"You don't pay your bills you get no protection": A Trump effect on NATO members' military expenditures?

Founta, Konstantina; Kollias, Chrēstos; Tzeremes, Panayiotis Article "You don't pay your bills you get no protection": A Trump effect on NATO members' military expenditures? Peace Economics, Peace Science and Public Policy (PEPS) Provided in Cooperation with: De Gruyter Brill Suggested Citation: Founta, Konstantina; Kollias, Chrēstos; Tzeremes, Panayiotis (2025) : "You don't pay your bills you get no protection": A Trump effect on NATO members' military expenditures?, Peace Economics, Peace Science and Public Policy (PEPS), ISSN 1554-8597, De Gruyter, Berlin, Vol. 31, Iss. 2, pp. 145-160, https://doi.org/10.1515/peps-2025-0024 This Version is available at: https://hdl.handle.net/10419/333341 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Konstantina Founta, Christos Kollias* and Panayiotis Tzeremes “You don’t Pay Your Bills You Get No Protection”: A Trump Effect on NATO Members’Military Expenditures? https://doi.org/10.1515/peps-2025-0024 Received March 18, 2025; accepted March 28, 2025; published online April 16, 2025 Abstract: Dispensing with the customary courteous diplomatic language, much of the 47th President’s rhetoric concerning the other NATO Allies echoes the valid US criticism that many of them freeride on the US protective defence umbrella without contributing their fair share to the costs of NATO’s collective defence. “You don’t pay your bills, you get no protection…”, he is quoted saying in a pre-election speech while is his Davos 2025 address, he mentioned, referring to the 2 % NATO guideline and the effect of his first presidential term, that “…most nations didn’t pay until I came along”. Using such statements as an impetus, the paper sets out to examine whether a Trump effect of his first presidency on NATO members’military spending is empirically traceable. Results reported herein seem to offer evidence in favor of a tentative Trump effect inference. Keywords: NATO; military spending; Trump; MDS; K-Means; structural breaks 1 Introduction January 20, 2025, marked the beginning of Donald Trump’s second term in the White House’s Oval Office. In an interview a few weeks before his inauguration as the 47th President of the USA he stated that he would “absolutely”consider taking the United States out of the Alliance unless members are “treating us fairly”, that is if allies “pay their bills”. 1 He was referring to the lower defence burdens of most NATO *Corresponding author: Christos Kollias, Laboratory of Economic Policy & Strategic Planning, Department of Economics, University of Thessaly, Volos, Greece, E-mail: [email protected]. https://orcid.org/ 0000-0003-2876-4304 Konstantina Founta, Laboratory of Economic Policy & Strategic Planning, Department of Economics, University of Thessaly, Volos, Greece Panayiotis Tzeremes, Department of Accounting and Finance, University of Thessaly, Volos, Greece 1https://www.barrons.com/news/trump-says-will-absolutely-consider-leaving-nato-if-us-nottreated-fairly-9d3163cf (accessed 4 January 2025). Peace Econ. Peace Sci. Pub. Pol. 2025; 31(2): 145–160 Open Access. © 2025 the author(s), published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0 International License. members compared to the US defence budget. Indeed, in his first speech to an international audience at the Davos 2025 World Economic Forum Annual Meeting, he stressed that he intended to ask allies to increase their defence spending to 5 % of GDP. 2 Around a year earlier, in February 2024, before his official presidential nomination at the Republican National Convention in Milwaukee in July 2024, he is quoted saying during his speech at a campaign event in South Carolina in a rather blatant and diplomatically unbecoming manner that if “You don’t pay your bills, you get no protection. It’s very simple”. 3 In an interview, once again dispensing with the usual diplomatic tactfulness towards the other NATO allies, he was more explicit saying that in the cases of NATO members that did not contribute enough funds towards the Alliance’s collective defence he would “encourage”Russia “to do whatever the hell they want”, indicating that these Allies should not rely on US supporting them militarily. 4 Even though such statements are plainly devoid of the expected typical courteous diplomatic language especially when addressed to allies, they nonetheless echo the long-standing, and for all intents and purposes valid US criticism that many of their allies freeride on the US protective defence umbrella without contributing their fair share to the costs of NATO’s collective defence. The issue of burden sharing between NATO’s members is by no means a new one. Hence the attention it has received in the relevant literature that has been expanding steadily over the years. Recent papers on the theme of burden sharing and interdependency among Western Allies include George and Sandler (2024, 2022), Kivimäki (2019), Pedersen, Ohrt, and Svendsen (2023), Kim and Sandler (2020), Mombelli (2024), Caruso and Di Domizio (2016), Sandler and Shimizu (2014), Zyla (2016), Becker (2017). Random examples of more vintage papers on this theme are those by Sandler and Hartley (2001), Hartley and Sandler (1999), Oneal (1990), Solomon (2004), Murdoch and Sandler (1984), Sandler and Forbes (1980). A recent systematic and comprehensive review of this literature can be found in Bogers, Beeres, and Bollen (2022) and in George and Sandler (2024, 2022). Consequently, for reasons of brevity, we refrain from engaging in a fundamentally similar discussion here. Moreover, the aim of the present paper is not to engage with the burden sharing issue per se. During his first Presidential term, Trump did not spare his stern criticism towards other NATO members, especially towards major European countries, for not shouldering the costs of their defence. For example, during the 2018 Brussels NATO 2https://www.euronews.com/business/2025/01/23/trump-at-davos-nato-5-push-tariff-warnings-foreurope (accessed 23 January 2025). 3https://abcnews.go.com/Politics/trump-russia-nato-defense-funds/story?id=107136736 (accessed 23 January 2025). 4ibid. 146 K. Founta et al. Summit it was reported that he gave something of an ultimatum regarding allies’ defence spending: “He said they must raise spending by January 2019, or the United States would go it alone”. 5 A statement very much contextually similar to the recent one cited above concerning a possible US withdrawal from NATO. In a similar vein, at the outset of that Summit he is quoted calling other NATO members delinquent in their defence expenditures and demanded that they should increase their defence budgets immediately, 6 while he has also asked European countries to buy American weapons to keep NATO strong. 7 Foreign sales are a very important source of revenues for the US defence industry, a substantial share of which are exports to other NATO members. For instance, during 2000–2024 US manufactured weapon systems imported by European NATO Allies accounted for almost a quarter of the US industry’s total exports 8 while, the revenues of the US defence industries are also affected by the US political cycles (Balestra and Caruso 2025) as well as by foreign interventions (Baronchelli and Caruso 2024). In his Davos 2025 speech when calling for a 5 % of GDP allocation to defence, President Trump also referred to his first term in office and the effect it had on allies’ military spending: “I’m also going to ask all NATO nations to increase defence spending to 5 % of GDP, which is what it should have been years ago. It was only at 2 %, and most nations didn’t pay until I came along”. 9 In what follows we examine the impact exerted by the first Trump Presidency on NATO Allies’military spending, trying to determine whether the bragged Trump effect cited above is in fact empirically traceable. The section that follows contains a brief overview of the empirical methodologies used to probe into the issue at hand. The findings are reported and discussed in sections three and four while section five concludes the paper. 2 A Bird’s Eye View of the Data and Methodologies Two metrics are used to examine empirically whether a Trump effect is identifiable on NATO members’allocations to defence: military spending as a share of GDP and 5https://www.reuters.com/article/world/trump-warned-nato-allies-us-would-go-it-alone-if-theydid-not-spend-sources-idUSKBN1K21KY/(accessed 26 January 2025). 6https://edition.cnn.com/2018/07/10/politics/donald-trump-nato-summit-2018/index.html (accessed 26 January 2025). 7https://www.bloomberg.com/news/articles/2025-02-13/trump-tells-europe-to-buy-americanweapons-to-keep-nato-strong. 8Data calculated from SIPRI’s Arms Transfers Database. 9https://www.weforum.org/stories/2025/01/davos-2025-special-address-donald-trump-presidentunited-states/(accessed 31 January 2025). You don’t Pay Your Bills You Get No Protection 147 equipment expenditure as a share in total defence spending. Data for the former are drawn from SIPRI’s database and for the latter from NATO. Since the 2014 NATO Wales Summit, both are the metrics used to determine the minimum thresholds that Alliance members should adhere to in order to meet their obligation to the quantity of resources allocated towards their military capabilities (inter alia: Kim, Sandler, and Shimizu 2024; Haesebrouck 2022; Blum and Potrafke 2020; Arvanitidis, Kollias, and Messis 2017). To probe into the Trump effect hypothesis, these two metrics are compared and analyzed during two Presidential periods: the Obama two term presidency (2009–2016) and the first Trump presidency. To allow for a possible hysteresis in a Trump effect, the first year of the Biden administration is included in the later period. Hence the second period in the analyses and empirical tests that follow covers the years 2017–2021. Albeit data is available for more years, we opt to have 2021 as the cut-offpoint in the series given the profound effect the full-scale Russian invasion of Ukraine in February 2022 had on NATO defence planning and budgets (inter alia: Buts, Christie, and Du Bois 2024; Magula, Rouland, and Zwack 2022). All current NATO members 10 are included in the empirical analyses conducted herein. However, due to data availability constraints in the case of the second metric, i.e. the share of equipment expenditure in total military spending, Montenegro, North Macedonia, Finland and Sweden, are not included in the sample and the estimations. Using boxplots, Figures 1 and 2 offer a bird’s eye view of the data during the two periods mentioned above. That is, the two Obama presidential terms (2009– 2016) and the first Trump presidency wherein the first year of the Biden administration that followed is included (2017–2021) as previously pointed out. Figure 1: Military spending as a share of GDP during the Obama and trump Presidencies. The military spending as a share of GDP data is drawn from SIPRI’s military expenditure databaseen. 10 Although Finland and Sweden joined NATO in 2023 and 2024 respectively, we opted to include them in the analysis. 148 K. Founta et al. For the first of the two metrics, the median value for the entire first period was 1.33 % and for the second period it was 1.43 % while the corresponding mean values of defence spending as a share of GDP were 1.49 % and 1.59 % respectively. During both periods the average defence burden of the USA was the highest among all the NATO members standing at 4.2 % during 2009–2016 and 3.4 % during 2017–2021. The second and third largest defence spending to GDP ratios were those of Greece (2.6 %) and the UK (2.3 %) during the Obama administration period. The average ratios for major European powers such as Germany, France, Italy and Spain were 1.2 %, 1.9 %, 1.4 %, 1.3 % respectively during the first period depicted in Figure 1. During the second period (2017–2021), the second and third largest defence spending to GDP ratios were those of Greece (3 %) and Türkiye (2.3 %) while the corresponding averages for the largest European economies were: Germany 1.3 %, UK 2 %, France 1.9 %, Italy 1.5 % and Spain 1.3 %. Turning to the second metric, i.e. equipment expenditure as a share of total military spending 11 depicted in the boxplots of Figure 2, the median value for the entire first period was 13.48 % and for the second it was 19.8 % while the corresponding mean values were 14.45 % and 21.52 % respectively. During the two Obama presidencies period, the highest average share of equipment expenditure was France’s 27.3 % followed by the USA (25.5 %) and Türkiye (25.3 %). The corresponding averages for the major European powers were as follows: Germany 14.7 %, UK 22 %, Italy 11.9 % and Spain 13.3 %. During the second period depicted in Figure 2, Türkiye (32.5 %), Lithuania (32.4 %), Hungary (29.9 %), Slovakia (28.9 %) and the USA (28 %) emerge as the five NATO members with the highest shares in equipment expenditure, well above the 2014 Wales Summit target. The corresponding averages during Figure 2: Defence equipment expenditure as a share of total military spending during the Obama and trump Presidencies. The defence equipment expenditure as a share of total military spending data is drawn from NATO’s databaseen. 11 As previously mentioned, Finland, Sweden, Montenegro and North Macedonia are not included due to data availability constraints. You don’t Pay Your Bills You Get No Protection 149 2017–2021 for the major European powers were: France 25.4 %, UK 24.1 %, Spain 21 %, Italy 19.7 % and Germany 14.6 %. To probe into the issue at hand, a battery of methodologies is applied. The empirical analysis that follows in the next section starts with the application of a multidimensional scaling (MDS) approach to map the relative position of NATO members for each of the two metrics used herein. Briefly, MDS is a statistical technique used to visualize the similarity or dissimilarity between data points in a lowerdimensional space. In our case here, in terms of military spending as a share of GDP and equipment expenditure’s share in the NATO members’annual defence budgets. The fundamental idea behind MDS is to represent objects as points in a Euclidean space such that the pairwise distances between them approximate their original dissimilarities. In this case study, we use “Classical MDS”, a type of MDS where the input is a matrix of Euclidean distances and eigenvalue decomposition is performed on a transformed distance matrix (Pawliczek and Dzwinel 2007). The process of MDS begins with a distance or dissimilarity matrix, where each element represents the perceived difference between a pair of objects. In classical MDS, the dissimilarity matrix is converted into a squared distance matrix, followed by double-centering to compute a matrix suitable for eigen decomposition. The largest eigenvalues and their corresponding eigenvectors are then extracted to determine the coordinates in the lower-dimensional space (Pawliczek and Dzwinel 2007). In multidimensional scaling the subject of investigation is data space RNdata vectors Xmapped to the destination space, usually Y∈R2. The analyzed distances are R ij =D(Xi,Xj) between Xiand Xjin RNand the distances r ij =d(Yi,Yj)inR2. MDS aims to find a map X→Y=M(X), minimizing global measures of the topographic fit, i.e. the difference between R ij and r ij (Naud 2006). According to Kruskal (1964), a critical component of MDS is the stress value, which quantifies the discrepancy between the observed distances in the original space and the distances in the MDSgenerated space. A lower stress value indicates a better representation of the original data. The most commonly used measure of stress is Kruskal’s stress formula, which is defined as the square root of the sum of squared differences between the original dissimilarities and the fitted distances, normalized by the sum of squared original dissimilarities. Mathematically, it is expressed as: stress =S= ∑i<jdij −δij () 2 ∑i<jd2 ij √ √ where d ij represents the distances in the lower-dimensional space, and δ ij represents the original dissimilarities. A lower stress value indicates a better fit between the 150 K. Founta et al. transformed and original data. The distances are calculated by Euclidian distance formula: dij =xi1−xj1 () 2+xi2−xj2 () 2+…+xin −xjn () 2 {} 1 / 2 Interpreting stress values involves general guidelines, with lower values signifying better representations. A stress value below 0.05 is considered an excellent fit, between 0.05 and 0.1 is an acceptable fit, between 0.1 and 0.2 suggests a fair representation, and values above 0.2 indicate significant distortion, meaning the MDS solution may not be reliable (Naud 2006). In practical applications, minimizing stress is a crucial step in MDS analysis. This is typically achieved through iterative optimization techniques. Despite its importance, stress value should not be the sole criterion for evaluating MDS solutions. Additional diagnostic tools, such as Shepard diagrams, which plot the relationship between original dissimilarities and distances in the MDS space, can provide deeper insights into the quality of the representation (Naud 2006). In summary, stress value serves as a key metric in MDS to assess the accuracy of lower-dimensional embeddings. A lower stress value is desirable as it indicates a more faithful representation of the original data relationships. Then, as the second step in the empirical analysis we opt to employ the K-means clustering technique. Briefly, K-Means analysis is a widely used clustering technique that organizes data into distinct groups based on similarity. K-Means clustering is based on an algorithm that is used to partition data into k clusters, where each data point belongs to the cluster with the nearest mean. The algorithm aims to minimize the within-cluster variance, ensuring that points within the same cluster are more similar to each other than to points in other clusters (Patel et al. 2022). The K-Means process starts by selecting k initial cluster centroids, which can be randomly assigned or chosen using specific initialization methods, i.e. K-Means++. Each data point is assigned to the nearest centroid based on Euclidean distance, forming k clusters. Once all points are assigned, the centroids are recalculated as the mean of the points in each cluster. The process iterates, reassigning points to the closest centroid and updating centroids until convergence is reached. Convergence occurs when the cluster assignments no longer change, or a predefined stopping criterion is met. To improve stability, k-means is often run multiple times with different initializations, and the best result is chosen based on the lowest total within-cluster variance (Leonov et al. 2019). Both MDS and K-Means serve critical roles in exploratory data analysis. MDS provides a meaningful low-dimensional representation of complex data relationships, aiding in pattern recognition and visualization. K-Means, on the other hand, is effective in segmenting data into meaningful groups, making it a popular choice for You don’t Pay Your Bills You Get No Protection 151 applications such as pattern recognition and anomaly detection. The two techniques can be used together, where MDS reduces dimensionality before applying k-Means clustering to enhance interpretability and efficiency. As a final step in the empirical analysis, we opt to examine for the possible presence of structural breaks in the time series of the two variables for each NATO member. Structural breaks in economic variables are invariably the outcome of significant changes, domestic and/or external, that exert a statistically traceable effect on the variables concerned. Clearly, in the case of military expenditures these include geopolitical developments that impose rapid adjustments to such public outlays, an unexpected major change in the global or regional strategic environment, a major national security related event, a shift in defence policy due to a governmental change, a fiscal crisis such as the recent sovereign debt crisis that results in significant reductions and cut backs in public expenditures. Evidently, the Russian invasion of Ukraine as well as the shockwaves of the new Trump presidency are the type of geopolitical developments that leave an indelible mark and cause rapid adjustments to defence expenditures. Building on the methodology of Bai and Perron (1998, 2003), we employ a multiple structural break model that does not require any predefined breakpoints. Instead, potential break dates are estimated directly from the data. Each series can be modeled separately with mbreaks (leading to m+1 distinct segments) using an equation of the form: Φτ=Ψj+ωτ,τ=Tj −1+1,Tj −1+2,…,Tj, where Φτ is the military expenditure as a share of GDP (or equipment share in total defense spending) at time τ,j= 1,2,…,m+1,Ψjis the mean level of Φτ in the j-th segment, and ωτ captures any additional disturbances, including serial correlation or heteroscedasticity. The unknown breakpoints (T1,T2,…,Tm) are found by minimizing a chosen criterion (e.g., the Bayesian Information Criterion, BIC, or the residual sum of squares, RSS). 3 The Findings from MDS Mapping and K-Means Clustering We begin the presentation and discussion of the findings yielded from the empirical analysis with the results from applying the MDS and K-Means methodologies to the two metrics used herein to probe into the question at hand. These are shown in Figure 3 in the case of the military expenditure to GDP ratio variable and in Figure 4 for the equipment percentage share in total military spending metric. In each figure the corresponding maps and clusters for each of the two periods are included. 152 K. Founta et al. Bogers, M., R. Beeres, and M. Bollen. 2022. “NATO Burden Sharing Research along Three Paradigms.” Defence and Peace Economics 33 (5): 534–47. Buts, C., E. H. Christie, and C. Du Bois. 2024. “Exploring the Speed of Change: European Defence Budgets in an Age of Danger.”Defense and Security Analysis 40 (3): 319–40. Caruso, R., and M. Di Domizio. 2016. “Interdependence between US and European Military Spending: A Panel Cointegration Analysis (1988–2013).”Applied Economics Letters 23 (4): 302–5. George, J., and T. Sandler. 2022. “NATO Defense Demand, Free Riding, and the Russo-Ukrainian War in 2022.”Journal of Industrial and Business Economics 49: 783–806. George, J., and T. Sandler. 2024. “A Spatial Analysis of NATO Burden Sharing at the Operational Levels.” Kyklos 77 (4): 1026–47. Haesebrouck, T. 2022. “NATO Burden Sharing after the Wales Summit: A Generalized Set Qualitative Analysis.”Defence and Peace Economics 33 (6): 637–54. Hartley, K., and T. Sandler. 1999. “NATO Burden-Sharing: Past and Future.”Journal of Peace Research 36 (6): 665–80. Kim, W., T. Sandler, and H. Shimizu. 2024. “An Expanded Investigation of Alliance Security Free Riding.” Global Policy 15 (4): 570–82. Kim, W., and T. Sandler. 2020. “NATO at 70: Pledges, Free Riding, and Benefit-Burden Concordance.” Defence and Peace Economics 31 (4): 400–13. Kivimäki, T. 2019. “Power, Contribution and Dependence in NATO Burden Sharing.”European Security 28 (1): 66–84. Leonov, P., V. Suyts, O. Kotelyanets, and N. Ivanov. 2019. “K-means Method as a Tool of Big Data Analysis in Risk-Oriented Audit.”In Big Data Innovations And Applications, edited by M. Younas, I. Awan, and S. Benbernou, 206–16. Cham, Switzerland: Springer. Magula, J., M. Rouland, and P. Zwack. 2022. “NATO and Russia: Defense and Deterrence in a Time of Conflict.”Defence Studies 22 (3): 502–9. Mombelli, S. 2024. “Defense Burden Sharing and Military Cooperation in the EU27: A Descriptive Analysis (2002–2023).”Peace Economics, Peace Science and Public Policy 30 (4): 423–53. Murdoch, J., and T. Sandler. 1984. “Complementarity, Free Riding, and the Military Expenditures of NATO Allies.”Journal of Public Economics 25 (1–2): 83–101. Naud, A. 2006. “An Accurate MDS-Based Algorithm for the Visualization of Large Multidimensional Datasets.”In Artificial Intelligence and Soft Computing –ICAISC 2006, 643–52. Berlin, Heidelberg: Springer. Oneal, J. R. 1990. “The Theory of Collective Action and Burden-Sharing in NATO.”International Organization 44 (3): 379–402. Patel, S. S., S. S. Patel, N. Kumar, N. Kumar, J. Aswathy, J. Aswathy, S. K. Vaddadi, et al. 2022. “K-Means Algorithm: An Unsupervised Clustering Approach Using Various Similarity/Dissimilarity Measures.” In Intelligent Sustainable Systems, edited by J. Raj, R. Palanisamy, I. Perikos, and Y. Shi, 805–13. Singapore: Springer. Pawliczek, P., and W. Dzwinel. 2007. “Visual Analysis of Multidimensional Data Using Fast MDS Algorithm.” Photonics Applications in Astronomy, Communications, Industry, and High-Energy Physics Experiments 6937 (2M): 755–65. Pedersen, R., A. Ohrt, and G. Svendsen. 2023. “Free Riding in NATO after the Rise of Russia: Cost Sharing, Free Riding and Selective Incentives in NATO from 2009 to 2019.”Journal of Transatlantic Studies 21 (1): 54–72. Sandler, T., and J. F. Forbes. 1980. “Burden Sharing, Strategy, and the Design of NATO.”Economic Inquiry 18 (3): 425–44. You don’t Pay Your Bills You Get No Protection 159 Sandler, T., and K. Hartley. 2001. “Economics of Alliances: The Lessons for Collective Action.”Journal of Economic Literature 39 (3): 869–96. Sandler, T., and H. Shimizu. 2014. “NATO Burden-Sharing 1999–2010: An Altered Alliance.”Foreign Policy Analysis 10 (1): 43–60. Solomon, B. 2004. “NATO Burden Sharing Revisited.”Defence and Peace Economics 15 (3): 251–8. Zyla, B. 2016. “Who Is Keeping the Peace and Who Is Free-Riding? NATO Middle Powers and Burden Sharing, 1995–2001.”International Politics 53 (3): 303–23. 160 K. Founta et al.