How to back-test operational risk: an empirical basic analysis

Measuring, controlling and managing operational risk have played an important role for bank industry endeavors since the publication of Basel II (2006) as the regulatory framework for the effective management and supervision of financial risks. More recently, due to the current financial crisis, the Basel Committee on Banking Supervision (henceforth, the Committee) has revised such international standards giving rise to Basel III (2010) in order to strengthen the regulation, supervision and risk management of the banking sector. Inherited from market risk, the application of the Value at Risk (VaR) concept to the Loss Distribution Approach (LDA) has been encouraged by the Committee for measuring operational risk. Moreover, complementary analysis has also been recommended to calibrate the soundness of such estimates by assessing the exceedances beyond Operational Value at Risk (OpVaR) forecasts. In this paper, we conduct an empirical back-testing on the LDA model by using an Internal Operational Losses Database (IOLD) provided by a medium-sized Spanish savings bank. For this purpose, we calculate daily OpVaRs to which we applied the basic analysis, based on the binary indicator and extremal index. Our empirical results reveal that the implementation of the LDA model has proven to be unreliable for the savings banks analyzed under the Basel regulatory framework.