The Triple Convergence Of Credit Valuation Adjustment (CVA)
By Alvin Lee, Director, Product Specialist, Markit
The Basel Committee on Banking Supervision (BCBS) issued a consultation document on July 1st 2015 to revise the current CVA risk framework under Basel III standards for counterparty credit risk. Under the existing standards, the CVA capital charge captures credit spread volatility but not exposure volatility, which is inconsistent with front office pricing and financial accounting.
The revision proposes two different frameworks:
- FRTB-CVA,which requires regulatory approval and consists of an internal model (IMA-CVA) and standardised (SA-CVA) approach.
- Basic CVA, calculated using the equation defined in the regulation and applicable to banks that are not using FRTB-CVA
Convergence of front office, accounting and regulatory CVA with industry best practices
The document recognises that the current regulatory CVA approach is outdated and proposes an update that converges front office and accounting CVA calculations.
Recognising that banks may use different methodologies for accounting CVA, the document outlines conditions for the accounting exposure models used.
The document’s observation that different banks may adopt varying approaches, and the explicit imposition of best practices, is an indication that simplified CVA approaches will no longer be sufficient for regulators. Although some bank auditors may still be tolerant of simplified methodologies, regulatory scrutiny is increasing and requirements are becoming more stringent.
Explicit requirements for credit spreads of illiquid counterparties
One key challenge in CVA calculation is the ability to determine the credit spread of illiquid counterparties.
The document outlines specific requirements CVA calculations for illiquid counterparties, stating that credit spread curves of illiquid counterparties should be estimated from credit spreads observed in liquid peers and should be based on “an algorithm that discriminates on at least three variables: a measure of credit quality (eg rating), industry, and region.”
The rules highlight a need for credit spread datasets that cover not only liquid counterparties, but also illiquid ones. That the BCBS is now explicitly defining these requirements in the consultation paper is an indication of closer scrutiny of CVA calculation for illiquid counterparties.
Modelling requirements for internal models and standardised approaches
Capturing both exposure and credit spread volatility in IMA-CVA will require risk systems that are faster and more sophisticated by several orders of magnitude.
The consultation paper’s new modelling requirements include:
“Expected shortfall for CVA risk must be computed on a daily basis… [using] a 97.5th percentile, one-tailed confidence interval”
“Calculation must be based on at least the following inputs: (i) term structure of market-implied probability of default (PD); (ii) market-implied expected loss given default (ELGD); (iii) simulated paths of discounted future exposure.”
Furthermore, market risk factors must now be simulated as stochastic processes for the appropriate number of paths, based on future time points, capturing effects of margining collateral including margin period of risk.
The paper also states that risk factors driving exposure should be simulated for time horizons equal to the liquidity horizons as specified in market risk FRTB requirements. In addition:
“The dynamics of market risk factors must be calibrated to a period of stress… 12-month time period… which should go back to at least 2005.”
“A bank must calculate its internally modelled CVA expected shortfall charge with no supervisory constraints on cross risk factor correlations… A bank must also calculate six partial expected shortfall values for each of the following [six] risk types”
To provide a simplified illustration of CVA calculations based on these new requirements, if we assume that CVA is calculated using 100 timesteps with 10,000 scenarios per timestep, 1 million simulations are required just to compute the value of CVA.
Calculating CVA risk would require 250 daily market risk scenarios over the 12-month stress period. CVA has to be calculated for each market risk scenario, resulting in 250 million simulations. These calculations have to be repeated across 6 risk types and 5 liquidity horizons, resulting in potentially 8.75 billion simulations.
View the complete paper here.
We’d love to hear your feedback on this article. Please click here