The Application Of Monte Carlo Simulation In Measuring Operational Risk Of Property Insurance Company

The operational risk has been a serious threat to the solvency of the propertyinsurance companies and how to measure operational risk has been the focus ofacademia and the industry. Due to the complex and extensive reasons causingoperational risk and the lack of data, this thesis uses Monte Carlo simulation tomeasure operational risk.Firstly, it establishes the operational risk influence diagram which manifests thewhole process from occurring to subsequent control and the final loss of operationalrisk. Secondly, it expounds the basic theory and processes of Monte Carlo simulation.In turn, the thesis analyzes specifically how to use Matlab software to simulate andcompound loss severity and loss frequency, and how to use the Easyfit software to fitdistributions to calculate the eventual loss severity distribution and total loss ofinfluence diagram for each part. And then it deals with the correlation of autoinsurance and non-insurance with the Copula theory in order to realize Monte Carlosimulation and calculate the influence diagram to measure the total month’s loss ofoperational risk. And then it estimates the economic capital with VaR.Finally, it’s the case study. Because of the limited of data, the thesis obtains thesubjective data weighted by confidence indices through questionnaires, and tests thevalidity and reliability of the questionnaire using SPSS software in the case analysis.And then, it realizes Monte Carlo simulation by using Matlab and Easyfit softwares.And then it achieves the measure of operational risks. It concludes that the eventualloss distributions have thick tails and right-skewed characteristics excepting theDNPQ path of non-insurance, which indicate the losses of tail can’t be ignored. Themean of finances’ loss severity distribution is much bigger than non-insurance’s,which is bigger than auto insurance’s. Both finances and auto insurance’s month totalloss distributions are Johnson SB, which become Lognormal distribution fornon-insurance in the influence diagram. They are thick-tailed realisticallydistributions. The mean of month total loss distribution of non-insurance is thebiggest, which is about ten times more than auto insurance’s and three times morethan finances. The month total loss distribution of operational risk is thick-tailedrealistically Gen.Extreme Value whose mean is560million RMB and whoseeconomic capital is128million RMB at the confidence level of95percent for Property Insurance Company.