Data Scaling for Operational Risk Modelling

In 2004, the Basel Committee on Banking Supervision defined Operational Risk (OR) as the risk of loss resulting from inadequate or failed internal processes, people and systems or from external events. After publication of the new capital accord containing this dfinition, statistical properties of OR losses have attracted considerable attention in the financial industry since financial institutions have to quantify their exposures towards OR events. One of the major topics related to loss data is the non-availability of a suficient amount of data within the Financial Institutions. This paper describes a way to circumvent the problem of data availability by proposing a scaling mechanism that enables an organization to put together data originating from several business units, each one having its specific characteristics like size and exposure towards operational risk. The same scaling mechanism can also be used to enable an institution to include external data originating from other institutions into their own exposure calculations. Using both internal data from different business units and publicly available data from other (anonymous) institutions, we show that there is a strong relationship between losses incurred in one business unit respectively institution, and a specific size driver, in this case gross revenue. We study an appropriate scaling power law as a mechanism that explains this relationship. Having properly scaled the data from different business units, we also show how the resulting aggregated data set can be used to calculate the Value-at-OR for each business unit and present the principles of calculating the value of the OR capital charge according the minimal capital requirements of the Basel committee