Integrating with the Global Economy: The Evolution of the Export Profiles of the European Union’s Eastern Periphery (2000–2021)

Abstract

How have Central and Eastern European countries integrated into global production networks since the early 2000s? Have European Union accession, the Great Recession of 2008 and the COVID-19 pandemic significantly altered their export profiles? Using cluster analysis for 11 economies between 2000 and 2021, we uncover persistent differentiation and deepening specialisation over time. Czechia, Hungary, Slovenia and Slovakia emerge as the manufacturing hub of the region. When broader exports are accounted for, these economies look similar to Poland, Croatia and Romania. Estonia, Latvia, Lithuania and Bulgaria form a separate group that relies on more labour- and resource-intensive manufacturing goods, commodities and services.

Acknowledgements

The authors wish to thank Hilary Appel, Dorothee Bohle, Rachel Epstein, Evelyne Huber, Tomasz Inglot, Sandra Joireman, Magnus Feldmann, Mitchell Orenstein, Besnik Pula, Jenny Pribble, John Stephens and two anonymous referees for their constructive feedback. All remaining errors are our own.

Supplemental data

Supplemental data: Supplemental data for this article can be accessed online at https://doi.org/10.1080/09668136.2024.2332671.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Countries in CEE, unlike their counterparts from the former Soviet Union, did not possess the high levels of natural resource wealth that would have encouraged the emergence of predatory local elites. In states such as Russia, Kazakhstan or Ukraine, this natural resource profile created ‘capitalism from above’, with a distinctly patrimonial character (King & Szelényi 2005). Dependence on natural resources, however, gave these states a semi-peripheral status in the international system (Lane 2010).

2 For an argument about Romania’s convergence with the Visegrád states, see Ban (2019).

3 This sentiment is currently embraced by anti-EU populist forces in the region, such as Hungary’s Fidesz or Poland’s Law and Justice (Prawo i Sprawiedliwość) parties.

4 Originally applied to the emerging economies of Asia, Africa and Latin America, the middle-income trap describes a scenario whereby economies cannot compete on wages with low-income countries and on innovation with high-income ones. Many see the middle-income trap as particularly fitting in the context of CEE economies, which became attractive destinations for foreign capital due to their geographic location and highly qualified, inexpensive labour force but were primarily treated as assembly platforms, with most innovation, research and development, and high value-added production occurring outside of their borders (Győrffy 2022).

5 Bulgaria and Romania acceded in 2007 while Croatia became a member of the EU in 2013.

6 Estonia, Latvia and Lithuania were annexed by the Soviet Union in 1944.

7 Partial exceptions were Croatia and Slovenia. Both were part of communist Yugoslavia which, nonetheless, was never a member of the Warsaw Pact and became an associate rather than a full member of the Council for Mutual Economic Assistance in 1964.

8 This is an accepted measure of export specialisation in the region (Myant & Drahokoupil 2012). A limitation to the use of the SITC classification is that ‘high-technology methods can be used for traditional products across all SITC categories and some exports of apparently high technology products may be the result of only simple assembly work’. Nonetheless, after checking against a number of other indicators, Myant and Drahokoupil affirmed the method’s appropriateness (Myant & Drahokoupil 2012, p. 6; see also Myant & Drahokoupil 2010, ch. 4).

9 The Standard International Trade Classification (SITC) Revision 4 codes for the four categories of manufacturing exports are presented in Appendix 1.

10 Adding ‘communication services’, which encompasses the broadcast or transmission of sound, images, data or other information, does not substantially change our results.

11 We chose to include foods as CEE countries had already developed specialisations in these sectors at the beginning of the postcommunist transition.

12 The full listing for these categories appears in Appendix 1.

13 When using both SITC and TiVA databases, we divide all export dollar amounts by each country’s GDP and normalise all indicators.

14 Data for service exports are not available for Slovakia and Slovenia in 2007.

15 Data for information services exports are not available for Lithuania in 2016. We rely on data for 2017 to maximise the consistency of our sample.

16 The Calinski, Duda-Hart and Silhouette solutions are not always definitive and occasionally point to more than two clusters, but the resulting score improvements are not large and might be driven by the fact that services data are not available prior to 2007. When solutions were not consistent, we focused on the dendrograms and drew on existing scholarship on the region. For the sake of parsimony, this often means that we chose the smallest number of clusters.

17 The results from the k-means analysis reveal greater instability over time. We suspect this is because the procedure might be more sensitive to missing data. The placement of Romania and Poland varies slightly over time, depending on the number of clusters.

18 Estonia boasts the highest level of MSTI and HSTI exports during the period of our analysis, with levels reaching 30% in 2000.

19 Data are not available for Slovenia and Slovakia in 2007, so the first dendrogram excludes them.

20 The Calinski–Harabasz index is calculated using the following formula: CH(k) = [B(k)/W(k)] × [(n−k)/(k−1)], where n is data points, k is the number of clusters, W(k) is the within-cluster variation and B(k) is the between-cluster variation. Higher values indicate more distinct clustering. Choosing the number of clusters that renders the highest Calinski index score is therefore equivalent to choosing the solution that maximises between-cluster variation.

21 The Duda–Hart index is the sum of squares in the two clusters, divided by the sum of squares in the combined cluster. It essentially compares the sum of squares in the next pair of clusters to be combined, before and after combining (Duda et al. 2000).

Additional information

Notes on contributors

Bilyana Petrova

Bilyana Petrova, Assistant Professor, Texas Tech University—Political Science, Lubbock, TX, USA. Email: [email protected]

Aleksandra Sznajder Lee

Aleksandra Sznajder Lee, Associate Professor, University of Richmond—Political Science, 231 Richmond Way, Richmond, VA 23173, USA. Email: [email protected]