Military spending and budget deficits: the impact of US military spending on public debt in Europe (1988–2013)

Abstract

The aim of this paper is to study the relationship between military spending and sovereign debt in a panel of 13 European countries. In particular, under the assumption of the interdependence of military spending between US and European countries, we analyse whether US military spending affected European sovereign debt in the period 1988–2013. The empirical estimation is based on different steps: (i) a unit root test; (ii) an Arellano–Bond panel estimation and a linear fixed effect model; and (iii) a FMOLS estimation to highlight the long run relationship between debt and relevant variables. General results highlight that debt burden of European countries is: (1) positively associated with US military burden and (2) negatively associated with average military burden of other European countries.

Keywords:

• Military spending
• Public debt
• Interdependence
• Arellano Bond
• Fixed effect
• Panel cointegration
• Fully modified OLS

Jel Codes:

• H56
• H63
• F52

Acknowledgements

Earlier version of this paper has been presented at the 2015 ASSA/AEA conference, at the workwhop ‘Costs of Destructive Entrepreneurship and Economic Performance’, Maison des Sciences Economiques, Paris and in a seminar at the National Bank of Poland. We thank all participants in these events and in particular Joanna Tyrowicz, Jan Hagemejer, Julien Malizard, Solomon Polachek, Zbigniew Polanski, Stergios Skaperdas, Carlos Seigile and Merdhad Wahabi. The authors also warmly thank Ron Smith, Javier Gardeazabal and Vincenzo Bove and an anonymous referee for helpful and valuable comments. This paper extends the results of Caruso and Di Domizio (2015).

Notes

¹ See the next section for a brief survey.

² We are not considering endogenous explanations for military spending, namely: the impact of military industrial complex and associated sliding door hypothesis. See Duncan and Coyne (2015).

³ Albeit indirectly, this could be considered also a test of the Hegemonic Stability theory which has been the paramount approach among political scientists.

⁴ Kindleberger (1991), p. 152.

⁵ On convergence of military spending, see also Caruso and Di Domizio (2016), Arvanitidis, Kollias, and Anastosopoulos (2014). On world military expenditures, see also Arvanitidis and Kollias (2016).

⁶ Other NATO members not included in the panel are: Luxembourg, the Netherlands and Turkey.

⁷ For some countries, (Denmark [1988–1992], Germany [1988–1991], Portugal [1988–1990], and Sweden [1988–1993]), data from IMF (2013)were supplemented by OECD information in order to have a balanced panel.

⁸ The first is a cointegration test based on residuals of a spurious regression carried out on I(1) variables, and it allows for heterogeneous intercepts and time trend among cross-sectional units. Pedroni proposed several methods of constructing the statistics to test the null hypothesis of no cointegration; in particular, he proposed two alternative hypotheses: the ‘within’ and ‘between’ dimensions. In the within dimension, the coefficient of the first lag of the residual is imposed to be equal (and <1) for all cross sections, while in the between dimension, the only restriction is that all the coefficients of the first lag of the residual are <1. The Kao (1999)Cointegration Test is based on the same approach as in Pedroni, but it considers countries’ specific intercepts and homogeneous coefficients on the first-stage regressors.

⁹ Drawn at the website http://www.ipw.unibe.ch/content/team/klaus_armingeon/comparative_political_data_sets/index_ger.html (last access December 2014).

¹⁰ The estimation has been run by means of Eviews 8.

¹¹ It ought to be noted that coefficients of mentioned estimations differ in magnitude. In this respect, it is clear that the long run effect of the Arellano–Bond estimation has been computed through an elaboration based on short-term coefficients, whereas the coefficient presented in Table VI descends from a punctual long run estimation. The latter – plausibly – does also capture general and latent economic dynamics which are removed from the Arellano–Bond estimated coefficients.