The set of Basel III rules published last December finalized the development of the regulatory capital framework’s post-crisis reforms, accompanied by a heated lobby battle by the financial industry. However, there is an element of the newly developed FRTB (Fundamental Review of Trading Book) regime, which carries on keeping the industry leaders awake during the nights: the new approach to treating the non-modellable risk factors (NMRFs). This topic is gaining prominence both due to the knotty nature of the non-modellable risk and the fact that it has been estimated that 29% of total market risk capital charges under FRTB would be attributable to NMRFs (Industry FRTB QIS Analysis - Global Financial Markets Association).Predictably, the industry met the regulations with a fine degree of opposition (ISDA and Afme. Industry Response to EBA RTS Discussion Paper on Market Risk & Counterparty Credit Risk framework). However, we argue that despite the differences in the treatment of hard-to-model risks, the existing framework could still be put to use while addressing the new FRTB requirements.
To begin with, it is useful to explore the old and the new approaches for defining the hard-to-model risks. Previously, the industry spent lots of resources on efforts to account for the hard-to-model risks in their risk management frameworks. The concept describing these risks was Risks Not in VAR (RNIV) – which under the UK framework are described as “any risks which are not adequately captured by … models”, including illiquid risk factors such as “cross-risks, basis risks, higher-order risks, and calibration parameters”. In its turn, the newly composed FRTB text similarly prescribes to classify risk factors that do not possess a history of continuously available real prices as NMRFs (i.e. insufficient data does not allow them to be included into an internal model). Moreover, the real prices for a risk factor are considered continuously available only in case there are 24 or more price changes observations per year and there is no more than a month between the consecutive spot prices.
Theoretically, RNIV could have simply fallen into the NMRF categorizations. Indeed, both definitions have a lot in common – the concepts aim to capture the risk factors which do not entail enough price data to be adequately measured through an appropriate risk model. However, the FRTB text has a more rigorous classification of hard-to-model risks, “hardcoding” them by specifying the liquidity requirements. In contrast, under the RNIV framework, the industry players had a chance to provide their reasoning for classifying a certain factor as RNIV, allowing the regulator to make a final decision on inclusion or exclusion of a factor from the VAR models.
Another difference between the two concepts is that the FRTB framework adopts a far more granular approach to risk identification. For instance, the identification of interest rate risk factors requires identifying specific points on each relevant yield curve, which are included as separate risk factors. In contrast, under the RNIV framework, a hard-to-model yield curve would be considered as a single, combined risk factor. Since FRTB does not allow correlation or diversification offsets, this difference leads to greater capital charges than in the case of RNIV. According to the industry associations’ survey, these charges are estimated to amount to 4.3 times those for a current framework (Industry FRTB QIS Analysis - Global Financial Markets Association).
The differences between RNIV and NMRF are not only limited to identification discrepancies – the FRTB framework adopts a much more conservative approach to capitalization of such risks compared to the current rules. RNIV can be capitalized in two ways: through using the same VAR metrics as those incorporated in internal models or through performing stress tests.
The first approach is used in case there is sufficient data to estimate the VAR number, and the underlying hard-to-model risk factor is either awaiting regulatory approval to be included into the internal model or is excluded from it due to the banks’ own reasoning. The second approach is used when there is insufficient data to estimate the VAR number. Such an approach allowed much room for industry players to use the loose classification of hard-to-model risks within RNIV in order to create favourable capital outcomes. In contrast, the FRTB requires that all NMRFs are capitalized using stress scenarios, which should be calibrated to be as accurate as the level applied for modelled risks – an expected shortfall loss needs to be computed at a 97,5% confidence level – over a period specifically designed for an appropriate liquidity horizon of the factor.
It is notable, however, that the differences in identification procedures, capitalisation and classification do not introduce any new methodology for hard-to-model risk quantification. Indeed, the identification discrepancies would force banks to reclassify some of the risks included in the internal model into NMRF bucket as well as increase the number of such risks (due to higher granularity of risk treatment), but in the medium term, as intended by the regulator, financial institutions are likely to decrease exposure to complex risks with high capital charges. Importantly, the stress scenarios and the expected shortfall calculations, previously used in RNIV could still be employed for calculations under NMRF.
Despite the fact that the cumulative effect of the regulations in regard to non-modellable risks is going to lead to enormous capital charges compared to the previous regime (Industry FRTB QIS Analysis - Global Financial Markets Association), the modern reality of tough regulations sets the scene for a more transparent and customer-centric financial landscape. The new approach to NMRF is designed to prevent the banks from the opportunism of hand-picking risk factors for inclusion into the internal models aiming to achieve the desired capital outcomes. Moreover, it creates incentives for the banks to refrain from bearing complex and illiquid types of risks. Finally, it is important to note that despite the described differences, the NMRF approach has been partially based on the RNIV, and therefore the organisations should aim for leveraging their experience of employing the old framework and tailor it to suit FRTB rather than spending resources on creating and setting up the new system from scratch.