CSSF Thematic Review: Validation of Value-at-Risk models used by UCITS for global exposure calculation – Results

The CSSF conducted a thematic review on the validation of VaR models used by UCITS for global exposure calculation.
The review was based on a representative sample of 20 Luxembourg IFMs managing Luxembourg UCITS that use VaR for global exposure calculation. The CSSF observed different approaches to perform the validation of the VaR model depending on the structural set-up of the IFMs. The validation was notably carried out by a third party (notably an advisory firm) most of the time. The CSSF recommended that the validation should be conducted by a party independent of the building process following initial development or any significant change to the model.
The CSSF also observed that the use of the square-root of time rule for scaling a 1-day VaR to a 20-days VaR was not challenged when the conditions set in paragraph 4 of Box 15 of the CESR Guidelines 10-788 were not met as a result of the modeling approach underlying the model. The validation exercise should cover the assumptions used by the model (e.g., the value of the decay factor, the shape of the return distribution, the threshold value for Peak-over-Threshold approaches when modeling extreme values, etc.) and the methodology used by some IFMs to scale VaR figures to different time horizons.
The risk management function should perform ongoing validation of the VaR model (this includes, but is not limited to back testing as laid down in Box 18) in order to ensure the accuracy of the model’s calibration. The review should be documented. Where necessary, the model should be adjusted. The questionnaire used in the thematic review covered various aspects, such as the organizational set-up of the validation, the aspects of the models covered by the validation, the tests performed, and the conclusions drawn by the IFMs from the validation exercise. In addition, the CSSF also based its observations on the VaR validation reports provided by the IFMs.