Econometric modelling of the Lithuanian economic indicators

Abstract

The paper is devoted to model relations between Lithuanian indicators of production, foreign trade, income and prices and to present short-term forecasts. The join behaviour of Lithuanian GDP, exports and imports of goods and services, money, salaries and prices is examined by the structural vector auto-regression models (SVAR). Striving for the larger accuracy, apart the aggregated indicators their components are analysed as well.
Economic literature and experience of practical work show that there is relation between GDP, foreign trade, money and income indicators and unemployment. It was refer to Blanchard's work where he uses SVAR models to analyse GDP, consumer price index, money M1, wages and unemployment indicators. Unemployment was refused from a vector of Lithuanian economic indicators because the data from the Lithuanian labour exchange are unreliable. At the first stage there was limited oneself the detailed analysis of GDP and foreign trade indicators.
After the Russian crisis some relations between economic indicators do not observe in Lithuania. The­refore, forecasts that we calculated by using previous VAR models were not correct, and we had to change these models to the others, to look for new methods. One of them – to decompose GDP into separate parts (components) and forecast them separately.
In Lithuania it is seen that economic indicators are strongly cointegrated at the highest level of aggre­gation. These relations survived stable in period of the Russian crisis. For instance, GDP is cointegrąted with export and wages. That is why other alternative for SVAR models would be vector error correction model (VECM).
Co-ordination of those two methods is possible. The primary forecast, which was calculated as the sum of forecasts of components, is made more exactly, taking into account the residual of cointegration.
At the end of this research some results are presented, and the forecasts of GDP that were calculated using three methods mentioned above are compared to each other.