Position Paper on the Commission Services Consultation Paper on Trading Activities Related Issues and the Treatment of Double Default Effects

Transcription

1 Position Paper on the Commission Services Consultation Paper on Trading Activities Related Issues and the Treatment of Double Default Effects May 2005

2 Contents Introduction... 3 The treatment of counterparty credit risk and cross-product netting... 4 Approval to adopt an internal modelling method under EDA... 4 Internal Modelling Method (IMM)... 5 The Alpha multiplier... 5 Effective EPE... 5 Risk-neutral vs. actual distributions... 6 Margin agreements... 6 Model validation... 7 Stress testing... 7 The integrity of the modelling process... 7 Standardized Method (SM)... 8 The Beta multiplier... 8 Credit Conversion Factor... 8 Collateral... 8 Other considerations... 8 The treatment of double default General matters Protection providers Obligors Correlation parameters Maturity effects Further considerations Double recovery effects Calculation of expected losses Stress testing Supervisory Review Process Short Term Maturity Adjustment Improvements to the current trading book regime Perimeter of the trading book Modelling event risk and default risk Treatment of less liquid instruments Other considerations A proposed capital treatment for unsettled and failed trades /20

3 Introduction In preparing the Italian banking industry s position on the Commission Services Consultation Paper on Trading Activities Related Issues and the Treatment of Double Default Effects, the Italian Banking Association gathered the various points of view of its members and the proposals concerning the matters treated in the document. Based on the comments received and the work of several ad hoc interbank working groups, ABI developed this position paper, which has been transmitted to the national supervisory authorities, the European Commission and, in a slightly modified version, to the Basel Committee. 3/20

4 The treatment of counterparty credit risk and cross-product netting While we share the regulators aim of fostering the development by financial institutions of increasingly advanced methods of risk assessment and analysis, special attention must be devoted to ensuring that this effort results in savings sufficient to justify the investment needed to produce the requisite analyses. In this regard, the calibration of the alpha and beta multipliers is of the greatest importance. Approval to adopt an internal modelling method under EDA The consultative document provides that, for OTC derivatives for which a bank is not given authorization to use its internal model, the bank can obtain approval to use the standardized method; otherwise it must use the current exposure method (CEM). The proposal envisages the possibility of using all three methods jointly at the level of the banking group, while precluding that option for the individual legal entity. This approach would not seem optimal in light of the fact that operating reality, especially for a commercial bank, does not involve uniform treatment of different portfolios. For example, the business of selling derivatives to customers differs substantially from that of treasury derivatives, hedges, or ALMs. Netting agreements as well are limited in number vis-à-vis corporate counterparties but more common in the interbank segment (and with financial entities in general). Considering that the approaches differ in computational complexity, it would be helpful to allow for greater flexibility in the application of the models and, while regulatory arbitrage should not be permitted, the use of different methods even within a single legal entity should be allowed. It would especially useful to allow combined utilization of models not only within a group of products but also with respect to the same counterparty. It is increasingly common, in fact, to find within a trading book OTC derivatives that are hard to model for purposes of calculating expected positive exposure (EPE) e.g., equity swaps in which one leg is an exotic option linked to a basket of assets. Given that ordinarily one has a number of different OTC derivative contracts of different types in place with a single counterparty, it would be preferable to allow the application of internal models to all the contracts that those models can cover and apply the standardized method (or the CEM) to the others, regardless of legal entity. 4/20

5 Internal Modelling Method (IMM) The Alpha multiplier The consultative document suggests an alpha multiplier of 1.4, with the option for individual supervisors to increase it; where the multiplier is estimated internally, a floor of 1.2 is set. This seems especially burdensome. The alpha multiplier of 1.4 is quite burdensome because it is estimated on the basis of hypotheses assuming low granularity, diversification of counterparties in portfolio and number of risk factors. These assumptions do not correspond to the properties of banks actual portfolios. In the event of development of an internal model to estimate the parameter, which requires the allocation of human and financial resources, the compulsory floor of 1.2 might make the costs of implementation unjustified, considering among other things that an internal model could very well result in a lower alpha factor. The main market operators (including ISDA) have initiated an interesting discussion on this floor. The burden imposed could be a disincentive for banks to refine their methodology for determining the alpha multiplier, and hence their economic capital. Moreover, since it is determined as the ratio between economic capital deriving from a full simulation (credit and market risk factors) of the counterparties exposure and economic capital based on EPE, the value of the alpha multiplier depends on portfolio composition, and in particular on the number of counterparties, the types of contract, and their maturities. In addition, as noted, the multiplier is also determined by the methodology used to estimate EPE. Its value takes account of the risk stemming from the granularity adopted in the estimate of the exposure. As a consequence, the greater the granularity adopted, the lower the multiplier needed to take account of the relevant risk. On these premises, in the view of the Italian banking industry it would be preferable not to set a floor for the alpha factor, leaving it up to supervisors, in validating the model, to authorize a level of alpha consistent with the individual bank s specific portfolio characteristics. Alternatively, it could be provided that the rules lay down distinct alpha values, each linked to definite characteristics of a portfolio and of the methodology used to estimate EPE, so as to leave some flexibility for individual banks. Effective EPE Requiring banks to estimate EPE daily is quite costly and of questionable usefulness. It would be better to interpret this provision as requiring calculation on a daily basis for the first few days of the time bucket but with less frequent reports. 5/20

6 The regulation calls for rigid identification of the time buckets for determining the EE profile. This is too constrictive, and not in keeping with best practice. In fact, in order to optimize the EPE estimate, the grouping of instruments into bucket classes should be appropriately based on the positions held and the bank s portfolio structure. Specifically, to take account of contracts with short maturities, one could use time intervals of various amplitude (e.g., a weekly t bucket up to three months, monthly up to one year, quarterly from one to two years, half-yearly up to five years, and then yearly). Increasing the number of short maturity buckets, one can better grasp the evolution of the EE profile for short maturities. The calendar for introduction is too stringent. In particular, in view of the deadlines envisaged for the introduction of the rules, it could be hard to meet the requirement that, for the model to be approved, it must be effectively utilized in the year preceding. It could be advisable to provide national supervisory authorities with a certain degree of discretion, in the transitional phase, in applying such requirement. In terms of collateral, the period of risk on collateralized counterparties has been taken as a minimum of two weeks. This period should instead be brought into line with market practice, where the normal time reference is ten days. Finally, the calculation of gross and net EPE i.e., the distinct indication of the effect of collateral appears hardly applicable to securities financial transactions, and repos in particular. What is more, conduct stress tests also for gross EPE entails an unjustified workload, considering the nature of securities financial transactions and transactions in collateralized derivatives. It should also be pointed out that estimating rollover risk by making rollover assumptions could result in unrepresentative estimates of EPE. We believe that the definition of Effective EPE as the weighted average of the peak values of EE is sufficiently prudent, capable of adequately taking account of rollover risk. Risk-neutral vs. actual distributions We confirm that it is not advisable for the regulation to specify which distribution to use but it should leave the choice between risk-neutral and actual distribution to the bank, with the prior approval of the supervisory authorities. Margin agreements If the netting set is subject to a margin agreement and the internal model captures the effects of margining when estimating EE, the model s EE measure may be used directly to calculate EAD. However, there is a misalignment between the calculation of EAD, which for the first year is based on the definition of Effective EPE, and the Accord s formula, which contemplates only the use of EE. Greater consistency would be helpful here, amending the general formula, or else a specific clarification on its use. 6/20

7 Model validation The requirement for validation of an internal model is its capacity to simulate variations in risk factors (interest rates, exchange rates, share indices, etc.) on time horizons longer than one year, potentially up to the maturity of the contract. This means for interest rates, for example an analysis of time series to estimate the mean reversion level and mean reversion coefficient vectors and selecting the stochastic process that best describes the movement of risk factors over time. In a word, this means adopting probability-based mathematical models, which could turn out not to be robust to variations in. This requirement could be a disincentive for internal models; it could also lead to methodological choices, and consequently EE profile evaluations, that are not fully representative. Think for instance of the possibility envisaged in ISDA contracts for OTC derivatives of break clauses allowing the counterparties to terminate the contract in advance. In these cases, the contract is considered up to the date of the possible early termination, regardless of the formal contract maturity. One might well consider a less stringent requirement, such as the ability to simulate variations in risk factors (interest rates, exchange rates, share indices, etc.) over time horizons longer than one year five years, say. t Stress testing Banks are required to have suitable stress testing processes in place, and within them the bank must evaluate its solvency target over the life of all contracts in each netting set. Considering that a given netting set may comprise contracts with widely differing maturities (from one to thirty years), we think clearer indications concerning what stress to use are in order. The integrity of the modelling process The indication is for time series of market data of at least three years for estimating volatility and correlation matrices over an entire economic cycle. But recent market values are considered more significant and representative, so it is suggested to allow the estimation of volatility and correlation matrices using one-year series and adopting methodologies such as EWMA, which by using weights that decrease backwards in time accord more importance to more recent observations, like the methodologies used in internal models to estimate generic trading book risk. 7/20

8 Standardized Method (SM) The Beta multiplier In general, it is worth underscoring the advisability of defining the beta factor on the basis of empirical evidence, so as to give an incentive for the development of an internal model for CCR risk assessment. We should accordingly recall the strong implicit prudence of the methodology, which precludes offsetting between different hedging sets (e.g., interest rates and exchange rates). The beta factor should also incorporate the effects of modelling error and bad state economy. Yet while modelling error may be negligible for delta-invariant portfolios, the effect of bad state economy is arguably already incorporated, at least partially, in the formula, which does not allow for considering the diversification effect between hedging sets. The beta multiplier is set at 2, higher than the 1.4 coefficient for alpha, to take account of the fact that Effective EPE estimated through the internal model could take a higher value than EAD. However, in determining beta, it would be good to also take account of the assumptions made in the Standardized Model, which are such that the estimate of EAD incorporate prudential traits in the risk assessment. We refer in particular to the fact that an open position for a short maturity, such as one day, remains open and invariant for the entire EAD estimate time horizon, and no risk mitigation due to margining is recognized. Credit Conversion Factor While we understand the need to standardize the treatment of derivatives, since the determination of the Credit Conversion Factors (CCF) depends on market conditions, if no space is to be allowed for an internal determination by applying the methodology, at least we must provide for periodic re-estimation of the coefficients to approximate them to actual market conditions. Collateral The consultative document gives no indications on the treatment of collateral in the Standardized Model. Given the importance of the issue, greater clarity would be welcome on the possibility of reducing EAD by computing under the formula for the market value of the collateral. Other considerations It is worth remarking, as to credit derivatives, that there is no detailed specification of the implementing procedures for recognition of a hedge when this is via an internal deal from banking to trading book and when it is negotiated also with more than one counterparty. In particular, it would good to have clarification on the conditions for recognizing a hedge hence for including it in double default and the criteria for identifying the protection provider. 8/20

9 Further on credit derivatives, we note that the provision for a 5% add-on for CDSs with qualified reference obligations is high and worsens the present situation. 9/20

10 The treatment of double default General matters ABI appreciates the suggestion of a methodology to measure double default; we are convinced that the present treatment does enable banks to benefit fully from the risk mitigation due to credit protection. The substitution ceiling approach set out in the Re-casting Directive 2000/12/EC of the European Parliament and of the Council of 20 March 2000 relating to the taking up and pursuit of the business of credit institutions and Council Directive 93/6/EEC of 15 March 1993 on the capital adequacy of investment firms and credit institutions of July 2004 left a number of important points unresolved, including: the absence of incentives for personal guarantees on credits with high creditworthiness; the lack of risk sensitiveness vis-à-vis credits with low standing; and inconsistency with the methodological principles of probability calculation. In any case, we consider the approach proposed by the European Commission to be too rigid in defining the scope of application and too conservative in its underlying hypotheses. Highly preliminary simulations show that, as currently defined, the approach is especially conservative as regards sub-investment grade counterparties and in some cases also for the lower-rank investment grades, for which the capital requirements are higher than in the substitution approach. 4,0% Graph 1. Guarantor AA. Sales of primary obligor >50 million euros 6,0% Graph 2. Guarant or A-. Sales of primary obligor >50 million euros 3,5% 3,0% 2,5% Hypotheses: -45% LGD f or guar antor and pr i mar y obl i gor -M = 1 -Guar antor i s AA (0.03% PD) -Sal es of pr i mar y obli gor > 50M eur os 5,0% 4,0% Hypotheses: -45% LGD f or guar antor and pr imar y obl igor -M = 1 -Guar antor isa - (0,17% PD) -Sales of pr imar y obl igor > 50 M Eur o 2,0% K_Sosti tuzione 3,0% K_Sostituzione 1,5% K_DoubleDef ault K_Doubl edef aul t 2,0% 1,0% 1,0% 0,5% 0,0% 0,0% AAA AA+ AA AA- A+ A A- BBB+ BBB BBB- BB+ BB BB- B+ B B- AAA AA+ AA AA- A+ A A- BBB+ BBB BBB- BB+ BB BB- B+ B B- 10/20

11 3,0% Graph 3. Guarantor AA. Sales of primary obligor <5 million euros 5,0% Graph 4. Guarantor A-. Sales of primary obligor <5 million euros 2,5% Hypotheses: -45% LGD f or guar antor and pr imar y obl igor 4,5% 4,0% Hypotheses: -45% LGD f or guar antor and pr imar y obli gor -M = 1 2,0% -M = 1 -Guar antor i s AA (0.03% PD) -Sales of pr i mar y obl igor <= 5 M Eur os 3,5% -Guar antor is A - (0,17% PD) -Sal es of pr imar y obl igor <= 5 M Eur os 3,0% 1,5% K_Sostituzione 2,5% K_Sostituzione K_DoubleDef ault 2,0% K_Doubl edef ault 1,0% 1,5% 1,0% 0,5% 0,5% 0,0% 0,0% AAA AA+ AA AA- A+ A A- BBB+ BBB BBB- BB+ BB BB- B+ B B- AAA AA+ AA AA- A+ A A- BBB+ BBB BBB- BB+ BB BB- B+ B B- Also, where the credit guaranteed is small business retail, using the corporate formula (page 38, Working Document) to calculate the capital requirement is excessively burdensome and is actually a disincentive by comparison with the substitution hypothesis. Protection providers The treatment of double default in the consultative document can apply only in the presence of (a) guarantees granted by financial institutions or (b) protection providers rated at least A- 1 (or an equivalent PD). The combination of these two requirements (type and rating of protection providers) is excessive, in that the good rating of the provider is guaranteed per se by the limitation of application of double default to types which for regulatory and credit access reasons are already bound to maintain an adequate rating. 2 Assuming that for prudential reasons one accepts the idea of imposing a further condition in addition to the provider-type requirement, we must note that the rating by itself is not a relevant criterion in determining the relationship between obligor and protection seller; other elements come into the picture. 3 Among other things, a recognition threshold obviously engenders excessive variability in the capital requirement as a function of the guarantor s rating (cliff effects), and it is not clear how one must behave in the event that a guarantor falls below the rating threshold. From the technical standpoint, it would seem preferable to set limits only in terms of PD (rather than tie oneself to the A- outside rating requirement), 1 Requirement (a) seems to be intended to minimize wrong-way risk (i.e., the existence of a correlation between guarantor and obligor due to dependence of the performance of both on common economic factors). Requirement (b) evidently assumes that the correlation between protection provider and prime obligor increases when the guarantor s rating is lower than A-. 2 If anything, this combination, albeit with the limits set forth in the paragraphs that follow, might be justified in the case of an extension of eligible protection providers (see below). 3 As regards credit derivatives, for instance, in general, as the riskiness of the protection seller increases, so does the degree of collateralization requested by the protection buyer. 11/20

12 because rating may be based on different criteria from bank to bank (and among different ECAI) and may represent different degrees of risk depending on the cyclical phase. All this premised, we propose to broaden the range of guarantors eligible for the double default provision to include all eligible guarantors identified for personal guarantee/credit derivatives 4 in the Standard and IRB Foundation approaches (without prejudice to the requirements set in those approaches). In particular, the a priori exclusion of recognition of protection accorded by sovereign states in addition to being excessively prudent would mean failure to accord full recognition to public credit guarantee instruments that in Europe have worked successfully to facilitate credit access for SMEs. 5 We hope that banks will be allowed to demonstrate to their own supervisory authorities that it is possible to recognize as eligible for double default treatment also sovereign states that show a correlation with the obligor no higher than that set in the consultative document (ρ og =0.6). In addition, the overall exclusion of corporate guarantors is hardly consistent with the criteria for detailed derivation of risk weights under the IRB approach, which, given equal PD, do not distinguish in terms of risk between corporate exposures and those vis-à-vis banks and sovereign borrowers. It might be justified to impose an additional requirement on corporate protection providers, 6 postulating that the correlation between protection provider and primary obligor increases when the guarantor s rating is lower than A-. Finally, as regards transactions in credit derivatives, we call for the elimination of the rating (or better, PD) requirement on corporate protection sellers if the contract contains clauses requiring the seller, in the event of a downgrade, to increase the collateral. Obligors Double default treatment can be applied exclusively for exposures vis-à-vis: - corporate borrowers (including SMEs but excluding specialized lending), and 4 For credit derivatives only, the principal counterparties of the banks buying protection are insurance companies and investment funds. Narrowing the range of eligible guarantors for these technical forms is too severe a penalty and fails to reflect the successful outcome of transactions of this kind in the credit derivatives market. 5 Just as examples, let us cite the FEI and Italy s SACE. 6 From the prudential standpoint, that is, it might be reasonable to set this requirement even on collateralized exposures, to counter the potential wrong-way risk between corporate guarantors and corporate obligors envisaged in the consultative document. 12/20

13 - small businesses classed as retail. In our view the exclusion of financial institutions as obligors is over-prudent. While in a domestic context there may effectively be a correlation among financial institutions, in a broader context this is questionable. For instance, in the case of exposures towards domestic financial institutions guaranteed by foreign financial institutions, the correlation between the two will not necessarily be strong, so the prerequisites for application of double default provisions may be met. The Italian banking industry accordingly calls for greater flexibility in recognizing obligors, allowing banks to demonstrate to their respective supervisory authorities that the correlation between the financial institutions involved is below the level set by the consultative document (ρ og =0.6), so that they are eligible for double default treatment. Correlation parameters The calibration of the correlation parameters, especially ρ g, is quite severe. In treating the double default effect it is set at 0.7, whereas the ρ o set under IRB methods is at most Even allowing the possibility that ρ g may be higher than the correlations envisaged in IRB, as it is estimated on a portfolio that includes not only financial but also corporate and sovereign counterparties, we feel that 0.7 is too high. We should also like to observe that, given the formula for calculating K DD, an excess of prudence, hence limited benefits in terms of regulatory requirements, may produce the paradoxical effect of a disincentive to hedge exposures to the riskiest counterparties. Our recommendation is therefore that the correlation parameters be revised, and in particular ρ g, ; this is all the more necessary in the event that the set of eligible protection providers is broadened, as we have proposed. Maturity effects In addition to the over-prudent aspects of the treatment discussed above, the adjustment for maturity effects is also conservative in considering the lower PD as between guarantor and obligor. Further considerations Double recovery effects The document utterly neglects the possibility of double recovery effects, i.e. the fact that for a guaranteed exposure LGD may be lower than that of the 13/20

14 guarantor. The paper holds that the practice of recovering from both guarantor and primary obligor is not common. Yet this is because the guarantor is usually able to pay; if this were not so, the banks would also proceed against the obligor. The relevant LGD, in effect, is that of the guarantor. The possibility of using that of the obligor is remote in view of the eligibility conditions, which already provide for recovery from the guarantor at first request. Instead, we call for allowing the possible recognition of the double recovery effect via internal estimates of LGD for the banks using the AIRB approach and a reduction of regulatory LGD under the FIRB method, reflecting the fact that in banking practice recovery is sought both from the primary obligor and against the guarantor and not from one party only, as the document now postulates. As an example, assume the guarantor-related LGD is 45% and so is that of the obligor. In this case, in the event of obligor default the actual LGD will be less, because the reimbursements from obligor and guarantor will be summed. Moreover, the linear relation between LGD and capital charge would permit a direct effect on the latter. The reason why this way of considering double recovery is not allowed (page 3, footnote 4 Explanatory Document) should be explained more clearly. Calculation of expected losses We agree with the proposal to assume as equal to zero the expected loss on the guaranteed portion of the exposure. Stress testing Stress testing of the treatment of the double default effect requires special care, because the PDs associated with financial institutions display a trend with significant discontinuities. Supervisory Review Process As to Supervisory Review Process, we find that implementation of the provisions of para. 3.3 of the Explanatory Document will be difficult, in that it is quite unrealistic to assume that the creditor bank can know whether the guarantor bank has a concentration of exposures towards companies operating in the same sector as the obligor. 14/20

15 Short Term Maturity Adjustment While we appreciate the attempt to institute a reduced regulatory requirement against some short-term exposures, we believe that the proposed maturity adjustment is not conceptually appropriate to capture the risk of exposures with less than a year s maturity. The short-term maturity adjustment as conceived by the European Commission, in fact, is a proxy to capture losses in case of a downgrade of exposures with maturity of longer than one year. This premised, and acknowledging that there is no easy way at this point in the proceedings to introduce an adjustment other than via maturity, we must note that the principle of non-relationship on short-term credits (when there is no pressure on the bank to concede a rollover of the exposure with the same client), which should cover transactions in both the first and the second group, is somewhat vague and hard to implement. In particular, it is worth noting that a bank s prompt knowledge of any deterioration in the counterparty s creditworthiness often depends precisely on its relationship with the client. Requirement a): It would be better to drop the non-relationship requirement and rely more heavily on specific triggers in the contract, such as the bank s effective ability to recover its credit rapidly if the counterparty s creditworthiness falls below a predetermined threshold. We agree with the list of major categories of transactions, but we would suggest explicit inclusion also of hot money. It is noted that assessment of a bank s commercial willingness to terminate a transaction if the obligor s creditworthiness deteriorates is difficult to determine objectively. 15/20

16 Improvements to the current trading book regime Generally speaking, the Italian Banking Association considers that for reasons discussed below the proposed new rules on the trading book may generate an increase in the economic capital required that is not commensurate with the real underlying market risk on the trading book. Moreover, some of the rules would appear to diverge from banks internal operational practices, which runs counter to the Commission philosophy of trying to foster convergence between the two aspects. Perimeter of the trading book The consultative document, while identifying on a general plane the positions that can be in the trading book, 7 later lists some exceptions to the general rule. That is, even when they conform to the general rule the document excludes certain specific types of transaction 8 whose capital ratio is to be calculated according to the rules on credit risk. In our view, the approach poses several problems. First, owing to the exclusion of some types of transaction, the trading book does not distinguish between totally covered assets for which both generic risk and specific credit risk are covered and those that are partially covered or uncovered. Second, the treatment of the excluded assets subject to separate risk weights deriving from credit risk or securitization rules produces a measure that is relatively insensitive to risk, because it neglects the effect of diversification of idiosyncratic risk and any hedging of issuer risk. Third, hedges of trading book transactions must be expressly included in calculating VaR consistent with the assets hedged. Fourth, there is no quantification of the concept of short term for purposes of excluding securitizations. 7 These include positions in financial instruments and commodities that are held to hedge other trading book assets or for trading purposes, where positions held for trading purposes means those intended for sale in the short term and/or those taken for purposes of short-term gain on price differences, or to make a profit from arbitrage between markets; they may include own positions or positions deriving from services to customers or from market making activity. 8 Securitizations of credits that are not securitized at short term, securitizations of credit assets that are unrated or rated below BBB- or deducted from capital, under the rules on treatment of securitizations, because of insufficient information to value their risk. 16/20

17 As a consequence, the Italian banking industry proposes to allow banks to count all trading book positions in calculating VaR, including those currently excluded. However, it should be provided that, if the bank cannot demonstrate that its VaR model is sufficiently robust for the purpose, the additional risk-weighted assets generated by the positions excluded must be at least equal to those calculated by the credit risk rules, where appropriate. Finally, the rule apparently gives no indication concerning the possibility of including in the trading book positions in hedge funds, or whether these are subject to credit risk capital charges. Modelling event risk and default risk The current rules allow a determination of the capital charge against market risks, divided into several components, using either standard methodology or an internal VaR model. The components that determine the overall requirement can be grouped into two main categories: generic risk and specific risk on equity securities and debt securities. The specific risk component should comprise both idiosyncratic and event risk. Considering that only the idiosyncratic risk component can be included in VaR, banks adopting an internal model are allowed to use a factor designed to substitute for the event risk component. To obtain validation of its internal model, the bank must demonstrate that it meets not only quantitative but also qualitative requirements, an essential one of which is the effective use of the VaR in operational management. The operational use of VaR is possible only when the risk figure reflects the same market conditions and assumptions used by the front office. In line with the above, most of the existing models of market risk management include the generic risk component, and for specific risk only the idiosyncratic component (credit spread risk). Normally in these models VaR is determined on the basis of a holding period of one day and a confidence interval of 99%. To calculate the capital charge, VaR so determined is appropriately adjusted and rescaled for a holding period of ten days. The new rules envisage a new methodology for calculating event risk, borrowed from the IRB approach, and the determination of the new capital charge against this risk on different assumptions from those now used for VaR. Event risk should be calculated for a holding period of one year and a 99.9% confidence interval. The supervisory authorities will accordingly have to introduce a new formula to calculate the overall capital requirement, and, if the banks opt to apply for validation of just one of the risk components, the authorities must ensure consistency of the individual components with the overall amount. To guarantee consistency with what the main financial institutions have already developed, there should also be provision for the possibility of decomposing the current 17/20

18 charge against the specific risk on debt securities into two distinct components: one for idiosyncratic risk, which could possibly be included within VaR; and one for event risk. In the light of the above, we might well consider the advisability of a block approach to different risks, separating market risks, idiosyncratic risk and event risk and allowing the use of standardized or advanced models for each. We should also like to point out that for a trading book it is unrealistic to consider positions to be unvarying for a period of up to one year. We accordingly suggest considering the time horizon for calculating the new event risk charge not as the time necessary to liquidate the asset but simply as the period of time on which to measure potential loss (assuming, that is, constant risk but changing positions). Technically, VaR should thus be a VaR calculated on a one-day or ten-day holding period rescaled to one year in order to calculate the charge. As to the proposed method for rescaling VaR to one year, using the square root of time, this is not appropriate to a trading book that will typically include assets whose yields are not normally distributed. One might accordingly insert a conversion factor table that takes account of fat tails of idiosyncratic market factors, thus incorporating default risk. The conversion factors would obviously have to be such as to create a measure large enough to cover default risk with the requisite confidence level. Treatment of less liquid instruments With special reference to the more prudent valuation of the liquidity of trading book assets, it should be stressed that the inclusion of this rule would introduce an additional factor of volatility in core capital. The factors to take into account in valuing the instruments subject to the new rule, in fact, depend on market conditions at the time the analysis is effected, but that are by nature unstable (e.g., estimate time to cover position, volatility of the bid/ask spread, etc.). On this question, we think that the rules need to offer detailed indications on the frequency with which the adequacy of coverage of less liquid instruments must be evaluated and on the possibility of setting internal standards to qualify less liquid assets. Other considerations Under the standardized method, new capital ratios for the specific risk of government and other issuers have been introduced. Here, there does not appear to be any justification for a capital ratio of 12% for issuers rated below B- and just 8% for issuers with no rating. As to internal models, they must be validated by independent third parties not involved in their development. Here, it would be advisable for the rule to offer 18/20

19 some further indications enabling the banks, in strictly organizational terms, to properly identify the structure that can validate the model. 19/20

20 A proposed capital treatment for unsettled and failed trades First, let us note that, while the proposed rules for treatment of unsettled trades would entail substantial revision of the current supervisory approach with higher costs in terms of capital charges and development costs for the implementation of information technology systems they are basically appropriate. Nevertheless, the full deduction of the risk exposure from capital for late-paying trades after five days for DvP transactions and two days for non-dvp is excessively severe. Full capital charge assumes that the failed trade signifies, with certainty, the default of the counterparty and hence the non-recovery of the entire exposure. At least in the experience of the Italian settlement system, all the failed transactions are in any case eventually settled. 9 More specifically, it is not clear whether, in order to make the distinction between normal and longer settlement lag, one must take account exclusively of contract terms and, therefore, whether longer settlement lag comprises all forward transactions (e.g., FRX Swap and Outright) with value dates longer than T+5. 9 In the Express II system, in fact, in the past six months uncollateralized failed transactions have all been settled in full from a minimum of six days (March 2004) to a maximum of 21 days (November 2004). The statistics are available at 20/20