Article 47
Assessment of accuracy and frequency of the calculation of the own fund requirement for default risk
1. When assessing whether an institution’s own funds requirements for default risk equal a value-at-risk number at a 99,9 % confidence interval level as required by Article 325bn(1), points (a) and (b), of Regulation (EU) No 575/2013, competent authorities shall:
|
(a) |
verify whether the estimator used by the institution to estimate the value-at-risk is accurate; |
|
(b) |
where the value-at-risk calculation is based on Monte Carlo simulations, verify whether the number of simulations ensures convergence towards stable results, and the randomness properties of the sequences used to generate the simulations; |
|
(c) |
verify whether, before calculating the changes in the portfolio’s value following issuers’ defaults, the value of the positions in the institution’s portfolios refers to the value-at-risk’s reference date; |
|
(d) |
verify whether, with the exception of the positions subject to the derogation referred to in Article 325bn(3) of Regulation (EU) No 575/2013, a one-year time horizon is used in the computation of the value-at-risk; |
|
(e) |
where the default risk is computed less frequently than daily, analyse the process used by the institution to determine the frequency of the calculation of the own funds requirements for default risk, and verify whether the calculation at a reduced frequency does not lead to underestimation of risk; |
|
(f) |
verify whether for equity instruments, prices are set to zero when simulating the defaults of those equity instruments, and whether that is systematically ensured by the internal systems, and may verify whether that is the case on a sample of equity positions. |
For the purposes of point (a), competent authorities shall verify how the institution chose the estimator and the analysis made to back such choice.
For the purposes of point (b), competent authorities shall review the tests performed by the institution to set the number of simulations.
For the purposes of point (d), competent authorities shall verify whether the rationale of the institution for applying that derogation is sound, in particular when the institution uses a time horizon of 60 days for some equity positions, and a one-year time horizon for some other equity positions.
For the purposes of point (e), competent authorities:
|
(a) |
shall, where the default risk is computed weekly, analyse the process used by the institution to determine the day of the week when the own funds requirements for default risk are calculated; |
|
(b) |
shall require the institution to calculate, where not yet available, daily jump-to-default exposures over a given period, and assess whether those exposures hint at a systematically lower risk profile on those days in which the own funds requirements are calculated; |
|
(c) |
may also use additional figures that may be computed daily by the institution for internal risk-management purposes, including daily sensitivities to the most material issuers. |
For the purposes of the fourth subparagraph, point (b), where there are hints of a systematically lower risk profile, competent authorities may complement their assessment by requiring the institution to calculate, on a daily basis and for a given period, its own funds requirements for default risk, and by analysing whether those measures are systematically lower on the days chosen by the institution.
2. For the purposes of paragraph 1, point (b), competent authorities may, where they deem the tests performed by the institution to set the number of simulations insufficient:
|
(a) |
require the institution to provide the Monte Carlo statistical error at 95 % confidence level, and verify whether the method employed to measure such statistical error is sound; |
|
(b) |
require the institution to calculate the value-at-risk measure with several different seeds, all other assumptions being equal, and verify that the method used to generate simulation does not create bias in the results; |
|
(c) |
assess whether the differences in the value-at-risk measures with a different seed, as calculated in accordance with point (b), are compatible with the statistical error referred to in point (a), and, where that is not the case, assess the root cause of such incompatibility and the number of simulations needed to ensure that the statistical error is below 5 %. |