Title: On a number of time series econometric issues of some importance
Authors: Menelaos Karanasos - Brunel University (United Kingdom) [presenting]
Alexandros Paraskevopoulos - University of Patras (Greece)
Abstract: A new innovative methodology for analysing time series models without the need to work with lag polynomials is presented. Following laborious research work, the literature contains a diversity of linear time varying models (i.e., smooth transition AR specifications, GARCH processes with time dependent coefficients, Markov switching models, generalized random coefficients AR processes, periodical and cyclical formulations) whose time series properties remain unexplored. Making progress in interpreting seemingly different models requires us to provide a common platform for the investigation of their time series properties. We develop a theoretical foundation on which work in synthesizing these formulations can be done. With the help of a few detailed examples, i.e., GARCH in mean models with abrupt breaks, we demonstrate how to encompass various time series models within our unifying theory. Our innovative methodology allows us to study stochastic linear difference equations of ascending order and handle time varying specifications of infinite order. An advantage of our methodology is that it can be applied with ease in a multivariate setting and provide a solution to the problem at hand without adding complexity. The main strength of our general solution and the way we have expressed it, is that researchers can use it for a multiplicity of problems. The significance of our methodology is almost self-evident from the large number of problems that can solve.