Exploring modelling techniques for non-financial risk and stress testing

By Alexander Von Felbert, Head of Risk Controlling, Airbus Bank

What, for you, are the benefits of attending a conference like the ‘Stress Testing Europe Summit’?

For me, the biggest benefit of attending a conference like Stress Testing Europe is the opportunity to get in touch with other risk managers, that is, to be able to talk to subject matter experts and to benchmark your company in many respects.

You can hear about latest developments in the area of risk management and stress testing and you can also learn about new ideas and best practices. Information that can be gathered from peers can be highly valuable for projects and your day-to-day work. Apart from that, it is a great opportunity to extend your personal network.

How can risk professionals best manage uncertainty and risk?

The Non-Financial Risk family include a variety of very different risks. They, nonetheless, tend to be driven by low-frequency and high-severity events generated by so-called fat-tailed distributions. Hence, a representative of the non-financial-risk family is hard to quantify and even more difficult to predict, in other words, the inherent degree of uncertainty is usually high. Financial institutions should therefore assess their specific Non-Financial-Risk profile carefully. To this end, it is essential to identify the corresponding channels of contagions to better understand the risk and its mechanics. Both might lead to a formulation of a non-financial-risk-appetite statement. Management should also clarify to what degree they are willing to accepted uncertainty.
Due to the variety of non-financial risks there is not one single best way to manage this manifold risk family and its inherent uncertainty. There are, nonetheless, general rules regarding the management of uncertainty and risk, which also apply to non-financial risks.

A decision-maker, for instance, needs to know about the key aspects of relevant (risk) models in order to avoid wrong decisions based on falsely interpreted model results. In general, an open and transparent risk culture (e.g. transparency) is one key ingredient to best manage risks and uncertainty. Model validation, the willingness to learn and adapt, a decent model governance framework and suitable organizational structures (e.g. three lines of defense) mitigate the inherent uncertainty and fosters an effective risk management.

Can you provide our readers with an overview on modelling techniques – properties and pitfalls?

One of the main reasons why the Operational Risk’ Advance Measurement Approach (AMA) has been widely criticized might be, that the structure of operational loss data is barely in line with the modeling assumptions. Hence, I tend to agree that for certain situations an internal operational risk model might not be appropriate.

Practitioner are, nonetheless, often faced with situations, where the available data sample is too small to draw any statistical significant conclusion. At the same time, however, it is also often the case that experts do have a profound knowledge in that particular area. Apparently, ignorance is never better than (subjective) knowledge. By applying subjective probabilities within the framework of Bayesian statistics, for instance, we can leverage on that knowledge. The Bayesian approach has grown in popularity in the last couple of years in finance. Other techniques from decision and game theory (e.g. expected utility, decision under risk and uncertainty) but also robust statistics can also add value in finance and, in particular, in the decision-making process.
Whereas a full quantitative approach may never be achieved for all members of the Non-Financial Risk family, some techniques presented in my session might prove valuable for certain representatives of this family and other risk management purposes.

What are the challenges of turning a scenario analysis into economic impact?

A qualitative understanding of the relevant Non-Financial Risks is a very important first step.
What is the root-cause and the source of that particular risk and how does it propagate until it hits a financial institution? Are there any feedback loops or side effects that might increase the risk in relevant situations? How can the risk be mitigated?

If the mechanics of the risk are qualitatively well understood, it is still a hard task to translate these external non-financial risk factors into relevant internal risk factors. Identifying and gathering all available information is the basis to actually determine the degree of uncertainty regarding the specific Non-Financial Risk. The degree of uncertainty along with the model purpose (among other aspects) should finally lead to a decision regarding the model choice.

The challenges, however, even do not end if a suitable model could be identified. To explain and to convince all stakeholders, that the proposed model does add value is another important aspect that should not be underestimated. Finally, model results can only create an economic impact if valuable information are provided for decision makers.

What do you predict as the key opportunities and challenges of the next twelve months within Stress Testing?

Non-Financial risks will become even more important in the next twelve month and the overall uncertainty will stay high. Improving the quantification approaches in a sensible way such that value is added constitutes a great challenge for many institutions. At the same time, this challenge can also be an opportunity to increase the transparency, to foster the collaboration across divisions and to ask the right questions during the decision-making processes.