Testing the export-growth hypothesis: some parametric and non-parametric for Ghana

To test the export-growth hypothesis for Ghana, four parametric causality models (the Granger model, the Sims model, the modified Sims model, the Akaike Final Prediction Error model) and the non-parametric multiple rank F-test model of Holmes and Hutton are estimated for a variety of variable specifications and lag structures. Consistent with other published research, causality conclusions are sensitive to the different causality tests used, to the lag specification of the variables and to the structure of the error terms. The interesting contribution of this research is in comparing results derived from parametric and non-parametric models and in showing that violation of the normality assumption leads the Granger model to the apparently incorrect conclusion that GDP causes exports.