Advanced modular solutions for risk management

For example, bond issuing companies need to demonstrate that their risk management is sound if they want to receive good ratings from rating agencies. In many cases flexible solutions for risk management designed to meet the specific needs of the company may be more useful than big software packages.

What is QRR?

Quantitative Risk Research is a spin-off of RiskLab-Madrid, a team of researchers affiliated to the Universidad Autónoma de Madrid (Spain). The QRR team has been designing and developing software doing consultancy and in-company training in quantitative finance and risk management, for the last ten years in numerous banks and companies.

Burning issues in operational risk

During this coming year, 2007, a large portion of the banks are involved in active management of their operational risk and working hard on its measurement. Some of them are deploying AMA models. Most have opted for the Standard Approach Model. Both approaches require implementing a very active risk management compliant with the sound practices defined by the Basel Committee.

As a natural consequence of this process an increasing amount of data on operational losses will become available to the institution. Keeping historical records on these data will permit the participation in consortia like ORX. In order to understand the risk profile of the institution and to develop effective risk management practices, it is necessary to extract information, intelligence, from those data (both internal and external).

A first relevant aspect for the measurement of operational risk is the statistical characterisation of the collected loss data. This analysis poses a big challenge. Operational losses appear to follow heavy-tailed distributions, where typically more than 90% of the capital charge can be explained by a very small number of events. Data points span several orders of magnitude; it is common that the largest loss is 30 (or more) standard deviations away from the mean.

This implies that the estimates of the distributions used to model these losses may be very unstable and, as a consequence, this analysis requires the use of very advanced statistical techniques and mathematical modelling. In addition, many authors estimate that AMA models may yield an amount of regulatory capital smaller than 10% of the gross income while the Standard Approach would typically amount to more than 14%. Therefore, banks that have collected a sufficient amount of data to comply with Basel II, need to be able to determine whether AMA models are the next step in their risk management process. To make this decision, the need for quantitative approaches to operational risk becomes even more acute.

Another aspect related to the use of data is the search for models connecting losses with more qualitative approaches like key risk indicators (KRI), scorecards or others. Even banks adopting a qualitative approach to operational risk, need quantitative statistical evidence to support the validity of their analysis.

QRR experience in operational risk

We, at Quantitative Risk Research, have been working for more than four years both on quantitative and qualitative approaches to operational risk.

The main result of our efforts is Opvision, a powerful software application for operational risk developed jointly with Indra. OpVision, is an advanced multi-entity tool with distinct user profiles and the capability to manage diverse databases for the quantification of operational risk. This software allows the visualisation of losses from internal and external databases, and the calculation of the economic capital in the Loss Distribution Approach (LDA) using the most advanced quantitative and computational tools: Monte Carlo simulation, variance reduction techniques, Fast Fourier Transform (FFT), iterative methods, and grid computing, among others. It is Basel II compliant and allows the use and comparison of the most advanced models including piecewise fitting of distributions and the modelling of dependence structures in the calculus of capital. Opvision also incorporates modules for incorporating inputs from a qualitative model, stress and what-if analysis and flexible reporting.

OpVison is the solution selected by BBVA and other institutions are evaluating it in pilot projects. A new company (IRB Operational Risk) was created by Indra, BBVA and QRR for the joint development and marketing of OpVision.

QRR also has previous experience as consultants with Banesto on a quantitative approach to operational risk. It has participated in the first project (2005-6) on operational risk (qualitative models) at CECA (the Spanish Confederation of Saving Banks). The focus of our activity in this project was the definition of methodologies (scaling of KRI’s, aggregation of KRI’s, …).

Other fields

QRR offers a variety of software solutions and methodologies for problems in quantitative finance:

• e-Portfolio: a web-based tool for the optimization of investment portfolios.
• MatRisk: VaR calculation tool based on of extreme value theory and mixtures of distributions.
• QRR-rating: Calculation of company ratings.

Some of these tools are currently in use in the three biggest Spanish banks.