Podcast: Gregory and Chung on wrong-way risk modelling

Quants discuss a better way to model wrong-way risk

Nazneen Sherif and Mauro Cesa

Photo: Monika Ghose

Bill Chung, a quantitative analyst at IHS Markit in Tokyo, and Jon Gregory, a London-based consultant specialising in counterparty risk and valuation adjustments, dialled in to talk about their new paper on modelling wrong-way risk.

Wrong-way risk, or the risk of the exposure to a counterparty being adversely correlated with the credit quality of that counterparty, has always been difficult to model accurately because of a lack of sufficient data points in history to model such a scenario.

“We are looking at data where there haven’t been defaults…often we are in a situation where empirical data doesn’t really reveal wrong-way risk relationships, because that data doesn’t really contain the sort of effects which may be adverse, such as a default or an adverse credit situation, so effectively empirical data doesn’t show up causality and so on,” said Gregory.

Given there hadn’t been any regulatory push so far for banks to report wrong-way risk, there were limited efforts in modelling the quantity more accurately, as the risk only materialises when there is an actual adverse scenario.

The recent Basel capital requirements for credit valuation adjustment (CVA) risk, however, may now push banks to look at modelling it more accurately, argued Gregory.

The Fundamental Review of the Trading Book (FRTB) CVA framework, due to go live in 2022, provides more granular treatment of how CVA risk – the risk of movements in CVA – should be capitalised. It also establishes higher standards for how a bank must run its CVA desk.

“[The rules] do specify that a bank must explicitly account for the dependence between exposure and counterparty credit quality in its CVA calculations, and if that’s not done, they would be facing an increased multiplier within the capital formula. So that’s really the first time where there is a very explicit requirement to either model wrong-way risk or face higher capital charges,” argued Gregory.

In Chung and Gregory’s new paper, the quants build a model that captures information from quanto credit default swaps (CDS) markets to measure wrong-way risk more accurately.

A quanto CDS is a type of CDS where the premium and protection payments are in different currencies, so the quants argue the prices of quanto CDS contracts can be used to extract useful information about the dependence between foreign exchange jumps and defaults. Wrong-way risk can then be modelled using a jump-at-default model.

“Conventional risk management models which rely on diffusion dynamics or correlations fail to capture these kinds of jumps,” said Chung. “In practice, I have seen a lot of jumps and volatilities in particular for emerging market currencies...these FX jumps can have a huge impact on bank counterparty risk management, and the resulting changes in counterparty exposure can be quite difficult to manage.”

Both Chung and Gregory will be continuing their work on counterparty credit risk in the future.

Gregory will be looking more closely at central counterparty (CCP) risk, as this can be considered similar to wrong-way risk. The CCP's default fund, which is intended to protect members, will only get hit in a very adverse scenario.

Chung, on the other hand, will be looking at ways to develop counterparty risk management using data sciences and machine learning.

Index

0:00 Introduction

1:34 Why do we need to model CVA wrong-way risk?

3:23 Reasons wrong-way risk gets mis-specified

4:41 Quanto CDS basis for wrong-way risk

6:00 Advantages of the new technique 

7:21 Hedging of wrong-way risk

8:05 Japanese CDS markets

11:26 Impact on systemically important firms

13:31 Findings of the study

14:11 Future research

To hear the full interview, listen in the player above, or download. Future podcasts in our Quantcast series will be uploaded to Risk.net. You can also visit the main page here to access all tracks, or go to the iTunes store or Google Podcasts to listen and subscribe.