ABSTRACT
This paper employs a panel vector autoregressive (PVAR) approach to investigate the relationship among military expenditure, investment, and economic growth, over the period after the enforcement of the Maastricht treaty (1994–2022) in 25 European countries that participate in the Permanent Structured Cooperation (PESCO). By using the Louvain community detection algorithm on the network links that have been established through defence partnerships in PESCO projects, two different country clusters emerge. Findings suggest that military expenditures can stimulate economic growth but the effects may not be common for all Member States, which might benefit from the involvement in joint defence projects to maximize the effectiveness of their defence spending.
Acknowledgments
I am grateful to Alexandros Bechlioulis, Alexandra Kechrinioti, Claire Economidou, Vasilis Monastiriotis, and the participants at the European Security and Defence College’s (ESDC) Common Module “Budget & Finance in the EU Defence Sector” for useful comments and insights.
Disclosure Statement
No potential conflict of interest was reported by the author.
Data Availability Statement
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.
Notes
1. As military burden is referred the share of military expenditure in GDP and indicates the priority that a country attaches to the military, in other words, how much of available output is devoted to defence (Ron Citation2017)
2. The recent study of Economidou et al. (Citation2024) uses insights of network analysis and employs a network GVAR model to analyze the dynamic interdependencies between military expenditures and the real economy, highlighting how changes in one country’s military spending can influence the spending behaviors of others within the network of top military spenders.
3. For further information on timeline of the EU cooperation on security and defence visit.
https://www.consilium.europa.eu/en/policies/defence-security/defence-security-timeline/.
4. Consolidated versions of the treaties currently in force can be found here.
https://eur-lex.europa.eu/collection/eu-law/treaties/treaties-force.html.
5. PESCO’s provisions are enshrined in Article 46 of the Treaty on the European Union (TEU) and Protocol 10 on permanent structured cooperation, established by Article 42(6) TEU.
6. Council Decision (CFSP) 2017/2315 can be found here.
https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:32017D2315&rid=6#d1e32-70-1.
7. Austria, Belgium, Bulgaria, Czech Republic, Germany, Estonia, Ireland, Greece, Hungary, Spain, France, Croatia, Italy, Cyprus, Latvia, Lithuania, Netherlands, Luxembourg, Poland, Portugal, Romania, Slovenia, Slovakia, Finland and Sweden. At the end of 2022, two countries did not participate: Denmark and Malta. United Kingdom, before Brexit, didn’t participate as well.
8. According to the article 1 of the Council Decision (CFSP) 2017/2315: ‘PESCO within the Union framework is hereby established between those Member States whose military capabilities fulfill higher criteria as referred to in Article 1 of Protocol No 10, and which have made commitments to one another in this area as referred to in Article 2 of that Protocol, with a view to the most demanding missions, and contributing to the fulfillment of the Union level of ambition’.
9. ANNEX II of the Council Decision (CFSP) 2017/2315.
10. One PESCO project was officially closed on 2020 and can be found here.
https://pesco.europa.eu/project/european-union-training-mission-competence-centre/.
11. According to the article 5 of the Council Decision (CFSP) 2017/2315: ‘participating Member States taking part in a project may agree among themselves to admit other participating Member States which subsequently wish to take part in the project’.
12. For more information on Graph Theory see Easley and Kleinberg (Citation2010).
13. For more information on heuristic procedures see Pearl (Citation1984).
14. For more information see Newman and Girvan (Citation2004) and Newman (Citation2006). Modularity takes values between −1 and 1. Deviation from randomness is indicated for values different than 0, while values greater than 0.3 indicate a significant community structure.
15. Authors describe optimal moment and model selection criteria following Andrews and Lu (Citation2001) and the widely used likelihood-based selection criteria BIC, HQIC, and AIC, while the GMM selection criteria are based on the J statistic for testing over-identifying restrictions. Results are not reported for brevity and are available upon request.
16. Although the GMM estimator’s advantages in dealing with endogeneity, efficiency gains from joint estimation, and handling unobserved heterogeneity are well documented (Abrigo and Love Citation2016; Arellano and Bover Citation1995; Holtz-Eakin, Newey, and Rosen Citation1988), it is not without challenges. One of the primary limitations is the potential for instrument proliferation, which can lead to overfitting and weaken the validity of the instruments, thereby compromising the reliability of the estimates (Andrews and Lu Citation2001; Newey and Windmeijer Citation2009). Furthermore, GMM estimators may not be asymptotically efficient, and their performance is heavily dependent on the choice of the weighting matrix, which can be challenging to optimize (Hansen Citation1982). Additionally, finite sample bias is a concern, given the specific period and number of countries analyzed, potentially leading to biased estimates in small samples (Roodman Citation2009). Furthermore, the GMM estimator assumes that the panel error term is identically and normally distributed, an assumption that may not always hold in practice (Lechner and Breitung Citation1996).
17. For robustness reasons, results with a different time horizon are also presented.
18. Data for each round of PESCO projects are taken from here.
First: https://www.consilium.europa.eu/media/33065/st06393-en18-council-decision-pesco_press.pdf.
Second: https://pesco.europa.eu/wp-content/uploads/2019/05/CELEX_32018D1797_EN_TXT.pdf.
Third: https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32019D1909&from=EN.
Fourth: https://www.consilium.europa.eu/media/53013/20211115-pesco-projects-with-description.pdf.
19. For more information see Persyn and Westerlund (Citation2008). For the implementation of the cointegration tests, a bootstrap approach with 400 iterations is applied, a constant is included, and the Akaike information criterion (AIC) is used to determine the optimal lag and lead length for each separate time series. Results are similar when a trend is included.
20. In evidence of cointegration, a panel vector error correction model (VECM) should be applied. The error correction term derived from the cointegration vector would capture deviations from the long-term equilibrium and adjust the short-term dynamics towards the equilibrium path, allowing for a comprehensive analysis of both short-term fluctuations and long-term relationships.
21. Graphs of eigenvalue stability conditions are not reported for brevity and are available upon request.
22. Results are not reported for brevity and are available upon request.
23. The equation becomes: Yit = Γ0 + Γ1(L)Υit-1 + Γ2Xit + μt + γt + εit, where Xit is a dummy variable that takes the value of 1 for years 2007 to 2009 and 2020 to 2021, and 0 otherwise.
24. Menon (Citation1996) points out that ‘post-Maastricht scholarly evaluations have tended to be positive about its implications for enhanced European defence co-operation and have stressed the fact that Europe seemed to be moving towards a European defence structure’
25. This comes out from the robustness analysis that shows an insignificant association of the variables of interest for the period before the establishment of the CFSP (1970-1993) for both clusters of countries (Panels B and C).
26. Examples of collaboration for countries like Belgium, France, Germany, Luxemburg, Spain, and Italy, include the three-nation Tornado, the four-nation Eurofighter Typhoon, and the seven-nation Airbus A400M airlifter (Hartley Citation2008). Further co-operative ventures are described in Menon (Citation1996).
27. For a relevant press release see here.