NATO Security Burden Sharing, 1991–2020

ABSTRACT

In contrast to much of the extant literature, the paper devises a composite security burden measure for the NATO alliance that accounts for three different contributions by allies to their collective security: namely, military expenditure (ME), foreign assistance, and UN peacekeeping spending. Generally, NATO defense burden sharing and free riding are judged solely based on ME even though foreign assistance and peacekeeping promote world prosperity, stabilize regimes, and quell conflicts that affect NATO’s collective security. Our parametric tests for free riding apply a spatial-lag panel model, which addresses the interdependency issues, to a broader security-spending measure that accounts for allies’ membership, contiguity, and inverse distance. In all spatial models, we uncover robust evidence of free riding where allies decrease their aggregate security spending in response to increases in the collective security spending of other allies. We apply a panel generalized method of moments (GMM) estimator to adjust for endogeneity concerns.

KEYWORDS:

• NATO
• Security burden sharing
• Peacekeeping and foreign assistance
• Spatial econometrics
• Endogeneity

JEL CLASSIFICATION:

• D74
• H56
• C21

Acknowledgments

We thank two anonymous reviewers and editor Khusrav Gaibulloev for helpful comments on an earlier draft. Full responsibility rests with the authors.

Disclosure statement

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

Notes

1. To include non-UN troop contributions as a fourth security activity, a researcher must devise a means for converting each ally’s contributed non-UN troops into an expenditure measure accounting for the value of deployed troops, which differs among allies greatly according to training and human capital investments (see Gaibulloev, Sandler, and Shimizu 2009). Given that the United States and the United Kingdom contribute the most non-UN troops, the inclusion of this fourth measure in future research will not overturn qualitatively the free-riding results found later in the paper.

2. At a later point, we choose not to use a fiscal-constraint control (e.g. debt share of GDP) other than GDP per capita since debt constraints are not germane to the United States, the United Kingdom, Canada, or other non-EU members of NATO. Also, such a fiscal constraint is generally correlated with GDP controls, which is often ignored in the literature.

3. The transnational terrorism coefficient does not appear in Table 3 for Model 1 to be significant at the 0.10 level because of rounding. The actual coefficient is 0.00139 with a standard error of 0.00083 so that the p-value is 0.095 and marginally significant.