Scaling Operational Loss Data and Its Systemic Risk Implications

In measuring its Operational Risk VaR, a bank needs to pay attention when including external data in its internal loss collection. In principle, these data should be scaled consistently to the specific nature of the bank's risk, but this is not done by the majority of institutions with advanced internal models, due to the lack of an established practice on how to do it. As highlighted in this paper this may have a negative effect on the capital requirement estimates both for the bank and at the systemic level. We propose a simple and viable scaling methodology to overcome this issue. After setting up a stylized banking system model we show, both analytically and through simulations, that failing to adopt a scaling methodology leads to underestimation of capital charges by the more risky banks and overestimation by the less risky ones. In other words, at a systemic level, the latter subsidize the former in term of regulatory capital. Moreover, the capital charges of the banks become correlated, therefore a systemic risk emerges when all banks need to raise capital at the same time to meet their increased regulatory requirements. We end by showing how our scaling methodology operates on simulated and real data to overcome these issues