|
Abstract:
|
Federal Reserve Bank (FRB) has been strongly recommending that financial institutions use macroeconomic factors to develop adverse scenarios for various types of risks. However, for operational risk, FRB expressed reservations about relying on macroeconomic factors for developing conservative stress scenarios. The paper presents an extensive feasibility study carried out to find out whether the macroeconomic factors suggested by FRB can be used to build robust and meaningful regression models for estimating time development of operational risk losses. We have developed a two-step approach for building libraries of repression models connecting operational risk losses and macroeconomic factors. First, build variety of so-called Base Models for different historical periods and lags - each model can be considered as an expert who learned under specific circumstances. Then combine Base Models into the Final Model - contribution of each Base Model to the Final Model is proportional to its ability to estimate accurately the latest house prices. We employed this approach to build the library of multivariate regression models for the major Basel types of operational risk events.
|
