Short-term accuracy vs long-term vision: What should risk managers focus on?

“It's really hard to look past the two defaults that happened in Q1” – Stuart Smith, Co-Head Business Development, Acadia, shared with us in an interview. The events were eye-opening, with many risk management teams talking about consumer behaviour models and updating them in order to better reflect reality. Looking 50 years ahead though, are models good enough to help us prepare for the future? We spoke to Stuart Smith about what’s important to bear in mind when working with counterparty credit risk models and why accuracy is not always the most important feature of a model. Watch the interview or read the transcript below.

Tell us about Acadia and what you do!

Acadia is really well known for its work in the collateral space, but that's been changed dramatically over the last few years due to the introduction of the UMR rules and going to a risk-based calculation. Acadia has gone through a huge pivot of starting in the collateral space and moving more into the risk space and been a key vendor in how we've deployed risk models out to a group of firms who've probably never done that type of calculation before. It's been interesting and we're now really excited about what the potential is for the future and how we can take that same model into different areas and see what it can do for the industry to move it forward again.

Risk is a massive and evolving space. It always is with how events are going. […] This year has been so eventful. From a counterparty credit risk perspective, what do you see as the biggest news?

It's really hard to look past the two defaults that happened in Q1 – Credit Suisse and SVB. Really, it reminds everybody that credit risk isn't just a book exercise. It's not something that's done on paper but that these are real events that do happen, and as interest rates continue to rise, it's almost certain we'll see further defaults. Whether it's in banking or other industries is yet to be seen.

The nature of those defaults, I think, has changed a little bit. We've gone from a period where, if you remember Northern Rock, people would queue up outside the bank to withdraw money. And there was a period of time where maybe everybody knew it was going to default, but it took a long time by comparison. Today, instant news, instant money transfers, a default can happen in minutes and days, rather than weeks and months. And that changes a little bit how CROs need to behave and how they need to react in those kind of scenarios.

What do you see as the key things to bear in mind now as a result? What is it that you think CROs need to keep focus on?

You've got to look at those changes in the dynamics of the marketplace. You've got to understand how defaults will happen today and how they're going to be different to how they happened in 2008. And then make sure that you're well set up to cope with that. There's an old adage that you can't outsource risk management. But it's important to really think about what that means.

It doesn't mean you can't outsource risk calculation. And that in many ways, there's huge benefits to doing that, to offloading a lot of the work to firms, to vendors, but then really focusing on the business of risk management and managing the firm, using the results, analysing and understanding what impact that could have for you.

Really, it's no different to how it's always been. The details are different, the structures are different, and the mechanisms are different. It’s important to understand them and make sure you're ready to act in that even quicker time period than you probably had to 10 years ago.

You will be talking about accuracy as being one of the greatest things that Acadia might be offering to firms. But there's always this sort of question around model risk management where accuracy takes time and whether that's efficient enough. So how do you see the happy balance here?

Yeah, it's a really interesting area. People can get very focused on things that are easy to measure. It's easy to measure if a swap model is accurate in pricing a swap or derivative model. In counterparty credit risk, at times, we're looking at a 50-year horizon calculation. The most important models that go into that are the simulation models that explain how the world's going to evolve over that period. And those models are incredibly hard to measure, incredibly hard to prove they're accurate. Really, they don't need to be 100% accurate. What you're looking for is a guide that helps you find that space.

So I think we need maybe a little bit more focus on those simulation models and understanding the impact of those limitations and what they mean. And maybe, certainly for counterparty credit, there's a little bit less focus on that short term accuracy, although it's easy to measure. It's not so material in that space as the simulation models, which are key.

Looking ahead, what is it that you're looking forward to discussing and learning about at RiskMinds International?

First off, technology. The technology that we work in continues to evolve. The advances in AI and ML. are fascinating and really prevalent to what goes on in risk as well. It'd be really interesting to see how people have picked up some of those technologies, how they're using them to take the industry forward.

But then also look at how the industry's coping with the regulations that have come out. We're coming off the back of a huge regulatory wave. It'd be good to understand where people see themselves having landed at the end of that and what's next. How are they going to take those systems they've put in place and really use them to evolve into this new space that we're in: fast defaults and different mechanisms driving what we see.