their own calculation for internal risk management purposes using their preferred parameters and a separate regulatory calculation using the standardised parameters. There will also have to be continuing discussion and debate about the appropriateness of the standardised parameters, which will need to be kept under review in the light of experience and changing market circumstances.
3.4 The problem of cross entity risk management
One of the main benefits from the use of value at risk models by securities firms is that it enables them to manage their risk on a portfolio basis across a number of different books and legal entities. Thus firms may, for example, aggregate all their European interest rate risk into a single portfolio with offsetting hedges recognised between different markets and different legal entities within their group. Moreover, since the use of models produces non-aggregative results (i.e. the value at risk in one larger portfolio will always be less than that in two smaller portfolios of similar instruments) regulators will need to consider how to respond to the request by firms to be able to benefit from such portfolio effects within any particular jurisdiction. In addition, this will create an incentive for regulated firms to minimise their regulatory capital requirements by moving their trading activities into regulated entities with larger portfolios where the benefits of the value at risk approach will be greatest. Nevertheless, regulators will need to recognise that each legal entity may be subject to differing regulatory and bankruptcy requirements, which must be accounted for in a regulatory capital regime.
Further consideration will need to be given to this factor, particularly in the context of group supervision.
3.5 Regulatory techniques and regulatory resourcing
The move towards using value at risk models for regulatory purposes will place significant additional pressures on securities regulators. In certain countries which have a tradition of relying extensively on external auditors, the regulatory commission may itself have to assume direct responsibility for oversight of the standards which will have to apply to firms using models, and to the models themselves. To the extent that regulators do not currently have expertise in this area it will have to be developed. Regulators will therefore need to give careful consideration to the implications for their recruitment and remuneration policies of the need to employ staff for this task before deciding to endorse a move in this direction. Some jurisdictions may wish to consider alternative techniques such as the employment of a specialised consulting firm or the development of a standard of best practice with their audit firms in order to perform the requisite independent model and control testing programme.