A regime-switching approach to model-based stress testing

ABSTRACT The current financial marketplace has seen the development of increasingly complex products and thus, proper risk management has become a challenging task. Accurate stress testing relies upon models that capture their essential features. Novel, powerful ideas have emerged regarding the dynamics of financial instruments, such as incorporating mechanisms to model regime-switching and non-linearity. This paper outlines the logistic mixture of the linear components (LMLC) model and illustrates how it can be used to perform stress testing on an asset that follows regimeswitching dynamics. Once the model is estimated, running a stress test is straightforward; expected loss estimates are produced from the fitting equation. A market stress test is run on a merger arbitrage hedge fund position. It is demonstrated that the LMLC stress test produces expected loss estimates that are more conservative than estimates produced from a linear factor model.