The risk factor modellability assessment is only one of a raft of changes to market risk capital under the Fundamental Review of the Trading Book (FRTB), but one that has attracted significant discussion due to the impact on capital. In fact, an industry survey conducted by Oliver Wyman has suggested that banks expect the non-modellable risk factor (NMRF) stressed capital add-on (SES) charge to account for 30 to 50% of their total internal market risk capital.

Given these sizeable figures, firms need to start developing a thorough understanding of the expected regulatory capital impact today, even though the go-live date for FRTB compliance is a number of years away. To support banks in doing this we've developed the Markit Modellability Model (M³), a rigorous methodology that allows firms to assess the modellability of the risk factors in their trading books.

The M³ methodology segments continuous market data objects (e.g. IR curves) into buckets so that real price observations can be counted to assess risk factor modellability. It allows the granularity of the buckets used for modellablity testing to be different from the risk factor granularity within the risk theoretical P&L. Subject to regulatory approval, larger buckets for determining modellability will increase the number of modellable risk factors and reduce the overall capital impact.

The technique can be used by firms for Quantitative Impact Studies, to accelerate Internal Model Approach waiver applications and to address the significant ambiguity in the regulatory text. Preliminary results from M³ using data across rates and credit for real-price observations in 2015 from MarkitSERV, our market-leading confirmation/affirmation platform, show that:

1. Highly liquid yield curves are modellable at short, medium and long tenors (as expected)
2. Medium liquidity yield curves (1,000 - 10,000 transactions per year) display a range of modellability results. Non-modellable sections of these curves are commonly due to seasonal trading patterns
3. CDS curves are generally modellable around the 5Y tenor point, but not across the whole curve

There is currently no one-size-fits-all risk factor taxonomy, given the specificities of each bank's internal model. Indeed, we are seeing more focus on product and instrument taxonomies. However, the purpose of M³ is to present the industry with a proposed bucketing structure as a starting point for modellability determination. Firms may either choose to use the proposed settings as default configuration across their entire portfolio; refine M³ settings for a subset of risk factors (e.g. changing thresholds); or further customise the approach to match specific internal model or taxonomy considerations.

The PLA / NMRF paradox

The FRTB P&L attribution (PLA) requirement, which incentivises banks to align their risk models with the more granular front-office models, may at first appear to conflict with the NMRF test.

If firms align their modellability buckets exactly with their IMA risk factors to improve their PLA performance - they will simultaneously allocate price observations more sparsely, thereby increasing the number of NMRFs.

However, it is possible to successfully mitigate this conflict by decoupling the two constraints.

Our preliminary results indicate that creating a rigorous framework solves the paradox by introducing sufficient transparency between the bucketing approach used for modellability determination and the internal risk models used for PLA.

While firms may therefore be inclined to use a more granular model closer to the point-wise approach to ensure they pass the PLA test, they will also more likely fail the NMRF test as a result.

Our research shows that banks can significantly improve their modellability assessment performance by decoupling real price observation bucketing from risk theoretical P&L taxonomies.

If you would like more information about this research and M³, you can download the full paper at the following link: http://events.markit.com/l/44362/2017-05-07/jcshqb

S&P Global provides industry-leading data, software and technology platforms and managed services to tackle some of the most difficult challenges in financial markets. We help our customers better understand complicated markets, reduce risk, operate more efficiently and comply with financial regulation.

This article was published by S&P Global Market Intelligence and not by S&P Global Ratings, which is a separately managed division of S&P Global.