PLA and the RFET: A Perfect FRTB Storm 

Despite the several global delays to FRTB go-live, many banks are still struggling to be prepared for the implementation of profit and loss attribution (PLA) and the risk factor eligibility test (RFET). As both tests have the potential to considerably increase capital requirements, they are high on the agenda for most banks which are attempting to use the internal models approach (IMA).  

In this article, we explore the difficulties with both tests and also highlight some underlying similarities. By leveraging these similarities to develop a unified PLA and RFET system, we describe how PLA and RFET failures can be avoided to reduce the potential capital requirements for IMA banks.

Difficulties with PLA

Since its introduction into the FRTB framework by the Basel Committee on Banking Supervision (BCBS), the PLA test has been a consistent cause for concern for banks attempting to use the IMA. The test is designed to ensure that Front Office (FO) and Risk P&Ls are sufficiently aligned. As such, it ensures that banks’ internal models for market risk accurately reflect the risk they are exposed to. To assess this alignment, the PLA test compares the Hypothetical P&L (HPL) from the FO with the risk-theoretical P&L (RTPL) from Risk using two statistical tests - the Spearman correlation and the Kolmogorov-Smirnov (KS) test. 

There are potentially significant consequences of trading desks not passing the test. At best, the desk will incur capital add-ons. At worst, the desk will be forced to use the more punitive standardised approach (SA), which may increase capital requirements even more. 

There are several difficulties with PLA: 

• No existing systems: As the test has never before been a regulatory requirement, many banks do not have suitable existing systems and processes which can be leveraged to identify the causes of PLA failures. Although the KS and Spearman tests are easy to implement, isolating the causes of PLA failures can be difficult. 
• Risk factor mapping: Banks often do not have accurate and reliable mapping between the risk factors in the FO and Risk models. Remediation of the inaccurate mapping can often be a slow and manual process, making it extremely difficult to identify the risk factors which are causing the PLA failure. 
• Data inconsistency: As the data feeds between Risk and FO models can be different, there can be a large number of potential causes of P&L differences. Even small differences in data granularity, convexity capture or even holiday calendars can cause misalignments which may result in PLA failures. 
• Hedged portfolios: Well-hedged portfolios often find it more challenging to pass the PLA test. When portfolios are hedged, the total P&L of the portfolio is reduced, leading to a larger relative error than that of an unhedged portfolio, potentially causing PLA failures. You can read more about this topic on our other blog post – ‘To Hedge or Not to Hedge: Navigating the Catch-22 of FRTB’s PLA Test’

Issues with the RFET

The RFET ensures that all risk factors in the internal model have a minimum level of liquidity and enough market data to be accurately used. Liquidity is measured by the number of real price observations which have been observed in the past 12 months. Any risk factors that do not meet the minimum liquidity standards outlined in FRTB are known as non-modellable risk factors (NMRFs). Similar to the consequences of failing the PLA test and having to use the SA, NMRFs must use the more conservative stressed expected shortfall (SES) capital calculations, leading to higher capital requirements. Research shows that NMRFs can account for over 30% of capital requirements, making them one of the most punitive drivers of increased capital within the IMA. The impact of NMRFs is often considered to be disproportionately large and also unpredictable.

There are several difficulties with the RFET: 

• Wide scope: The RFET requires all risk factors to be collected across multiple desks and systems. Mapping instruments to risk factors can be a complicated and lengthy process. Consequently, implementing and operationalizing the RFET can be difficult. 
• Diversification benefit: Modellable risk factors are capitalised using the expected shortfall (ES) which allows for diversification benefits. However, NMRFs are capitalised using the stressed expected shortfall (SES) which does not provide the same benefits, resulting in larger capital. 
• Proxy development: Although proxies can be used to overcome a lack of data, developing them can be time-consuming and require considerable effort. Determining proxies requires exploratory work which often has uncertain outcomes.​ Furthermore, all proxies need to be validated and justified to the regulator. 
• Vendor data: It can be difficult for banks to quantify the cost benefit of purchasing external data to increase the number of real price observations versus the cost of more NMRFs. Ultimately, the result of the RFET is based on a bank’s access to real price observation data. Although two banks may have identical exposures and risk, they may have completely different capital requirements due to their access to the correct data.  

The interconnectedness of both tests 

Despite their individual difficulties, there are a number of similarities between PLA and the RFET which can be leveraged to ensure efficient implementation of the IMA: 

• Although PLA is performed at the desk-level, the underlying risk factors are the same as those which are used for the RFET.​ 
• Both tests potentially impact the ES model as the PLA/RFET outcomes may instigate modifications to the model in order to improve the results. For example, any changes in data source to increase the liquidity of NMRFs (which is a common way to overcome RFET issues) would require PLA to be rerun.​ 
• Ultimately, if any changes are made to the underlying risk factors, both tests must be performed again.​ 
• Hence, although they are relatively simple tests (Spearman Correlation and KS, and a count of real price observations for the RFET), banks must develop a reliable architecture to dynamically change risk factors and efficiently rerun PLA and RFET tests. 

Zanders’ recommendation 

As they greatly impact one another, a unified system allows both components to be run together. Due to their interdependencies, a unified PLA-RFET system makes it easier for banks to dynamically modify risk factors and improve results for both tests.​  

• In order to truly have a unified PLA-RFET system, the PLA results must also be brought down to the risk factor level. This is done by understanding and quantifying which risk factors are causing the discrepancies between RTPL and HPL and causing poor PLA statistics. More information about this can be found in our other blog post ‘FRTB: Profit and Loss Attribution (PLA) Analytics’. 
• Once the risk factors causing PLA failures have been identified, a unified approach can prioritise risk factors which, if remediated, improve PLA statistics and also efficiently reduce NMRF SES capitalisation. 

Conclusion 

While PLA is crucial for IMA approval, it presents numerous operational and technical challenges. Similarly, the RFET introduces additional complexities by enforcing strict liquidity and data standards for risk factors, with failing risk factors subject to harsher capital treatments. The interconnected nature of both tests highlights the need for a cohesive strategy, where adjustments to one test can directly influence outcomes in the other. Ultimately, banks need to invest in robust systems that allow for dynamic adjustments to risk factors and efficient reruns of both tests. A unified PLA-RFET approach can streamline processes, reduce capital penalties, and improve test results by focusing on the underlying risk factors common to both assessments.  

For more information about this topic and how Zanders can help you design and implement a unified PLA and RFET system, please contact Dilbagh Kalsi (Partner) or Hardial Kalsi (Manager).