Analysis of the Economic Effects of Defence Spending in Spain: A Re-Examination Through Dynamic ARDL Simulations and Kernel-Based Regularized Least Squares

ABSTRACT

Spain’s economic growth over the last few decades constitutes a rather distinctive case in post-World War II Europe, which has attracted international academic interest. This paper examines the relationship between Spain’s military expenditure and economic growth during the period 1954-2021 to shed further light. Using the recently developed method of Dynamic Autoregressive Distributive Lag (DARDL) simulation proposed by Jordan and Philips (2018) and the Kernel-Based Regularized Least Squares (KRLS) technique which accounts for potential non-linearities, interactions, and heterogeneous effects, this paper tests the short and long – run equilibrium relationship between military expenditure and economic growth in Spain. Furthermore, with the cumulative sum (CUSUM) and cumulative sum of squares (CUSUMSQ) tests, we examine the stability of the above relationship. The results show that there is a positive short and long – run relationship between military expenditure and economic growth in Spain during the period under study. However, when we stratify our sample (before and after Spain’s democratization), the positive relationship is sustained only during the years of the non – democratic regime. This has further policy implications as policymakers in Spain need to carefully balance national security concerns with the need for sustained economic growth.

KEYWORDS:

• Spain
• economic growth
• military spending
• DARDL, KRLS

JEL CLASSIFICATION:

• C22
• H30
• O40

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. As Frieden (2020, 7) puts it: ‘The economy [is all about] … high politics, and much of politics … [is] about the economy’.

2. For a detailed analysis see: Duch-Brown and Fonfría (2014).

3. Spain’s particular geographical location between Europe and North Africa and between the Mediterranean and the Atlantic and as a State that includes the mainland, the Balearic and Canary archipelagos and the two autonomous cities of Ceuta and Melilla.

4. For a detailed analysis of Spain’s global security strategies see: Government of Spain (2013).

5. The key differences between DARDL, ARDL, and Johansen’s approaches lie in their specifications of lag structures and the types of models used for testing cointegration. DARDL allows for a more flexible and dynamic lag structure, while ARDL allows for the inclusion of both stationary and nonstationary variables. Johansen’s approach involves estimating a VECM using maximum likelihood estimation, which allows for the modelling of multiple cointegrating relationships. See Jordan and Philips (2018) for a detailed discussion on DARDL vs. ARDL by Pesaran, Shin and Smith (2001) vs. Johansen (1991, 1995) approaches to cointegration testing.

6. Each autonomous community is composed of provinces, each of which has a government institution with autonomy to manage its own interests. They are: the Andalusia, Aragon, the Principality of Asturias, the Balearic Islands, the Canary Islands, Cantabria, Castilla-La Mancha, Castilla and León, Catalonia, Extremadura, Galicia, La Rioja, the Community of Madrid, the Region of Murcia, Navarre, the Basque Country and the Valencian Community (Smith and Heywood 2000).

7. See Casanova and Andrés (2014) for a detailed account of the Spanish political history.

8. WE is a population weighted average of Austria, Belgium, Denmark, Finland, France, Germany, Italy, the Netherlands, Norway, Sweden, Switzerland, and the UK.

9. For a detailed discussion regarding the differences in structural change between Spain and other WE nations from the mid-twentieth century until 2015 see: González-Díez and Moral-Benito (2019).

10. For more see De la Escosura (2000, 2007).

11. For a detailed analysis of the 1959 liberalization plan in Spain see Martínez-Ruiz and Pons (2020).

12. ^Spain suffered from high unemployment since 1994 (about 15%) reaching 20% in 2013. See more at Ortega and Peñalosa (2013).

13. Spain joined NATO in 1982 but did not align its military structure until 1999.

14. Major military transformation policies were: ‘the Law 20/1981, which reduced the number of officials, … the General Plan for the Modernization of the Army (META, Spanish acronym) in 1983, the Plan for the Reorganization of the Army (RETO) in 1990, the Plan for the New Organization of the Army (NORTE) in 1994, and … the reorganization of the army set by the Royal Order 416/2006’ Sabaté (2016, 392). Additionally, on the evolution of defence policy in Spain and its relationship with the economy, see Velarde (2000).

15. Spain 2004 defence transformation’s initiative was to be harmonized with both NATO and European Union plans and programs.

16. For a general overview of the Spanish contemporary security policy, see Maxwell (1991) and Ortega (2008).

17. See more in Fonfría (2013).

18. The so-called guns vs. butter debate.

19. The lagged values model the dynamics of the spending process to allow a hangover from the previous years’ spending (Dunne, Smith, and Willenbockel 2005, Heo 2010).

20. For a thorough discussion see Martí and Pérez (2015).

21. For a detailed analysis of the policies associated with the restructuring of the defence sector and military reforms, see Molas –Gallart (1992, 1997) and Brzoska and Lock (1992).

22. For a detailed discussion regarding the choice of the Barro-style model and variables, see: Dimitraki and Menla Ali (2015).

23. The graphical representation of our main series has been also done but not included in the paper – available upon request.

24. To select a smoother, less complicated function the KRLS method minimizes the squared loss of a Tikhonov’s regularization problem. See Hainmueller and Hazlett (2014) for a detailed discussion.

25. For comparison purposes, the ARDL model has been estimated as well, which provides similar results since similar specification is used (e.g. the DARDL estimates rely on fitting the most appropriate ARDL model – For more see Jordan and Philips 2018). However, DARDL is preferred, in the current paper as it allows for a more dynamic adjustment process. The results are not reported in this paper but are available upon request.

26. Data are available at https://www.sipri.org/databases/armstransfers.

27. For a detailed analysis see Sabaté (2016).

28. Wang, Hou, and Chen (2022) use the variable Demo from the Polity IV dataset which provides levels of democratic development via regime classification, ranging from − 10 (hereditary monarchy) to 10 (consolidated democracy).