I'm trying to specify a model with government expenditures and economic growth. My VECM passes all the robustness checks, when variables are included without transformations, but the coefficients are abnormally high, because I'm including GDP per capita data in absilolute values and expenditures data in percentages. If I take all variables in log transformations, then the coefficients make sense, however, the model has serial correlation and heteroskedasticity problems (only in logs). Can I proceed with the VECM model without log transformations?
These coefficients show the speed of convergence towards long-run equilibrium. In the short run, the variables can deviate from long-run equilibrium or cointegration relationships. The adjustment coefficients depict how these deviations will be corrected
- Badges
- Science topic