We propose a new methodology for the analysis of impulse response functions in VAR or VARMA models. More precisely, we build our results on the non ambiguous notion of innovation of a stochastic process and we consider the impact of any kind of new information at a given date $t$ on the future values of the process. This methodology allows to take into account qualitative or quantitative information, either on the innovation or on the future responses, as well as informations on filters. We show, among other results, that our approach encompasses several standard methodologies found in the literature, such as the orthogonalization of shocks (Sims (1980)), the "structural" identification of shocks (Blanchard and Quah (1989)), the "generalized" impulse responses (Pesaran and Shin (1998)) or the impulse vectors (Uhlig (2005)).
Caroline Jardet, Alain Monfort and Fulvio Pegoraro
Classification JEL : C10, C32.
Keywords : impulse response functions, innovation, new information.
Updated on: 06/12/2018 11:00