I tried to study the relationships amongst funding, liquidity and DVA. You can find the paper here.I came up with a new definition of DVA and a new way to consider it in the balance sheets of banks.It seems to me that this new way is more consistent and sensible than the current practices. Anyway, I am eager to receive your comments.Thank you.AC.

I will read it with interest and pleasure next week I'll post my comments after that.

Great paper. It captures the essence of credit/counterpary risk in my opinion. Just a technical question: given that I have a run of traded CDS on the counterpart (ranging from 6m-10Y), and I'm looking to value the CVA on a swap contract. Should I match the probability of default of each cash-flow to the corresponded PD? Should I use flat recovery rate (say 40% across the tenors)?Tnx.

Thank you VolMaster. As for the calculation of CVA, I would bootstrap the term structure of implied PDs from CDS prices, given the standard assumption if 40% recovery (for some CDS you may need to change the recovery rate to get sensible PDs). I would not use the 40% recovery for swaps though, since you have to consider the different claimant's priority on default for the counterparty of a swap contract and for a bond's holder.

Net receivables under an ISDA master agreement are generally thought to be pari passu with senior unsecured bonds, so the same 40% (or whatever) may be appropriate, although some of the things that were thought to be true regarding the treatment of swaps in bankruptcy are being challenged in the mother of all bankruptcy cases (Lehman).

Isn't the standard 40% recovery just a plug agreed on the market, to be able to convert par spreads into an upfront payment that is required after the CDS Big Bang? Recovery is a stochastic variable and it is agreed on the dealers poll after the default occured. The only thing traded in CSD is default intensity times LGD (1 - recovery). You will never know what recovery to use. Currently I am busy with other topics, but I will try to come back with comments if I manage to have anything interesting to say. For the moment, I think it is worth checking what are the changes in IFRS 13 related to fair value measurement published in May, which replaces some points of IAS 39. Tha main change is perception of the fair value as the exit price from the transaction.

I agree that mindlessly plugging in 40% across the board is unlikely to be optimal, although as an unconditional estimate (not knowing any specifics of the underlying name) it is not too far from the long term historical average recovery rate for senior unsecured bonds. For whatever that is worth.

Just to be clear, in my paper I was not investigating the way to compute the DVA and the CVA, or the problems relataed to the correct PD or LGD to use. My aim was at understanding what is the real nature of the DVA, which is different from the nature of the CVA, seen from the same counterparty.If one accepts the new notion of DVA I give (and hopefully convincingly prove), then the consequences are rather huge:1) banks made phoney profits in the last two years by including the DVA in their balancesheets as a reduction of the liabilities,2) current accounting standards (also in the new documents) are still allowing for this misleading practice,3) the DVA must included in the balance sheet of a bank not as a liabilities' reduction, but as a reduction of the equity.This means that also banks that are not considering the DVA and think to be conservative, are still not totally so (although it is better than deducting the DVA from the liabilities).Also Basel III, that just considers the CVA (thus believing to be conservative) is actually not taking into account the costs (ie: losses) implied by the DVA. The new notion I propose should have relavant impacts also in this case.

After a few months from the first version of my work (see the thread's inception), the BIS has published a consultative paper you can find here on how to deal with the DVA of derivative contracts. The regulator's view fully accepts my point, although they are still messing things up with bonds' issuances. I think that the necessary consequences of the reasoning in my work would eventually lead to the same treatment by regulator also for the DVA of bonds. The consultative paper is also a refusal of all the frameworks lately circulating that claim the replicability of the DVA (and hence justify its inclusion in the balance sheet as a reduction of the liabilities). It is quite clear why DVA cannot be replicated and hedged from my paper, but I will write another note where it will be even clearer, and it will also show why these frameworks are dangerous since the startegies they suggest subtly imply a slow but constant destruction of the bank's franchise.

QuoteThe regulator's view fully accepts my point, although they are still messing things up with bonds' issuances.Can you please elaborate on this point? Would be most interested in hearing your views!

Freediemac, the BIS accepts my view because it explicitly forces banks to derecognize the DVA at the inception, and not only its variations. This means, exactly as I suggest in the paper (please read it for more details, the link is at the inception of this thread) , that the derivative contract has to be valued ?as if? the bank were default risk-free, and in this way the DVA becomes a cost that has to be deducted from the CET1 (the regulatory economic capital) right from the beginning of the contract. By the same token, all the differences between the ?risk-free? value (which includes the CVA anyway) and the value including the DVA is deducted also at future dates. In this way the variations of the DVA are simply used to allocate the total DVA at the start of the contract amongst the several periods until the expiry. The total of all the variations will still be the DVA at the start, since it collapses to 0 at the maturity. This view is shown to be in my paper the only consistent with the going concern principle, the only principle useful to run a company. Under this perspective the DVA is a cost that a bank has to pay for not being a default risk-free operator, although from its own perspective the fact that it can go defaulted is simply immaterial.The concept of DVA extends to any kind of contracts, also for bonds. This means that also for bonds the difference between what the bank gets at the start (likely 100) and the corresponding risk-free value (above 100) is a cost that should immediately be deducted from the CET1. Unfortunately for bonds the consultative paper of the BIS still seems to allow to recognize the DVA at inception (so the bond does not produce any loss at the issuance) although subsequent variations are not recognized (so, as for derivatives, banks cannot make phony money when the default risk increases). This is clearly inconsistent with a unified treatment and I believe that eventually also for bonds the initial DVA will be deducted from regulatory capital. Hope this clarifies better my point in the paper. 

This is a question for Antonio and others. I am working on a model to calculate CVA and DVA in monte carlo framework that can even price IR exotics. I am trying to model several features in the model for example collateral, margining, default, stochastic recovery, funding and investment cost. But my question is that if we adopt this credit approach to derivatives portfolio pricing, we will have to say goodbye to more market oriented stochastic basis approach in which credit and liquidity are not directly modeled but a stochastic basis curve is modelled as a random curve. Or may be we should model a totally credit approach in which everything will be calibrated considering default probabilities and funding and investment costs among others and calibrated directly to credit of the counterparty and funding and investment cost of the issuer. Or there could be a hybrid model that could be somewhere in between? This is more of a philosophical question than discussion of particular techniques employed in modeling.

Amin,the problem you are working on are at the moment of great interest for many others (including me).Ses for example : http://www.damianobrigo.it/funding.pdfAs for the basis modelling or direct modelling of the single risk factors, clearly it is much more realistic the second you suggest, but I think sometimes effective factors (stochatic basis for the EONIA/3M Libor for example) can do a good job. For banking book products, I would definetly adopt the modelling of the single factors (funding, def prob, recovery, behavioural optionalities, etc.).

Thank Antonio. I read several of yours papers when I was learning about CVA etc and found them very interesting. I would like to exchange result of my current work when it has clear results.

Antonio>Apologies, I'm not familiar with the details behind accounting treatment (net equity vs liability). I assume the liabilities argument is a PV/mark-to-market of liabilities approach, whereas the net equity has to do with a known (or expected) loss at maturity, which has effectively been realised in the contract. I'd welcome some clarification here.Secondly, regarding the treatment, it seems the main issue sits around the hedging of DVA. If the contract was between 3 parties, a completely risk-free counterparty X, and two counter parties A and B, and counterparty X entered into a contingent derivatives contract with A, where X's obligations were extinguished should B default, then, X would pay A compensation for allowing into the contract the possibility of default by B. Furthermore X would be placed to hedge this default, and so, potentially make profit from this contractual provision. In this case, there would be a DVA/CVA relationship.But, for a similar contract between B and A, there is a loss they will realise on any derivatives contract valued for a risk-free counterparty, because they themselves are not risk-free. This is a cost that they have realised as a result of entering into the contract, and cannot hedge/trade (this has to do with the going concern provision). For this reason, the DVA is locked in at inception of the trade and realised (possibly undeterministically, due to spread movements etc.) over the life of the trade?One last question. If A and B enter in to a derivatives contract, with B paying DVA to A. On the following day, the risk-free value of the contract itself does not change, but B's credit is downgraded. Because B is now ostensibly more risky, A will exit the contract for less than it would have the previous day, as it avoids the additional default risk. So, under conventional views of DVA, B could lock in a profit by settling the contract at this point. I think what your method is doing is realising the day 1 DVA as a loss, and will realise increments of DVA up until T, so that no pnl benefits can occur from shifts in the spread. Is this a fair assessment?