Measurement of Operational Risks with Stochastic Models in Turkish Banking System

Abstract

English Abstract: Work on the scale and scope, as well as direct and indirect impacts of operational risk, endorses the need for the measurement and management of this risk. In order to manage operational risk, it must be quantified and measured properly. Measurement of operational risk requires considerably different and more sophisticated quantitative methods and techniques than the ones currently used in the measurement of financial risks.

Literature on operational risk measurement is still in its infancy. Therefore, there is a potential to improve and develop the methodologies used in the quantification and measurement of the operational risks. This research best to the author’s knowledge is the first study in Turkey on the application of both LDA and EVT on operational risk measurement within the context of stochastic modeling of operational risks. The main purpose of this paper is to contribute to the literature on operational risk measurement still in progress both in Turkey and in the world.

Operational risk has unique features in comparison to other measurable and manageable risks. For this reason, measurement of operational risk requires considerably different and more sophisticated quantitative methods and techniques than the ones currently used in the measurement of financial risks.

This study has two main purposes. The first one is to develop a methodological framework of Loss Distribution Approach (LDA), which originated from actuarial mathematical models. During the research, the LDA is developed and turned out to be suitable for the measurement and management of operational risk and capital allocation.

Within this context, in this research, after a comprehensive literature review and a discussion of the theoretical background of the LDA, the extent and methodology of the research were given and data issues were handled. In order to represent the unique features of operational risks, a measurement model was constructed by two stochastic processes namely “severity” and “frequency” of loss events. These two processes modeled separately and then brought together to form an aggregate loss model. Using this model, as a risk measure Operational Value at Risk (VaR) and Operational Expected Shortfall (ES) have been estimated. Then, operational VaR and ES estimates have been back tested in order to determine the accuracy and reliability of the aggregate loss models.

Second main purpose of this study is to improve the estimation capability of measurement models in modeling tail probabilities of loss distributions that is at the very heart of operational risk management. Due to the heavy-tail property of operational risk data, LDA models are unable to model tail probabilities of loss distributions accurately. Within this context, in order to overcome these difficulties Extreme Value Theory (EVT) has been employed in the modeling tail probabilities of operational loss distributions. Through application of EVT to actual operational loss data, operational risk measures (i.e.Operational VaR and ES) have been estimated.

Date of Award	2005
Original language	English
Awarding Institution	Gazi University
Supervisor	Serdar Kilickaplan (Supervisor)