Electronic Trading (E-Trading) activity including Trading Algorithms, E-trading Platforms and Direct Market / Strategic Access continue to be the dominant way in which Firms trade and continues to increase in proportion to voice trading. Given the pace of change within E-Trading, risk management has largely been driven by First Line and the industry standard for second line oversight and wider risk management framework is still evolving. In this blog, we focus on the challenges posed by managing the risk associated with the models in E-Trading activities as firms respond to CP6/22.
Model risk management (MRM) and its associated principles has been a consideration for long time within the risk and control framework of E-Trading businesses. However, approaches to and embedding of MRM in E-Trading, remains inconsistent across the industry, which may result in some firms falling short of PRA expectations by not being able to demonstrate adequate oversight and effective challenge across all models within E-Trading.
What are the current challenges in involving Model Risk within E-Trading?
There continues to be considerable variation in the responses of Firms to the challenges presented by MRM in E-Trading. Several Firms have responded by creating bespoke approaches when applying MRM to E-Trading activity. These are typically manual overlays to existing Algo trading processes, as the model lifecycle (i.e., Identification, Development, Validation, Implementation, Use / Deployment and Monitoring) are aligned to the E-trading lifecycle. However, where E-trading fundamentally differs is in the rapid pace of change in the E-Trading lifecycle, the shorter time to market and the extent and speed at which firms must adapt in response to rapidly changing market conditions. Additionally, model risk is not always the primary or only risk impacting E-trading activity. The PRAs CP6/22 sets out five principles for managing Model Risk and some of the key challenges with potential responses to embed model risk within E-Trading activity include:
1) Model Identification and Model Risk Classification
Challenges: Firms need to establish clear processes around model identification within the E-Trading inventory consistently without creating manual or expensive overlays, and be able to incorporate key components of the model inventory without duplicating the Algo Trading Inventory. Additionally, Firms need to define what is material model risk as Algo trading can impact multiple risks, trying to isolate or measure model risk appropriately can be challenging and can result in a disproportionate application of the relevant controls and oversight. Please refer to our blog published recently providing additional insight into Model Identification.
Industry Insights: A number of Firms have started working with Second Line risk teams to identify models within the E-trading Inventory. In some instances, Firms have gone a step further towards meeting the PRAs principles by creating a process to identify models at the development and approval stage to link the E-Trading and model inventory. Most Firms may decide to do both i.e., identify models within the existing E-Trading Inventory, as well as create a process to identify models when approved.
2) Governance and Oversight
Challenges: Firms continue to find integrating and embedding second line oversight of E-trading activities as part of the Algo lifecycle. Some common challenges in implementing appropriate second line oversight are:
- Multiple second line teams have oversight responsibility across the different risks impacting Algo trading activity, including but not limited model risk which could result in duplication and inefficient processes.
- Accountability for business risk is usually across different product classes and jurisdictions resulting in a fragmented oversight structure across first and second line functions.
- Lack of E-trading expertise especially within second line teams, leading to reliance on the first line.
Industry Insights: Some Firms have incorporated Model risk oversight as a part of existing E-trading governance forums which interact with entity or global risk management oversight structures. Depending on the level of MI available on Model risk metrics, some Firms have created manual overlays for reporting and escalation.
3) Model Development, Implementation and Use
Challenges: Traditional model risk management has evolved for models (i.e., valuation, capital models etc.) which can take months unlike Trading Algorithms to develop, test, deploy and validate. In such circumstances, the underlying documentation may not change for years. However, the E-trading product lifecycle moves at such a pace that trying to directly clone existing model approaches are likely to be costly to the business and impractical from an operational perspective as it will not match the pace required for E-trading. This presents a few key challenges, which include:
- Complex and/or fast rate of change in data, scope, structure and use required within E-Trading (e.g., significant volume, variety and variable quality of market data inputs, frequent small changes, or multiple impacts of one change).
- Defining appropriate use cases and its scope for E-trading (e.g., applying execution models developed in one asset class in other asset classes).
- Documentation for development and testing presents two major challenges. Firstly, availability of documentation in time to enable effective testing and challenge. Secondly, depth and breadth of coverage of documentation. The rapid creation and change in models combined with short lifecycle requires more automated approaches to documentation generation and validation.
- Controlled deployment and implementation i.e., in E-Trading, implementation and release is frequently carried out using deployments and not necessarily at a single point in time.
Industry Insights: This is one of the more challenging areas for Firms to implement, specifically with regards to integrating E-Trading and Model risk lifecycle (including underlying documentation) with appropriate second line review and challenge. Some firms have responded by creating bespoke or additional documentation via risk templates commonly used by second line risk teams. However, such an approach may not be scalable or cost-effective in the longer-term as E-Trading continues to dominate trading and as the use of Machine Learning (ML) models in E-Trading increases and poses further challenges around the explainability and transparency of models.
4) Independent Model Validation
Challenges: The key challenge is calibrating the level of analysis that is proportionate in E-Trading, whilst also defining the extent and timing of independent validation required. Traditional model risk validation is usually conducted fully independently where the testing, development and implementation can be validated through either, reviewing the mathematic/code and/or recreating testing scenarios. However, such an approach can be extremely challenging in an E-trading environment where the volume of data, pace of change and use of technology is often far more demanding. This is especially challenging for second line teams because cost constraints often mean that they do not have the same technology, skills, and market access as that of First line teams.
Industry Insights: Many firms are at differing stages of establishing dedicated second line teams to provide independent validation for E-Trading models. Although responses are often constrained by both cost and operational factors such as computing resources and availability of appropriate quantitative skills, firms are increasingly devising ways to deal with these challenges such as resource offshoring and use of cloud to augment computing resources.
5) Model Risk Mitigants
Challenges: Firms need to define Model Risk materiality and establish policies, procedures, and processes to use for monitoring model performance and identifying any model adjustments. The main challenge in appropriate second line oversight and ensure that these policies and procedures are embedded appropriately is to ensure that the second line teams keep pace with the evolving nature of the first lines E-trading strategies and application development.
Industry Insights: Some Firms have created bespoke approaches to assess materiality for the models identified within their E-trading inventory, which has informed them on the level of oversight required from second line teams. For example, some Firms have isolated the risk emanating from model risk and excluding other risk impacts (i.e. market abuse); whereas, some Firms have looked at the Model risk based on the outcome of the independent model validation and materiality of any issues identified.
PRA publication of CP6/22 sets out a consistent and broad definition of model that brings many components of Algo Trading activity in scope. A new Supervisory Statement is expected to be released in Q1 2023, and firms usually have 12 months thereafter to assess the readiness to comply with these requirements including:
Identify any E-trading models that should be brought under your MRM framework/inventory
Complete a self-assessment against the 5 proposed principles in CP6/22
Identify any gaps for remediation to address any issues and establish a timeline for completion.
Other Recent Deloitte publications relevant to this topic which you might be interested in: