Table 1 in the attached file contains an estimation of the number of available single-name CDS quotes for each category identified by EBA. The data source is from MarkIt, as of beginning of August 2013. To comply with the liquidity criteria, only entities with at least 2 independent contributions for 1Y, 5Y and 10Y tenors has been taken into account.
Following the UCG’s internal methodology, only buckets with at least 5 contributions (highlighted) would be considered reliable for the generation of a proxy spread quote. As a result, the buckets for sectors “Insurance” and “Other Financial Services” are poorly represented for all regions and the sector “Banks” is fairly populated only for Europe.
Based on the obtained statistics, we discourage the use of Financial sub-sectors “Banks”, “Insurance” and “Other Financial Services” in favor of a unified Financial sector, which would greatly improve the representativeness of the generated proxy spreads in all considered regions.
The existence of liquid CDS quotes for the vast majority of the traded instruments generally implies a limited recourse to the use of proxy CDS curves in the risk management activities and hence the absence of an established market practice within this regard.
To the purposes of CVA VaR calculation, we support the use of the single-name curve of a parent undertaking as proxy for the subsidiaries.
For companies not in the same group, an institution should be allowed to use credit spread indices based on liquid CDS quotes and encompassing the level of granularity by sector, rating and region as described in Article 3 of the RTS and taking into account the amendment to the considered sectors suggested in the answer to Question 1.
We deem that the use of indices: provides representative proxy curves and is a solid ground for the adoption of common practices across industry; would increase the transparency and the efficiency of the mapping activity; would remove the ambiguity inherent in the use of single-name proxies when more than one curve satisfies the sector/region/rating criteria; by averaging across spread quotes in the same bucket, would intrinsically account for rating and liquidity adjustments of the underlying curves, thus preventing instability of the proxy associations.
In order to produce reliable proxy series, careful consideration is required for the use of externally calculated indices because of potential data quality and interpolation issues. The combination of robust controls and use of expert judgment is needed to support the integration of CDS proxy indices within the on-going market risk process. For this reason, institutions should be allowed to compute CDS indices internally according to a disclosed methodology compliant with the liquidity criteria.
With respect to the definition of the rating attribute in point (b) of Article 3(1), we believe that institutions should be allowed to use internal ratings only in order to provide an homogenous evaluation of the counterparties’ creditworthiness.
The chart in Figure 1 in the attached file summarizes the distribution of rating’s dispersion within an industrial group, based on Bloomberg data as of beginning of August 2013 for companies relevant for UniCredit Groups’ positions. More than 600 industrial groups have been considered, for all sectors excluding Government. The ratings have been clustered into letter-grade classes (AAA, AA, A, BBB, BB, B, CCC). The analyses shows that in the 71% of the cases all companies in a group belong the same rating class (range=0). In the 93% of the cases, all companies in the group belong to at most 2 adjacent classes (rating=0,1).
Based on the above statistics, we deem that the suggested approach would be appropriate.
Covered Bonds are backed by pools of high-quality, low-risk assets generally composed by public sector debts, commercial mortgages, residential mortgages and other loans, whose credit risk is mostly GDP related. The highest protection granted by the access to dual recourse and other forms of credit enhancement make Covered Bonds closer in credit-worthiness and more correlated to sovereigns.
Moreover, the exclusion from bail-in would further differentiate the credit risk level of Covered Bonds from the one of typical senior unsecured financials.
Based on the considerations above we suggest to allow, for counterparties representing Covered Bonds issuers, the use of the sovereign CDS curve of the country the issuer is established in.
We deem that the criteria set out in points a. and b. of Article 5(1) of the RTS do not represent sound tests of the achieved IMM coverage of the institution. Such granular limits are seen as too prescriptive since they potentially steer the extension plan of the IMM approach onto directions that deviate from the institution’s own view of proper management of resources and risk allocation.
In particular, the criterion on the number of transaction set out in point a. of Article 5(1) is not representative of the real exposure of the institution and possibly reflects a misleading measure of the IMM coverage.
From UniCredit’s perspective, the general criterion on the total size of the non-IMM portfolios set out in point c. of Article 5(1) is a sufficient and adequate incentive for an institution to extend the IMM coverage according to its own development plans.
We seize here the opportunity to highlight some shortcomings we perceive in the definition of ‘size’ contained in Article 2(3) of the RTS. We support the idea to rely on the exposure calculated with the mark-to-market method as a common benchmark for IMM and not-IMM positions. However, the exclusion of collateral proposed in the RTS may unnecessary return a distorted picture of the real exposures of the institutions, especially considering that collateral is one of the key components in the general framework of the counterparty risk. Furthermore, the chosen measure does not weight adequately the exposures with the actual risk of incurring into losses implied by such exposures.
We think that a more sound definition for the ‘size’ of a portfolio, addressing the shortcomings mentioned above, would be represented by the CVA VaR of the portfolio, based on the portfolio’s exposure calculated with the mark-to-market method (including the effect of collateral).
Hence we strongly suggest that only the test set out in point c. should remain as criterion of a limited number of smaller portfolios, together with the alternative definition of ‘size’ proposed above.
Based on preliminary estimates as of end of June 2013, with the defined thresholds and the ‘size’ definition contained in the RTS, UniCredit would be forced to redesign its IMM extension plan.
We do not agree with EBA expectations in terms of number of counterparties. As of end of June 2013, about 450 counterparties (equivalent to 68% of the overall list) receive a proxy spread. They account for 15% of the Group’s CS01 and mainly stem from transactions with small/medium financial institutions across Europe.
We agree about the fact that the proposed set of rules would help harmonizing practices across Member States. However unexpected costs may arise in order to calculate the size of portfolios as specified in Article 2(3) of the RTS, mainly due to the exclusion of collateral.
The costs for the institutions would be even much higher if the proposed proxy methodology is meant to affect also the currently implemented market risk framework. Indeed, the definition of proxy spread set out in Article 2(4) seems to imply the adoption of new guidelines that would also affect the proxy methodology of internal market risk VaR. If this interpretation was confirmed, the RTS would result into a significant and unexpected additional amount of effort in terms of time and resources for the institutions.
We deem that the EBA mandate contained in the CRR should be interpreted as if the new guidelines act as an extension of the current VaR proxy methodologies that applies to counterparties for the calculation of CVA VaR, without altering the currently established rules for market risk VaR.
As a side remark, it is worth mentioning that the upcoming Fundamental Review of the Trading Book seems to be a more comprehensive and adequate framework to address and review the market risk VaR methodologies.