RDP 9408: The Supervisory Treatment of Banks' Market Risk 6. Equity Risk

The equity risk proposal envisages a capital standard to cover the risk of holding positions in traded equities and related derivative instruments. The requirements will also apply to all instruments which exhibit market behaviour similar to equities. Equity instruments which can be treated as traded debt (some forms of preference shares for example) would be covered by the guidelines for interest rate risk.

Consistent with the building block philosophy, two areas of risk are identified as arising from traded equity positions:

• specific risk, which deals with the risk of a price movement confined to one particular equity or derivative product linked to it (credit-related risk); and
• general market risk, which concerns price movements that are unrelated to any specific equity.

6.1 Specific Risk

It is proposed that the specific risk of a bank's portfolio under the proposal is measured against the gross equities position, that is, the sum of long and short equity positions. The approach assumes the specific risk associated with any single equity instrument is unrelated to the specific risk of any other equity.

The main focus of debate has been the appropriate level of capital to be applied against specific risk. The equities proposal recommends that a capital charge of 4 per cent be applied where diversified books of liquid portfolios are involved. Where less liquid equities or portfolios are concerned, an 8 per cent charge is recommended. This differs from similar market-risk guidelines put forward by the European Economic Community through the latest Capital Adequacy Directive^[19] which regard a 2 per cent capital charge as sufficient where a liquid portfolio is concerned. As in the previous case, 8 per cent is viewed as the appropriate charge where less liquid equities or portfolios were concerned. It is viewed as the responsibility of the respective supervisor to determine which equities, or portfolios, would be deemed ‘highly liquid.’

6.2 General Market Risk

The general market risk of the portfolio is measured by the net equities position. That is, short positions in one set of equities can be used to offset the general market risk arising from long positions in other equities. The implicit assumption here is that there is perfect correlation between movements in equity prices. This assumption is acknowledged as unrealistic, but is adopted in the interests of simplicity.^[20]

A capital charge of 8 per cent of the net position is proposed to cover market risk.

The capital charges applied to an equity portfolio may be illustrated by the following example:

Example 4

	Position $m
	Long	Short
BHP ordinary shares	230
GIO Australia	100
Western Mining Corp		120
General Property Trust Income		75
Woolworths	50
TOTAL	330	195

For this portfolio, the gross position is $525 million and the net position, $135 million. The total capital charge, assuming this is considered a liquid and diversified portfolio, would be $31.8 million; $21 million against specific risk and $10.8 million against general market risk.

6.3 Empirical Analysis of the Equity Risk Proposals

The previous sections discussed the proposed structure of the capital charge to cover equity position risk. In this section, the performance of that capital charge is measured against actual share price movements in the Australian equity market.

Australian banks' exposures to equity risk is not as significant as exposures to price risk on debt securities and foreign exchange risk, mainly as a consequence of regulatory restrictions on banks' equity holdings. As a result, data on typical equity portfolios are not readily available. However, a number of banks have stockbroking subsidiaries whose exposures will attract a capital charge. A small sample of actual portfolios was obtained from banks in that position. These actual portfolios were supplemented by a large number of randomly generated portfolios.

The randomly generated portfolios were created from a selection of around 60 shares traded on the Australian Stock Exchange. The equity portfolios were created assuming that the size of banks' long and short positions in each stock followed a uniform distribution between $0 and $100 million. Three hundred portfolios were generated in this way.

The portfolio was revalued at fortnightly intervals using actual prices for each of the stocks making up the portfolio. The losses and gains on the overall portfolio were calculated from the differences in the value of the portfolio over a five year period from January 1988 to February 1993. A change of two standard deviations in the value of these portfolios was used as a benchmark for evaluating the performance of the capital charge.

The capital charge associated with each hypothetical portfolio was calculated using each of the following three methods:

• the standard capital charge proposed by Basle; 4 per cent charge against specific risk and an 8 per cent charge against general market risk (BASLE);
• the alternative capital charge proposed by the Economic Community as part of their capital adequacy directive which sets 2 per cent against specific risk and 8 per cent against general market risk (CAD); and
• an alternative methodology used by the Securities Exchange Commission in the US which is different to both of the above (SEC). The SEC rules require that capital of 15 per cent be held against the gross position of the portfolio. Hedging of long or short positions is permitted to the extent of 25 per cent of the longer side of the portfolio, and is deducted from the capital charge on the gross position. The SEC approach does not attempt to separate risks into specific and general market risk.

The coverage of the capital charges on the actual portfolios over the simulated losses were not as conservative as the results from the debt or foreign exchange proposals. The standard method covered around 85 per cent of all portfolio movements (i.e. both positive and negative movements). In other words, around 30 per cent of losses were not covered by the capital charge. The SEC comprehensive method performed somewhat better covering about 94 per cent of all changes in portfolio value while the E.C. approach covered only 79 per cent of all changes.

The results from the simulation analysis were consistent with the results from the actual portfolios. Coverage of the standard method was 91 per cent of all changes, while coverage of the SEC and E.C. methods were 99 per cent and 79 per cent respectively. Figure 3 below indicates that there is no strong linear relationship between the losses incurred and the capital charge obtained under the standard method. This is confirmed by the results of the linear regression analysis.

Regression results for equity capital charges:

# indicates that the variable is significant at the 1 per cent level. Standard errors are shown in brackets.

The results show that the capital charges have a highly significant regression coefficient but a very low coverage of potential losses. The R² statistic shows that none of the three methods are satisfactory in tracking the losses. Further, the coefficients on the dependant variables which may be interpreted as the ‘amount’ of the capital charge which should be applied, indicate that the standard approach recommended by Basle and the approach recommended by the European Economic Community are not adequate to cover two standard deviations in changes in portfolio values. In the case of the CAD method, the regression results indicate that twice as much capital is required, on average, to cover the losses observed.

The performance of the capital charges can be shown to change with different assumptions regarding the distribution of long and short positions held in the portfolio. The assumption underlying the results described above is that a bank is equally likely to hold a long position as a short position; that is in a diverse portfolio the average net position is zero. Altering these assumptions, to bias the portfolio towards long positions, significantly affects the efficiency of the capital charge.

Regression results for equity capital charges – bias towards long positions:

# indicates that the variable is significant at the 1 per cent level. Standard errors are shown in brackets.

The difference in explanatory power between the two sets of results can be interpreted as follows. When portfolios are constructed to have an average net position of zero, capital charges are too low, since the capital charge is a function of the net position. These capital charges are not well correlated with actual loss experience. When portfolios take on non-zero net positions, thus generating larger capital charges, the linkage between the capital charge and the losses improves. In other words, the use of net positions to measure general market risk is unsatisfactory. Risk does not appear to decline as net positions take on smaller values. The assumption that equity prices are perfectly correlated does not hold, and has a significant impact on the ability of the capital charge to predict losses.

Results from this analysis indicate that the proposed capital charge for equities does not cover as large a proportion of losses as the capital charges covering foreign exchange or debt securities. However, the problem is not that the size of the capital charge is not sufficiently high, but rather that the method used to measure the risk is not sufficiently accurate. An inefficient method of achieving greater coverage of losses would be to merely increase the amount of capital which is required to be held. Alternatively, the simple approach to measuring equity risk could be replaced by a more sophisticated method which tracked risk more closely. However, this latter approach would be at odds with the Committee's objective of keeping calculation of the capital charge as simple as possible.

Footnotes

These guidelines are due to be implemented by the EEC in 1996. [19]

It will be recalled that in the debt securities proposal, full offsetting was permitted as a means of calculating the general market risk of a portfolio but disallowance factors were introduced to modify the outcome. Effectively, only partial offsetting was allowed. No such disallowance factors are introduced in the equities proposal. [20]