ABSTRACT
Given its significant socioeconomic implications, the nexus between military expenditures and economic growth has been the subject of an extensive theoretical and empirical debate. The present study focuses on NATO vs. non-NATO member countries (i.e. 141 countries) as a case study to empirically examine the aforementioned complex relationship during the period 1992–2020. Using annual data, we employ a panel model, as well as spectral preliminary analysis that takes into account the nonlinear behaviour of our series, positive or negative causal relationships and their coherence characteristics across both time and frequency domains. Findings reported herein from an extended panel of several states suggest that military expenditure enhances economic growth under the NATO alliance case, whereas it becomes harmful for growth in the case of non-NATO alliances such as the SCO & CSTO. Our study contributes to the understanding on the crucial role of a military alliance membership in the formation of the economic growth and military spending relationship. Hence, our findings can be beneficial for policymakers to account for the existence of possible spillovers that may arise from country participation in a military alliance.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Disclaimer
The views and opinions expressed in this paper are those of the authors and do not reflect their respective institutions.
Notes
1. Appendix A provides a list of countries included in the sample.
2. Real GDP is expressed in per capita terms to account for changes in total population that may arise across the time period considered in the sample from 1992 to 2020. However, our results remain qualitative unchanged when real GDP growth is considered instead of per capita real GDP.
3. See Appendix C for a brief explanation of the three alliances.
4. The accession year differs for each member state.
5. Wavelet coherence coefficients between time series x and y only provide values between zero and one due to their inherent squared nature. As a result, we cannot differentiate between negative and positive correlations.
6. The analysis was carried out utilizing R Studio, drawing on a suite of packages to facilitate data processing, statistical modeling, and visualization.
7. To validate whether a fixed or a random effects model is more appropriate we performed the Hausman test, which is distributed as χ2. The results indicate that the random effects model is rejected at any level of significance with χ2 = 189.68 (p-value = 0.00).