By Azar Khurshid, Director, Global Risk Management, Mizuho International
Interview ahead of the Model Risk Management Europe Summit
Can you please tell the Risk Insights readers a little bit about yourself, your experiences and what your current professional focus is?
I am currently leading the development of the Global FRTB solution for Mizuho securities group. I am also the product owner of the existing internal model (VaR based) risk platform. I have more than 10 years of experience in the financial risk management, during which time I have covered both credit and market risk areas. I have also been involved in setting up appropriate governance and operating models across the risk discipline. Before my life in finance, I was working in the area of Neuro-science and developing technologies to leverage the computational models of learning, for example voice and natural speech recognition. I did my PhD in Neuroscience, developing computational models of speech perception in the brain.
The current challenges we face are establishing a target operating model that is fit for purpose of internal risk management as well as regulatory requirements. This is further complicated by the fact that we operate in multiple jurisdictions, each with their own established structures. FRTB requires a significant re-alignment in terms of how one quantifies the model quality and model risk (very specific measures for P&L attribution tests for IMA), alignment of pricing and risk models and data, as well as potential revision of desk structures and policies and procedures for banking book/trading book classification and operation of internal risk transfer desks. Further challenges come from trying to look at overlaps between different regulations and how best to seek synergies (IFRS-13 and RFET based classification, as well as managing the LIBOR replacement related changes within potentially similar time-scale).
What for you are the benefits of attending a conference like the Model Risk Management Summit and what can attendees expect to learn from your session?
I see attending a conference like the Model Risk Management summit for everyone serious about learning about the most current and pertinent challenges that face this area and various approaches being adopted in meeting these challenges. It gives participants an opportunity to get a broader perspective of the industry trends as well as sharing ideas.
Within our industry, there is a huge potential for learning from pears and adopting best practice in order to face common challenges – internal prudent management of model risk, as well as external regulatory requirements in terms of capital. I find participants are very open about challenges they face and approaches that might have worked, as well as (and just as useful) approaches which don’t work.
The agenda for the conference is very relevant to the challenges facing us as an industry. I expect the participants will find more clarity on the size of these challenges, as well as industry recognised common approaches to addressing them. Benchmarking your own solutions against industry peers is a great source of confidence, or a clear indication of where improvements need to be made. Talks on diverse topics like validation of traditional and non-traditional models, stress testing of models and developments in global regulations should provide with wide coverage of topics.
What are the implementation challenges and short comings of the most recent regulations within MRM?
While it’s not possible to list all the big challenges thrown by the most recent regulations, the one that is most likely to be most difficult to address satisfactorily is about the independence of risk models as tools to challenge and provide an independent view, as well as ensuring we can derive as much benefit in terms of managing risk effectively, given the huge investment going into this area to fulfil all our regulatory obligations.
For the new internal models usage for regulatory capitals, banks need to be able to prove that the model generated P&L is close to the actual P&L for each desk. In order to achieve this, the necessary alignment of risk and pricing models is the most viable approach. However, the inputs into the model also need to be aligned. Traditionally, the risk model and data divergence was not a problem, and provided a valid challenge and reference point to measure pricing model performance. Therefore, this alignment entails a more reliance on independent price validation process which will be even more key process in the future. The hard headed review of models to determine if they can be used for regulatory capital is a huge challenge. Additionally, implied changes in the organisational policies and procedures, and the operating model are key implementation challenges from a non-technology perspective.
From a data handling and computational requirement perspective, the IMA calculations are very demanding and are leading to a total rethink of legacy infrastructure and systems. While some institutions have already updated their systems to be scalable, other institutions which have been lagging behind, these requirements are becoming prime drivers for a full system overhaul.
How can we look to effectively harmonise regulatory change across suite of models?
There are several approaches which could be used for this purpose. Most of the common elements of each approach are:
- Establish a global, comprehensive model inventory
- Standard methodology to measure model performance – i.e. development of suitable metrics
- Creation and maintenance of a heat map of models, identifying areas of weakness. This could be multi-dimensional or specific to each regulation/regulatory area
- Formalisation of model risk within the organisation and senior management engagement.
How is TRIM impacting MRM?
The biggest impact currently is around increased investment in the review exercises, inventory creation etc. In the medium to long terms, hopefully we will see a more confident model risk quantification approach which allows us to use the results of the model more effectively, as well as to drive improvements in the models.