Estimation of operational risks using a semi-parametric approach

This paper employs a semiparametric framework for estimating 99.9 per cent operational risks (ORs) and expected shortfalls (ESs) of some US businesses. The latter is a coherent risk measure, while the former is not. Since the sever- ity loss distributions are heavy right-tailed, the peaks over threshold method, which is known as POT, is used. To implement this method, we estimated the threshold parameter by adopting the approach proposed by Huisman et al. [ 2001 ]. The Generalised Pareto Distribution (GPD) is then tted to the data in the right tail. In contrast to the previous studies that estimated the population mean by the simple sample mean, this paper used the method proposed by Johansson [ 2003 ] which exploits the property of the tail index to estimate the population mean. The estimates of 99.9 per cent ORs and ESs of ve business line and three event type losses, range from $ 1,361 mln to $ 12,547 mln and from $ 1,683 mln to $ 20,350 mln, respectively