Economic activities and indicators of a small open country are crucially influenced by the dynamics of export volume. This paper examines the Lithuanian export trends by different countries using a modem non-stationary time series and econometric theory.
To avoid spurious regression, time series are modelled as an integrated process. The stationarity of Lithuanian export and the main explored exogenous variables (GDP, import values, exchange rates and CPI) are analysed using the augmented Dicky–Fuller test. Most of the indicators are integrated of order one. Further investigated modelling alternatives are the vector autoregression (VAR) and the error correction model (ECM). Although according to the Granger representation theorem each finite order VAR has a respective ECM, it is known that, in general, co-integrated (in level) process does not admit a pure VAR representation in first differences. Hence VAR is used, when there is no statistical co-integration evidence. In eleven of fourteen cases ECM proved to be appropriate for the Lithuanian export modelling. VAR turned out to be needed only in three cases.
The forecasted value of total Lithuanian exports in year 2000 is LTL 14293 mn. Export to the EU, the CIS, and the CEFTA groups are forecasted to be LTL 7006 mn, LTL 2420 mn, and LTL 1043 mn, respectively. The forecasts of Lithuanian exports to the three different country groups are formed assuming constant shares between the analysed and the remaining countries in a certain group. Conditional export forecasts are based on the exogenous variables forecasts. These are modelled using the SAS/ETS software as minimum RMSE trend, exponential smoothing or ARIMA model.
This work is licensed under a Creative Commons Attribution 4.0 International License.