Why is it so difficult to incorporate KVA into derivative pricing?

The KVA requires knowledge about economic capital and regulatory capital frameworks. Many of the quants, which traditionally have been involved in derivatives pricing, are lacking such skill set. For me it was easier to embrace the KVA since I had worked with capital issues for years before I started to price derivatives.

The KVA is also problematic due to overlaps and inter-dependencies with other valuation adjustments. For instance, there is an overlap with the CVA which you have to take into account in order to avoid double counting. The capital requirement for counterparty credit risk is intended as a buffer for unexpected credit risk (i.e. the risk that the losses become higher than expected). The CVA complements the KVA nicely since it instead focuses on expected credit losses. However, a CVA based on quoted CDSs may also contain the market’s risk-premium and this could be substantial. This risk-premium represents the market’s worry that the credit losses become higher than expected. Thus, the risk-premium in CVA overlaps with the KVA. Furthermore, the CVA should be deducted from the exposure at default (EAD) which is an important component in the KVA calculation. This means that there is a clear dependency between the CVA and KVA.

The KVA also requires forecast of future capital requirements. This is not trivial since you need to take a bet on how future capital legislation  will evolve. You also need to forecast the credit quality development of your counterparty (rating migration) as well as the development of other important input factors.

There are also plenty of technical challenges. Basel III introduced a stressed EAD. This means that you need to run a so called nested Monte Carlo simulation (a Monte Carlo simulation inside another Monte Carlo) in order to forecast the stressed expected exposure curve at a future point in time. This is either time-consuming or requires you to rely on a proxy instead.

Can KVA be treated in the same way as the other XVAs or does KVA need to be treated differently?

I think that the KVA shares many characteristics with other valuation adjustments. For instance, many XVAs are basically calculated by multiplying a rate with the present value of some expected quantity. When it comes to KVA it is the cost of capital rate that is multiplied with the expected amount of capital that is required in the future. This is similar to the FVA which is calculated as a funding spread, times the expected amount that has to be funded. Furthermore, the CVA can simply be explained as a credit spread times an expected exposure amount.

Some XVAs (such as the CVA) try to assess the cost of hedging a risk that has not been addressed in the basic valuation model. I think that the KVA is different in this respect. A pre-condition for a capital requirement is that there is a risk. Thus, the KVA assumes that the risks inherent in the trade will not be hedged, at least not to 100%. The no arbitrage assumption, which is the foundation of many valuation models, is not always realistic. Some positions cannot be hedged, or are actually intended to be warehoused.

It is not yet clear if and how you should reflect the KVA in the official accounting. I am aware of many banks (especially here in the Scandinavian countries) that account for the KVA on new trades. This means that they are not booking the KVA as an up-front profit at the trade date but they rather distribute the KVA over the trade’s life-time.

How is KVA moving towards point-of-trade calculation?

My perception is that most tier 1 banks now have a KVA procedure in place, at least for pricing of large trades. I know that also some tier 2 banks are able to compute the KVA for individual trades. However, few (if any) front-office systems have this feature built-in. Thus, most traders or sales managers have to rely on the XVA desk or separate KVA/XVA calculators or pricing grids. We rolled-out our first real KVA calculator to all derivatives desks in February 2012 but it has been improved significantly since then, especially when it comes to assuring that the tool is fast and easy to use. 

We have introduced pretty good tools in Nordea so that people can calculate and price most XVAs themselves. However, we also have a centralised support of pricing and structuring of trades. Such a service is often  needed when pricing complex or large transaction.

How can you manage accounts to increase the efficiency of KVA calculations?

We have introduced several initiatives which are intended to increase the capital velocity. Transfer pricing gives an incentive to work with the existing portfolio in order to reduce counterparty credit risk, capital consumption or to release some other XVAs.

We monitor pricing and XVAs in several ways. For instance, a centralised post-trade calculation of the XVAs (including the KVA) complements the pre-trade calculations used by individuals. The monitoring of won and lost trades also gives good insights of how our pricing model stands towards our competitors’ pricing. 

The most important way, I think, to achieve capital efficient trades, is to have tools that enable people to play around with different features of a potential trade and see how they affect the KVA. This would support an efficient trade selection and enable us to structure trades that fit both the clients’ demand for hedging and our possibility to deliver such solutions.

What would you like to achieve by attending the 9th Annual Practical KVA and CVA Forum?

I think this is an excellent way of staying up-to-date with the developments in the industry. The XVA framework has improved considerably over the past years but I think that there are still many unresolved issues. I hope to be served with many new ideas and thoughts.