ABSTRACT

There is hardly a concept in econometrics that is more mystifying and challenging than that of exogeneity in particular and its types in general. The study under considerations is an endeavor to argue that exogeneity is rather a simple concept, readily definable in terms of standard econometric models and deterministic relationship, and that the conceptual mystery primarily stems from improper usage of statistical vocabulary in a dynamic framework. Since, one purpose of econometric analysis is to device models for policy implications primarily relating to the concept of super exogeneity (SupExt, hereafter) sensing the underlying causal relation. There is an ample amount of tests that are available in literature for testing SupExt. The study compares the performance of SupExt tests on the basis of their size and power using Monte Carlo simulations combines with recently developed techniques of selecting data driven breaks or location shifts i.e. Indicator Saturation (like; Impulse Indicator Saturation (IIS), Step Indicator Saturation (SIS) & Trend Indicator Saturation (TIS)). To the best of our knowledge the performance of SupExt tests under stationary, non-stationary and dynamic settings using Monte Carlo Simulation has not been compared particularly combined with above mentioned break selections. Also, the study is not just limited to testing the performance of SupExt tests under IIS, SIS & TIS separately but it further extends its horizons to gauge the size and power of SupExt tests taking all these type of breaks into account jointly as well. While this is the main theoretical contribution of our study, from applied side; the study uses time series data to testify the hypothesis regarding SupExt of putative regressors using IIS, SIS & TIS in money demand function in case of Pakistan. The relevance of the empirical study to model money demand using tests of SupExt further strengthens the argument on a growing literature that empirically refutes the Lucas Critique for the determination of specific macroeconomic variables, resulting in to ensure the validity of policy simulations based on conditional model alone. After a detailed stability analysis in terms of SupExt testing of putative regressors in estimated money demand model, the empirical exercise concludes that the dynamic VECM is stable against the relevant class of interventions. Hence, estimated parsimonious model can be used for policy simulations. While taking power and size of these tests into account, one can use the optimal testing procedure to confirm the existence of SupExt of contemporaneous conditioning variables in their estimated model. Therefore, on the x basis of detailed simulations, the study concludes that whatever is the type of break whether it is IIS, SIS & TIS or we used all these at a time (jointly), the test like IBTest and RB-Test outperforms the performance of other SupExt tests. At the end, the power of the tests is increased significantly using all breaks at a time. Therefore, we recommend while testing SupExt using all breaks at time is more informative and useful than what is found to be in individual scenarios. The comparison of such SupExt testing procedures and selecting the best test out of these further blurs the demarcation of the critique raised by Lucas. Lastly, as far as the economic significance is concerned, the application of SupExt tests is not limited to test the Lucas critique but also help policy makers identify the existence of famous Ricardian equivalence indirectly as well.