Exogeneity in error correction models (Q1308648)

This research report is concerned with the concept of weak exogeneity in Error Correction Models (ECM). In recent applied econometric literature, the ECM as a dynamic specification has become very popular, both for single equations and multivariate models. In chapter 1, some historical background to the modeling of economic time series is given. Chapter 2 presents some alternative approaches to cointegrated systems with special emphasis given to those systems on which the empirical part of the book is focussed. These systems rely on some form of ECM. A brief overview of different interpretations is given which make some difficulty to distinguish with incomplete models. Particularly, in the case of cointegrated systems it is difficult to distinguish between full-system and subsystem approaches. Chapter 3 points out that the use of a single ECM requires substantial hypotheses both upon the long run properties and on the weak exogeneity status of the variables. The single equation ECM in reduced form is derived and the conditions for an appropriate assumption of weak exogeneity are characterized. An analogous procedure for ECM in structural form with consequences for inference is discussed. The result is that the particular form of an ECM implies different conditions of exogeneity, and that the orthogonality condition is not sufficient for weak exogeneity. Some empirical studies from literature are investigated with respect to weak exogeneity in single equation models. Chapter 4 deals with exogeneity and the incomplete simultaneous equations model (SEM). It is shown that in the case of cointegrated systems a certain form of SEM -- among different representations of the incomplete SEM -- reveals as more useful for empirical analysis. The results for limited simulation experiments in the presence of cointegrated variables (Monte Carlo study) show the empirical power of orthogonality tests. The asymptotic behaviour of Hausman tests is investigated. In chapter 5, the modeling of cointegrated systems is studied within an empirical study of aggregate imports of Belgium over the period 1964-1990 for a full-system and a partial approach. Two alternative approaches, one based on a complete simultaneous model, another based on an ECM in structural form are tested on exogeneity. Weak exogeneity is not rejected for both the long-run and short-run case.