Dennis stresses the importance of real-time, pre-trade analytics that are capable of modelling all these costs through time to determine how the trade will behave, and how profitable it will be into the future. Dennis also explains the compute complexity of these calculations and the use of cloud technology to manage cost.
Sadak: Well that’s a very complicated question. So as you mentioned the debate still goes on. And the debate still goes on, not only on the technical level in terms of the infrastructure but the debate still goes on a theoretical and mathematical level. That’s from those two sides. And then you have the third side where accountants are debating with traders where is the right place to mark to market these derivatives. So on one side we have traders that are saying hey, we need to incorporate all those charges into my P&L because that’s the true cost of putting these trades onto the banks’ balance sheet. Which is a fair argument.
On the other hand, we have the accountants. Saying what we really should be marking to market these trades is the average cost of the market. In order words, where would it take us to unwind these trades and if you’re going to unwind these trades, then the actual cost of this is the average cost on the entire market. So this debate is not yet settled. However as we saw, quite recently, some of the banks are already reporting some of these adjustments on their balance sheet. So this indicates that at least on the traders versus accountant’s argument, this argument is being won mostly by the traders right now.
Jockle: Alright so 1-0 traders on that point. Let’s talk about the calculations themselves. You said it best, true P&L. We have the regulatory overlay, we’re bringing into all these adjustments, it’s no longer mark to market and greeks. Right it’s the total risk profile. Let’s talk about some of that cost. So IT’s involved, so we’ve got the desk and IT. Tell me about some of the debates and challenges that those groups are facing.
Sadak: Right so what’s happening now is the technology requirements in order to compute all these adjustments are immense. So before the calculation hurdles were about calculating greeks for the entire portfolio, which are much easier to measure, which is the measurements of today. The XVA measures are more of the future measures. So we have to calculate the exposures for the entire time. And what this entails is we have to take all the trades. We have to calculate the exposures for each one of those trades. On each one of the observation dates, we have to also model the collateral which has to behave following the practices of the underlying risk factors. So before, in order to price the trades, one trade, we had to take only a few market risk factors into account, now we have to take all of them.
So this entails into building out these large, multi-factor hybrid models. That have to capture thousands and thousands of risk factors – model them all within one simulation consistently and therefore produce these exposures. Once these exposures are produced for the trades, we have to take into account the collateral. Once the collateral is taken into account all these exposures are being aggregated that of collateral and then appropriate discount rates are used to calculate these XVA measures.
So as you can see, the calculations are very complex and they require immense computational power. So now, what is happening is that in order to maintain the computational frameworks capable of producing these numbers, customers and banks are building out these server farms. And that is a very costly process given that technology is moving forward in leaps. What you build today is going to be outdated tomorrow. So a lot of questions and a lot of debates reside in whether it’s more beneficial for customers to actually move on and try to push over this task to somebody who is more efficient in this like Google and Microsoft.
Jockle: So let’s separate the two though right? And we can get to the compute and cloud momentarily. But the question I really want to focus on is post-trade versus pre-trade. Right, so if we think about all the incorporation of P&L, what is changing in the pre-trade environment in terms of understanding execution prior to that trade that’s going to impact? Because now we have central clearing, we have to understand margining, at what point are you doing the calculation on the XVA’s? How is that changing?
Sadak: Right so you’re absolutely right. Before I put on the trade – so for example, I have a task where I have to off load some of my interest rate risk onto the market. Now there’s multiple ways I could do that. I could go to the OTC market and try to off load it in that market. I could go into the CCP’s and put on a more simple swap through a CCP. Or on another hand that could go to an exchange and put on a series of interest rate futures and off load the same risk. So how do I make that decision?
Obviously each market has its own cost associated with it. So while OTC markets now are having these CVA, DVA costs associated with it they have to be reported on the banks’ balance sheet. We also have the capital charges that are being imposed by the regulators. So that’s on the OTC side.
Then on the CCP side, what we have is the initial margin that has to be posted there, and in order to protect the central clearers in the event of the default. So there is a cost associated with that because we have to fund the initial margin from somewhere. There’s a very similar argument that arises on the exchange traded side where all the trades are marked on a daily basis and on top of that, all the initial margin has to be posted there as well.
So what one has to do is try to model all these costs through time, and try to understand that in order to make the decision, how the trade will behave, how profitable the trade will be, once they put that on. So, we’re not only looking at this trade today, we have to look at the behavior of this trade, and behavior of all these market risk factors through time, in order to understand how much collateral and in order to model all these costs. Once we model all these costs we’ll be able to make the decision of which market to utilize to put on this trade - to off load that risk.
Jockle: Well Dennis, I mean clearly the debate is getting much more complex. A lot of different stakeholders within the groups whether it’s treasury risk, the traders themselves, and clearly this is going to be something that continues to evolve. And I want to thank you for shedding some of the light of where we are today around this debate. And of course we want to talk about the topics that you want to talk about. You can follow us on twitter @nxanalytics or on our blogs and video blogs. And we definitely want your feedback to help shape our agenda. We’ll catch you next time. Thanks.
