Community
The Prudential Regulation Authority (PRA) recently held an IRB wholesale roundtable, introducing new guidelines for firms using or considering the Internal Ratings Based (IRB) approach for wholesale exposures.
In this post, we’re delving into the 6 key takeaways:
To start, the PRA's latest guidance sets forth several crucial expectations regarding model risk differentiation. These include:
Model redevelopment requirements
Use of historical downturn periods
Approach for selecting the model’s target variable
Assessment of the model’s performance
To align with these expectations, we recommend that firms:
Evaluate portfolio characteristics: Assess your portfolio's default level and external rating coverage. This evaluation will guide you in selecting the most appropriate target variable approach. This could be either the default predictor model or the shadow rating model (using external agencies or expert rankings).
Ensure meaningful risk differentiation: Thoroughly assess your models' compliance and performance. Your rating system should demonstrate; Clear differentiation of risk across various segments, robust performance during downturn periods, and consistency across all material segments of your portfolio.
Low default portfolios present unique challenges in risk modelling. The PRA has acknowledged the challenges firms face in aligning with SS11/13 12.4, which addresses firms with low internal default levels. Plus, situations where reliable PD estimates cannot be derived from external default data sources
A key industry-recognised approach to address low default challenges:
The Pluto-Tasche method
Provides a more statistical foundation for deriving PDs
Contrasts with mapping obligors to external or expert rating scales
For firms dealing with low default portfolios, we recommend:
Consider the Pluto-Tasche approach: This method offers a robust statistical foundation for PD derivation.
Explore multiple methodologies: Evaluate the Pluto-Tasche method alongside other relevant statistical approaches.
Justify your chosen approach: Whichever method you select, ensure you can provide a comprehensive justification for your choice.
Document your process: Maintain detailed records of your methodology selection, including; comparative analysis of different approaches, reasons for selecting or rejecting each method, and evidence supporting the effectiveness of your chosen approach.
The PRA has also highlighted a critical issue regarding Long Run Average (LRA) PD calibration. There are significant differences in cycle length selection across firms. These differences can lead to model miscalibration.
Key PRA observations:
Firms need to reassess their selected cycle length for LRA PD
Several factors should be considered when selecting an appropriate cycle length
Stress testing requirements must be taken into account
To address the PRA's concerns, we recommend that firms:
Evaluate cycle length comprehensively: Ensure your selected cycle length aligns with all PRA considerations, including; capturing a mix of good and bad economic years and complete economic cycles.
Use peak-to-peak or trough-to-trough evaluations: These methods can support the identification and inclusion of complete cycles.
Regularly review and update: Given economic changes, periodically reassess your cycle length to ensure it remains appropriate.
The PRA has provided guidance on the use of Master Rating Scales for calibration. The key takeaway is that Master Rating Scales remain appropriate for calibration. However, the number of risk grades requires careful consideration.
Key factors to consider when determining the number of risk grades:
Concentration
Model performance
Model uses
To optimise your Master Rating Scale, we recommend that firms:
Assess obligor volume per grade: Ensure each risk grade contains a sufficient number of obligors to support robust calibration.
Evaluate risk homogeneity: Within each risk grade, confirm that obligors share similar risk drivers and comparable performance characteristics.
Analyse grade expansion impact: Before increasing the number of grades; assess the potential impact on risk discrimination, and implement additional grades only if there's a clear improvement in discrimination.
Balance granularity and stability: Consider the trade-off between more granular risk differentiation (with more grades) and rating stability over time (which may decrease with more grades).
The PRA has also highlighted concerns regarding the use of MoCs. MoCs are often not used in line with their intended purpose and highlight that MoCs should not be used to mitigate fundamental data and methodological deficiencies.
Key PRA expectations:
Robust governance should be in place to regularly review the use of MoCs
To align with the PRA's expectations on MoCs, we recommend that firms:
Conduct a comprehensive MoC review: Identify all MoCs currently applied to your models and assess the purpose and appropriateness of each MoC.
Address underlying issues: Fix any observed data or methodology issues and apply MoCs only to reflect uncertainty and human judgement in adjustments.
Develop an action plan for unresolved issues: Where immediate fixes are not possible, create a plan to rectify deficiencies and reduce estimation errors. Also, set a reasonable timeline for implementation and consider the materiality of estimation errors in the rating system.
Implement robust governance: Establish a process for regular review of MoCs. Ensure ongoing assessment of MoC appropriateness. Document justifications for maintaining or adjusting MoCs.
Maintain transparency: Clearly document the rationale behind each MoC. Ensure traceability between MoCs and the specific uncertainties they address.
The PRA has identified several issues with corporate IRB models, specifically regarding model scope for corporate exposures. This includes concerns about the appropriate categorisation of obligors and issues with potential overlap across PD models.
Obligors must be categorised appropriately
There should be no overlap across PD models
To address the PRA's concerns and optimise corporate IRB models, we recommend that firms:
Assess model segmentation: Evaluate the volume of obligors in each model segment. Review the range and relevance of key risk drivers for obligors within each segment. Plus, ensure segments are neither too broad nor too niche.
Optimise segment granularity: Strike a balance between having enough obligors for statistical significance and maintaining homogeneity within segments. Plus, consider creating sub-segments if risk profiles within a segment are too diverse
Prevent model overlap: Review the model build and implementation process. Then, implement checks to ensure no obligor can be assigned to more than one PD model.
This content is provided by an external author without editing by Finextra. It expresses the views and opinions of the author.
Nick Green Director at Purple Patch Broking Ltd
22 August
Oleg Stefanet Chief Risk Officer at payabl.
20 August
Venkatesh P Co-Founder and Director at Maveric Systems
19 August
Welcome to Finextra. We use cookies to help us to deliver our services. You may change your preferences at our Cookie Centre.
Please read our Privacy Policy.