Direct and Indirect Causality Between Exports and Economic Output for Bangladesh and Sri Lanka: Horizon Matters

The extensive body of research that examines for (Granger, 1969) causality from exports to output for developing countries, including Bangladesh and Sri Lanka, using vector autoregressions and/or vector error correction models, is limited in only examining for one-period ahead or direct causality; the exception is in bivariate systems. This (usually unrecognized) focus on one-period causality in multivariate systems has often led to conclusions that exports do not Granger-cause economic output. We show that moving to Granger-causality at longer horizons, in a commonly used multivariate system, leads to bidirectional causality between exports and output, even when there is not one-period causality; the longer horizon causality arises indirectly through one or more of the auxiliary variables.