Hany Farag, Senior Director, Head of Methodology and Analytics, Capital Markets Risk Management, CIBC
Brought on by concerns after the 2008 financial crisis, the final version of the FRTB is now out, and it needs to be understood and adhered to by risk managers. In this article, Hany Farag, Senior Director, Head of Methodology and Analytics, Capital Markets Risk Management at CIBC, explores the origins of the FRTB, and how the fundamental review of the trading book will impact our existing standardised and internal models approaches, and the business.
We give a rapid and simplified introduction to the Fundamental Review of the Trading Book (FRTB), which represents the new regulatory standard for market risk capital.
The FRTB is the biggest and most sweeping change in market risk in over two decades, and affects all banks with trading books. Simply put the FRTB is the strategic response to the financial crisis, with Basel 2.5 being the tactical response to the same.
Indeed, Basel 2.5 reforms of 2009 were a reflexive response to plug the holes in the 1996 standard which was in place at the time of the crisis. It acknowledged and addressed to some degree several issues. First the procyclicality of VaR required the introduction of Stress VaR (SVaR), which is calibrated on a stress period to avoid holding low capital at the peak of the economic cycles. It also accounted for a 10-day liquidity horizon.
Next, default and rating migration risks for issuers of debt or equity was not previously captured adequately, and therefore required the introduction of the Incremental Risk Charge (IRC).
Then securitisation products were considered to possess too much model risk and were consequently excluded from the internal models approach altogether (a.k.a advanced models approach at the time).
This tactical response, however, left many other issues unaddressed, which the Basel Committee on Banking Supervision (BCBS) identified but required a comprehensive overhaul, hence the fundamental review of the trading book.
The FRTB defines a more precise boundary between the Banking Book (BB) and the Trading Book (TB), to avoid capital arbitrage. Furthermore, it restricts the so-called Internal Risk Transfers (IRTs) that allow hedging of BB risk via TB activities/trades.
The committee recognised that the existing standardised approach was too risk-insensitive and was defined coherently on bank-level only. It therefore proved difficult to enforce in the case of poor performance of the modelled approach on some products only. Instead, the FRTB defines almost all its measures (and tests) on a prescribed notion of desks, precisely to allow some desks to be taken off the IMA based on performance.
The FRTB-SA is risk-sensitive and calibrated for linear products and vanilla options, using measures such as Delta (for linear risks), Vega (for vanilla optionality), and Curvature, which is a full-revaluation version of gamma risk. It also defines a standardised version of the Jump to Default risk (JTD). Finally, for exotic options, or options on exotic underlyings, it defines a Residual Risk Add-On (RRAO) based on notionals, which is simply a catchall bucket for anything not well-captured by the rest of the SA. We note that the SA addresses correlation risks via correlation scenarios, constrained diversification between the measures and the risk classes, and is calibrated on stress market conditions.
In the most updated version of the FRTB regulation, the BCBS also addressed concerns for small banks, for which the above version of the SA is deemed too onerous for their limited trading activities or infrastructures. To that end, they have now allowed such banks to use a more conservative (scaled) version of the Basel 2.5 standardised approach.
To capitalise tail risk more adequately, the FRTB uses Expected Shortfall (ES) instead of 99% VaR/SVaR. The calibration targets a neutral impact for distributions with benign tails. To account for the different degrees of liquidity for different types of risks, the FRTB introduces several categories of risk factors and assigns them Liquidity Horizons (LH), with ten days being the minimum. Their contribution to ES is then appropriately scaled by these liquidity horizons.
To capitalise the default risk of issuers of debt and equity and to reduce the effect of diversification between migration and default, which was seen to be acute in the IRC model, the Default Risk Charge (DRC) now excludes migration risk. The latter is now incorporated in ES for credit spreads, using more extreme shocks. Furthermore, the allowed classes of models for DRC have been restricted to address significant variability among banks’ models that was observed in the IRC as well. To address correlation risk, the measures are averaged out between a constrained correlation version of ES and another without such constraint.
Beyond the above technical enhancements, the committee introduced some far-reaching innovations to the standard that go beyond the modelling principles described above. These innovations have to do with the quality of the historical data consumed by the risk models, or the quality of the output of pricing models which are subsequently also consumed by the risk models. The first class pertains to the use of data proxies which are ubiquitous in risk models. Currently, a risk factor that has little reliable history is “mapped” to a proxy in some fashion. The committee introduced a class of risk factors referred to as Non-Modellable Risk Factors (NMRFs) which do not possess sufficiently reliable observations over the last 12 months. These are then capitalised (still in the IMA) but outside ES calculations, using a more conservative treatment.
The second innovation is a control on the Profit and Loss (PnL) measurement produced by the risk system and a requirement to produce the PnL distribution for ES. The committee introduced the Profit and Loss Attribution Test (PLAT). It is this test, in particular, that requires FRTB programs to spend substantial investments to align risk pricing with the official front office pricing models. Simplified risk representations of complex derivatives (e.g. Taylor approximations) are no longer viable under FRTB.
Stuart Nield, Global Head of Solution Management in Financial Risk Analytics, IHS Markit, on FRTB adoption challenges
Tighter governance and controls are certainly required for FRTB just like most regulatory initiatives. However, the impact from FRTB goes much further. Constraints on IRTs require some tightly controlled processes for compliance. More fundamentally however, the SA and IMA bring in substantial processes that banks are not used to. Indeed, banks currently on the SA with simple reliance on notionals have to calculate an extensive set of sensitivities from the front office systems to produce the new SA capital numbers. A great amount of reference data will also be required. IMA banks will have to align their finance, front office, and risk systems as well, in order to pass the PLAT for every IMA desk. Banks will also have to pass several versions of backtesting for every such desk as well. Banks have to produce clean automated infrastructures and diagnostic tools in order to manage these challenges on an ongoing basis.
The FRTB has undergone several revisions based on feedback from the industry to the regulators. In fact, from the first publication of the standard in January 2016 until the final version in January 2019, the framework underwent several calibration adjustments. While the final impact is yet to be determined over the next year or two via regular Quantitative Impact Studies (QIS), it appears that regulators remain on course to a close-to-neutral impact on IMA, and around 30%-50% impact on SA. While there can be a spectrum of views on this topic, my view is that the regulatory estimates are fairly reasonable at this point. Perhaps the one product category that got the biggest impact is securitisation, and particularly Correlation Trading Portfolios (CTP) where the impact appears closer to 100%.
There remains, however, one area that causes challenges for banks pursuing the IMA; namely the Capital Output Floors. It has been agreed among the regulators to control the excessive Risk Weighted Asset (RWA) variability among banks, and to set a 72.5% floor level. This is the percentage of total SA capital for operational, credit, counterparty credit, CVA, and market risks combined, even if a bank is on the IMA for all but operational risk and CVA. The latter are no longer allowed a modelled approach.
Recently however, banks observed that their credit risk IMA numbers (a.k.a AIRB or Advanced Internal Ratings Based models) are closer to approximately the 60% mark than the 72.5% and, at the same time, dominate the capital number calculations. As a result, credit risk benefit cannibalises the benefit for the other classes in some artificial sense. This can be argued as purely an issue of allocation of cost by the bank. Indeed, the cost of trading from a market risk perspective for an IMA bank should be viewed as 72.5% of the SA for the same. To be charged 100% of the SA instead, as a penalty, simply due to the size of the bank’s credit risk or the calibration of the SA for it, does not seem to make much sense. It is a distorted representation of the economics of the line of business.
Nonetheless, it would be much better if the regulators address this issue to maintain a strong incentive for this sophisticated industry to continue to model its market risk and to avoid systemic risk from trillions of dollars being managed to a standardised approach that will always have its limitations. As the industry competes on sophistication, such limitations will only become more pronounced over time.
Remark: views represent the author’s personal views and not those of his employer.
Get the latest updates at one of our upcoming events
RiskMinds Americas
9-11 September 2019 | Fairmont Copley Plaza – Boston, Boston
RiskMinds Asia
21-23 October 2019 | Harbour Grand, North Point, Hong Kong
RiskMinds International
2-6 December 2019 | Hotel Okura, Amsterdam