We calculate and measure the regulatory and economic capital for operational risk using the internal Advanced Measurement Approach (“AMA”) methodology. Our AMA capital calculation is based upon the loss distribution approach (“LDA”). Gross losses from historical internal and external loss data (Operational Riskdata eXchange Association (“ORX”) consortium data), adjusted for direct recoveries, and external scenarios from a public database (IBM OpData) complemented by internal scenario data are used to estimate the risk profile (that is, a loss frequency and a loss severity distribution). Thereafter, the frequency and severity distributions are combined in a Monte Carlo simulation to generate potential losses over a one year time horizon. Finally, the risk mitigating benefits of insurance are applied to each loss generated in the Monte Carlo simulation. Correlation and diversification benefits are applied to the net losses in a manner compatible with regulatory requirements to arrive at a net loss distribution at Group level, covering expected and unexpected losses. Capital is then allocated to each of the business divisions and both, a qualitative adjustment and an expected loss deduction, are performed.
The qualitative adjustment (“QA”) reflects the effectiveness and performance of the day-to-day operational risk management activities via KRIs and self assessment scores, focusing on the business environment and internal control factors. The qualitative adjustment is applied as a percentage adjustment to the final capital number. This approach makes qualitative adjustments transparent to the management of the businesses and provides feedback on their risk profile as well as on the success of their management of operational risks. It thus provides incentives for the businesses to continuously improve the management of operational risks in their areas.
The expected loss (“EL”) for operational risk is based on historical loss experience and expert judgment, considering business changes denoting the expected cost of operational losses for doing business. To the extent it is considered in the divisional business plans, it is deducted from the AMA capital figure within certain constraints. The unexpected losses per business division (after QA and EL) are aggregated to produce the Group AMA capital figure.
Regulatory and economic capital for operational risk is calculated on a quarterly basis. The used internal data are captured in a snapshot at the beginning of the quarterly production cycle and undergo a quality assurance and sign-off process. Therefore, the complete history of previous quarters’ internal losses is taken into account in the calculation of the capital figures. ORX external data is submitted by the ORX members and also undergo quality assurance and sign-off. These data are recognized in the capital calculation at the earliest after six months. For the additional external loss data sourced from the IBM OpData (formerly named OpVantage), we use the data that are available twice a year (in the first and third quarters).
Economic capital is derived from the 99.98 % percentile, allocated to the business divisions, and used for performance measurement and resource allocation purposes, providing an incentive to manage operational risks, and optimizing the utilization of economic capital. The regulatory capital for operational risk applies the 99.9 % percentile. Economic and regulatory capital are calculated for a time horizon of one year.
Continued Operational Risk Framework Development
The AMA model is subject to continuous validation and enhancement, as we attempt to adequately reflect our risk profile. As part of the continuous enhancement and validation of our model we submitted model changes to BaFin and are awaiting approval from the joint supervisory team (BaFin and ECB). These model changes include an improved validation and recalibration methodology for insurance recoveries, changes to the modeling of the loss frequency as well as an enhanced scoring mechanism for the self assessment results in our AMA model.
Further, we have submitted an additional model change request to BaFin to replace the € 1 billion economic capital safety margin, which we have continuously applied since its implementation in 2011. This model change, which adds increased forward looking aspects to the AMA model, will result in higher economic capital even after we remove the safety margin. This change will make our model more risk sensitive by including reasonably possible litigation losses in our “Relevant Loss Data” set. Reasonably possible litigation losses may result from ongoing and new legal matters which are reviewed quarterly and are based on the judgment provided by our Legal Department.
While our dialogue with the joint supervisory team on these model enhancements is on-going, management has decided to recognize the impact of these model changes in the second quarter 2014 wherever they will lead to an increase in the capital requirement over the models that have previously been approved by BaFin.
Our Operational Risk Management Stress Testing Concept
We conduct stress testing on a regular basis and separate from our AMA methodology, to analyze the impact of extreme macro scenarios on our capital and the profit-and-loss account. In 2014, ORM took part in all firm-wide stress test scenarios and assessed and contributed the impact of operational risk to the various stress levels of the scenarios. The impact of operational risk on macro stress test scenarios has been moderate and remained in the expected range in regards to capital.
Our AMA Model Validation and Quality Assurance Review Concept
We independently validate all our AMA model components such as but not limited to scenario analysis, KRIs and Self Assessments, Expected Loss and internal loss data individually. The results of the validation exercise are summarized in validation reports and issues identified are followed up for resolution. This promotes enhancement of the methodologies. The validation activities performed in 2013 detected areas of improvement in our AMA model regarding the estimation of the loss frequency and the use of reasonably possible litigation losses. The results of this validation lead to model enhancements which increased the economic capital.
Quality Assurance Reviews are performed for the AMA components requiring data input provided by business divisions and result in capital impact. The AMA components data and documentation is challenged and compared across business divisions to help us maintain consistency and adequacy for any capital calculation.