The impact of data censoring on estimation of operational risk by LDA method
The impact of data censoring on estimation of operational risk by LDA method

Summary/Abstract: The article raises the problem of determining the quantification of operational risk by LDA method for left-censored observations. In literature that concerns the modelling of operational risk, the term censoring can be found, but in practice it relates to the truncated observation. This article distinguishes truncated and censored data. Different properties of these processes are analysed by means of simulation, in the context of their impact on the size of operational risk estimates by the LDA method. In the world literature, the application of left-censoring process in operational risk has not been sufficiently analysed. Data censoring, which is a compromise between a full report of losses and collection of observation of only above mentioned level, consists in counting these observations that are below the accepted threshold. Such operations affect directly both the loss severity process and the process of loss occurrence frequency, which are components of the LDA model. What is very important in case of modelling is the uncertainty of the estimates in risk models.

• Details
• Contents