Risk in a Recovering World:
Highlights From “Risk Americas 2021”

In the last week of May, I attended Risk Americas 10th Annual Virtual Event organized by CeFPro.

The virtual event was an excellent opportunity to learn first-hand from risk experts about the future of risk and compliance. Throughout the event, attendees were encouraged to participate and exchange ideas with the panelists. I highlighted five ideas that caught my attention from the event below:

  • Model capabilities through recovery—Reviewing Model Reactions to COVID-19 and Capabilities Through Recovery

  • Building models—Building Models Based on Historic Data and Impact on Forecasting

  • Simplification and Automation of the Model Development and Documentation Process

  • Developing ESG strategy—Developing ESG Strategy and Aligning with Future Regulatory Expectations

  • Elevating model validation—Elevating Model Validation Towards Holistic MRM Discipline: Learnings, Challenges, and Important Aspects

Reviewing Model Reactions to COVID-19 and Capabilities Through Recovery 

During this discussion, affected models that risk teams encountered due to COVID-19 included credit, trading, regulatory capital, pre-provisional net revenue, rating, and liquidity models. Each model has its problems, associated solutions, and impact that risk teams would need to account for. When addressing these model problems, Jing Zou—Managing Director at the Royal Bank of Canada, suggested that model developers and business owners analyze the model to determine remediation actions. These actions can fall under tentative solutions such as model overlays and model adjustments or strategic solutions such as model updates and new models. Zou emphasized that tentative solutions are deployed during crises, so that model owners have more time to work on strategic solutions for the long run.

After the pandemic, institutions are looking for ways to manage models and model risk with continued uncertainty. Building a strong model risk culture that enables institutions to identify model problems quickly, establish a robust governance process, and enhance model risk efficiency will be the key to tackling this uncertainty. According to Zou, building a model risk culture means educating the first line of model risk defense, deepening their understanding of model risk policies/cycle/roles, and testing expectations. A crucial aspect to a robust model risk culture is finding an efficient way to identify model problems. To do this, creating the first line of model risk defense is crucial for quickly identifying counterintuitive results. Additionally, an ongoing model performance process helps institutions identify potential model concerns during a crisis.

A strong model culture needs a transparent and closely governed overlay process. Specifically, a model risk policy guides the entire model overlay process, including submitting, reviewing, and approving a model overlay. During crisis where institutions can experience large volumes of model overlays, I found this to be vital in ensuring quality and consistency. Finally, Zou suggests that risk managers leverage model risk culture education, automation, shared code libraries, and agility to enhance their model risk efficiency. With automation becoming a crucial aspect to facilitating communication, enabling project management, and generating reports/testing results, I found this point to be especially relevant.

Building Models on Historic Data and Impact on Forecasting

In a data-driven world, many believe that having data is good. However, during this panel discussion, Aruna Joshi—Vice President of Model Risk Management at Visa, highlights how data can be good or bad depending on the situation since it is still a garbage in, garbage out process. Applying this data when building models can be especially tricky, since some models can be more affected than others. Data from the 2008 crisis helped enrich data, but the latest crisis is not the same since it has not affected all of the segments. Emre Balta—Senior Vice President of Financial and Compliance Risk Model Validations at US Bank, explained how the COVID-19 crisis was a unique event that unfolded issues such as the restrictions on the movement of people (thus restricting business) and the increasing role of government interventions. Using historical data is a must, but we have to use it carefully since the data is limited. Another important aspect to remember, noted by Adam Behrman—Head of Model Risk at Investors Bank, is how the frequency/use of data plays an important role in building models. In the current crisis, some models were less impacted than others.

Behrman stated how government intervention significantly impacted liquidity and interest rates since there was greater sensitivity towards these macroeconomic variables compared to another crisis. Government support for loan payment and other financial aid has also skewed financial model performance resulting in very few defaults. Hence, the model segment has been benign for certain sectors. In the end, models forecast based on data, but these forecasts do not align to reality due to skewed data, thus not showing an accurate picture of the economy. Balta said that Machine Learning could help with the accuracy of forecasting during extreme event scenarios. With alternative data sets such as high-frequency datasets, the forecasting of models can be increased and improved.

Simplification and Automation of the Model Development and Documentation Process

The Year 2020 witnessed an unprecedented global pandemic, followed by a financial crisis that left an indelible impact on financial institutions. This also resulted in structural changes in financial markets, which compelled risk managers to recalibrate their models to price their positions accurately. The model methodology and validation teams in global banks have spent long hours redeveloping their pricing, risk, and stress-testing models. In addition, some of the regulators requested additional scenarios in the 3rd Quarter of 2020, which has further added burden on already stretched risk teams. We have also seen certain events of crisis like Archegos in the recent past. These warranted active risk management actions & model recalibration for specific portfolios of the financial institutions.

Risk Americas 2021 has a panel discussion session on this topic moderated by Steve Lindo (Course Designer and Lecturer, Columbia University MS in Enterprise Risk Management). The panel members included Michael Scarano (Director of Compliance, Project Management & Process Improvement – Risk Capital & Model Development, BMO Financial Group), Aradhana Hugar (Senior Manager, Risk Capital and Model Development, BMO Financial Group), and Amit Inamdar (Senior Quantitative Analyst, Director, Evalueserve). Michael laid an outline of the concept and mentioned key steps that were implemented to simplify and automate model development and documentation processes. This exercise aimed to gain operational efficiencies or go faster in end-to-end model development and documentation processes. While implementing this concept, they understood the pain points/themes in the process and discussed them with the various stakeholders in the bank. Aradhana further mentioned that model owners and model validation worked in parallel to expedite the process. This has been executed in an Agile manner – Parallel Validation – Sequence / Artifacts focusing on the technical (Code calculation review, Analytical tests are completed e.g., Backtesting, Sensitivity, etc.) then you start documenting. This resulted in a 50% reduction in the development and validating of models.

Amit further mentioned that while working with financial institutions; we observed that quantitative modeling groups are working on several initiatives to simplify and automate, and key details are – a) elevate Golden Source of Data into curated data sets for Model development b) Risk team in the banks are trying to get common code libraries and arranging them through micro or platform-based automation c) Risk teams have identified and separated upstream base models and product-specific down-stream models so that common tests can be run for base models and results can be used in downstream product-specific models d) Robotic Process Automation bots have been deployed in several places to automate models risk process.

Developing ESG Strategy and Aligning with Future Regulatory Expectations

With ESG becoming an increasingly important topic in the risk space, developing an ESG strategy that aligns with future regulatory expectations is becoming a priority for many institutions. Lourenco Miranda, Head of ESG & Climate Change at Societe Generale, introduced ESG dimensions, instruments, and financial risks that come with creating an ESG strategy. There are multiple points of view when looking at ESG Dimensions. For example, corporations focus on the corporate & social responsibility aspect of ESG while regulators focus on mandatory disclosures and prudential regulations. Keeping note of these different perspectives will allow institutions to create an ESG strategy that fits their own needs and focus. Miranda introduced sustainability-linked instruments and green bonds & green loans that are products used to finance sustainable and responsible businesses. Sustainability linked instruments include loans, bonds, and derivatives that are each ESG focused. Green bonds and green loans are earmarked green projects with ESG impact.

With the introduction of ESG into institutions’ strategies, there are also financial risks that come with it. The first aspect is inherent risks associated with the ESG instruments, while the second aspect is financial risks associated with climate change, environmental risks, and social responsibility. These two aspects need to be accounted for and carefully considered when creating an ESG strategy. Institutions also need to be cognizant of other risks such as reputation, greenwashing, and more. With COVID-19 speeding up the urgency of climate change agendas and ESG, institutions need to speed up the creation and development of their ESG strategies. With ESG becoming an integral part of financial institutions, they will also have to ask themselves how they will make it happen.

Elevating Model Validation Towards MRM Discipline: Learnings, Challenges, and Important Aspects

COVID-19 has greatly impacted the advancement of regulations, brought scrutiny to the current guidance given for internal models, and has heightened regulatory expectations for institutions’ governance frameworks. Darius Grinvaldas—Head of Model Risk & Validation at Luminor Group, covered key considerations for model risk and governance, the future of model ownership, and questions institutions should focus on. Grinvaldas introduced a checklist that institutions should use when elevating their model validation that consisted of finding ways to improve model lifecycle routines, assessing model risk for unvalidated models, preparing organizations for essential concepts, managing model risks, creating model risk governance expectations, and launching state-of-the-art frameworks. Another aspect that institutions need to take note of is the question of model ownership. The ECB Regulatory guidance regarding ownership still has space for interpretation by institutions. Ownership can be interpreted three ways: owner-controller/method owner, owner-user, or a combination of both.

The combination of both options, what Grinvaldas referred to as “Option 1+2” combines the first two interpretations of model ownership into one approach. Despite the enhanced coverage, this third option needs four questions: How will the responsibilities be divided? Is this a static phenomenon that is occurring? How does this help in daily and practical situations? What are the key obstacles to effective model inventory without this optimal split? As we edge into the future, finding ways to make MRM a holistic discipline will continue to be an important goal for institutions. When COVID-19 first hit, the focus was mainly on finding ways to fix models and keep up with the disruption. Now, focus has shifted to making the model risk framework sophisticated, which will start paying dividends when modeling tools and technology expands.

Last year was full of extraordinary challenges and circumstances. Risk Americas was a chance to look ahead and see what is yet to come. To learn more about model risk management, contact MRMsolutions@evalueserve.com or speak with an expert.

Amit Inamdar
Head of AI ML Innovations Labs Posts

Latest Posts