Measuring systemic risk: A Markov switching regression approach
This thesis aims to contribute to the debate on systemic risk. The methodology developed in this thesis is applied to a small set of US bank holding companies using publicly available data. Specifically, we investigate the performance of the financial system conditional on the bank holding company being under nancial distress. To this end, we propose a Markov switching regression model to infer the regime in which the bank holding company was at any historical date. Conditional on the bank holding company being in a particular regime,we model the returns of a global Bank Index as an affine function of macro-economic state variables and the bank holding company’s stock price returns. Systemic risk contribution is assessed using correlations, changes in parameter estimates and a measure of regime comovement.The results indicate that correlations of returns increase during times of turmoil.The results are, however, mixed regarding the change in the fraction of shocks that are on average transmitted from the bank holding company’s stock price to the global Bank Index.We also briefly consider a Markov switching regression model with time-varying transition probabilities.