NATO DEFENSE EXPENDITURES: COMMON GOALS OR DIVERGING INTERESTS? A STRUCTURAL ANALYSIS

Abstract

By testing for structural breaks in defense expenditures, the dates of change in the pattern of defense expenditures for the NATO allies are determined. If NATO members are responding to a common threat, the breaks should be similar, in both direction and dates, for defense expenditures. The breaks should occur during major NATO strategy shifts. The results of the structural analysis tests suggest that NATO allies do not have an integrated response to NATO‐specific defense issues. It appears that NATO members, in general, adjusted their defense spending according to economic imperatives, political issues, and ally‐specific defense agendas.

Keywords:

• Defense expenditures
• NATO
• Multiple structural changes
• Defense burden

JEL Codes:

• C22
• H56
• H77

Notes

¹ Iceland maintains no military but its strategic location allows it to host NATO bases.

² Dunne et al. (2005) and Yakovlev (2007).

³ Dates and events are summarized from NATO updates available online at http://www.nato.int/docu/update/index.htm.

⁴ US Congressional Budget Office (2000).

⁵ US Congressional Budget Office (1996, 2000).

⁶ Hartley and Sandler (1999) discuss developments in NATO’s burden sharing debate resulting from the changes brought about by NATO’s enlargement and changing mission.

⁷ Article 5 of North Atlantic Treaty of 1949.

⁸ See Khanna and Sandler (1996, 1997) and Sandler and Murdoch (2000) for a discussion on NATO doctrine.

⁹ The Bai and Perron formulation allows for general forms of serial correlation and heteroskedasticity in the errors. It also allows lagged dependent variables, trending regressors as well as different distributions for the errors and the regressors across data segments. For additional details on the test, please refer to Appendix. A.

¹⁰ Our interest is the presence of a structural change in the mean of the series. As such, we apply the following multiple linear regression with m breaks and m+1 regimes for a sample of T observations:

for j = 1,…, m + 1 where y_t is the observed dependent variable at time t; x_t (p × 1) and z_t (q × 1) are vectors of covariates and β and δ_j are the corresponding vectors of coefficients; µ_t is the disturbance at time t. The break dates occur at {T₁,…,T_m }. For this study, z_t = {1} and q=1, for the case of changes in the mean of a series and p=0 for a pure structural change model where all the coefficients are subject to change.

¹¹ The upper bound of breaks is set equal to that specified for the sequential procedure.

¹² Bai and Peron (1998) show in simulation experiments that a small value for trimming such as 5% can lead to tests with substantial size distortions when sample size is small and the variances of errors across the data segments are different or when serial correlation is permitted. This is because the researcher is trying to estimate quantities using very few observations and could end up estimating for some segments, values such as the variance of the residuals, using only several observations. Bai and Perron recommend trimming values of 15% or higher in this case and a maximum number of breaks based on the trimming value, five breaks for a trimming value of 15%. The maximum number of breaks is set to int[1/ε] − 2 breaks where ε is the trimming size.

¹³ In this study, defense expenditures serve as a proxy for national defense. Testing for changes in NATO allies’ defense behavior by analyzing structural breaks in defense expenditures is one way of explaining collective action and does not necessarily reflect internal reconfiguration of forces in response to doctrinal changes.

¹⁴ It is difficult to predict the speed by which defense expenditures react to changes in doctrine due to budget cycles, political systems, and domestic considerations. A country may anticipate a doctrine change and modify its expenditures preceding a formal pronouncement of change in response to internal domestic pressures such as the expectation of a peace dividend following the end of the Cold War. In addition, the change in expenditures depends on the time required for the implementation of budget cycles. NATO members such as Greece, Turkey, and Portugal, which suffered a great deal of internal turmoil and had authoritarian regimes, may have responded rather quickly to doctrine changes before the transition to democratic rule, as opposed to countries where budget modifications are subject to an established political process. For the purpose of this study, the direction of the change (increase or decrease) and timeframe of several years preceding and following a doctrine change are recognized as potentially resulting from NATO doctrine change.

¹⁵ Devarajan et al. (1996) include a detailed discussion of the effects of decentralization on defense expenditures.

¹⁶ See Kollias (1996), Brauer (2002), and Sezgin and Yildirim (2002) for a discussion regarding the level of Turkey defense expenditures over time.

¹⁷ Less than 24 hours after the attack on September 11, 2001, NATO invoked Article 5 for the first time. However, this was seen as a gesture of political support rather than a call for the allies to take collective military action. The alliance’s military role and commitment has been minimal so far. CRS (2006) reports that because the US and NATO have greater unity of purpose in Afghanistan than other regions, the mission in Afghanistan is seen as a test of the alliance. However, it appears that the allies are not shouldering their share of the burden of military operations in Afghanistan. The report refers to an imposition of national caveats that restrict the use of forces, the crisis between the allies regarding the observation of international law, and the reluctance of the allies to provide troops and resources. The report concludes by stating that the US forces bear a disproportionate share of the most dangerous tasks.

¹⁸ The data used in this study do not extend beyond 2004. It is not yet apparent what the effects of the commitments by NATO allies to the wars in Iraq and Afghanistan may be.

¹⁹ http://www.nytimes.com/2007/04/27/world/europe/27russia.html?ref=europe.

²⁰ Thanks to Col. Michael Meese and an anonymous referee for bringing this explanation to my attention.