The short-run and long-run dynamics of GDP and trade in a seemingly unrelated regression framework

Abstract

While most time series studies consider country specific or restrictive panel models to study the short-run and long-run dynamic relationships of economic variables, we pursue an unrestrictive system framework to explore these relationships. We show that instead of estimating traditional VAR and VECM models in a one-country setting, using a Seemingly Unrelated Regression (SUR) system estimation can help gain efficiency in coefficient estimates and standard errors. We study the long-run and short-run dynamics between GDP and Trade in a reduced form VAR and VECM setting for Canada, the USA and Mexico. We estimate the models for each country separately and compare the estimates with those obtained in a SUR system allowing for cross-country contemporaneous correlations. We find that the results change considerably when the models are estimated in the SUR system, in terms of Granger-causality as well as the long-run adjustment parameters.