In this paper, we propose a new approach for characterizing and testing Granger-causality, which is well equipped to handle models where the change in regime evolves according to multiple Markov chains. Differently from the existing literature, we propose a method for analysing causal links that specifically takes into account Markov chains. Tests for independence are also provided. We illustrate the methodology with an empirical application, and in particular, we investigate the causality and interdependence between financial and economic cycles in USA using the bivariate Markov switching model proposed by Hamilton and Lin . We find that financial variables are useful in forecasting the aggregate economic activity, and vice versa.
|Data di pubblicazione:||2015|
|Titolo:||Granger-causality in Markov switching models|
|Rivista:||JOURNAL OF APPLIED STATISTICS|
|Digital Object Identifier (DOI):||http://dx.doi.org/10.1080/02664763.2014.993367|
|Appare nelle tipologie:||2.1 Articolo su rivista |
File in questo prodotto:
|granger-2.pdf||Documento in Pre-print||Accesso chiuso-personale||Open Access dal 01/01/2025|