Response to consultation on draft RTS on extraordinary circumstances for continuing the use of an internal model
Go back
However, often there are symptoms of extraordinary circumstances that are more properly captured in metrics other than volatility measures. Below we outline indicators that should also be considered.
1. Correlation indicators are very important. Indeed, stress periods are characterised by markets dislocations, which are characterized by increased correlations and heightened systemic risk. In principle, higher implied correlations can be assessed for instance from the increase of index implied volatilities over individual constituents average implied volatilities. Alternatively, implied (credit) correlations could be assessed from the relative price of cash securitisation tranches, with equity or mezzanine tranche prices decreasing relative to the price of senior tranches;
2. Liquidity indicators (e.g. jump of risk free rates (RFRs)/overnight indexed swap (OIS) indicators) to identify a major regime shift associated with a level of stress similar to a situation of significant cross border financial market stress (e.g. a liquidity crisis);
3. Unusual deviations in the markets:
i. Unusual spread deviations between safer and riskier assets. For example, a sharp widening of spreads between developed and emerging markets caused by a sudden flight to quality triggered by a country-specific or regional debt crisis.
ii. Unusual deviations between cash and derivatives markets. For example, credit spreads on bonds and CDS for same or similar reference entities may experience an unusual divergence (a symptom would be for example a sudden large deviation between the iBoxx EUR High Yield indices and the ITRAXX crossover index spreads). Another example would be large differences between spot and futures/forward prices.
4. Restrictions on trading or delivery of financial instruments or commodities. For example, restrictions on convertibility of a certain currency. In other example, the restrictions faced by the Russian ruble during the Russia-Ukraine war. Another example occurred during the COVID crisis on the ability to move gold across markets that triggered unusual differences of spot/futures prices across markets.
5. In addition, an unusual number of backtesting overshootings in a short period of time (eg, more than 2 overshootings in a monthly period) being notified by several banks as well as the volume of contracts referencing those indicators and / or their sudden moves could also naturally complete the indicators themselves;
Every crisis is unique and cannot be defined ex ante. Thus, we believe that the list of indicators should not be fixed nor exhaustive as they will naturally evolve through time.
The industry recommends:
1. At the very least, to remove (b) of paragraph 2 of Article 1 as there should not be reference to absolute volatility levels observed during the global financial crisis or the COVID-2019 pandemic. Also, we note that a sudden jump of volatility after a long period of low volatility may reference a period of financial stress or regime shift (the level of which being in relative terms high compared to the low level volatility but not necessary as high in absolute terms compared to the reference of the global financial crisis or the COVID-2019 pandemic);
2. Ideally, we recommend removing the whole paragraph 2 of Article 1 to align with the Basel principle.
Besides, we understand that extraordinary circumstances are meant to address situations of systemic stress, i.e. expected to have an impact on several types of portfolios and several types of banks (whether using an internal model or not), and for those banks using internal models, this should not be dependent on the model used. This overarching principle should be more clearly addressed in the final RTS. Therefore, it should be clarified that what matters is the number of banks affected if they were using an internal model on typical portfolios rather than the number of affected banks. This clarification is important to be made, given the limited number of banks expected to have a validated internal model.
2. PLAT failures could occur in periods of low volatility, mainly due to the Spearman correlation (as correlation assessment between the RTP&L and the HP&L is generally blurred in a situation of continuous low level of volatility).
2. As a side note, the US NPR provides a possibility to exempt events due to technical issues unrelated to the internal model (in the context of BT requirements at desk level). This flexibility is crucial in the European Regulation too, as resources tend to focus on other priorities (rather than remediating technical production issues) in period of financial stress, as it was the case during the COVID-19 pandemic. Hence, during such stressed periods, technical issues leading to BT overshootings shall not be treated as related to some flaws in the model but rather as a consequence of the extraordinary circumstances and hence discarded. Similarly, failing the PLAT as a result of technical issues during extraordinary circumstances should be disregarded as a consequence of the situation rather than a model deficiency.
Q1. Besides volatility indices like, for example, the VIX and VSTOXX, are there any other factors or indicators, in your view, that could be used to identify situations of significant cross-border financial market stress or of a major regime shift?
We believe that volatility indices such as the VIX in the US and the VSTOXX in Europe are natural indicators than can be used to identify situations of significant cross-border financial market stress or of a major regime shift. They could be naturally supported by other market implied volatility indices in Asia or linked to non-equity asset classes (e.g. market implied volatility indices for rates) or realised volatility indicators (although those indicators will identify these stress situations with some delay, by construct of the indicators that are computed as averages based on historical measure); measures are important indicators.However, often there are symptoms of extraordinary circumstances that are more properly captured in metrics other than volatility measures. Below we outline indicators that should also be considered.
1. Correlation indicators are very important. Indeed, stress periods are characterised by markets dislocations, which are characterized by increased correlations and heightened systemic risk. In principle, higher implied correlations can be assessed for instance from the increase of index implied volatilities over individual constituents average implied volatilities. Alternatively, implied (credit) correlations could be assessed from the relative price of cash securitisation tranches, with equity or mezzanine tranche prices decreasing relative to the price of senior tranches;
2. Liquidity indicators (e.g. jump of risk free rates (RFRs)/overnight indexed swap (OIS) indicators) to identify a major regime shift associated with a level of stress similar to a situation of significant cross border financial market stress (e.g. a liquidity crisis);
3. Unusual deviations in the markets:
i. Unusual spread deviations between safer and riskier assets. For example, a sharp widening of spreads between developed and emerging markets caused by a sudden flight to quality triggered by a country-specific or regional debt crisis.
ii. Unusual deviations between cash and derivatives markets. For example, credit spreads on bonds and CDS for same or similar reference entities may experience an unusual divergence (a symptom would be for example a sudden large deviation between the iBoxx EUR High Yield indices and the ITRAXX crossover index spreads). Another example would be large differences between spot and futures/forward prices.
4. Restrictions on trading or delivery of financial instruments or commodities. For example, restrictions on convertibility of a certain currency. In other example, the restrictions faced by the Russian ruble during the Russia-Ukraine war. Another example occurred during the COVID crisis on the ability to move gold across markets that triggered unusual differences of spot/futures prices across markets.
5. In addition, an unusual number of backtesting overshootings in a short period of time (eg, more than 2 overshootings in a monthly period) being notified by several banks as well as the volume of contracts referencing those indicators and / or their sudden moves could also naturally complete the indicators themselves;
Every crisis is unique and cannot be defined ex ante. Thus, we believe that the list of indicators should not be fixed nor exhaustive as they will naturally evolve through time.
Q2. Do you agree with the approach presented in the RTS? If not, please clarify which alternative approach could be used or which additional aspects should be taken into consideration.
We appreciate that Option 1b has been chosen as the preferred option and the Draft RTS will set more general criteria that should be taken into account to recognize the significant financial market stress or major regime shift. Nevertheless, the EBA RTS mentions precise indicators and factors that reflect the nature of the financial stress or regime shift whereas none of them are listed in the Basel text (MAR 32.45).The industry recommends:
1. At the very least, to remove (b) of paragraph 2 of Article 1 as there should not be reference to absolute volatility levels observed during the global financial crisis or the COVID-2019 pandemic. Also, we note that a sudden jump of volatility after a long period of low volatility may reference a period of financial stress or regime shift (the level of which being in relative terms high compared to the low level volatility but not necessary as high in absolute terms compared to the reference of the global financial crisis or the COVID-2019 pandemic);
2. Ideally, we recommend removing the whole paragraph 2 of Article 1 to align with the Basel principle.
Besides, we understand that extraordinary circumstances are meant to address situations of systemic stress, i.e. expected to have an impact on several types of portfolios and several types of banks (whether using an internal model or not), and for those banks using internal models, this should not be dependent on the model used. This overarching principle should be more clearly addressed in the final RTS. Therefore, it should be clarified that what matters is the number of banks affected if they were using an internal model on typical portfolios rather than the number of affected banks. This clarification is important to be made, given the limited number of banks expected to have a validated internal model.
Q3. What kind of regime shifts would you expect to render the outcome of the back-testing/PLAT inappropriate?
1. A value at risk (VaR) measurement is likely to capture an extreme market shift with a time lag, as VaR is calibrated using 1y of the most recent historical data and would need some days to adjust to the new market conditions (i.e., there would need to be 2-3 extreme market observations for the shift to be incorporated in the 1% confidence interval metric). This time lag may lead to BT overshootings as, in contrast to VaR, the hypothetical and actual P&Ls capture an extreme market shift immediately.2. PLAT failures could occur in periods of low volatility, mainly due to the Spearman correlation (as correlation assessment between the RTP&L and the HP&L is generally blurred in a situation of continuous low level of volatility).
Q4. How do you expect the PLAT results to be affected or to deteriorate during a period of financial stress or a regime shift, and what are the reasons for your expectation?
1. On PLAT, although this is more difficult to assess, during a financial stress period or a regime shift, the impact from some risks which are not in the model and are usually small can become much bigger and hence affect the PLAT. Therefore, more desks failing the PLAT can be expected during such periods;2. As a side note, the US NPR provides a possibility to exempt events due to technical issues unrelated to the internal model (in the context of BT requirements at desk level). This flexibility is crucial in the European Regulation too, as resources tend to focus on other priorities (rather than remediating technical production issues) in period of financial stress, as it was the case during the COVID-19 pandemic. Hence, during such stressed periods, technical issues leading to BT overshootings shall not be treated as related to some flaws in the model but rather as a consequence of the extraordinary circumstances and hence discarded. Similarly, failing the PLAT as a result of technical issues during extraordinary circumstances should be disregarded as a consequence of the situation rather than a model deficiency.