Counting bodies,preventing war: Future conflict and the ethics of fatality numbers

Abstract

With conflict prevention as a commonplace agenda of international organisations, numerous instruments for gathering knowledge about potential armed conflict have emerged. This article focuses on the mode of knowing war through quantification in the form of fatality statistics or ‘death counts’, which are taken by analysts and policymakers to indicate the severity and extent of conflicts. Drawing on official documents and interviews, I argue that fatality numbers are productive of the reality of violent conflict as they shape what counts as conflict and what does not. In the reporting by prevention actors, such numbers indicate past and future trends of armed violence and, in this way, bolster the imperative to prevent by creating quantified futures of conflict. However, fatality numbers also normalise deadly violence as a baseline criterion, thus also limiting the scope of what is known as future conflict and omitting lived experiences from the abstractions behind such numbers.

Introduction

There is a universal commitment to preventing violent conflict among international actors today, according to the joint Pathways for Peace report by the United Nations (UN) and the World Bank (2018: xvii; see also Bartelson, 2018: 15; Hathaway and Shapiro, 2017: 9; Mueller, 1989). This commitment, the report states, is enshrined within a ‘global architecture for peace and security’ (UN and World Bank, 2018: xvii), personified by the pledge in the preamble of the UN Charter to ‘save succeeding generations from the scourge of war’ (UN, 1945). After the bloc confrontation that had hampered cooperation in the Cold War period eased after 1990, conflict prevention transformed into an international policy agenda with the aim of averting the outbreak, escalation, continuation and recurrence of violent conflict (UN and World Bank, 2018; Zartman, 2015). Since then, almost all international organisations (IOs), as well as non-governmental organisations (NGOs), commissions and fora have emphasised the need for preventing violent conflict (Lund, 2008).

Prevention, understood as an intervention into the present to avoid an undesired outcome in the future, is an inherently forward-looking rationality of governing war and armed conflict. It targets events that have not yet happened – and might never happen. The future is the ‘cause and justification’ for preventive action ‘in the here and now’ (Anderson, 2010: 778). Thus, the conception of what conflict prevention entails within policy documents often revolves around the temporality of violent conflict, both regarding its first occurrence or further escalation (e.g. Boutros-Ghali, 1992; Carnegie Commission on Preventing Deadly Conflict (CCPDC), 1997; UN and World Bank, 2018). The dominant understanding of the agenda within the policy sphere encompasses primary prevention (‘outbreak prevention’ or prevention strictu sensu) before the first onset of armed conflict, secondary prevention (‘escalation prevention’ or mitigation and containment) while it is already underway, and tertiary prevention (‘relapse prevention’ or non-recurrence) in its aftermath (Björkdahl, 1999; Call, 2012; Höglund and Orjuela, 2011; Melander and Pigache, 2007; Ramcharan and Ramcharan, 2020). Calls for limiting the definition of prevention to first outbreaks notwithstanding (e.g. Wallensteen and Möller, 2004; Woocher, 2009), all conceptions have in common that they rely on the idea of a causal chain that can be manipulated in the present to avert (further) violent escalation in the future. To do so, prevention as a policy agenda and course of action relies on an – implicit or explicit – delineation of war and conflict as that which is to be averted.

In this special issue contribution, I examine the role of quantified indicators in the delineation of war and armed conflict within the international agenda that aims at preventing future conflict. A particularly potent way of enumerating war is to count or estimate fatalities. ¹ Scholars, analysts and policymakers use such numbers to indicate the severity and extent of conflicts. I argue that fatality statistics are constitutive of the reality that IOs represent about war and conflict as they shape and entrench what counts as past and future conflict within the discursive space of prevention. As a result, they are more than mere retrospective numerical descriptors of armed conflicts. Fatality numbers not only describe previous dynamics of conflict but also work to create urgent quantified futures – actors’ visions of the future that are constructed and disseminated through metrics (Berten and Kranke, 2023) – that call on a wide range of actors including state governments, civil society and IOs to act preventively. They do so by defining conflict as producing large quantities of deaths, suggesting these will only grow in the face of inaction. Below, I show how the quantification of violent conflict increases its salience as a policy issue through the reduction of complexity, emphasises its scope as a global problem through the construction of cross-context categories such as ‘major’ or ‘low-level’ conflict, and underscores the urgency of violent conflict as a global problem through invoking implicit and explicit thresholds and trends to indicate future deterioration in the absence of policy interventions. ²

Furthermore, I argue that the production of quantified futures of conflict bears ethical consequences. The widespread use of death counts and estimates to represent war, especially within prevention advocacy, normalises physical and lethal violence as the definitional criterion for violence to count as war and conflict. In this way, fatality numbers deprioritise other types of violence. Where they function as indeterminate thresholds to assess whether a situation is at risk of escalating into open conflict, they produce an epistemological and ethical dilemma for analysts who aim to produce knowledge to inform prevention policy. If violence needs to cross a certain numeric, if unspecified, threshold of lethality before qualifying as an armed conflict, this means that the conflict is already underway and thus the window of opportunity for primary prevention is already closing.

While a sizable body of scholarship exists that investigates fatality statistics, it has so far principally been concerned with developing and evaluating methodologies of counting and estimating (civilian) conflict deaths (e.g. Daponte, 2007; Fischhoff et al., 2007; Jewell et al., 2018; Spagat et al., 2009), elucidating the associated challenges and biases (e.g. Aronson, 2013; Dawkins, 2021; Krause, 2013; Krüger et al., 2013; Price and Ball, 2015; Sloboda et al., 2013; Tabeau and Bijak, 2005), as well as discussing the often highly politicised nature of such undertaking (Aronson, 2013; Greenhill, 2010; Nettelfield, 2010; Rappert, 2012), but less so with tracing how death metrics shape the understanding of (future) conflict for international agenda-setting and decision-making. I expand this literature by analysing the discursive function of fatality statistics for a specific policy field – the prevention of armed conflict. Drawing on existing scholarship that has established the role of quantification systems as ‘fundamental to the project of global governance’ (Merry, 2016: 155; see also Bandola-Gill, 2022; Berten and Leisering, 2017; Freistein, 2016; Grek, 2009; Löwenheim, 2008; Martin de Almagro, 2021; Rottenburg and Merry, 2015), I offer a novel conceptualisation of fatality numbers as indicators of ‘quantified futures’ of conflict. Appreciating the discursive construction of war and violent conflict through numbers is important because it has epistemological and ethical consequences for what counts as war and conflict (or not), and thus falls within or outside the purview of prevention efforts.

To support my argument, I draw on textual outputs of actors in the international prevention sphere, as well as a small set of 10 semi-structured elite interviews with current or former staff in international organisations, NGOs and government departments that I conducted in February and March 2019. ³ I selected the respondents due to their role of gathering and developing anticipatory knowledge on imminent crises and conflicts within organisations that are advocating for or implementing primary, secondary and tertiary prevention activities. These include the UN as the largest intergovernmental organisation (IGO), regional IGOs, international NGOs, as well as UK government departments. ⁴ I identified relevant individuals based on their job titles and role descriptions, which include analysts, (senior) consultants, country coordinators, programme advisers, conflict leads and UN officers. All interviewees are ‘elite’ in the sense that they responded in a professional capacity, occupy a high-level position and/or strategic role within their organisation (as opposed to local, more operational stakeholders), and are citizens of, or primarily based in, the Global North (the United States, the United Kingdom, Canada).

This article proceeds as follows. In the next section, I first explain the role of quantification, and fatality statistics or ‘death counts’ in particular, for governance and advocacy. I then discuss their use in the conflict prevention agenda, with a particular view on how threshold criteria imported from academic research shape what organisations of the prevention sphere define as ‘war’ and ‘conflict’. In the subsequent section, I outline my argument that fatality numbers work as a forward-looking technology for governing future war. I then discuss the ethical implications of using death counts for knowing and governing future war, after which I conclude this article with a summary of the argument and a brief outlook on future research avenues.

Quantifying deaths for governance

Measurement, aggregation and calculation are the cornerstones of making societies and their activities legible to modern rulers and governments. As countries transformed into large territorial states and expanded their lands through colonial and imperial conquest, various societal, political and economic dynamics needed to be translated into terms that could be read and manipulated by a distant ruling elite (Cohn, 1996; Foucault, 2007: 67–79; Scott, 1998).

The centrality of quantification for steering and overseeing is ever more acute in today’s international governance culture that relies on evidence-based decision-making and audit-centred management (Merry, 2016: 4). Statistics and indices have proliferated in international agenda-setting and reporting over the last decades (Zapp, 2022). Global reports nowadays not only cite numerous scientific sources but also make extensive use of standardised indicators. Within the trend of ‘science-based global governance’, international actors’ authority and legitimacy stem from them producing and using ‘disinterested and policy-relevant scientific knowledge’ (Zapp, 2022: 457), which then informs particular visions of the future that orients discourses and actions (see also Berten and Kranke, 2022).

Quantification delineates the magnitude of problems, defines them in comparable terms and facilitates their classification, categorisation and prioritisation (Merry, 2016: 4). Statistics are particularly appealing to policymakers wanting to evade the accusation of partisanship, as they promise to be politically unbiased (Merry, 2016: 3–4). However, as instruments for formulating and managing problems, advocating for policy changes and mobilising resources, numbers are political vehicles in themselves. They organise and simplify knowledge, which then facilitates decision-making. The translation of a problem into quantitative and measurable terms makes it accessible to governance as a ‘series of technocratic interventions’ (Berten and Kranke, 2023) by suggesting that specific policies can manipulate the problem’s numbers – from influencing population dynamics such as mortality or birth rates to gaining or suppressing votes in democratic elections (Foucault, 2007; Rose, 1991). In international politics, for example, quantification helps actors to exert social pressure to ‘shame’ states into compliance with international norms such as human rights (Kelley and Simmons, 2015). Poor indicators that signal inadequate governance, such as institutional dysfunction or political instability and unrest, ‘invite[] international intervention and management’ (Merry, 2016: 5).

Quantification also facilitates the governance of war and armed conflict. The metric of fatalities is used widely because ‘[d]eath is final and corpses are easier to count than the wounded’ (Fazal, 2014: 122). Fatality statistics are metrics to capture the number of combatants and non-combatants killed and, in this way, help to make war’s devastation comprehensible and comparable across contexts. ⁵ The challenges of counting and estimating notwithstanding, death data are fundamental to scholars’ and practitioners’ understanding of how and for which reasons violent conflicts unfold (Krause, 2019: 129).

The enumeration of lives lost is a long-standing practice that can be traced back to medieval chronicles. As Espeland and Stevens (2008: 405) argue, medieval death numbers that indicated the sum of people killed from the Plague, for example, were not precise counts but rather served as iconic and didactic vehicles. These numbers were unimaginable for readers at the time and thus served to signal the great extent of suffering rather than to represent a precise measure (Espeland and Stevens, 2008: 406). Arguably, the same applies to current-day counts and estimates of war fatalities. Often in the thousands, ten thousands or even millions, the amounts of slain bodies are equally unfathomable to the minds of the 21st century. Nevertheless, the sheer volume gives weight to the policy imperative of preventing further violence that would add to an already high number. In the form of the ‘human costs’ of war, fatality numbers thus help to make the case that their incurrence is best prevented (Annan, 2001: 7).

Beyond making war and armed conflict legible and commensurable as a global phenomenon of a significant scale, fatality numbers have another, more implicit, function for prevention advocacy. They construct prevention as a way to maintain international security and call upon various authorities to live up to their duty to do so. As Auchter (2016: 47) explains, death is inevitable, but it is also uncertain and thus ‘the ultimate human insecurity’. Consequently, to secure is to either create or prevent death. A breach of this security in the form of death by violence or inadequate governance thus constitutes a violation of the social contract (Auchter, 2016: 39). Similarly, the understanding that resonates in the use of fatality counts and estimates for prevention advocacy is that such numbers make visible – or embody – the failure of the state as the principal protector, and the ‘global architecture for peace and security’ (UN and World Bank, 2018: xvii) as the secondary one, to secure its subjects against the lethal force of armed conflict.

The homogenisation inherent in aggregating separate instances into a singular phenomenon stabilises the idea that they can be approached using the same set of policies. In Appadurai’s (1993: 317) words, it creates the ‘illusion of bureaucratic control’ over a governance problem. An oft-cited example of this illusion is the British census in India as a technology to oversee colonies at a distance (Porter, 1995; see also Scott, 1998). The enumeration of the Indian population in the form of censuses was not only a mere extension of the logic and purpose of counting populations in the metropole but stabilised and homogenised categories such as caste, which were fluid and contextual prior to the British colonial project (Cohn, 1996). Fixing these categories helped to put the reality of the colonial experience into terms that were legible for British governing, thus creating the ‘sense of a controllable indigenous reality’ while flattening existing nuances and contexts (Appadurai, 1993: 317). Similarly, death counts and estimates flatten differences and nuances of armed conflict, as they rest on the assumption that war manifests at different times and in different spaces worldwide but its outcomes – such as the war dead – can be consolidated into one sum. That is, even if its circumstances differ and even if it occurs within the territorial bounds of a sovereign state, war becomes a genuinely global problem that warrants governing by international actors.

Measuring a phenomenon quantitatively not only acknowledges its existence as ‘fact’ ⁶ but also helps to signal its importance to policymakers (Andreas and Greenhill, 2010b; Seybolt et al., 2013). Estimates and counts of the number of people killed by war inform decisions on, for example, the need for secondary prevention efforts in the form of military operations, tertiary prevention in the form of peacekeeping, humanitarian interventions or the allocation of international aid and relief efforts (Auchter, 2016: 40). Quantification is key to ‘actionable’ knowledge as it not only allows to define and compare a problem but also to evaluate the effectiveness of any action taken in response to that problem. As Espeland and Stevens (2008: 432) note, measuring phenomena in numbers ‘can help us see complicated things’ – such as war and its future development – ‘in ways that make it possible to intervene in them productively’, such as with prevention policy. In this way, the enumeration of war as death counts and estimates contributes to the constitution of conflict as a governable entity. Especially in contexts where a conflict situation seems particularly complex and intractable, simple narratives and numbers that ostensibly speak for themselves make it possible to ‘identify salient issues [and] dictate urgent action’ (Autesserre, 2012: 208). The idea of ‘intervenability’ conveyed by numbers reinforces the notion of the situation being a case of concern, as ‘a bad condition does not become a problem until people see it as amenable to human control’ (Stone, 1989: 299).

Then again, such statistics are not always compiled out of purely benign motives. As Gregory (2022) argues, the coalition in Afghanistan took up tabulating civilian casualties not principally out of genuine concern about harming civilians but to manage the strategic and reputational issues arising from ‘collateral damage’. The statistics revealed patterns and trends of civilian casualties, which then helped coalition forces to ‘calibrate the violence they inflicted’ to maximise the effectiveness of counterinsurgency operations (Gregory, 2022: 481). The ensuing reduction in civilian harm was thus more a by-product of compiling casualty statistics rather than its central aim. That is, the coalition exploited civilian casualty numbers as a weapon of war, making them complicit in creating the violence they were designed to enumerate.

The delineation of conflict in the prevention agenda

To bolster the case for prevention, organisations frequently use cumulative fatality estimates. The UN’s 1992 Agenda for Peace report notes that since 1945, more than ‘100 major conflicts around the world’ resulted in approximately 20 million deaths (Boutros-Ghali, 1992: 3). Five years later, the Final Report of the Carnegie Commission for Preventing Deadly Conflict sets the death toll at over 100 million since the beginning of the 20th century (CCPDC, 1997: 11). In 2011, the World Development Report entitled ‘Conflict, Security, and Development’ notes that the number of battle deaths fell by 75% from 164,000 per year in the 1980s to fewer than 42,000 after 20 years, as both ‘major’ and ‘minor’ civil wars became ‘less violent’ and declined in number (World Bank, 2011: 52). However, these statistics represent civilians only when they died as a result of fighting counted as battle, so that they only tell part of the story of war’s lethality. Furthermore, while the report characterises the decline of wars and their associated battle deaths as a positive trend, it simultaneously warns that violent conflict often comes in repeated cycles, which severely undermines countries’ development progress (World Bank, 2011).

However, only a few years later, the aforementioned Pathways for Peace report notes inverse trends in 2018, with an increase in wars and ‘lower-intensity’ conflict, as battle deaths in civil wars have almost tripled and risen by 60% in low-intensity conflicts in the decade before the report was compiled (UN and World Bank, 2018: 13–14). While the latter constitutes the most recent comprehensive document by the UN and the World Bank on the prevention agenda at the time of writing, other organisations such as the International Crisis Group (ICG) also use fatality counts to emphasise the case for prevention. In 2022, their mission to ‘prevent wars and shape policies that will build a more peaceful world’ is justified by the claim that the failure to prevent conflicts has resulted in 81,447 deaths in 2020. ⁷

The claims in the 2011 World Development Report and Pathways for Peace, respectively, differentiate between ‘major’ and ‘minor’ conflict or (civil) war and ‘low-intensity conflict’. These distinctions rest on a fatality threshold that originates from the Uppsala Conflict Database Programme (UCDP), on which both documents rely for their fatality reporting. In the UCDP codebook, armed clashes have to reach a minimum intensity level of 25 or 1000 battle-deaths per year to be coded as a minor conflict or war, respectively. ⁸

The introduction of the 1000-deaths threshold is widely attributed to J. David Singer and Melvin Small (1972; Small and Singer, 1982), who popularised it through the Correlates of War (COW) project (for attributions, see, for example, Geller, 2004; Mueller, 2009; Sambanis, 2004). COW compiles an extensive database cataloguing the frequency, participants, duration and (combatants’) battle-connected deaths of all wars since the end of the Napoleonic era in 1816. ⁹ While the original authors seem to have abandoned the threshold of 1000 deaths in later coding, it has since become a frequently used cut-off value in conflict databases and large-N comparisons of (civil) wars (Krause, 2017; Sambanis, 2004). As Mueller (2009: 298) writes, recalling a personal conversation with Singer, ‘the 1,000 figure more or less fell out of the analysis when other aspects of what could be considered warfare were assembled’. However, as Heuser (2022: 61) notes, the number was already used by Quincy Wright (1942) in A Study of War, who, in turn, adopted it from the Militär-Historisches Kriegs-Lexikon by the military historian Gaston Bodart. ¹⁰

The criterion of battle deaths for the categorisation and ordering of wars by severity and intensity relies on a set of, often implicit, assumptions about the nature of armed conflict and how to know and represent it. ¹¹ First, it defines war and conflict as essentially lethal, which means that other types of violence that do not result in death are not defined as conflict, as I explain in more detail below. Second, where battle deaths concern only armed personnel, such as in the COW coding, the measure requires that the latter can be clearly distinguished from non-combatant deaths. However, this distinction might not be discernible in the reality of conflict, as people switch between roles or categories of civilian and combatant might intersect, such as in the case of child soldiers (Andreas and Greenhill, 2010a; Gade, 2010; Krause, 2013). Yet, as such numbers proliferate in organisations’ reports, agendas and mission statements, the interpretive decisions behind them are usually not acknowledged (Merry, 2016: 1). Third, even where battle deaths include civilian fatalities resulting from armed clashes, they do not account for the effects of war beyond battle – lethal and non-lethal – like the destruction of shelter, food and health infrastructures. The latter can drive conflict escalation in complex feedback loops (Hendrix and Brinkman, 2013), thus affecting the prevention of conflict recurrence. The omission of conflict-related but indirect fatalities is often explicitly addressed both in academic and policy outputs when they note that ‘battle deaths do not tell the full story’ of armed conflict (e.g. Lacina and Gleditsch, 2005: 158; UN and World Bank, 2018: 27). However, this limitation risks to be obscured where the battle deaths are the definitional criterion of what is a war and what is not. ¹² Finally, the battle death indicator is an ahistorical measure when used to compare conflicts over time, as it does not account for the changing character of war-fighting and improvements in preventive care, battlefield medicine, evacuation processes and body protection (Fazal, 2014). As a result of these changes, wars become less lethal while the wounded-to-killed ratio increases so that the value of the ‘primary measure to count wars’ (Fazal, 2014: 96) decreases. Due to these shifts, fewer conflicts meet battle-death criteria. This tendency has led some to conclude that wars are on the decline (e.g. Goldstein, 2011; Horgan, 2012; Lacina and Gleditsch, 2013; Mueller, 2009; Pinker, 2012), while conflict scholars have challenged this claim (see, for example, Braumoeller, 2019; Fazal, 2014; Gohdes and Price, 2013). ¹³ Thus, where reports of the international prevention architecture make claims of positive conflict dynamics based on fatality statistics, they risk underestimating the magnitude of armed conflict as a global problem and signalling decreased urgency in addressing it (e.g. World Bank, 2011).

When policy reports rely on databases like the UCDP, they internalise the cut-offs that coders apply to pragmatically categorise armed hostilities, which then – inadvertently but rarely explicitly – shape what kinds of conflict come to be understood as ‘war’ or ‘low-intensity conflict’ in the agenda-setting of international organisations. In turn, when armed violence falls below a fatality threshold or when violence is not considered conflict-related, actors might miss important dynamics leading to conflict escalation and the associated opportunities for preventing the outbreak, further escalation or resurgence.

Two examples of such violence that falls below common fatality thresholds or is not considered battle-related but nevertheless contributed to the deterioration of militarised conflict are the case of the Rohingya, as I explain further below, or the practice of cattle raiding in South Sudan. Due to herders’ intermittent but decade-long integration into various armed forces, cattle raids are now heavily militarised and often lethal. Political opponents have systematically exploited the practice, so that cattle-raiding is regularly reigniting the South Sudanese conflict (Da Costa et al., 2022: 235; Wild et al., 2018). Despite cattle-raiding becoming increasingly deadly, it falls outside the purview of violence in the focus of prevention efforts, as it often does not meet fatality thresholds and is not considered battle-related. As the flagship document of the international prevention agenda, the Pathways for Peace report discusses South Sudan as one of the central concerns for conflict prevention several times throughout. While it notes in an info box that conflict-related deaths occur outside battle, including due to raids, mentions of the South Sudanese case remain focused on higher-level efforts at pacification and political settlement among political elites (UN and World Bank, 2018: 27).

Retrospective numbers and urgent futures

When used as thresholds to ascertain which instances of violence to categorise as armed conflict, death counts work as numerical sieves that either catch certain instances and define them as belonging to the category of conflict and war, or they fall through. Such thresholds constitute the basis of claims about conflict trends such as in the examples of the World Development Report or Pathways for Peace. In this way, they affect which instances of violence are cast as relevant for prevention agenda-setting. However, the quantification of conflict and the internalisation of (battle) death thresholds for the definition and categorisation of conflict are not restricted to reports, mission statements and other textual outputs. While usually implicit and unspecified, fatality thresholds also play a role in conflict monitoring undertaken by organisations, as they set and entrench the expectation of conflict as inherently deadly.

In the previous section, I have shown how war is rendered knowable and governable through quantification. Quantification, in turn, is often a function of larger endeavours of social and political steering, in the sense that it is ‘work that makes other work possible’ (Espeland and Stevens, 2008: 411). Specifically, quantification makes phenomena such as war and violent conflict amenable to statistical operations of all kinds, including the calculation of trends such as the likelihood of (further) violent escalation. By enumerating the future to come, such as through probabilities and trends, numbers render complex phenomena actionable as ‘an object of scientific enquiry and political intervention’ (Anderson, 2010: 748; Aykut et al., 2019: 2; Berten and Kranke, 2023).

As it aims at averting the escalation of (further) violence in the future, prevention is a form of anticipatory governance (see also Berten and Kranke, 2022). It is inherently forward-looking since ‘one has to know what is coming [. . .] in order to prevent it’ (Zartman, 2015: 7). Preventive action thus requires a knowledge production apparatus that provides decision-makers with analyses of conflict dynamics and trends (Boutros-Ghali, 1992: 6). This knowledge might come in the form of risk assessments and forecasts to buttress structural prevention efforts in the longer term, or in the form of early warning systems that target short-term, operational prevention activity (Carment and Garner, 1999). All of these rely on the idea that the future of conflict can be known probabilistically as they produce claims about how a situation is likely to turn out if nothing is done (Muggah and Whitlock, 2022: 10). In this sense, early warning systems, forecasts and risk assessments construct potential dangerous futures (see Amoore, 2013; Aradau and van Munster, 2007; De Goede, 2008). These futures are rendered present through statistical models and trends, narratives as well as tangible artefacts like written reports and visualisations (see also Anderson, 2010: 783).

Early warning mechanisms are crucial components of anticipatory knowledge production for prevention activities. The former are processes of systematic data collection and analysis to inform and alert policymakers about imminent outbreaks to avert the first outbreak (primary prevention), further escalation (secondary prevention) or recurrence (tertiary prevention) of armed conflict (Organisation for Economic Co-operation and Development (OECD), 2009: 22). A manifold range of early warning systems has developed since the 1970s that operate on several levels, spanning systems run by intergovernmental organisations such as the UN, various regional organisations (such as the European Union, African Union or the Intergovernmental Authority on Development), as well as state departments, intelligence agencies, NGOs and think tanks, academia and the private sector. ¹⁴ In this sense, ‘early warning’ is an umbrella term that captures a broad range of tools using a variety of data-driven approaches and methodologies. These data can but do not need to be quantitative. While fatality numbers might play a role in their predictions, fatalities are not the only – or even central – metric on which such systems rely. ¹⁵ It would thus be too simplistic to suggest that early warning systems are ‘triggered’ once clashes cross a certain fatality threshold. ¹⁶ Instead, many of my interviewees reported that they mostly rely on their experience and personal judgement to assess what is worthy of further inquiry and what is not, due to the lack of formalised rules. That is, analysts have to decide intuitively which types of violence, and how many instances of it, count as being ‘relevant enough’ to be taken as a sign of imminent escalation.

As one NGO consultant put it, there is no clear cut-off for when the situation within a country ‘gets bad enough to cover it’ and frame it as significant, so that it receives greater attention (interview with NGO consultant, telephone, February 2019). Within their organisation, an international think tank aimed at conflict resolution and prevention, this consultant’s task was to provide expertise on a South Asian country through a mix of desk and field research, which would then inform the organisation’s in-house early warning tool. Assessments of how relevant specific cases are for the larger policy work of the organisation thus required knowledge of the particular context of violence within a focus country. However, one crucial component of such judgements of relevance and severity, this interviewee noted, is usually that ‘a fair amount of people got killed’ (interview with NGO consultant, telephone, February 2019). In other words, ‘war and conflicts are deemed to begin only when a certain number of dead bodies appear’ (Krause, 2017: 102).

Following Merry (2016: 5), who posits that rather than revealing the truth, quantitative indicators create it (see also Appadurai, 1993; Porter, 1995), I thus argue that rather than revealing the future of conflict, quantitative markers such as the number of battle-related deaths produce it. They do so through their capacity to define, categorise and commensurate. That is, although fatality numbers are retrospective indicators of past conflict, they create ‘truth’ around future conflict as they influence which instances and which extents of armed violence count as a ‘war’ or ‘conflict’ in the first place. In this way, the backward-looking measure of fatalities has a forward-looking effect, as it affects which kinds – and especially which extents – of violence are (not yet) considered relevant to future escalation because they (do not yet) meet certain fatality criteria.

As the example of the South Asia analyst illustrates, even where early warning mechanisms rely on qualitative, contextual and country-specific analyses, they can include implicit quantitative reasoning for determining whether and when to sound the alarm. In this sense, the quantification of war in the form of fatality statistics shapes imaginations and expectations of war and armed conflict. Dead bodies are easier to quantify than other harm inflicted by conflict (Fazal, 2014: 122). At the same time, Merry (2016: 219) notes that ‘things that are more easily counted [. . .] tend to be those counted in the future’, while those that are harder to enumerate ‘disappear from view’. The implicit but unspecified threshold of lethal violence to qualify as relevant mentioned by the NGO consultant goes to show how quantification, even where it manifests as unspecified thresholds, has a ‘tendency [. . .] to remake what it measures’ (Espeland and Stevens, 2008: 432), as decisions on what to count – and to count as what – translate into decisions on what matters.

As became apparent in my interviews, political shifts and attacks in the form of direct violence are the main focus of what analysts look for when monitoring (potential) crises and conflicts. For example, I spoke with a former employee of a philanthropic organisation aimed at alerting policymakers in Washington D.C. of tensions and bringing them together to act swiftly, and ideally preventively, to curb imminent violence. During their time at this NGO, this interviewee was monitoring an East African country, which was largely done by scanning local news outlets with targeted keyword searches. The central keywords, as they explained, were ‘attack’ and ‘dead’ (interview with former employee in policy NGO, Skype, February 2019). Yet, none of my interview respondents stated that the focus on direct violence was an explicit requirement. That is, the taken-for-granted references to lethal violent incidents as the ones of interest suggest that it is an epistemological routine stemming from the unquestioned assumption that policymakers, that is, potential ‘end-users’ of knowledge production outputs, require evidence of physical violence.

Then again, the focus on spectacular violence is already built into reporting mechanisms that determine what analysts who scan newswires, such as the one in the example above, end up seeing in the first place. ‘Stories’ that promise to be relevant and interesting to an international audience are of high priority, while ‘slow’ violence such as famine presents as less attractive for media and advocacy attention (see also Dawkins, 2021). These pressures influence which kind of violence data ends up in newswires and, in further consequence, in the databases and policy reports that rely on news sources. Such data often already privileges lethal violence, which further entrenches the expectation of war and conflict as principally manifesting in the form of deaths.

As death numbers inform which violent events become defined as wars and conflicts, they not only matter to show war’s historical and recent effects but also its potential consequences in the form of quantified futures of war. The latter, as outlined in the previous section, are then leveraged by international actors to make the case for prevention. Following Anderson (2010: 778), who argues that the future is not separate from the present but rather constructed in the latter, I argue that retrospective fatality numbers create urgent futures of conflict through conflict trends suggesting the prospect of (further) escalation. That is, death counts, and fatality thresholds specifically, constitute ‘present futures’ (Berten and Kranke, 2022: 157) of conflict.

The following example further illustrates how implicit but unspecified lethality threshold shapes perceptions of future conflict risk. One of my interviewees was monitoring a country for a large international research and advocacy NGO, which aims at providing early warnings and facilitating early action to prevent violent conflict. This analyst’s focus country did not experience active conflict (of any level) at the time, and any other occasional fatalities, such as those resulting from protests turning violent, were ‘comparatively low in number’ (interview with NGO analyst, telephone, February 2019). Due to these characteristics, the respondent remarked that this country is not one of the ‘typical violence cases’. In other words, the absence of an explicit trend of escalating violence in the form of battle deaths meant that this country was not considered at risk of a conflict outbreak. Although, according to the interviewee, the history of conflict justifies the continued interest, the case is somewhat of an ‘outlier’ in the country portfolio of their organisation that focuses on preventing conflict. From a viewpoint of relational interviewing (Fujii, 2018), this interviewee’s characterisation is an almost apologetic justification of the country’s case, which has not seen significant battle death numbers in recent years, as ‘relevant enough’ to the organisation’s mission of monitoring conflicts to avoid (further) escalation in the future.

As a definitional criterion, death counts normalise lethal violence in that they conform to an implicit ideal-type of armed conflict after which others of the same kind are expected to be modelled (see also Espeland and Stevens, 2008: 416). This model comes in the form of thresholds or rubrics that suggest that for a string of violent events and armed clashes to count as a full-scale war or low-level conflict, it should normally be deadly on a large scale. That is, fatality thresholds – be they clearly determined or unspecified – have a normative-prescriptive character (see also Foucault, 2007: 57), in that they not only describe past conflicts but also delineate the discursive possibilities of what can be counted as ‘future conflict’.

What counts? Implications of death counts as markers of conflict severity

My interviews with analysts in government departments, IOs and NGOs further indicated that fatality thresholds have ethical implications for prevention advocacy. As I have argued in the previous section, they normalise the expectation of physical violence as the baseline for what makes a situation count as an active or imminent conflict. In this way, fatality numbers create an ethical dilemma as there needs to be at least some suffering before the definitional criterion is fulfilled, thus limiting the scope of what falls within the purview of ‘future conflict’. Death numbers as benchmarks for severity omit non-lethal types of violence and oppression that might contribute to the escalation of armed conflict, such as torture, forced disappearance, disenfranchisement and displacement (Krause, 2017). As a former senior UN official noted, the 2015 elections in Myanmar were celebrated for remaining peaceful. However, the focus on electoral violence shifted attention away from the disenfranchisement of the Muslim minority, which turned out to escalate tensions in the decade-long military conflict, including mass displacement and genocide of the Rohingya (interview with former UN staffer, Skype, March 2019). In this case, the privileging of immediate and lethal violence over structural violence led the international prevention architecture to neglect how the latter can develop into the former.

In a similar vein, another respondent drew attention to the gendered dimension of the focus on lethal violence, lamenting the ‘masculinist obsession’ with fatalities (interview with conflict adviser for large international disaster relief organisation, Skype, February 2019). While men are more likely to die from fighting or deliberate killing by a conflict party, women are likelier to die due to the indirect effects of armed conflict outside combat and in post-conflict phases (Ormhaug et al., 2009). Thus, as the interviewee points out, the use of battle deaths as a severity indicator, such as in the databases and reports cited above, obscures female fatality rates as a result of conflict. At the same time, battle death numbers do not pay adequate attention to other types of violence, such as sexual and gender-based violence, that might not be lethal but influence conflict dynamics and inflict immense pain and trauma. ¹⁷ This interviewee’s remark echoes a critique from feminist theory that has excavated implicit hierarchies of harm in the narration of war and conflict through battlefield violence. Feminist theorists have long advocated for understanding war and conflict as unfolding along a continuum of violence, which captures various harms that are a result of conflict but take place beyond the battlefield or during ostensibly peaceful ‘post’-conflict episodes (e.g. Cockburn, 2004; Confortini, 2006; Enloe, 2000; Reardon, 1993).

Where fatality numbers work as proxies for determining whether the situation at hand counts as a conflict such as in the analysts’ examples I described in the previous section, the expectation of direct, physical and attributable violence resulting in death has an ethical implication. Applying a somewhat arbitrary set of fatality thresholds means that there needs to be a certain level of suffering before action can be taken. Even when their particular assessment frameworks lead analysts to conclude they are seeing signs of an impending outbreak of armed conflict, it can prove difficult to use such knowledge effectively (interview with conflict adviser for large international disaster relief organisation, Skype, February 2019). Donor and policy mechanisms often only kick in after the realisation of direct violence. For example, the UN Security Council (UNSC), which is the body with the highest authority in terms of deciding on international efforts for conflict prevention, is unlikely to mandate substantive action and allocate resources when evidence of significant violence is absent (Call and Campbell, 2018: 71).

It is not a new insight, however, that high-level bodies’ interventions are slow and late or absent altogether. As UN Secretary-General Annan (2001: 12–13) notes in his 2001 report Prevention of Armed Conflict, the focus of the UNSC remains ‘almost exclusively on crises and emergencies, normally becoming involved only when violence has already occurred on large scale’. Thus, preventive deployments remain rare, as the international community is reluctant to expend resources ‘without the clear case for deployment that is made by open conflict’ (Annan, 2001: 21). However, the expectation of lethal violence as the starting point to recognise a situation as imminent conflict, as illustrated by the above examples, creates a dilemma in which analysts produce knowledge to inform prevention policy but begin their inquiry at the occurrence of quantifiable deadly violence, thus after what is to be prevented is already taking place and the window of opportunity for primary prevention is already closed or closing. Where conflict expertise depends on violence reaching certain levels of lethality, it omits other preceding manifestations of violence and repression, which might be systematic and widespread, but not deadly, yet often contributing to the escalation of conflicts.

Fatality statistics offer a language that ‘enables us to talk about this violence [of war] without ever confronting the devastating effects of what happens to the people involved’ (Gregory, 2021: 205). In this way, death counts and estimates have a paradoxical effect of being an indicator of how intense and severe a conflict was, currently is or might become, as outputs such as the reports cited above use them to make the case for prevention. At the same time, the increased ‘scientificness’ of global reports that is underpinned by statistics and indicators risks reducing complex relations to neat figures (Zapp, 2022: 469), which can only give an abstract and sterile image of war. Simplified claims and aggregate numbers tend to be repeated in various outputs and subsequently turn into reference points for policy deliberations (Paris, 2011: 63). Then again, as measurement renders complex issues intervenable (Appadurai, 1993; Espeland and Stevens, 2008), it also tends to prioritise things that can be easily quantified, such as bodies and deaths. At the same time, it risks overlooking other ways in which war manifests that might not be easily put in numbers, such as pain, suffering and grief.

Enumerating requires categorising, and categorising shapes problem definitions, which might then diverge from what is important to those affected. Taking the example of domestic violence, Merry (2016: 27) notes that incidence counts do not account for the networks of kinship and gender norms. However, these are exactly the dimensions that determine how the person that is abstracted into a number experiences domestic violence. Affected women’s perceptions of the severity of the violence they endured depend on the specificities of the relationship and a long chain of events, both violent and non-violent. Indeed, Merry (2016: 83) finds that the ‘act of physical violence was less important to these women than the violations of a sense of self, repeated insults and humiliations, threats to children and pets, and excessive demands for money’. Similarly, where (future) war is quantified as separate incidents tallied up in numbers of deaths, its context, conditions, relations and histories that led to the killings are omitted. Aggregations cannot account for the pain and suffering inflicted on the victims of war (Gregory, 2021: 200). The dehumanising effect of aggregating deaths intensifies where fatality counts are appropriated towards strategic ends or even weaponised to inflict violence more effectively (Gregory, 2022: 481; see also Emery, 2020).

At the same time, being included in statistics can help to build identity, create community, strengthen victims’ agency and facilitate access to resources, security, and justice for those affected by violence. As Nelson’s (2015: 82) study of counting victims in the aftermath of the Guatemalan civil war shows, being part of such numbers can also be generative for victims, survivors and their loved ones. Aggregation thus can also restore humanity and give hope that, by being counted, those who died through violence might eventually be accounted for (Auchter, 2016: 40; Nelson, 2015; Sloboda et al., 2013: 54). Viewed from this angle, it is not being counted that is dehumanising (see also Copeland cited in Auchter, 2016: 42).

Adopting a feminist approach to resolving and preventing war, Gregory (2021) thus suggests rethinking the production and consumption of death counts and estimates as representations of war. Such numbers cannot account for violence beyond battle and risk underestimating the killing of civilians. Consequently, they do not capture the full reality of armed conflict as it is experienced by those subjected to it, yet it is often used as the primary measure to represent, monitor and compare armed conflicts and impending crises (Fazal, 2014: 96). The abstracting and dehumanising effect of fatality numbers that strip war off the lived experiences when transforming it into a statistic is not resolved by more accurate and precise modes of counting and reporting war deaths, as they provide ‘no guarantee [. . .] to disrupt the martial logic that renders this killing permissible in the first instance’ (Gregory, 2021: 208). To counter the dehumanisation inherent in abstracting the deaths of the victims of war into numbers, Gregory (2021: 201) suggests naming the victims and telling their stories – ‘not just stories about their grisly demise but also stories about who they were as individuals and the lives that they left behind’. After all, while the creation of commensurable categories of conflict and thresholds is necessary to define, recognise and assess conflict so that it can be prevented, adding stories to fatality numbers not only re-humanises the people behind those statistics but also concretises what exactly it is that the policy agenda of prevention is aiming to avoid: death, loss and suffering.

Conclusion

In this special issue contribution, I have analysed the role of conflict death counts and estimates as a specific type of ‘quantified futures’ to show how they ‘underpin the construction of challenges as salient rather than negligible, as global rather than local, and as urgent rather than deferrable’ (Berten and Kranke, 2023). I have argued that by making violent conflict measurable and comparable, fatality numbers constitute war and conflict as knowable and governable entities. In the reporting of IOs and other actors in the prevention sphere, fatality numbers indicate trends in the past and future development of the world’s conflict landscape. That is, death counts and estimates not only serve as scientific evidence that bolsters the imperative of prevention of organisations’ agendas. Drawing on the work of Merry (2016) and others showing how numbers create what they purport to measure, I have argued that they are productive of the reality of violent conflict as they influence what counts as conflict and what does not.

In so doing, numbers not only shape how conflict is understood in the present but also create urgent futures upon which prevention actors draw in their advocacy and agenda-setting. More specifically, I have argued that the backward-looking measure of fatality numbers is forward-looking at the same time. They shape which kinds and extents of violence count as being indicative of future conflict as the proliferation of fatality numbers creates the expectation of conflict as inherently and quantifiably deadly. Within increasingly evidence-based and scientific reports, agendas and programmes of organisations of the prevention sphere, academic standards of conflict are imported into the realm of practitioners and policymakers. As my interviews have shown, this is not only the case where fatality numbers feature as implicit thresholds for the categorisation of conflict in conflict databases that get replicated in reports. In the programming and monitoring of international actors and organisations, analysts and conflict advisers also apply implicit and undefined thresholds of lethality as a criterion for violence to qualify as imminent or future escalation. In this way, fatality numbers create quantified ‘present futures’ (Berten and Kranke, 2022: 157) of conflict that signal urgency to policymakers. The imperative of preventing conflict relies centrally on claims about how global dynamics of violence will turn out in the absence of preventive action (Muggah and Whitlock, 2022: 10). That is, fatality numbers suggest that conflict prevention is needed, and if nothing is done, those numbers will only grow.

While the power to shape policy and power relations ‘by affecting how resources, status, knowledge and opportunities are distributed’ (Espeland and Stevens, 2008: 412) is a feature of quantification more generally, I have focused here on fatality numbers specifically – not only because they are widely, yet often only implicitly, used as a definitional criterion for war and conflict but because their use has important ethical implications. As the above examples illustrate, death numbers cited in policy reports, briefs and mission statements do not need to be accurate to fulfil their purpose of indicating the severity of conflicts, creating a focal point for advocacy and signalling urgency to policymakers. Rather, they primarily need to be large enough to warrant attention and action – the higher the number, the more significant the calamity (see also Auchter, 2016: 40–41; Espeland and Stevens, 2008: 405–406). Fatality numbers thus bolster the imperative to prevent, as they underline that war and armed conflict not only kill but that they kill at a significant scale – so much so that it can often only be expressed in eerily round numbers that will increase in the case of inaction.

Within this special issue’s focus on the production of quantified futures in international politics, the case example of conflict prevention provided only a first glimpse into the role of quantitative indicators of armed conflict, and fatality numbers specifically. More research is needed to understand their production, purpose and effects on the idea and practice of conflict prevention in more detail. For instance, future studies could drill down into the nuances of how organisations and individuals understand the task of prevention in light of the lethality expectation I outlined in this article, and how it shapes decision-making processes around preventive action. To do so, further research could ‘follow the indicator’ through immersive data-collection within international organisations, similar to the approach taken by Merry (2016), to trace where quantified indicators enter decision-making processes and what their effect is on action taken within specific organisations and their early warning mechanisms.

Finally, I have argued that fatality numbers, by definition, require violence to be deadly to count as war or armed conflict, thus deprioritising other types of violence that can play a role in the escalation of disputes into armed hostilities. This baseline expectation of war’s lethality, even if not explicitly quantified as specific thresholds, limits the scope of what comes to be understood as imminent or future conflict. As a result, this expectation produces an ethical dilemma in which a certain number of people need to die first before cases of violence become a concern for preventive action. Numbers of conflict such as death counts and estimates thus provide analysts and policymakers with a repertoire to measure, compare and communicate conflicts, yet at the same time limit the scope of what can be understood as ‘future conflict’. In addition, as they aggregate suffering and death into statistics and indicators, death numbers omit the specific circumstances of the killing and the life stories of the dead. In short, they decontextualise and dehumanise.

However, I do not argue that quantitative indicators and thresholds are right or wrong per se. Instead, as modes of governance, they shape decisions concerning international politics through mechanisms of knowing and applying power (Merry, 2016: 33). In their study on narratives of terrorism, Homolar and Rodríguez-Merino (2019: 564) argue that definitions of terrorism ‘deserve our attention [. . .] because they provide the discursive setting in which the interpretation of an event as [a] terrorist act takes place’. Similarly, with this article, I have intended to show how numbers, specifically those indicating deaths, provide quantitative-discursive frames for defining and interpreting (potential) violent events as wars and armed conflicts. I have thus joined Gregory (2021) in arguing for contextualising such numbers to account for violence beyond battle and the resulting fatalities. Putting quantification into context not only counters sterile and abstract representations of (future) war but also underlines the principal purpose of prevention by making visible the violence that is to be prevented.