Explicit Duration Markov Switching Models PDF Download

Multivariate regime-switching presents an efficient way of jointly modeling the cyclical behavior of financial time series. Standard regime-switching models thereby a priori determine the relationship between the regime-switches of individual assets. These switches are usually assumed to be either perfectly synchronized or fully independent. However, neither assumption seems realistic in practice. This thesis develops a multivariate Markov regime-switching model to infer the actual degree of synchronization from the underlying data. This flexible model allows subgroups of assets to be driven by individual Markov chains. At the same time, these Markov chains underlie a dynamically changing degree of synchronization. In comparison to most existing solutions, this model is not restricted to bivariate analysis. To keep the model traceable, a novel factorization algorithm for the regime-dependent correlation matrix is formulated. This algorithm scales down the increase in parameters and presents an efficient way of ensuring positive semi-definite correlation matrices. The structure of the flexible regime-switching model is motivated by the initial synchronization analysis conducted in this thesis. The analysis of univariate regime-switching results shows that neither perfectly synchronized nor fully independent regime cycles are empirically observable. The synchronization of regime cycles tends to dynamically change over time. Some assets, however, might show more contemporaneous switching dynamics and can therefore be governed by a joint regime process. The empirical results for a sample of six international equity markets confirm the assumptions underlying this thesis. The flexible model reveals a stable synchronization factor, marked by one particular change in synchronization. The estimated parameters of this model closely cover the individual dynamics of their underlying assets and confirm the model's validity. Moreover, in some.