The following table provides statistics of available liquid CDS curves per region/rating buckets as of July 2013.
Only curves that are deemed compliant with the liquidity criteria enclosed in Article 3(5) of the RTS have been retained (at least 5 Markit contributions).
From the above table, it clearly appears that granting flexibility with respect to the proposed granularity is essential as some buckets aggregation / dimension collapse will have to be performed in order to adequately calibrate proxy spreads.
As mentioned above, there is no market standard regarding the use of single-named proxies to derive proxy spreads within the scope of specific risk internal models as very few names have a market spread available.
It is however common practise to use such single-named proxies for the management of the incurred CVA. We are therefore favourable to the introduction of such single-named proxies for the sake of computing; as long as it remains an option (and not an obligation) as currently drafted in the RTS proposal.
We finally advocate that CDS on single-named proxies should be recognized as eligible CVA hedges with respect to exposures mapped to such proxies. It would consistently reflect the way incurred CVA is hedged by CVA desks.
We think that proxying the spread of a subsidiary by the spread of its parent company even where no rating is available generally constitutes a reasonable proxy. Actually, we even expect it produces in most cases a more accurate proxy than a generic proxy built on rating/industry/region attributes.
However, it can happen in some situations that such a proxy is not suitable. As a result, proxying the spread of a subsidiary by the spread of its parent company should remain an option subject to appropriate justification.
Same answer as for question 3.
Proxying the spread of a regional government or local authority by the spread of the relevant sovereign seems to us a reasonable proxy including cases where no rating is available for the regional government or local authority. However, it must remain an option.
We have no particular suggestion at this stage but we advise the EBA to leave the door open for other single named proxies to be used provided they are justified and authorised by competent authorities. Their use should be recognized in instances where they undeniably constitute more relevant proxies than the generic ones.
We reiterate our view that the threshold in number of portfolios is irrelevant and should be removed.
Only the thresholds in terms of size should be kept.
Otherwise, this will generate undesirable situation like the one where a single portfolio accounting for 10% of the total size of exposures will be eligible to advanced treatment while portfolios representing 20% in number but less than 10% in size would not.
We strongly disagree with the EBA expectation. Taking into account all exemptions from the CVA capital charge enclosed in CRR, we find that as of June 30th 2013 more than 90% of counterparties included in the CVA capital charge need to receive a proxy spread for the sake of computing the CS01 formula which represents more than 50% of the total CS01. Main contributors are commercial banks and insurance companies, fund related activities (regulated and hedge fund) and financial companies.”
Again, provided that the draft RTS only applies to the computation of si in the CVA CS01 formula, then we have no major objection with the EBA costs and benefits analysis.
On the contrary, should the RTS call into question the way credit spread shocks are determined within internal VaR models, then the impact in terms of implementation cost and risk analysis would be drastic. Concretely, a large majority of banks would face the overwhelming challenge to revisit in depth their credit VaR models within a very tight timeframe without prior assessment of the impacts both in terms of capital charge and risk management.
Finally, the cost/benefit assessment currently ignores the burden to switch between the advanced CVA charge and the standard CVA charge as a consequence of the requirement introduced in CRR Article 383(6) to fallback to standard method in case proxy spreads are deemed not compliant. Indeed, in major institutions, standardized methods are under the Finance function responsibility while advanced methods are under the Risk function responsibility. As a result, switching from a method to the other will be burdensome in terms of workflow and aggregation of results.