Publication Type

Master Thesis

Publication Date



Markov switching models with time-varying transition probabilities address the limitations of the earlier methods in the early warning system literature on currency crises. Most of the Markov switching models in the literature are largely based on univariate models of exchange rate fluctuations. In this thesis, the components of the index of speculative pressure are modeled using the Markov Switching VAR with time-varying transition probabilities of Martinez Peria (2002). Two approaches, both of which are derived from this model, are taken to determine the probability of a currency crisis: the probability of a turbulent regime and the expected value of the index of speculative pressure. This study shows that the Markov Switching VAR model with time-varying transition probabilities is a good method to use in building an early warning system of a currency crisis. Results show significant improvement on predicting the Asian Financial Crisis by signaling its occurrence at an earlier period with a higher probability when the probability of a turbulent regime approach is employed. It is also more sensitive in detecting turbulent periods that are not necessarily currency crises and therefore renders itself useful in short-term forecasting of speculative pressure episodes. The leading indicators of the Asian Financial Crisis identified in this study are real effective exchange rate, export growth, GDP growth, real domestic credit, M2 ratio, deposits to M2 ratio and non-FDI flows.


autoregressive modelling, economic impact, Markov-switching models, maximum likelihood, recessions, time-varying transition probabilities

Degree Awarded

MSc in Economics


Econometrics | Finance


MARIANO, Roberto