Assessment of proxies
1. When assessing whether an institution’s internal model is implemented with integrity as required by Article 325bi(1) of Regulation (EU) No 575/2013 in relation to requirements on the use of proxies, competent authorities shall verify whether:
(a)
the institution has established, as part of the internal policies referred to in Article 325bi(1), point (e), of Regulation (EU) No 575/2013, criteria outlining:
(i)
when a risk factor is proxied;
(ii)
how a risk factor would be proxied if subject to a proxy approach.
(b)
the internal policies referred to in Article 325bi(1), point (e), of Regulation (EU) No 575/2013 cover all proxy approaches employed by the institution, including, where used:
(i)
factor models;
(ii)
beta approximations;
(iii)
mapping of risk factors to benchmarks, including names representative of the sector and region or indices;
(c)
for non-modellable risk factors, there is a clear rationale for using a proxy approach, even though the number of returns N in the time series for the risk factor resulting from Article 7 of Delegated Regulation (EU) 2024/397 would allow for using the historical method or the asymmetrical sigma method referred to, respectively, in Articles 8 and 9 of that Delegated Regulation;
(d)
the approach used to proxy the risk factor is appropriate and ensures, as required by Article 325bh(1), point (g), of Regulation (EU) No 575/2013, a conservative calibration of the scenarios of future shocks for modellable risk factors and of the extreme scenarios of future shock for non-modellable risk factors;
(e)
for risk factors for which proxy data are used only for specific periods in the time series, there are no anomalous jumps between the parts of the time series that are proxied and the parts of the time series that are not proxied.
2. For the purposes of paragraph 1, point (c), competent authorities shall, on a sample of risk factors that are proxied, verify whether:
(a)
the proxy approach used for those risk factors is the approach described in the internal policies as referred to in paragraph 1, point (a), and the proxy used is economically meaningful;
(b)
the basis risk between that risk factor as proxied and other risk factors is duly captured, including where different risk factors are proxied by mapping them to the same risk factor;
(c)
there are no cases where, as a result of the proxy, the specific risk is not duly captured.
When applying that assessment method, competent authorities shall choose a sample of risk factors reflecting a variety of proxy approaches, including, where used, factor models, beta approximations, and mapping of risk factors to benchmarks, including names representative of the sector and region or indices.
3. For the purposes of paragraph 1, point (c), competent authorities shall, on a sample of risk factors for which data in the last 12-month period have been proxied:
(a)
require the institution to provide the time series of the proxied risk factors as used in the internal risk-measurement model and the time series of the corresponding pricing factors as used in the end-of-day valuation process;
(b)
verify that the volatilities of the two time series referred to in point (a) do not substantially diverge;
(c)
verify that the two time series are highly correlated.
When applying that assessment method, competent authorities shall choose a sample of risk factors reflecting a variety of proxy approaches, including, where used, factor models, beta approximations, and mapping of risk factors to benchmarks, including names representative of a given sector and region or indices.
4. For the purposes of paragraph 1, point (c), to test the conservativeness of proxy approaches, competent authorities shall select a sample of approaches and apply, for each proxy approach, all the following steps in following order:
(a)
require the institution to provide the time series of a sample of risk factors that are not proxied and that, if proxied, would follow the proxy approach being assessed;
(b)
require the institution to provide the time series that would be used by applying the proxy approach being assessed to the risk factors’ time series referred to in point (a);
(c)
for both time series, obtain the volatilities of the risk factors in the stress period and in the last 12-month period, and verify that the volatility resulting from the proxy time series referred to in point (b) does not underestimate the volatility resulting from the time series referred to in point (a).
When applying that assessment method, competent authorities shall choose a sample of risk factors reflecting a variety of proxy approaches, including, where used, factor models, beta approximations, and mapping of risk factors to benchmarks, including names representative of a given sector and region or indices.
5. For the purposes of paragraph 1, point (c), competent authorities shall, on a sample of non-modellable risk factors for which proxy data have been used in the stress period despite the number of returns N in the time series for the risk factor resulting from Article 7 of Delegated Regulation (EU) 2024/397 would allow for using the historical method or the asymmetrical sigma method referred to, respectively, in Articles 8 and 9 of that Delegated Regulation:
(a)
require the institution to provide the original time series for the risk factors before any proxy approach has been used;
(b)
require the institution to provide the time series used for the proxied risk factors;
(c)
compare the upward and downward calibrated shocks as resulting from the application of Article 8 and 9 of Delegated Regulation (EU) 2024/397 to the time series referred to in points (a) and (b) of this paragraph, and verify that shocks resulting from the proxied time series are not systematically less conservative than the shocks obtained by using the original time series.
When applying that assessment method, competent authorities shall choose a sample of risk factors reflecting a variety of proxy approaches, including, where used, factor models, beta approximations, and mapping of risk factors to benchmarks, including names representative of the sector and region or indices.
Assessment of the data quality
1. When assessing whether an institution’s data standards meet the minimum standards for the internal risk-measurement model to be considered reasonably accurate in measuring risks as required by Article 325bi(1), point (f), of Regulation (EU) No 575/2013, competent authorities shall verify whether:
(a)
the institution documents, as part of its internal policies, any methodology used to fill in time series with missing data points, and whether such documentation contains sound analysis showing that those methodologies do not affect the risk factors’ volatilities and correlations;
(b)
the institution has established objective criteria setting out which methodology to fill in time series is used, where more than one methodology is available, and has documented those criteria in its internal policies;
(c)
the institution has established, as part of its internal policies, the process to be followed whenever the values in a time series are changed, and whether such process includes the documentation of the performed changes;
(d)
the institution does not perform filtering of data, including flooring, capping and exclusions of outliers, unless the institution is able to demonstrate that the excluded data point relates to erroneous or stale data, and the institution documents such an exclusion;
(e)
the institution performs periodic quality checks on the time series used for the computation of the expected shortfall risk measure, and documents those checks and the corresponding results are documented;
(f)
the institution analyses, as part of the checks referred to in point (e), the effect that missing or replaced data and the methodology used to obtain the time series have on the risk factors’ volatilities and correlations;
(g)
the data quality of the time series used by the institution is appropriate.
For the purposes of point (e), competent authorities shall verify whether the institution monitors, as part of the checks referred to in that point, for each time series:
(a)
the number of days for which data points were initially missing and were then filled in using a particular methodology;
(b)
the number of days for which data points were initially available and have been replaced using a particular methodology;
(c)
the number of days with no daily changes;
(d)
the maximum number of consecutive days with no daily change.
2. For the purposes of paragraph 1, point (g), competent authorities shall:
(a)
require the institution to provide an overview of their time series used in
,
,
,
,
and
, as referred to in Article 325bb of Regulation (EU) No 575/2013, and in the calculation of the stress scenario risk measures, as referred to in Article 7 of Commission Delegated Regulation (EU) 2024/397, and to include in that overview for each time series used:
(i)
the total number of days in the historical observation period used to calculate the
,
,
,
,
and
, as referred to in Article 325bb of Regulation (EU) No 575/2013, and the stress scenario risk measures, as referred to in Article 7 of Commission Delegated Regulation (EU) 2024/397;
(ii)
the number of days with missing data in the time series before the institution introduces any adjustment;
(iii)
the number of days without any daily change in the time series before the institution introduces any adjustment;
(iv)
the maximum number of consecutive days without any daily change in the time series before the institution introduces any adjustment;
(v)
the number of days for which data were initially available in the time series but that the institution excluded or changed before being used in the calculation of the
,
,
,
,
and
, as referred to in Article 325bb of Regulation (EU) No 575/2013, and the stress scenario risk measures, as referred to in in Article 7 of Commission Delegated Regulation (EU) 2024/397;
(b)
based on the overview referred to in point (a), identify those times series used for risk factors that may be affected by low data quality;
(c)
based on the overview referred to in point (a), select a sample of time series that are characterised by a high number of data points initially missing, and apply the following steps in the following order:
(i)
require the institution to provide the time series with the initial data points only, and the time series after they have been filled in;
(ii)
verify that the time series have been filled in in accordance with the methodologies envisaged in the internal policies as referred to in paragraph 1, points (a) and (b), and that such methodologies are appropriate for the case at matter;
(d)
based on the overview referred to in point (b), select a sample of time series characterised by a high number of data points that were initially available but have been substituted by other data points, and apply the following steps in the following order:
(i)
require the institution to provide the time series with the initial data points only, and the time series after data points in the time series have been substituted;
(ii)
verify that the data points have been replaced in accordance with the methodologies envisaged in the internal policies as referred to in paragraph 1, points (a) and (b), and that such methodologies are appropriate for the case at matter.
For the purposes of point (b), competent authorities may use, where appropriate, as a basis for the identification referred to in that point, the following indicators:
(a)
time series with less than 10 % of initially available data points;
(b)
time series with 20 consecutive business days without any daily change;
(c)
time series with more than 20 % of days with no changes;
(d)
time series for which more than 50 % of the initially available data have been changed.
Competent authorities shall require the institution to justify the use of those time series and, where applicable, the reason why the corresponding risk factor is included in the reduced set of risk factors as referred to in Article 325bc(2), points (a) and (b), and Article 325bc(3), points (a) and (b), of Regulation (EU) No 575/2013.
3. For risk factors for which proxy data are used, competent authorities shall perform the assessment referred to in paragraph 2 on the proxy time series as used in the calculation of the
,
,
,
,
and
, as referred to in Article 325bb of Regulation (EU) No 575/2013, and on the stress scenario risk measures, as referred to in Article 7 of Commission Delegated Regulation (EU) 2024/397.
