The final version of the Basel BCBS “Minimum capital requirements for market risk” (still generally referred to under the previous working title of the Fundamental Review of the Trading Book, or FRTB) directive was issued in January. The latest document
added additional clarifications and a few, mostly welcome, changes. The revised go-live timeframe is now January, 20201 (as opposed to the original date of January 2018, which was arguably bordering on impossible, given the scope of changes).
To date, the bulk of efforts for banks tended to be focused on capital and business impact of FRTB, understandably. Consultations, Quantitative Impact Studies (QISs) and internal projects were around the impact on firm and trading desk capital, under various
Internal Model Approval and Revised Standard Approach scenarios, and the assessment of Internal Model eligibility.
Although there is still uncertainty in the extent of the capital impact, it is more tightly bounded, as neatly expressed in the recent ISDA FRTB QIS4 Refresh summary slide, which puts capital increase factor in the range 2.4 (all desks on Standard Approach)
and 1.5 (all desks successfully on IMA).
So now that the “what” is known to a better extent, we now turn to the “how?” question. Exploring the “how?” question, firms are beginning to assess what process and system changes will be required.
Whether to reval or not to reval
Previously, the FRTB draft has left no doubt as to the BCBS’s preference for full revaluation. Interestingly, the final version of the FRTB text makes no mention of full revaluation, but, naturally, continues to insist that option non-linear behaviour and
vega risk are appropriately modelled/captured. Although most firms have continued to work towards a full revaluation solution, some are planning to build incrementally on existing sensitivity-based Taylor Series approximation solutions. In principle, Taylor
approximations can accommodate this, but there are many complications when this is extended to more extreme scenarios (e.g. worst over 10-year period).
Firms whose systems use predominantly sensitivity-based approximations typically have also invested heavily in large support and “data scrubbing” teams, sometimes running to hundreds of staff. Many are not convinced that the large sensitivity data scrubbing
factory approach is the right one for FRTB.
Paying back the technical debt
The ever-present temptation to resort to tactical applications, work-arounds and significant non-strategic modifications of existing systems remains a threat. The technical and process debt incurred over the years since the introduction of VaR-based market
risk measures for regulatory capital is significant, and should not be augmented to achieve FRTB compliance. Many firms will have a significant technical and process debt around:
- Trade and sensitivity data collection, transformation and quality assurance
- Time series data analysis, management and reporting
- Ex ante trade and marginal risk analysis
- Model validation and backtesting
- Stress scenario calibration for rare events
These are but a few of the areas where desk-level model approval, operation and validation processes will require significant new functionality and flexibility. There is no getting away from the stark fact that the FRTB is the biggest change in market risk
regulation since the introduction of Value at Risk. Firms would do well to completely re-assess their market risk and valuation infrastructure, processes and tools, in some cases following through with comprehensive systems re-platforming and overhauling of
Here’s to liquidity
One of the changes in the recent FRTB revisions relates to the reduction of the liquidity holding periods used in the Expected Shortfall under IMA as well as in the sensitivity based standardised approach (SA). Over the five bands previously articulated,
the maximum liquidity horizon is now 120 days. Among the highlights are that some Credit exposures have come down from 250 to 120 days, equity small-cap vol and equity-other have come down from 120 to 60 days.
Traffic signals for backtesting
The January update links the multiplier directly to backtesting performance at a 1-day 99% confidence interval, now floored at 1.5, as per the table. The multiplier ranges from 1.50 to 2.0, depending on performance. Moreover, exceeding these upper bounds
(12 exceptions for 99% or 30 for 97.5%), will result in the desk in question reverting to the standardised approach.
Even more challenging, and complex to compute, are the P&L attribution requirements. If a trading desk experiences four or more breaches within 12 months in respect of either of the two P&L attribution metrics, then it also reverts to SA. Concerns are widespread
about backtesting and P&L attribution, especially since performance must be evidenced for 12 months’ running prior to approval. This was also highlighted in key findings of the Joint Associations’ FRTB QIS Analysis.
While revised aggregation rules for non-modellable risk factor charges under IMA and a certain risk factor differentiation for residual risk under SA have brought some capital relief, the pressure for comprehensive quality controls of models, market data
and time series remains high.