Cross–Country Dependence, Heterogeneity and the Growth Effects of Military Spending

ABSTRACT

Until today, the great majority of the relevant literature is dominated by growth models that assume common structural parameters and Total Factor Productivity (TFP) evolution across countries. In the context of the modified Solow convergence equation, the role of the present paper is to highlight the importance of incorporating cross–country differences, both in observables and unobservables, in the investigation of the defense–growth nexus. Special attention is paid to the heterogeneity in TFP growth and the cross–country dependence induced by global shocks. Furthermore, it analyzes the econometric issues involved by providing a brief review of various heterogeneous dynamic panel estimators. The presented estimators are then applied to two alternative panel specifications of the Solow model. Overall, the empirical application revealed that the Dynamic Common Correlated Effects (DCCE) estimator seems to be the most reliable option among the various dynamic panel estimators employed.

KEYWORDS:

• Defense spending
• economic growth
• dynamic panel estimation

JEL CLASSIFICATION:

• C23
• H56
• O23

Acknowledgments

We are grateful to an anonymous referee for valuable comments and suggestions.

Disclosure Statement

No potential conflict of interest is reported by the authors.

Notes

1. Yildirim and Öcal (2016) addressed the issue of cross–country dependence in the augmented Solow model via the inclusion of spatial spillovers. This approach, however, focuses only on ‘weak’ factors, thus omitting global shocks.

2. The average reduction in military spending shares during the period 2010–2013 amounted to – 7.4%, – 4.4%, – 2.5% and – 2,9% in the cases of Greece, Ireland, Italy and Cyprus, respectively.

3. The defense burden of the NATO allies reached, on average, 4.3% during the period 1955–1959.

4. The amount of global defense outlays for 2019 was estimated to have been $1,922 billion or 2.2% of Gross Domestic Product, 9.4% higher than in 2009.

5. For a detailed discussion on that topic, see Eberhardt and Teal (2011).

6. In the constructed panel, the average number of time observations is approximately equal to 34.