Skip to main content

Podcast: Fabrizio Anfuso on computing for Archegos-like event exposures

BoE quant discusses top-down counterparty risk framework using Gaussian distributions and copulae

Fabrizio Anfuso in the Infopro studio

With the default of Archegos in 2021, risk management of highly leveraged counterparty exposures shot right up banks’ priority lists.

The aggregate $10 billion mauling that followed the family office’s default proved a persuasive argument for quants to seek out suitable solutions for the hard-to-model problem.

One of the most active researchers in the space has been Fabrizio Anfuso, a senior technical specialist at the Bank of England. Together with Dimitrios Karyampas, a visiting lecturer at Bocconi University and the University of Zürich, Anfuso has developed a framework for stress-testing the exposure to tail events.

In this episode of Quantcast he discusses his most recent contribution.

 

Anfuso’s solution is a combination of his own experience in modelling wrong-way risk (WWR) and the use of previous ideas and tools proposed by other researchers to address the problem.

One of them is a Gaussian copula to model the WWR correlation between counterparty creditworthiness and a portfolio’s performance. The other is a mixture of Gaussian distributions that allows managers to closely capture the probability distribution of credit exposures.

Anfuso explains that his approach is inspired by senior Federal Reserve economist Michael Pykhtin’s model and one proposed by Matthias Arnsdorf of JP Morgan.

“My contribution is […] merging these two approaches using a new tool. Michael proposed the usage of a copula to filter the scenarios that drive the exposure conditional upon default. Matthias [provided] the correct intuition that it’s not just a matter of selecting a severe scenario,” he says. “It’s also a matter of having scenarios that are generated by a heavy-tailed distribution.”

Anfuso adds in mixture models to replicate the heavy tails – because, he says, it’s an off-the-shelf solution that is well understood and more flexible than single distributions, thus easily adaptable to the complex distribution of credit exposures.

Throughout the podcast, Anfuso implies that modelling leveraged counterparty risk isn’t yet a purely scientific process. Some steps are somewhat ‘finger-in-the-air’, guided by the practitioner’s own experience and perception of risk. The calibration of the copula coefficient is one of them. It’s an essential element in this approach, but – as in Pykhtin’s model – there isn’t a unique correct value for it.

Similarly, there isn’t a precise recipe for how many Gaussian distributions need to go in the mixture. There should be just enough to achieve good explanatory power without resulting in overfitting.

The objective of the framework is not just to come up with an estimate of potential future exposure, but also to generate stress scenarios. “This is a stress-testing model,” explains Anfuso. “Stress testing is pretty much the main tool to monitor this type of exposure.”

So far, the proposed methods build on two fundamentally distinct principles. Anfuso describes his earlier approaches as bottom-up – they would rely on granular information on the credit portfolio to assess the bank’s exposure to a given counterparty. He describes most other known approaches as top-down, by contrast – they don’t attempt to model the idiosyncratic circumstances of one counterparty, but aim instead to depict a bigger picture of the credit exposure.

A top-down approach tends to be less sensitive – though not immune – to the lack of data coming from clients, who are notoriously reluctant to disclose their positions. Anfuso discusses the possible solutions to this problem and what role third parties might play in improving transparency.

With this latest framework, Anfuso veers towards a top-down approach, which also better aligns with his intention to extend his work to a wider credit portfolio model – and which is ultimately what a bank needs to manage its book.

INDEX

00:00 Introduction

05:52 An innovative tool added to previous approaches

10:03 Mixture models and fat tails

17:31 Concentration, leverage and WWR

19:34 Model output

24:31 Calibration of parameters for broker-dealers and hedge funds

27:45 Poor disclosure as an obstacle

31:46 Approach compared to banks and feedback from the industry

34:12 Future extensions

Disclaimer: Any views expressed are solely those of the author and so cannot be taken to represent those of the Bank of England or to state Bank of England policy. This conversation should therefore not be reported as representing the views of the Bank of England or members of the Monetary Policy Committee, Financial Policy Committee or Prudential Regulation Committee.

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 Spotify, Amazon Music or the iTunes store to listen and subscribe.

 

Only users who have a paid subscription or are part of a corporate subscription are able to print or copy content.

To access these options, along with all other subscription benefits, please contact info@risk.net or view our subscription options here: http://subscriptions.risk.net/subscribe

You are currently unable to copy this content. Please contact info@risk.net to find out more.

Most read articles loading...

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here