Causality Tests, Interdependence and Model Selection: Aplication to OECD countries 1960-97

This paper compares several methodologies for analysing unidirectional and bi-directional causality between Consumption and GDP in 25 OECD countries during the period 1960-95. For analysing unilateral causality a comparison is made between cointegration tests and joint regression on alternative explanatory variables, with 100% of correct results in the case of joint regression and lower percentages of success for cointegration approach. Bilateral causality is analysed comparing Granger´s test, a modified version of Granger´s test here suggested, TSLS, Hausman´s causality test and other approaches. The main conclusion is that the modified version of Granger´s test performs rather well and that Hausman´s test is very often useful for reinforcing the conclusions of multiple equations models with contemporaneous interdependence. Regarding the bilateral relationship between Consumption and GDP we conclude that there is a moderate degree of contemporaneous relation, with a high degree of dependence of Private Consumption on GDP and a lower dependence in the case of the reverse relation, because GDP is more dependent on supply side conditions than on demand side. This result is relevant for economic policies in less developed countries where very often emphasis is made more in the reverse relations than in the main ones.