ORX Analytics

The core purpose of ORX is to advance the measurement and management of operational risk in the global financial services industry. To support this objective ORX supports research into operational risk on the basis of the ORX Global Database. The objective is to undertake collective analysis of the ORX database with the purposes of increasing Member understanding of the use and utility of the ORX database, to develop common analytical tools and to propose amendments to ORX data standards or practices. 

The ORX Analytics work programme is overseen by the ORX Analytics Working Group. To undertake research on behalf of the Working Group ORX contracted with IBM Research Zurich, as Analytic Agent, to provide technical support in May 2007. Since that date research has proceeded on issues such as:

• Homogeneity;
• Scaling;
• Correlation;
• Capital Modeling.

The homogeneity of data is to establish whether the ORX data can be broken down into groups and sub-groups that have common statistical characteristics.  Given the data collected the dimensions that could be used in creating homogeneous clusters included Business Line, Event Type and Region.  The analysis found that homogeneous clusters of loss data can be created.

For banks using ORX data, whether for internal benchmarking or inclusion in various capital calculations, an issue is how to scale external data.  The starting point for the scaling was the clusters of homogeneous data.  The outcome is that scaling can be applied using an exposure indicator to the location of the data distribution (the mean) and in some circumstances to the shape parameter (equivalent to the standard deviation).  The end result of this analysis was the distribution of a scaling tool to ORX Members.

An aspect of operational risk modeling, especially for capital purposes, that has been discussed is correlation or co-dependence between events.  However, an issue for individual firms has been the amount of available data; this can be alleviated by performing the analysis at the ORX level.  The result of the analysis showed that correlation does exist, however it is very low.

ORX has undertaken to publish the results of this analysis and the following papers and presentations provide an overview of the work described in brief above:

• Operational Loss Scaling by Exposure Indicators: Evidence from the ORX Database (7 October 2008)
• Observed Correlations and Dependencies Among Operational Losses in the ORX Consortium Database (27 November 2008)
• Challenges in Measuring Operational Risk from Loss Data (9 September 2009)
  
   

The ORX Analytics research programme will continue and, as appropriate, ORX will publish results through this website and industry media and journals.