Operational Risk Measurement: An Optimal Timescale Selection Method Based on Information Entropy

Abstract

Operational risk capital is typically allocated on an annual timescale in extant research, while ignoring that the dependence between operational risk events may differ on different timescales, such as semi-annual and quarterly. This study examines whether operational risk dependence structures differ on multiple timescales and proposes an optimal timescale selection method based on information entropy theory to obtain more reasonable operational risk measurement results. Different dependence modeling methods, including linear correlation and copula functions, are employed to analyze the dependence structures on multiple timescales. The information changes between different timescales are calculated to select the optimal timescale. The empirical analysis is based on 2,024 operational risk events from 1994 to 2017 in the Chinese Operational Loss Database (COLD), which is one of the largest external operational risk datasets for the Chinese banking industry. The results reveal significant differences in operational risk dependence at different timescales which affect the operational risk measurement results. Crucially, the quarterly timescale should be considered for more reasonable measurements. The findings provide important insights for considering a more reasonable allocation of operational risk capital under multiple timescale dependence.