2002(a) the data inputs used in their risk measurement model accurately reflect the volatilities of and correlations between risk factors that are included in the risk measurement model; and
2004(b) any transformations applied to data inputs shall not have the effect of reducing the accuracy of the volatility of and correlations between risk factors that are included in the risk measurement model.
2012An institution shall perform at least quarterly analysis to compare prices series in point (a) with the alternative price series in points (b), (c) and (d) as follows:
2020(d) price data used in the independent price verification process in accordance with paragraph 8 of Trading Book (CRR) Article 105 including daily and intra-month data where this is collected.
2028(a) shall compare the levels, volatilities and correlations of price series from these four alternative price series for the purpose of highlighting differences between the sources that are material in terms of their impact on the measurement of the expected shortfall;
2030(b) shall, where the four alternative price series are derived from overlapping underlying data, explicitly reflect this in the analysis. The institution shall give due considerations to price uncertainty; and
2032(c) shall combine all available information, including information about intra-day movements, to derive a statistical test or tests that monitor price series referred to this paragraph to assess whether the price data used in the risk measurement model results in an understatement of the measurement of the expected shortfall,
2034provided that, for the purposes of any analysis involving the price series in point (c) of paragraph 7, the institution may perform the assessment on a best efforts basis.
2040An institution shall appropriately review and escalate the methodologies and results of the analysis in this Article. Where a potential understatement of ES is detected, an institution shall consider at least one of the following actions:
2054(a) the data inputs used for their risk measurement model are updated at least weekly; provided that by way of derogation from this requirement, an institution may update certain data inputs for their risk measurement model less frequently than weekly but not less frequently than monthly, where the institution is able to demonstrate that less frequent updates are appropriate or necessary;
2056(b) where it uses regressions to estimate model parameters for their risk measurement model, it re-estimates such parameters with sufficient frequency and at least fortnightly. By way of derogation from this requirement an institution may re-estimate certain model parameters for their risk measurement model less frequently than fortnightly if the institution is able to demonstrate that less frequent re-estimation is appropriate or necessary and this is specified in the institution’s IMA permission;
2058(c) its risk measurement models are calibrated to current market prices which are of the same observation period as the calibration of front office pricing models;
2060(d) it has a workflow process for updating the sources of data that allows it to obtain alternative data sources in a timely manner where the data sources presently used cease to be available; and
2070An institution shall ensure that the data inputs used for their risk measurement model for the purpose of calculating the partial expected shortfall calculations referred to in Article 325bc(2) are determined directly from market prices in the period of significant financial stress identified in accordance with point (c) of article Article 325bc(2); provided that, by way of derogation from this requirement, where the fundamental characteristics of a certain risk factor now differ from the characteristics of that risk factor in the identified period of significant financial stress and the institution is able to empirically justify each instance where the derogation is applied, an institution may determine stressed data inputs from market prices other than those in the identified period of significant financial stress.
2076Where a risk factor did not exist in the identified period of significant financial stress, an institution may determine data inputs from market prices other than those relating to that risk factor in the identified period of significant financial stress, subject to the following requirements:
2078(a) it shall be able to empirically justify that the data inputs used are consistent with the level of changes observed in similar risk factors in the identified historical period; and
2080(b) it shall not include the idiosyncratic component of name-specific risk factors in the subset of modellable risk factors chosen in point (a) of Article 325bc(2), unless specified otherwise in its IMA permission;
2082provided that, where an institution is unable to empirically justify that the data inputs used are consistent with the level of changes observed in similar risk factors in the identified historical period, the risk factor shall not be included in the subset of modellable risk factors chosen in point (a) Article 325bc(2) and specified in the institution’s IMA permission.
2100Where an institution uses a proxy to represent a risk factor in the risk measurement model, it must use the value of the proxy rather than the risk factor itself for calculating the theoretical changes in portfolio value for the P&L attribution requirements in accordance with Article 325bg. By way of derogation from this requirement, an institution may use the value of the actual risk factor for calculating the theoretical changes in portfolio value for the P&L attribution requirements in accordance with Article 325bg, subject to meeting the following conditions: