ENERGY-CONSUMPTION, REAL INCOME AND TEMPORAL CAUSALITY - RESULTS FROMA MULTICOUNTRY STUDY BASED ON COINTEGRATION AND ERROR-CORRECTION MODELING TECHNIQUES

Unlike previous studies on the causal relationship between energy cons umption and economic growth, this paper illustrates how the finding of cointegration (i.e. long-term equilibrium relationship) between these variables, may be used in testing Granger causality. Based on the mos t recent Johansen's multivariate cointegration tests preceded by vario us unit root or non-stationarity tests, we test for cointegration betw een total energy consumption and real income of six Asian economies: I ndia, Pakistan, Malaysia, Singapore, Indonesia and the Philippines. No n-rejection of cointegration between variables rules out Granger non-c ausality and imples at least one way of Granger-causality, either unid irectional or bidirectionial. Secondly, by using a dynamic vector erro r-correction model, we then analyse the direction of Granger-causation and hence the within-sample Granger-exogeneity or endogeneity of each of the variables. Thirdly, the relative strength of the causality is gauged (through the dynamic variance decomposition technique) by decom posing the total impact of an unanticipated shock to each of the varia bles beyond the sample period, into proportions attributable to shocks in the other variables including its own, in the bivariate system. Re sults based on these tools of methodology indicate that while all pair -wise relationships shared common univariate integrational properties, only relationships for three countries (India, Pakistan and Indonesia ) were cointegrated. For these countries, temporal causality results w ere mixed with unidirectional causality from energy to income for Indi a, exactly the reverse for Indonesia, and mutual causality for Pakista n. The VDCs were not inconsistent with these results and provided us w ith an additional insight as to the relatively more dominant direction of causation in Pakistan. Simple bivariate vector-autoregressive mode ls for the three non-cointegrated systems did not indicate any directi on of causality, significantly in either direction.