CONVERGENCE OF EXPORT-IMPORT FLOWS AND ECONOMIC DEVELOPMENT IN THE CENTRAL AND SOUTHEAST EUROPEAN UNION

: The study presents the long-and short-term relationship between international trade flows (exports/imports) and economic development (GDP) as the main driver of international economic trade. Panel data econometric models emphasize the fixed effects of the eight Central and Southeast European Union countries. The cointegration condition is met to identify the existence of long-run equilibrium. The error correction model is the iterative short-run adjustment solution to the long-run relationship for both exports and imports. Regional trade convergence is achieved by observing the cointegration of the analysed variables; it is described by their average levels for all countries.


Introduction
The economic development of the member states of the European Union is different.The accession to the EU of the states of Central and Southeast Europe provides the framework for a similarity in the dynamics of their macro-indicators.
The economic openness of the countries in the Central and Southeast of the EU is the basis of economic development to ensure economic and social security, and also the well-being of the population.
The two international commercial flows: exports and imports, are highly interdependent.Imports involve payments and exports involve receipts.The coverage ratio is the rapport of exports to imports; if it is greater than 100%, it shows a positive state of the foreign trade balance, otherwise a negative one because the country has to pay more for its imports, and its exports become cheaper as a result of the changes in the exchange rate.Developed countries have higher export volume than import volume, meaning a coverage rate of exports higher than 100%.
Fig. 1 shows the evolution of the proportions of exports (p_X) and imports (p_M), calculated as ratios in GDP at current prices, at the EU-27 level and for each EU CSE country.We see that in Romania, throughout the analysed period, the proportions of imports were higher than those of exports, standing between 40-50%.Only Croatia recorded higher import proportions than export proportions, but very close until 2019.The impact of the 2020 pandemic can be seen for all countries, including the EU-27.

Objectives
The objective of our study is to analyse the long-run relationships separately between exports and GDP and between imports and GDP for EU CSE countries.
The membership aspect of the CSE region can confer a resemblance of long-term equilibrium, taking into account country specificities, levels of economic development, and in accordance with each country's export/import policies.

Data
The annual data of volume macro-indicators of GDP, imports and exports are expressed in millions euro 2010.In Fig. 2 we can see the evolution of GDP and international flows of exports and imports for each CSE EU country.We can see the very close evolution of imports and exports in Poland and Croatia until 2019.Slovenia, Slovakia, The Czech Republic, Poland and Hungary had positive trade balance for the entire period 2011-2022.Romania and Bulgaria had a foreign trade deficit for the entire period.The export volumes are very close to GDP in Hungary and especially in Slovakia, with proportions higher than 90%, in Figure 2.

Methodology
We consider GDP as explanatory variable firstly for exports (X) and then for imports (M).If the variables are non-stationary and they are integrated of the same order, they are cointegrated if they admit a stationary combination.The long-run equation offers the error correction term for the short-run equation.
The panel data models assumes to establish the significance of random effects and of fixed effects for each and then for both dimensions: cross-sections and periods.
After establishing the appropriate panel data model for the long-run model, the error correction term (ECT) is established.The residuals (ECT) become the cointegration term in the short-run equation (eqn.1).The ECT variable must be stationary.The lag ECT is used in the short-run model which is the Error Correction Model (ECM).
(1) If the coefficient β2 of ECT in ECM is negative and significant, then there exists a longrun equilibrium and the coefficient is just the speed of adjustment during one year.
Then we repeat the analysis for imports.The findings highlight the convergence of GDP influence on the two international trade flows in the EU CSE region.

Results
Pairwise Granger causality tests show that GDP is a Granger cause of exports (X).Pairwise Dumintrescu Hurlin causality tests support the null hypothesis that both X and GDP do not homogeneously cause the other variable.
In Table 1, the descriptive statistics of exports (X) and GDP volumes show the ascending order of the evolution of these indicators for the eight analysed countries.We opt to use the logarithmic values of exports (LX) and for GDP (LGDP), to explain the influence of the independent variable (LGDP) on the explained variable (LX) in relative terms.

The economic development and export policies in the Central and Southeast EU countries during 2011-2022
To test non-stationarity, we use the unit root tests for each of the two variables LX and LGDP; they are nonstationary in levels, but stationary in 1 st differences.We conclude they both are integrated of order 1, I(1).
The Pedroni residual cointegration test with individual intercept for each cross-section presents 11 test statistics with the associated probabilities, in Table 2.There are 6 test statistics which reject the null hypothesis of no cointegration and accepting the majority, we decide that the variables are cointegrated.The Johansen Fisher panel cointegration test with linear trend (option 3) and maximum 1 lag for both variables is presented in Table 3. Except Poland, for all the other CSE countries, the Johansen Fisher cointegration test rejects the null hypothesis of no cointegration, so they are cointegrated and they all have at most one cointegration equation.In Table 1, we can see that Poland is the most developed country in the Central and Southeast region of the European Union.To choose the best panel data model for long run equation, we build the long run model with the pool OLS regression, which is not recommended because it does not consider the heterogeneity of countries.The pooled OLS models is used only to test the existence of the random effects and the dimensions for which they may be significant.
The Lagrange Multiplier tests for random effects applied on the pooled OLS long run model find them as significant for cross-sections and for both cross-sections and periods, as seen in Table 4.We build the FE panel data model, in Table 5.The long run model with fixed effects of countries and GLS weights as Cross-section SUR solves the dependence of residuals in cross-sections.We build the series of residuals, named ECT_long_fe.
The unit root tests show the series of residuals as being stationary and we conclude that the variables LX and LGDP are cointegrated, meaning that they have a significant long run relationship.
The ECT_long_fe term is normally distributed.The null hypothesis of no cross-section dependence (correlation) in weighted residuals is accepted.
The long run equation with fixed effects of cross sections rejects the null hypothesis of redundant fixed effects tests and it finds them as significant.As we can see in Table 5, the too high value of R-squared and the too low value of Durbin-Watson statistic is a sign of a spurious regression.It is important to build an Error Correction Model (ECM).
The long run equation with random effects of cross sections rejects the null hypothesis of Hausman test of correlated random effects and the fixed effects model is better, as seen in Table 6.
The series of residuals of random effects cross section model, named ECT_long_re, is not stationary.Verifying the autocorrelation coefficients in levels and in 1 st differences with global tests Q-statistic of stationarity lead us to the same conclusion.
The random effects model is not good to be considered.We use the residuals of the fixed effects cross-section model, ECT_long_fe, as the lag term of ECT in the Error Correction Model (ECM), in Table 7.The ECT coefficient is negative and significant, as expected.The conclusion is that there exists a long run equilibrium of the export and economic development policies of the CSE EU countries.The coefficient ECT represents the speed of adjustment to the long run equilibrium.The system is running back with 45.69% annually towards the equilibrium running from GDP to exports.
The long-run equation is from Table 5, the FE cross-section model: The coefficient of LGDP is significant at a probability value less than 1%.The interpretation of this coefficient is that at 1% increase in GDP, the exports (X) increase by 1.772%.
The ECM from Table 7 is: In Table 7, we can see the significant short run coefficient of GDP at less than 1% Pvalue.At 1% increasing in GDP, exports increase by 1.8546% on average in the short run.The discrepancy between 1.77% in the long run and 1.85% in the short run is corrected each year by 45.69%.Each country has a specific coefficient comprised in the fixed effects, which affects the intercept of each country model.

The economic development and import policies in the Central and Southeast EU countries during 2011-2022
We repeat the same analysis for the relationship between imports and GDP.Both GDP and imports are Granger cause of each other, and neither homogeneously causes the other.
We check the non-stationarity of imports, and LM (using the logarithms of M) is nonstationary and integrated of order 1.The LM and LGDP are cointegrated, as seen in Table 8; the null hypothesis of no cointegration is rejected for 8 of 11 test statistics: In Table 9, the ECM for imports and GDP shows the negative and significant coefficient of ECT shows that the system is coming back towards equilibrium with a speed of adjustment of 30.23% in one year.We find that there is also a long-run equilibrium between the imports and GDP, for the EU member states in the Central and Southeast Europe in the period 2011-2022.

Conclusions
We can see in Table 10 a summary of the coefficients of the long run and short run equations for both international commercial flows.
In the short run, GDP influences more the exports than the imports.
In the long run, the influence of GDP is higher on imports than on exports.The speed of adjustment is higher for exports during one year.Analysing the long run coefficients, we conclude that the imports should be higher than the exports, because the influence of 1 % of increase in GDP is higher for imports of about 1.9458% compared with 1.772% for exports.
The convergence of export-import flows with economic development is sustainable in the long term for the Central and Southeast region of the European Union.
In the short term, we can see the greater influence of GDP on exports as a stimulating factor.For imports, it is less encouraging than for exports.Imports are a source for final consumption, and especially for household final consumption, and it is important by increasing the standard of living of the population.The same conclusion of export stimulation goes for explaining and interpreting the higher speed of adjustment for export than for import.

Fig. 1 .
Fig. 1.Proportions of exports/imports in GDP at the EU27 level and in the EU CSE countries in the period 2011-2022

Fig. 3 .
Fig. 3. Evolution of GDP, exports and theoretical exports with ECM

Fig. 5 .
Fig. 5. Average evolution of GDP, imports and theoretical imports in the CSE EU region