Export diversification across countries and products: Do Eastern European (EE) and Commonwealth of Independent States (CIS) countries diversify enough?

Abstract

Despite the importance of geographical and product diversification of exports, this question has not got enough attention in the literature. We look at country and product diversification of exports from Eastern Europe (EE) and Commonwealth of Independent States (CIS), two groups of countries that both substantially increased trade openness since the beginning of transition, but took different paths in terms of product and geographical composition of exports, and compare with the export diversification predicted by the gravity equation estimated on a large sample of countries in 2001–2007. The results demonstrate substantial deviations of the actual diversification levels from the levels predicted by the gravity model for the CIS countries, while the EE countries' levels of diversification are much closer to the levels predicted by the model and consistent with the data. All CIS countries lag behind the region leaders – Czech Republic and Poland – in terms of the degree of export diversification. In particular, the CIS countries extensively engaged in the export of raw materials have the most concentrated exports in terms of their product composition. In terms of geographical diversification, Belarus has the least diversified exports among all transition countries.

Keywords:

• gravity model
• trade potential
• trade policy analysis
• diversification

JEL Classifications:

• F12; F14; F17

Notes

 1. The sample of EE countries includes Albania, Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Macedonia, Poland, Romania, Serbia, Slovakia, and Slovenia.

 2. The sample of CIS countries includes Armenia, Azerbaijan, Belarus, Georgia, Kyrgyz Republic, Kazakhstan, Moldova, Russia, and Ukraine.

 3. We define expansion in exports at the intensive margins as growth of the value of exports of goods that has been previously exported to the same countries, while export of new products or export of existing products to new countries is expansion at the extensive margins.

 4. Suppose that countries are indexed j = 1, 2, … , N and products are indexed k = 1, 2, … , K. The country- and product-HHI are computed according to the following formulae. Denote export from country i to country j in product category k in year t as . The country-HHI in product k in year t is , where . The figure presents country-HHI averaged across products and over time. The product-HHI is computed as , where . The figure presents product-HHI averaged over time.

 5. Imbs and Wacziarg (2003) identified the turning point at the level of 13,000 of constant 2000 USD (real GDP per capita in Ireland in 1992), the level that had not been reached by any of the CIS countries by 2007.

 6. A standard trade model without fixed trade costs and heterogeneous firms would not be able to generate patterns with asymmetric exports and zero trade flows, predicting more diversified exports.

 7. We consider a partial equilibrium model with fixed capital during the period being investigated. Labor is the only input that is perfectly mobile across industries but immobile across countries.

 8. Based on WDI data on GDP in nominal US dollars, .

 9. In the following discussion, the analysis is carried out at a product level of aggregation. The product index k is dropped to simplify notation. At the same time, the time dimension is introduced to capture the dynamic nature of exports.

10. There is evidence that the probability of exporting at year t depends on theprobability of exporting at year t − 1 (Carrere 2006), or the length of the relationship (Besedes and Prusa 2006). We model the persistence of trade relationships coming from its impact on fixed costs, because a firm entering a foreign market presumably needs to spend more on researching local conditions, adjusting products to local requirements, establishing a distributional network relative to a firm that is already present at the foreign market.

11. Alternatively, both sides of equation (9) are divided by σ_ω . Both procedures lead to the same outcome in terms of predicting the probability of positive trade.

12. The number of source and destination countries is not balanced due to poor export data availability for Afghanistan, Angola, Liberia, Libya, Turkmenia, Uzbekistan and due to negligible export flows of some island-states (Bahamas, Bermuda, and Cayman). Still, import data to those countries are recorded and are quite substantial to be ignored in the analysis. Twenty-three transition economies were excluded from the sample at the estimation stage.

13. We first acquired data for 63 SITC sectors. Further aggregation to 10 product categories is carried out for presentational convenience. We performed a robustness check on the impact of the level of aggregation of export data on the main results and found no effect on computation of geographical diversification indices, while some changes in the levels of product diversification. The changes, however, have not altered the ordinal ranking of countries in terms of product diversification indices.

14. We have checked how this finding influences our main results. We replaced the official common language variable by the ethnic common language in the selection equation. We also included the official common language in the gravity equation. Our conclusions are robust to the alternative specifications of the model.

15. A formal econometric test on poolability of the data is strongly rejected.

16. Helpman, Melitz, and Rubinstein (2008) use the common religion variable asthe excluded variable based on the statistical argument – it is not significant at the second stage of their estimation. They also use the common language variable as an alternative and find no significant changes to their results.

17. We have checked whether the results are sensitive to the choice of the threshold, by imposing a stricter threshold of 0.75, which led to essentially the same conclusions.

18. A plausible explanation to this observation is that the EU membership has the different impact on exports relative to the average effect of other regional trade agreements. As a robustness check, we have run a model that controlled for the EU membership as an additional determinant of exports. This modification, however, had very little impact on the predictions of the model, leaving our conclusions unchanged.

19. It can be argued that exports of natural gas from Russia are distorted due to political motives, since it is used as a tool of political pressure on neighboring countries. Armenia and Azerbaijan, Georgia and Russia are examples of country-pairs whose trade is well below potential due to political conflicts.