The Defense Investment Pledge agreed to by NATO allies in 2014 is reaching its decade-long finish line. The Alliance’s own data indicate that not all allies will cross that line, as many still spend less than the equivalent of 2 percent of their gross domestic products on defense and several still devote less than 20 percent of their defense budgets to acquisition and related research and development. Nonetheless, some allies like the United States are advocating to increase the 2 percent target. This is sure to run into resistance. How can the United States and like-minded allies successfully negotiate higher targets? They might start by agreeing to portray NATO burden- and risk-sharing more accurately. Although some argue that inputs like defense spending tell us a lot about outputs like contributions to Alliance operations, recently available data indicate this is not necessarily the case: New statistical analysis shows that whether or not a country has met the 2 percent spending target doesn’t tell us whether or not they’re contributing equally to the Alliance’s mission. If burden- and risk-sharing could be portrayed more accurately, those opposed to increasing the input targets might be more willing to reconsider. Even if they do not, improving how NATO depicts burden- and risk-sharing would benefit lawmakers, analysts, academics, and the public. Recommendations on how to achieve this follow the statistical analysis.
NATO has been negotiating new defense spending targets.1“NATO Countries to Discuss Defence Spending Target—Stoltenberg,” Reuters, January 3, 2023, www.reuters.com/world/europe/nato-countries-discuss-defence-spending-target-stoltenberg-2023-01-03. In 2014, the allies agreed to a Defense Investment Pledge (DIP) comprised of two related burden-sharing commitments. The DIP committed each ally to work toward spending the equivalent of 2 percent of their gross domestic products (GDPs) on defense and 20 percent of their defense budgets on acquisition and related research and development (R&D). Allies gave themselves ten years to achieve these goals. With the end of this timeframe just around the corner, some allies like the United States, Poland, and Estonia are pushing for more ambitious defense spending targets while others would be content to see the DIP disappear altogether.2Robbie Gramer, Amy Mackinnon, and Jack Detsch, “Eastern Europe Wants NATO to Beef Up Defense Spending,” Foreign Policy, February 2, 2023, https://foreignpolicy.com/2023/02/02/eastern-europe-nato-defense-spending-ukraine-russia-poland-estonia/. Clearly, the negotiations have been difficult.
Burden-sharing disagreements within the Alliance are nearly as old as NATO itself. Today, though, transatlantic tension regarding burden-sharing is heightened by a complex security landscape. Increased Russian aggression since 2014 has required reinvestment in conventional military capabilities, while asymmetric threats including cyberattacks, global warming, pandemics, mass migration, and terrorism continue to threaten allied security. Equitable burden-sharing among allies is not only a matter of principle, but also has operational and even tactical implications. Without equitable burden-sharing, the allies will quickly prove unable to meet their security commitments.
The latest allied defense spending figures show that while progress has been made on both elements of the DIP, not all allies will achieve the 2 percent and 20 percent goals by 2024. How then can the United States and like-minded allies successfully negotiate even higher targets? They might start by agreeing to portray NATO burden- and risk-sharing more accurately, as a way of ameliorating the naming and shaming that comes with not achieving the 2 percent and 20 percent goals. Although some argue that “inputs” like defense spending tell us a lot about “outputs” like contributions to Alliance operations, recently available data indicate this is not necessarily the case. If burden- and risk-sharing could be portrayed more accurately, those opposed to increasing the input targets might be more willing to reconsider their opposition. Even if they do not, improving how NATO depicts burden- and risk-sharing would benefit lawmakers, analysts, academics, and the public.
This issue brief begins by examining whether defense inputs really do tell us much about defense outputs, particularly contributions to allied operations. The brief does this first qualitatively by analyzing allied troop contributions to the International Security Assistance Force (ISAF) in Afghanistan in 2011, at the height of the troop surge. It then employs a statistical analysis to examine allied troop contributions to not just ISAF in 2011 but also to the Kosovo Force (KFOR) and the Enhanced Forward Presence (eFP) mission in the Baltic states. The issue brief concludes with recommendations on how NATO might more accurately depict burden- and risk-sharing, as a way of ultimately reaching higher DIP targets.
What do defense inputs tell us?
NATO relies on eleven metrics to measure burden-sharing.3Douglas Lute, remarks delivered during “The Cost of European Security” panel event sponsored by Carnegie Europe and held in Brussels, Belgium, on September 17, 2015. Transcript available at https://carnegieendowment.org/files/CE_Transcript_Cost_of%20European%20Security.pdf, and video available at https://www.youtube.com/watch?v=kw_jJ4jNqyY. Two of the metrics focus on so-called inputs and nine focus on outputs. Inputs refer to the funds that member states allocate to defense, and these formed the basis for the 2014 DIP. The DIP was noteworthy not because allies had committed to the 2 percent/20 percent targets—in fact, the 2 percent target had existed for many years prior—but rather because the commitments were made by Alliance heads of state and government. In the past, Alliance defense ministers were the ones who had committed to the defense spending targets.
Regarding burden-sharing outputs, most of the data on these NATO metrics are classified for a variety of political and operational reasons. For example, publicly acknowledging that some Alliance forces are not as ready to respond to a crisis as claimed could prove politically embarrassing, provide adversaries like Russia with useful intelligence, and undermine deterrence. However, the fact that not all burden-sharing metrics are public leaves some decision-makers, legislators, experts, and allied citizens without a complete picture of who in the Alliance is doing their fair share.
Some experts argue that the inputs correlate “a lot” with defense capabilities and capacity and hence provide a good enough picture of the outputs.4Jordan Becker, “Clearing the Air on Transatlantic Burden-Sharing, Part 2: You Gotta Give (Inputs) to Get (Outputs),” War on the Rocks, May 31, 2017, https://warontherocks.com/2017/05/clearing-the-air-on-transatlantic-burden-sharing-part-2-you-gotta-give-inputs-to-get-outputs/. This may appear somewhat intuitive—you get what you pay for. However, such claims are problematic on at least two counts. First, these claims ignore risk-sharing, an important yet often overlooked component of the broader concept of burden-sharing. The examples of Greece and Denmark illustrate this point well. Greece routinely spends more than the equivalent of 2 percent of its GDP on defense and recently it has spent well above 20 percent of its defense budget on equipment and related R&D (38.8 percent in 2021). Greece, therefore, may appear to be a model ally. In contrast, Denmark routinely spends less than the equivalent of 2 percent of its GDP on defense, averaging just 1.23 percent since 2014, and less than 20 percent of its defense budget on equipment, averaging 13.8 percent since 2014. Clearly, the Danes appear to be burden-sharing laggards.
Subscribe for events and publications on transatlantic security
Sign up for updates from the Atlantic Council’s Transatlantic Security Initiative, covering the debate on the greatest security challenges facing the North Atlantic Alliance and its key partners.
However, during NATO’s operation in Afghanistan, Danish forces operated in southern Afghanistan, a region that saw some of the heaviest fighting. As a result, the Danes had one of the highest per capita casualty rates of allied troop-contributing countries.5Ian S. Livingston and Michael O’Hanlon, “Afghanistan Index,” Brookings, March 31, 2012, https://www.brookings.edu/wp-content/uploads/2016/07/index20120331.pdf. Meanwhile, Greek forces served primarily at Kabul’s international airport and in allied medical facilities in Afghanistan. These troops performed important, necessary missions, but they did not face the same risks as Danish troops.
The second reason to doubt claims that NATO’s defense input measures tell us most of what we need to know about defense outputs is that merely possessing capabilities is no guarantee that those capabilities will be used in allied operations, regardless of the risk-sharing factors noted above. Again, referring to the cases of Greece and Denmark, the contributions of the former to Alliance operations in Afghanistan6“Afghanistan Troop Numbers Data: How Many Does Each Country Send to the NATO Mission There?” The Guardian, June 22, 2011, https://www.theguardian.com/news/datablog/2009/sep/21/afghanistan-troop-numbers-nato-data. (1.2 percent of all allied troops during the 2010-12 surge) and Kosovo7NATO, “Kosovo Force (KFOR): Key Facts and Figures,” August 2022, https://www.nato.int/nato_static_fl2014/assets/pdf/2022/8/pdf/2022-08-KFOR-Placemat.pdf. (2.8 percent of the total Kosovo Force as of August 2022) were and are somewhat low, given that Greek troops constitute 3.34 percent of total NATO forces. In contrast, at the height of the surge in Afghanistan, Denmark had deployed 0.56 percent of all allied troops even though its forces constituted just 0.51 percent of total NATO forces in 2011. Similarly, today Denmark contributes 0.93 percent of all allied troops in Kosovo even though its troops constitute just 0.52 percent of total NATO forces. In other words, when it comes to sending troops into harm’s way, Denmark appears to do more than what might be expected of it.
It is possible, though, that the cases of Greece and Denmark constitute idiosyncratic burden- and risk-sharing behavior. To explain, perhaps most allies at or above the 2 percent/20 percent thresholds make contributions that are what would be expected of them or greater—given their relative sizes within the Alliance. Conversely, it may be the case that most below the 2 percent/20 percent thresholds make undersized contributions, again relative to their sizes within the Alliance.
This can be tested by looking at three recent NATO operations—the International Security Assistance Force in Afghanistan in 2011; the Kovoso Force over the last decade; and the Enhanced Forward Presence mission in the Baltic states (Poland, Slovakia, Hungary, Romania, and Bulgaria) from 2017 until the present.8There are a variety of reasons why a country may contribute troops to a particular mission, including the role of threats and their proximity, domestic political calculations, relations with other allies, or any number of other factors. The analysis in this issue brief is agnostic on which of these matters most for each ally; instead, the emphasis herein is on whether, from a broader Alliance-wide perspective, there exists a positive relationship between defense inputs and defense outputs like troop contributions. These three operations were selected for two reasons. First, they are primarily land operations, and nearly all allies have land forces that could play a role in Afghanistan, Kosovo, and Eastern Europe. This permits a more broad-based analysis than, for instance, examining NATO’s Baltic air policing operation, since not all allies have fighter aircraft. Second, the operations represent a good mix of typologies—crisis response/counterinsurgency (ISAF), peacekeeping (KFOR), and defense/deterrence (eFP)—even if the needs of the Alliance today likely skew toward defense and deterrence.
The specific time periods for each operation were selected for the following reasons:
- NATO’s presence in Afghanistan was at its height in 2011, during the surge of allied forces there. Allies were under pressure from both Washington and Brussels to participate.
- KFOR is a long-term peacekeeping operation and examining it over the last decade includes the period when NATO was consumed with Afghanistan, the period when NATO was increasingly focused on deterring and defending against Russia, and the few years in between.
- The eFP mission in the Baltics is tied to the very heart of the Alliance—the Article 5 mutual defense clause of the NATO treaty—especially since Russia’s second invasion of Ukraine in February 2022.
By examining each ally’s contribution to the three operations relative to that ally’s military forces within the entire Alliance, one can get a sense of whether the success in achieving the goals of the DIP really does yield better burden-sharing.9Some have instead sought to focus on the allied contribution to each mission per 1,000 citizens as a measure of burden-sharing. This measure misrepresents the burden carried (or shirked) by those allies that have smaller militaries relative to their populations but that make proportionally larger contributions to allied military operations, like Canada, the Czech Republic, or Latvia, as well as those allies that have large militaries relative to their populations but that make below-average contributions to allied military operations, like Greece and Turkey. For example, if a particular ally met either the 2 percent or 20 percent threshold, and if inputs tell us a lot about outputs, it is reasonable to expect that that ally would contribute a proportional share of its military forces (relative to its forces within the Alliance) to the three operations.
Table 1 shows the data for ISAF in August 2011. Although the Defense Investment Pledge would not be made for another three years, the second and third columns indicate whether allies were above or below the 2 percent and 20 percent thresholds at the time—cells are highlighted green when above either threshold and red when below. Table 1 then shows a comparison between, on the one hand, each ally’s contribution to ISAF as a percent of all ISAF forces and, on the other hand, their military forces as a percent of NATO’s total military forces. (The figures on defense spending as a percent of GDP and the figures on percent of defense spending dedicated to equipment and R&D are lagged by three years because it takes time for money spent to have impact on capabilities.)
Table 1: NATO in ISAF, August 2011
If defense inputs tell us a lot about defense outputs, one might presume that if a country has sufficient inputs (i.e., it spends at least 2 percent on defense and meets or exceeds the 20 percent threshold for equipment and related R&D), it would contribute a percentage of ISAF’s forces proportionate to its share of total NATO forces. The last column of the table shows whether and to what degree allies’ contributions to ISAF were over (positive number) or under (negative number) their proportion of the Alliance’s total military forces. If an ally failed to spend 2 percent on defense and failed to meet the 20 percent threshold for equipment and related R&D, we would expect this number to be negative—that is, it fielded a percent of ISAF’s total force at a level below that which its forces represent within NATO more broadly. Conversely, if it spent more than 2 percent on defense or crossed the 20 percent threshold for equipment and related R&D, we would expect this number to be positive—that is, it fielded a percent of ISAF’s total force at a level above that which its forces represent within NATO more broadly. Exceptions—either over or under expectations—are highlighted in yellow. Of the twenty-four allies listed,10The list of allies includes Canada and all European allies that had been in the Alliance for at least three years as of 2011 and that have military forces. Hence, Iceland is not listed because it does not have military forces, and Albania and Croatia are not listed because they did not join the Alliance until 2009. (Montenegro joined in 2017, North Macedonia joined in 2020, and Finland joined in 2023.) eight did not contribute troops in a way one might expect, either positively or negatively, while sixteen did. More specifically, seven out of twenty-four allies met all or part of the DIP and yet failed to meet expectations for providing defense outputs in ISAF at the height of the surge.
Seven laggards out of twenty-four is not too bad. If roughly 70 percent of the time allies met expectations, it may be fair to conclude that defense inputs result in expected outputs, at least in terms of contributions to current operations. However, when analyzed statistically and when considering other operations like KFOR and eFP, the evidence of correlation between defense inputs and outputs is more mixed.
Table 2 shows the pairwise correlations among non-US allied defense spending as a percent of GDP, equipment spending as a percent of overall defense spending, “other” spending (which includes operations and maintenance, or O&M) as a percent of overall defense spending, and each ally’s defense budget as a percent of NATO’s overall defense spending on the one hand and the percent of ISAF troops on the other. Pairwise correlations allow for comparisons between a pair of variables, to shed light on whether a linear relationship exists between them. The results of a pairwise correlation analysis are expressed as correlation coefficients, ranging from -1, meaning a perfect negative correlation, to 1, meaning a perfect positive correlation. The closer the correlation coefficient is to either end of the scale (-1 or +1), the closer the relationship between the two variables. If, for example, a correlation coefficient between two variables is 0.75, this means that 75 percent of the time, if one of the variables moves in one direction (e.g., increasing), we can predict that the other variable moves in the same direction (e.g., also increasing). In social sciences generally, a coefficient from 0.7 (or -0.7) to 1 (or -1) indicates a strong positive (or negative) correlation; 0.3 (or -0.3) to 0.69 (or -0.69) indicates a moderately positive (or negative) correlation; and coefficients between 0.29 (or -0.29) and 0 indicate a weak or no positive (or negative) correlation.11Subsequently, a test for statistical significance can indicate whether a strong correlation between the two variables is due to random chance. Note that correlation does not mean or imply causation.
In the case of ISAF in 2011, the percent of troops contributed is poorly correlated with most key defense spending input measures, including the DIP and the percent of defense spending on O&M.12The conclusions drawn from the pairwise correlations shown in this issue brief stand up even when run through a bivariate regression analysis. The only variable that shows any strong correlation with the percent of troops contributed to ISAF is each ally’s defense spending as a percent of total NATO defense spending.13It is worth noting that this last correlation—percent of ISAF with percent of NATO’s overall defense spending—is the only one that is statistically significant.
Table 2: Correlations among Defense Input Measures and ISAF Troop Contributions, 2011
Moreover, statistical analyses of non-US troop contributions to other Alliance operations show similarly mixed correlations at best or no correlation at all. Table 3 features a similar pairwise correlation analysis for KFOR from 2012 through 2020. Given the nine-year timeframe considered here, the analysis includes over 240 data points. As seen in the table, there are no strong or moderate correlations, positive or negative, between any of the inputs addressed here.
Table 3: NATO in KFOR, 2012-2020
Table 4 features statistical analysis of non-US allied contributions to eFP from 2017 through 2022. The table is based on data for allied contributions to eFP in the eight host countries—Estonia, Latvia, Lithuania, and Poland from 2017 to the present, and Bulgaria, Hungary, Romania, and Slovakia since 2022. Data here include contributions from non-US allies each year from 2017 through 2022, resulting in 140 data points. As with the case of ISAF, the only defense input strongly correlated with percent of troop contributions to eFP is the percent of total allied defense spending. It is unclear why percent of total NATO defense spending is the only defense input that strongly correlates to both ISAF and eFP contributions, but it may reflect differing allied attitudes toward those operations on the one hand versus attitudes toward KFOR on the other.
Table 4: NATO in EFP, 2017-2022
Alternative measures, or just repackaging?
If defense inputs do not tell us much about the outputs, and if NATO does not make public the nine output measures it tracks, how can legislators, academics, other experts, and average citizens understand whether burden-sharing within NATO is fair? NATO burden-sharing is an inherently subjective and political issue. Given this, allies will likely try to manipulate whatever figures or data are used to measure burden-sharing to depict themselves in the best light. In short, no set of measures nor burden-sharing assessment process is perfect.
Nonetheless—and although it may be impossible to keep countries from using burden-sharing assessments as a way of burnishing their images—NATO can and should do more to include outputs in its publicly accessible annual reports. Doing so may offer all allies a fairer depiction of whether and how they are sharing burdens and risks, which could help lessen opposition from some allies to increasing the defense spending targets.
To achieve this, NATO should better leverage the data it already has available publicly—this will be far easier than trying to convince allies to declassify other information. For example, the Alliance should include in its annual defense expenditures report the troop contributions of each ally to ongoing operations. Information on allied contributions to particular operations is already publicly available, but it is not aggregated with the annual defense spending figures. This is sure to displease allies that contribute to other, non-NATO military missions in Europe and beyond including European Union or United Nations operations. To appease these allies, their contributions to non-NATO missions ought to be at least acknowledged, assuming those non-NATO operations are deemed to contribute to security in the transatlantic space.
Additionally, NATO should consider recrafting how it depicts the data it already makes public. For example, the Alliance staff charged with portraying the annually collected fiscal data should add an average trend line for all defense spending categories in its annual defense expenditures report. This is especially important for the personnel, infrastructure, and other (O&M) spending categories, which are not currently part of the DIP. Although there are no recommended minimum spending allocations for these categories, as there is for acquisition and related R&D (20 percent), simply showing a NATO average across all spending categories would provide a sense of which allies may need to recalibrate their spending.
Finally, the Alliance should think about ways to better portray risk-sharing. For instance, the Alliance should consider including per capita casualty figures in its reporting on current operations. For example, none of the “placemats” that the Alliance makes available for its operations in ISAF, KFOR, or eFP list casualty figures. This is politically risky and of course there have been few, if any, casualties in KFOR or eFP, given the nature of operations in Kosovo and across Eastern Europe. Additionally, the Alliance should continue to produce and release maps depicting in gross terms where allies (by country flag) are deployed in a particular operation. For instance, with an understanding that Alliance operations in southern Afghanistan were generally more challenging and dangerous than those in the north, a map depicting Danish, British, Canadian, Romanian, and American flags in the south provides a good sense of which allies are bearing greater risk.
The burden-sharing inputs used by NATO, specifically the Defense Investment Pledge agreed to at the 2014 Wales summit, do not correlate well or sometimes at all with one of the most important outputs—namely, relative troop contributions to NATO operations. Moreover, the inputs tell us nothing about risk-sharing, an important yet often overlooked component of broader burden-sharing considerations.
All of this means that those outside closed-door Alliance meetings—such as the general public but also most legislators—have little insight into whether allies are sharing responsibilities equitably. As the Alliance debates whether and how to replace the 2014 DIP, it should also consider ways to more accurately report on and portray burden- and risk-sharing. Doing so would permit a better informed transatlantic discussion on how allies should equitably share responsibilities, may lessen opposition toward increasing defense spending targets, and would ultimately improve transatlantic security in an era of strategic competition.
Dr. John R. Deni is a research professor of security studies at the US Army War College’s Strategic Studies Institute, a nonresident senior fellow at the Atlantic Council, and a nonresident associate fellow at the NATO Defense College. The views expressed are his own. The author would like to thank Matthew Woessner, Jordan Becker, and two anonymous peer reviewers for their invaluable comments on earlier drafts. Additionally, the author is grateful for research assistance from Chelsea Quilling, Nate Forrest, Max Haseman, and Sean Sanko.
The Transatlantic Security Initiative, in the Scowcroft Center for Strategy and Security, shapes and influences the debate on the greatest security challenges facing the North Atlantic Alliance and its key partners.