Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
Robust Credit Stress Testing Through a Cointegrated Framework
Tiziano Bellini

Last modified: 2013-07-02

Abstract


Stress testing has become an important topic since the development of the risk management and the enforcement of banking international supervisory requirements. International standards, however, do not define how to stress risk parameters. Our research is devoted to develop a robust cointegrated framework where we investigate the mechanics linking banking credit risks and macro-economic variables. Remarking that the cointegration analysis is particularly sensitive to outlying observations, we exploit a robust approach to carry out the study. Comparing the pseudo maximum likelihood (PML) estimators and the forward search (FS) cointegration analysis, we analyze real Italian time series showing the effectiveness of the proposed approach in detecting atypical units and data structures.

Full Text: PDF