An internal fraud model for operational losses : an application to evaluate data integration techniques in operational risk management in financial institutions

• Autor: Paredes Leandro, Rocío Margaret
• Formato: tesis doctoral
• Fecha de Publicación: 2016
• Institución: Pontificia Universidad Católica del Perú
• Repositorio: PUCP-Tesis
• Lenguaje: inglés
• OAI Identifier: oai:tesis.pucp.edu.pe:20.500.12404/7998
• Enlace del recurso: http://hdl.handle.net/20.500.12404/7998
• Nivel de acceso: acceso abierto
• Materia: Administración de riesgos; Instituciones financieras; https://purl.org/pe-repo/ocde/ford#5.02.04

The handling of external operational loss data by individual banks is one of the longstanding problems in risk management theory and practice. The extant literature has not provided a method to identify the best way to combine internal and external operational loss data to calculate operational risk capital. Hence, to improve the knowledge and understanding of internal-external data combination in operational risk management, this study applied a simulation-based evaluation of well-known data combination techniques such as the scaling, the Bayesian, and the covariate-base techniques. This research considered operational losses arising from internal fraud in retail banking within a group of international banks that share data through an operational loss data exchange. One of the key elements of the simulation-based statistical evaluation was the development of a dynamic internal fraud model for operational losses in retail banking. The internal fraud model incorporated human factors such as the number of employees per branch and the ethical quality of workers. It also included the extent of risk controls set by bank managers. There were two sets of findings. First, according to the simulation-based evaluation, the scaling technique was by far the less useful for estimating the appropriate operational risk capital. The Bayesian and the covariate-based techniques performed best. The Bayesian technique was the best for higher percentiles while the covariate-based technique was the best at not so extreme quantiles. The choice of technique therefore depends on the risk appetite of the financial institution. The second set of findings relates to the model validation with hard data. Losses generated by the model in the banks across the world were associated with GDP growth and the corruption perception of the country where banks were located. In general, internal fraud losses are pro-cyclical and the corruption perception in a country positively affects the occurrence of internal fraud losses. When a country is perceived as more corrupt, retail banking in that country will feature more severe internal fraud losses. To the best of knowledge, it is the first time in the operational risk literature that this type of result is reported