Nonlinear Cointegration and Nonlinear Error-Correction Models: Theory and Empirical Applications for Oil and Stock Markets

Recently national markets around the world have become more interdependent due to the increasing process of economic globalization as well as to the rapid development of internet and telecommunication technologies. One of the major facts of the global finance is that asset price movements in one market can now spill over easily and quickly to another market. For this reason, one must consider the joint behavior of financial variables to better understand the dynamic nature of market interrelations. Traditionally, the relationship between any two variables can be investigated within the causality framework of Granger (1969) which bases itself upon on the predictability of time series. That is, if forecasts of a variable, say Y, using both past values of Y and those of another variable, say X, are better than forecasts obtained by using only past realizations of Y, then X is said to cause Y in the sense of Granger causality. Of course the direction of causality can be inversed. Notice also that the vector autoregressive (VAR) model popularized in econometric literature by Sims (1980), coupled with block exogeneity tests (or also referred to as Granger causality tests) is central to situations where the analysis requires a simultaneous equations framework rather than a single equation.