The last part of the new market risk regulation are rules capitalising CVA risk.
CVA is where most of the losses suffered by the banks in 2008-2009 happened, yet Basel 2 had no requirement to capitalise it. A patch, Basel 2.5, came in pretty quickly on the heels of the crisis to cover this glaring hole, but as any hot-fix, it had problems of its own.
There were two main issues.
The first was that B2.5 recognised only credit spread hedges, such as single name CDSes (on the very specific entity), or in some cases, Index CDSes. Not only it meant inability to hedge well some names (that had no single-name CDSes, and were weakly correlated with indices), but worse yet, it ignored the other part of CVA calculation entirely.
CVA is calculated as a product of probability of default and exposure at default. The credit-spreads allow for hedging of the former, but market risk factors drive the latter. But if an institution hedged the latter (for example with swaptions to hedge swap related exposure), to Basel 2 and 2.5 it looked like a naked, risk taking position, requiting additional capital. Prudent behaviour was all of sudden expensive and penalised by regulations. This is the so called split-hedge problem.
In addition to that, both regulators and accountants had a look at CVA, and each came up with slightly different rules. Hence the world now has so called “regulatory CVA” and “accounting CVA” (as IFRS 13). I do not wish to cover the differences in detail here, but there are two major ones.
Firstly, accounting CVA requires that it is “fairly valued”, at an commonly agreed exit price. This means the institution must use risk-neutral, market prices. Regulatory CVA does not specify whether the calculation should use risk-neutral or real world measures. In practice, risk-neutral measures lead to higher CVA, but at the same time, you can’t hedge real-world measures with risk-neutral ones. On the other hand, using the market measure means one’s RWAs (and capital) are much more volatile, especially in times of stress.
The other issue is that accounting CVA takes into account DVA (banks own credit risk) – but regulatory doesn’t as it would imply that the worse the bank is, the less regulatory capital it needs (talk about paradox).
BCBS decided that the CVA warrants its own rules. After all, CVA is more complex, more computationally intensive, and much more model dependent (and sensitive) than classic market risk. BCBS 325, currently in consultation stage, is the result.
BCBS 325 would apply to all derivative transactions not cleared through Central Counter Party (regardless of whether they are in a trading or banking book), and also securities financing transactions.
As mentioned in my first post, BCBS 325 allowed for three options. Basic Approach (BA-CVA) and two approaches standard apprach broadly consistent with FRTB-TB one (described in more detail here) – Standard Approach-CVA (SA-CVA) and Internal Model Approach-CVA (IMA-CVA). The latter approach (IMA-CVA) was discarded in March 2016, and will not be discussed below.
BA-CVA is just that – a basic approach. It has a formula looking very similar to the current Standard Approach (CVA), with a few modifications. The major difference is that BA-CVA now allows also (n addition to perfect single name hedges and index hedges) imperfect single-name hedges, such as CDSes on legally connected entities, and region/sector hedges (with decreasing correlations).
BA-CVA still does not allow any exposure hedges, but chances are any institutions using BA-CVA will not hedge exposures anyways, as that requires an extra level of sophistication and infrastructure, and so is not much of a problem.
Unlike FRTB-TB, where one does not have to apply to use Standardised Approach, with CVA using FRTB approach is subject to regulatory approval. There are three basic criteria (and final version may include more) that must be met to even consider this approval:
In my opinion, a bank with well functioning CVA desk should meet all three criteria.
One of the major benefit of being FRTB-CVA accredited is the ability to include one’s exposure hedges in the CVA capital charge calculations, and do away with the split-hedge problem.
SA-CVA follows very similar philosophy to the FRTB-TB.
Calculations for SA-CVA would be very recognizable to everyone who looked at FRTB-TB, although there are few major differences.
Firstly, SA-CVA does not include default charges, for obvious reasons.
Secondly, as it is very computationally intensive, some of the measures (gamma for example
) are excluded, and replaced by conservative aggregation rules (such as disallowing perfect hedging under SA-CVA).
Of note is that any instrument that is not allowable under IMA-TB, is disallowed from FRTB-CVA. Hence no securitization or tranched CDS are valid hedge instruments under FRTB-CVA.
Lastly, FRTB-CVA also includes a new asset class – counterparty credit spreads.
Of course, as we saw with FRTB-TB, it’s entirely possible that the regulation will change again dramatically before it’s actually released. Thus I will hold any comments on its impact until such a time.
This article was first published on Vlad Ender’s Linked-in profile, it is included here in an edited version.