Counterparty Credit Risk Management CVA Demo - Marginal Exposure and Delta CVA

Counterparty Credit Risk Management CVA Demo – Marginal Exposure and Delta CVA

In the previous post, we described the netting concept in CVA. This post describes the marginal exposure and delta CVA concepts in our CVA demo, and show related benchmarks.

Counterparty Credit Risk Management: Marginal Exposure and CVA

The marginal exposure of a trade is defined as the exposure of the trade’s netting node, to which we subtract the exposure of this netting node computed as if our trade did not exist. To put it simply, this measure is the exposure of this trade’s netting node minus the exposure of this trade’s netting node without this trade.

Therefore in counterparty credit risk management, the marginal exposure of a set of trades is defined as follows: for each of the netting nodes involved, we compute their exposures with and without the trades in our set. We take their difference for each node, then take the positive sum of all these differences to compute the marginal exposure. Finally, the marginal CVA, another counterparty credit risk management measure, is simply the CVA computed with the marginal exposure instead of the regular exposure.

Counterparty Credit Risk Management: Delta CVA

To better understand the counterparty exposure, we measure the CVA sensitivity to Credit Default Spread (CDS). Delta CVA is the variation in CVA when the CDS varies by 1 basis point. As we consider the CVA formula presented previously, it appears that CVA variation due to CDS is that of the default probabilities (D variable), the other variables are not subject to change as the spread changes. We are then able to calculate the Delta CVA on the fly based upon the counterparty’s CDS or its credit rating.

The Delta CVA is dynamically aggregated at the netting nodes based upon the netting rules with the same logic as the CVA is aggregated.

Benchmarks

Hardware: Dell server running 2 Intel Xeon X55 CPUs:- 16 cores 48 GB RAM 20 time buckets x 1,000 simulations by time buckets

Scenario 1

300,000 trades (300 CP, 1,000 trades per CP)

  • memory footprint: After GC 28GB, 32GB Regular usage
  • Loading time: 191s – total CVA: 400ms
  • CVA for 1 CP group (with our without members drill-down): instantaneous (< 100ms)
  • CVA for all trades within a netting node: 800ms
  • Marginal CVA for one counterparty by trade: 1.6ms
  • CVA matrix country / rating: 1.6sec

Scenario 2

150,000 trades (1000 CP, 150 trades per CP)

  • Memory footprint: After GC 17GB, 23GB Regular usage
  • Loading time: 86s
  • Total CVA: 1.8sec
  • CVA for 1 CP group (with our without members drill-down): instantaneous (< 100ms)
  • CVA for all trades within a netting node: 370ms
  • Marginal CVA for one counterparty by trade: 450ms
  • CVA matrix country / rating: 6.5sec

Scenario 3

300,000 trades (1000 CP, 300 trades per CP)

  • Memory footprint: After GC 32GB, 35GB regular usage
  • Loading time: 210s – total CVA: 1.9sec
  • CVA for 1 CP group (with our without members drill-down): instantaneous (< 100ms)
  • CVA for all trades within a netting node: 800ms
  • Marginal CVA for one counterparty by trade: 1,1s – CVA matrix country / rating: 7-12s

In the next post we describe the Quartet value at risk software demo and dive into its cube structure.

Related resources

eBook – Product Control: Explain and Reconcile PnL in a Single System
ActiveViam’s Market Risk Accelerator
ActiveMonitor
FRTB Accelerator
The Fundamental Review of the Trading Book (FRTB) – Tackling A New Approach For Market Risk
ActivePivot Under the hood
Managing business performance and detecting outliers in financial services
ActiveLedgers – Facilitating and shortening the accounting period close
Collateral Optimization
Optimizing Liquidity Management
Real-time Credit Value Adjustment
eBook – Intraday Liquidity Management
eBook – Liquidity Management

GET IN TOUCH

Questions ? Want to know more about our solutions?




Follow us: linkedin twitter