Conflict, Peacekeeping, and Humanitarian Security: Understanding Violent Attacks Against Aid Workers

International Peacekeeping ISSN: 1353-3312 (Print) 1743-906X (Online) Journal homepage: https://www.tandfonline.com/loi/finp20 Conflict, Peacekeeping, and Humanitarian Security: Understanding Violent Attacks Against Aid Workers Kristian Hoelscher, Jason Miklian & Håvard Mokleiv Nygård To cite this article: Kristian Hoelscher, Jason Miklian & Håvard Mokleiv Nygård (2017) Conflict, Peacekeeping, and Humanitarian Security: Understanding Violent Attacks Against Aid Workers, International Peacekeeping, 24:4, 538-565, DOI: 10.1080/13533312.2017.1321958 To link to this article: https://doi.org/10.1080/13533312.2017.1321958 Published online: 17 May 2017. Submit your article to this journal Article views: 1383 View Crossmark data Citing articles: 3 View citing articles Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalinformation?journalcode=finp20

INTERNATIONAL PEACEKEEPING, 2017 VOL. 24, NO. 4, 538 565 https://doi.org/10.1080/13533312.2017.1321958 Conflict, Peacekeeping, and Humanitarian Security: Understanding Violent Attacks Against Aid Workers Kristian Hoelscher, Jason Miklian and Håvard Mokleiv Nygård Peace Research Institute Oslo (PRIO), Oslo, Norway ABSTRACT What factors explain attacks on humanitarian aid workers? Most research has tended to describe trends rather than analyse the underlying reasons behind attacks. To move this agenda forward, we present to our knowledge the first peer-reviewed cross-national time-series study that identifies factors related to violent attacks on humanitarian aid workers. Our theoretical framework explores two sets of potential explanatory factors: dynamics of conflicts; and the politicization and militarization of humanitarian operations. Using a global sample at the country level from 1997 to 2014, our results suggest that: (i) the presence and severity of armed conflicts are related to increased attacks on aid workers; (ii) aid workers do not appear to face greater risks even where civilians are targeted; (iii) the presence of an international military force does not appear to add to nor decrease risks to aid workers; and (iv) the effects of peacekeeping operations upon humanitarian security are varied. We discuss this in light of the ongoing challenges facing humanitarian organizations to provide security in fragile and conflict-affected areas. 1. Introduction In October 2015, an AC-130 gunship operating for the US Air Force made five bombing runs over a nondescript hospital in Kunduz, Afghanistan. Loaded with heavy military weaponry, the AC-130 hit and destroyed the hospital s emergency wing. It quickly became clear that this was not a military success, but one of the worst attacks on humanitarian aid workers in history. Fourteen Médecins Sans Frontière (MSF) staff members were killed, along with at least 24 patients and four caretakers. MSF called the attack proof of the global erosion of the rules of war, and many decried the event as evidence that the lives of aid workers are ever-more threatened. 1 Yet while this and numerous other examples show how on-ground conditions for humanitarian workers are becoming increasingly perilous, humanitarian aid funding is going through a golden age. Global aid spending by CONTACT Kristian Hoelscher hoelscher@prio.org Supplemental data for this article can be accessed at https://doi.org/10.1080/13533312.2017.1321958. 1 Miklian, Hoelscher, and Nygård, What Makes a Country. 2017 Informa UK Limited, trading as Taylor & Francis Group

INTERNATIONAL PEACEKEEPING 539 governments and private actors has increased 400% since 2000, amounting to $25 billion USD in 2014. 2 The number of international non-governmental (INGO) aid workers 3 has also quadrupled in the past 25 years, with 115,000 INGO aid workers in 1997, 210,000 in 2008, and 450,000 in 2014. 4 These increases reflect greater global societal instability, as some 58 million people were displaced in 2014 the highest total ever recorded. 5 Yet this international goodwill has also created hazardous side-effects. Aid delivery areas tend to be in conflict or crisis zones, increasing operational insecurity and occasionally blurring motivations for aid disbursement. Furthermore, as humanitarian, development, political and military goals becoming increasingly aligned, 6 these new models of humanitarian engagement may impact upon aid worker security. Figure 1 shows the locations of global aid worker attacks between 1997 and 2014, with bubble sizes corresponding to the number of attacks in each country. Four hundred and sixty-one aid workers were attacked in 2013, representing the most violent year ever in terms of absolute numbers, 7 and occurred disproportionally in conflict-beset countries such as Afghanistan, Pakistan, and Syria. This picture drives a narrative that aid work is becoming more dangerous 8 as conflict actors fail to adhere to maxims of humanitarian neutrality. 9 While the dynamics of conflict and the changing nature of humanitarian intervention are often regarded as factors that might explain why aid workers are attacked, we lack an evidence base that can support or challenge this. 10 This knowledge gap has important implications for security and risk protocols in humanitarian organizations; 11 and for donors and researchers 2 ALNAP, Humanitarian System and GHA, Global Humanitarian Assistance. 3 Aid workers are defined as the employees and associated personnel of not-for-profit aid agencies (both national and international) that provide material and technical assistance in humanitarian relief contexts. This includes both emergency relief and multi-mandated (relief and development) organizations: NGOs, the International Movement of the Red Cross/Red Crescent, donor agencies and the UN agencies belonging to the Inter-Agency Standing Committee on Humanitarian Affairs (FAO, OCHA, UNDP, UNFPA, UNHCR, UNICEF, UN-Habitat, WFP and WHO) plus IOM and UNRWA. The aid worker definition includes various locally contracted staff (e.g., drivers, security guards, etc.), and does not include UN peacekeeping personnel, human rights workers, election monitors or purely political, religious, or advocacy organizations. Humanitarian Outcomes, Aid Worker Security Database, np 4 ALNAP, Humanitarian System and Fast, Aid in Danger. 5 GHA, Global Humanitarian Assistance. 6 Carmichael and Karamouzian, Deadly Professions. 7 Humanitarian Outcomes, The New Normal. 8 Stoddard, Harmer, and Haver, Providing Aid; Humanitarian Outcomes, Unsafe Passage; and Brooks, Humanitarians under Attack. 9 Put into perspective, international aid workers are killed at a rate of approximately 50 deaths per 100,000 workers, below that of loggers (108/100k) or pilots (64/100k) in the United States according to BLS, National Census; and comparable to the homicide rate of violent urban centres such as Caracas, Venezuela (100/100k), and Kingston, Jamaica (50/100k) according to The Guardian, The 10 World Cities. 10 Hoelscher, Miklian, and Nygård, Understanding Attacks. 11 For example, see Fee and McGrath-Champ, The Role of Human Resources.

540 K. HOELSCHER ET AL. Figure 1. Number of aid worker attacks, 1997 2014. Note: Size of bubbles is proportional to number of attacks. Countries in grey have seen at least one year of internal armed conflict (as defined by Gleditsch et al., Armed Conflict ) over the same period. engaging with humanitarian insecurity. In response, this paper aims to better understand the spatial and temporal distribution of attacks on humanitarian workers, and how this relates to the dynamics of conflict and the nature of humanitarian engagement. The following section reviews the state-of-the-art on humanitarian security, and how conflict dynamics and the politicization of humanitarian engagement may underlie attacks on humanitarian workers. Section 3 outlines our theoretical framework and hypotheses, and Section 4 outlines our data and empirical strategy. Section 5 presents results and discussion, with Section 6 discussing limitations and caveats. Section 7 considers implications and concludes. 2. Background 2.1. Measuring and clarifying humanitarian insecurity An emerging literature has examined aid worker attacks over the past decade. Stoddard and co-authors, 12 Fast, 13 and Wille and Fast 14 have all made important contributions, and the Aid Worker Security Database (AWSD) is a considerable resource, a global compilation of reports on major security incidents involving deliberate acts of violence affecting aid workers. 15 The 12 Stoddard, Harmer, and Haver, Providing Aid; Stoddard, Harmer, and DiDomenico, Providing Aid. 13 Fast, Mind the Gap ; Fast, Aid in Danger. 14 Wille and Fast, Security Facts. 15 Humanitarian Outcomes, Aid Worker Security Database, np.

INTERNATIONAL PEACEKEEPING 541 Figure 2. Yearly aid worker attacks and total number of battle deaths, 1997 2014. AWSD has provided annual data since 1997 on lethal and non-lethal attacks on aid workers, and these data are instrumental in highlighting modes of aid worker deployment and patterns of aid worker attacks, including modality of attack, and institutional affiliation of the target. 16 Stoddard et al. 17 suggest that security protections have deteriorated, particularly for local UN staff. Wille and Fast 18 compare targeting of international and national staff, suggesting that fatalities of national staff have increased proportionately over time, particularly in national divisions of United Nations and Red Cross agencies. The key implication is that it has become more dangerous to be an aid worker in the field, 19 particularly in conflict-affected regions. Figure 2 shows the total number of global aid worker attacks from 1997 to 2014, and the number of battle-related deaths globally in the same period. 20 The figure presents a general upward trend in attacks on aid workers in recent decades and a correlation between aid worker attacks and conflict battle deaths. Two factors are likely responsible for the reported increase in attacks, particularly in recent years. First is the increasing number of aid 16 Data quality and coverage regarding attacks on humanitarian personnel are challenging. Agencies are highly protective of such information, and even more difficult to estimate and accurate number of global humanitarian personnel deployed each year. Other caveats include potential temporal biases in coverage and under-reporting of events, particularly non-lethal attacks on national staff. Despite this, the AWSD dataset represents the most complete record of aid worker attacks available. For a broader discussion of conceptual and data issues, see Fast, Aid in Danger; Weissman, The Meaning of Measuring ; and Neuman and Weissman, Saving Lives. 17 Stoddard, Harmer, and Haver, Providing Aid. 18 Wille and Fast, Security Facts. 19 Fast, Aid in Danger. 20 UCDP, Dataset.

542 K. HOELSCHER ET AL. workers in the field, with the period between 1997 and 2014 having seen a steady increase in the number of aid workers deployed. 21 While overall attacks have increased, per capita rates of attacks infer that the overall risk has stayed more or less constant since the late 1990s and potentially back to the mid-1980s. 22 Second, the increased number of total attacks appears driven by a small number of countries registering an above average number of incidents. Afghanistan and Syria, for example, have seen many more recent aid worker attacks than other countries. 23 While description of trends is important, one drawback of existing humanitarian security research is that it lacks a focus on explanation of underlying causal or proximate factors and relies primarily on anecdotal or ad hoc evidence to interpret or explain attack trends. While this type of analysis is important, such a focus may inadvertently encourage incomplete narratives about why aid workers are targeted, obscure important explanatory factors that may not be identified with descriptive data analysis, sideline issues related to addressing legal protection gaps and disparities in staff vulnerability, 24 and/or bias certain policy actions or security considerations in response. Given the changing role of aid in the dynamics of war 25 and the changing role of conflict and humanitarian engagement in the twentyfirst century 26 a closer examination is needed to better understand (and ultimately work to reduce) violence against humanitarian workers. 2.2. Understanding attacks on humanitarians With INGO agencies increasingly operating as close to the front lines as possible, employees often assume that protections are greater than they actually in dangerous situations. For example, a July 2016 attack on aid workers in Juba, South Sudan, left one dead and dozens more assaulted or raped 27 with neither UN Peacekeeping Forces nor local embassies responding to urgent requests for help. The event prompted calls for reform and increased accountability for attacks from both attackers and allies, 28 more deeply securitizing aid work as fundamentally off limits to combatants. Following, humanitarian experts have called for improved communication and risk management from the UN and from their own agencies to help staff leave before the conflicts escalate. 29 21 Humanitarian Outcomes, Aid Worker Security Database. 22 Sheik et al., Deaths among Humanitarian Workers. 23 The area to the right of the vertical line in Figure 2 and the shaded areas in Figure 3 suggest that attack increases are being driven by a small number of countries. 24 On this point, see Brooks, Humanitarians under Attack. 25 Wood and Sullivan, Doing Harm. 26 Donini et al., Humanitarian Enterprise. 27 Grant, South Sudan. 28 For example, McIlreavy, Enough Is Enough. 29 Grant, South Sudan.

INTERNATIONAL PEACEKEEPING 543 Figure 3. Expected aid worker attacks as conflict intensity increases (from Table 1). With these new operational imperatives to work at an increasingly dangerous frontline, uncertain support from peacekeeping forces, and an erosion of the perception of the apolitical aid worker, how can we better understand the current terrain of humanitarian insecurity? Two somewhat distinct lines of argument are used to explain why humanitarian workers come under attack. The first suggests that proximate factors related to the dynamics of conflicts that aid workers operate in and around can explain how they are targeted. Some link attacks to severity or intensity of fighting, noting that the vast majority of attacks occur in the countries where civil conflicts are most severe. 30 There is also evidence that humanitarian intervention itself may also intensify risks to aid workers in conflict zones by prolonging or exacerbating conflict dynamics. 31 However, beyond merely the presence or intensity of fighting, other characteristics of a conflict may also be important. Broadly, evidence suggests that conflict dynamics can influence the risks that humanitarian workers face. For instance, aid workers may be exposed to greater risk due to the vulnerable populations they work with, such as refugees or displaced populations. 32 Further, there also may be operational or strategic motives that make aid workers more likely to be attacked by rebel or government forces. While attacks 30 Humanitarian Outcomes, Unsafe Passage. 31 Narang, Assisting Uncertainty ; Nunn and Qian, US Food Aid. 32 Links between concentrations of displaced persons and the spread of conflict, terrorism, and instability have been emphasized by several authors including Choi and Salehyan, No Good Deed ; and Lischer, Dangerous Sanctuaries.

544 K. HOELSCHER ET AL. against aid workers are likely costly both in financial and reputational terms, goals of intimidation, civilian control or other strategic or operational logics may increase risks to workers in the field. 33 Logics of rational insurgent violence against civilians 34 may also mean that humanitarians face similar risks. Second, the politicization of humanitarian aid has been linked with increased risk. INGOs have a long history of seeing themselves as removed from but working in parallel to the conflicts they operate within. They are now working in more remote field settings and undertaking a wider variety of tasks 35 that demand a new range of security protocols. 36 However, as aid agencies have expanded from humanitarian work to more significant development portfolios that tend to be more political in nature, risks may be increasing as professed impartiality confronts the local politicization of humanitarian action. Reflecting this, there are growing similarities between corporate, military, and INGO operations and security strategies in fragile and conflict areas. 37 The impact upon humanitarian space includes the potential loss of perceived neutrality and impartiality important components for access and protection in contested areas. Embedded aid and integrated missions have re-ignited debates over the effectiveness and local perceptions of using armed escorts for aid workers, 38 particularly in heavily-militarized conflict zones such as Afghanistan. 39 Critics 40 argue that INGO links to Western governments to operationalize peace and security has eroded local perceptions of the value and neutrality of humanitarian space. Some believe this places INGO staff at additional risk as (t)he integration of politics and humanitarian action has been a major reason behind the attack on humanitarian aid workers and their inability to deliver aid to the neediest. 41 Some tie humanitarian insecurity to the militarization of aid, 42 and military-embedded humanitarian operations; 43 though recent empirical evidence 44 calls this into question. 45 Overall, there is a growing 33 Narang, Biting the Hand ; Crost, Felter, and Johnston, Aid under Fire ; Narang and Stanton, A Strategic Logic. 34 Wood, Understanding Strategic Motives. 35 See, for example, Miklian, Past, Present and Future ; and Sandvik and Hoelscher, War on Drugs. 36 Barnett, Humanitarianism Transformed ; Pringle and Lambrechts, The Risk of Humanitarianism. 37 Avant and Haufler, Transnational Organizations. 38 For example, Harmer, Integrated Missions. 39 Olson, Fighting for Humanitarian Space. 40 Important criticisms include: Chandler, Military Humanitarianism ; Mills, Neo-humanitarianism ; and Duffield, Macrae and Curtis, Editorial. 41 Abiew, Under Fire, 208. 42 Lischer, Military Intervention. 43 Barry and Jeffries A Bridge Too Far. 44 Mitchell, Blurred Lines. 45 The assumption that the aid NGO community was ever truly neutral or apolitical has been questioned, particularly as governments have securitized aid programmes for over 120 years according to Barakat, Deely, and Zyck, Tradition of Forgetting. The harshest critiques (Duffield, Challenging Environments ; Richmond, Emancipatory Forms ) view contemporary INGO work as little more than attempts to socially engineer societies based on western ideals in the process threatening the continued validity of the humanitarian enterprise (Donini et al., Humanitarian Enterprise). Studies of integrated UN missions

INTERNATIONAL PEACEKEEPING 545 consensus that the politicization of aid and its embeddedness within military operations may be contributing to greater humanitarian insecurity; 46 with shrinking humanitarian space, increased targeting or exposure of aid workers to frontline fighting, or threats of terrorism or anti-western sentiment increasing attack risks. 3. Theoretical framework and hypotheses Our theoretical framework focuses on two sets of factors explaining attacks on humanitarian workers: (1) the nature of conflicts that humanitarian agencies operate within and (2) the politicization of humanitarian operations. 3.1. Conflict dynamics We first assess assumptions that humanitarian security is related to conflict dynamics. Humanitarian agencies engage in myriad conflict and post-conflict environments, and travel and operational security for aid workers is more difficult in these situations. 47 Qualitative evidence suggests conflict dynamics play a significant role in how and where aid workers are targeted. 48 Quantitatively, for a small sample of countries with a high presence of aid workers, attacks were greater in countries with interstate wars and under conditions of civil violence but less likely where civil wars occurred. 49 We thus anticipate that the presence and severity of conflict in a country influences how severely humanitarian workers are targeted. Hypothesis 1a: Attacks on aid workers will be greater where conflicts are present in a country. Hypothesis 1b: Attacks on aid workers will be greater where conflicts are more violent. Further, the type of conflict may have an effect on how aid workers are targeted. In particular, in conflicts where insurgents are seeking secession, rebel groups might be more suspicious of the presence of aid workers. 50 We suggest: Hypothesis 2: Attacks on aid workers will be greater where conflict actors control territory or aim to do so. that merge military, humanitarian and political action are more nuanced (Combaz Integrated Missions ; Ferreiro, Blurring of Lines ; Donini, Between a Rock ). 46 Hammond, Protective Principles ; Spang, The Humanitarian Faction. 47 Humanitarian Outcomes, The New Normal; Humanitarian Outcomes, Unsafe Passage. 48 Fast, Aid in Danger. 49 Stoddard, Harmer, and Haver, Providing Aid. 50 See Hammond, Protective Principles, on the potential strategic reasons for targeting aid workers, particularly as related to targeted attacks serving as performative violence.

546 K. HOELSCHER ET AL. The strategic use of violence during a conflict may also influence how aid workers are targeted. 51 In examining this, we principally consider the extent to which combatants target civilian populations as a factor that may potentially increase aid worker risk. Broadly, where civilians make up a large percentage of casualties in a conflict, combatants typically use violence less discriminately against non-combatants, 52 and may therefore have fewer reservations against targeting neutral parties such as aid workers. Similarly, aid workers may often be at increased risk of collateral damage as they operate to ameliorate direct civilian humanitarian need. 53 However attacks of international aid workers in particular may represent a higher long-term reputational cost for combatants despite their possible short-term benefits. Supporting this, rebel group structure and organization may differentially incentivize attacks on civilians or aid workers. 54 Variations in respect for International Humanitarian Law 55 may also influence the degree of attacks upon aid workers. Despite these emerging findings, there is considerable evidence that violence against civilians and aid workers cooccurs. 56 Therefore, we assume that risks of attack for aid workers will increase where non-combatants and civilians are more frequently targeted. We propose the following hypothesis: Hypothesis 3: Attacks on aid workers will be greater where conflict actors are actively targeting civilians. 3.2. Humanitarian operations A second set of explanations study the nature of humanitarian INGO operations, particularly how aid agency linkages with military and development actors may create conditions that place aid workers at risk. Some argue that conducting humanitarian operations alongside the military places aid workers in greater danger, either by blurring lines between military actors and humanitarian agents, 57 or due to humanitarian workers being perceived of being biased towards a particular conflict party. 58 Moreover, certain conflict actors may be opposed to the real or perceived political, developmental, and humanitarian agendas of aid organizations. In such cases the presence of actors such as the United States military or NATO may encourage extremist 51 For example, Hammond, Protective Principles. 52 Eck and Hultman, One-Sided Violence. 53 See Ferreiro, Blurring of Lines on this issue and the shrinking of humanitarian space. 54 Narang, Biting the Hand. 55 Fazal and Konaev, When Do Rebel Groups. 56 Fast, Aid in Danger. 57 Duffield, Macrae, and Curtis, Editorial ; Mitchell, Blurred Lines. 58 Abiew, Under Fire.

INTERNATIONAL PEACEKEEPING 547 groups to attack humanitarian actors perceived to be agents of great powers. 59 We consider this in hypothesis four: Hypothesis 4: Attacks on aid workers will be greater where NATO or U.S. ground operations are present. Similarly, humanitarian security may be affected by operational integration between aid INGOs and United Nations peacekeeping operations (PKOs). UNPKOs are assumed to create space for secure humanitarian entry into post-conflict areas, though some argue this also leaves humanitarian workers vulnerable to attack given their international and/or Western associations. 60 Given the evidence that UNPKOs are generally peace-positive and create conditions that reduce likelihood of conflict and conflict diffusion, 61 our general assumption is that UNPKOs make aid workers safer, and that both the presence and size of UNPKOs will have an effect. 62 We propose that: Hypothesis 5a: Attacks on aid workers will be lower where UNPKOs are present. Hypothesis 5b: Attacks on aid workers will be lower the larger the UN peacekeeping force. The type of mandate employed may also be important. Broadly, traditional peacekeeping and peace enforcement missions with mandates of protection of civilians are shown to more effectively reduce violence against civilians. 63 Moreover, the type of mandate or action undertaken by PKOs appears to matter. More robust UNPKOs appear to reduce civilian targeting in conflicts, 64 reduce battle deaths, 65 and reduce the likelihood of conflict recurrence. 66 Similarly, transformational mandates those designed to address the roots of the conflict, such as economic reconstruction and institutional transformation (i.e. reform of police, army, judicial system, elections) are most effective at curtailing violence. 67 Based on this, we assume robust PKOs reduce likelihood for violence against aid workers, and hypothesize that: Hypothesis 5c: Attacks on aid workers will be greater where UNPKOs have traditional mandates rather than transformational mandates. 59 For example, Fast, Mind the Gap. 60 InterAction, A Humanitarian Exception. 61 Doyle and Sambanis, International Peacebuilding ; Beardsley, Contagion of Armed Conflict ; and Mealander, Where Murderers Lurk. 62 See Ruggeri, Gizelis, and Dorussen, Managing Mistrust. Collier, Hoeffler, and Söderbom, Post-conflict Risks, also show that increasing PKO budgets have positive effects for conflict reduction. 63 Kreps and Wallace, Just How Humanitarian and Hultman, Keeping Peace. 64 Hultman, Kathman, and Shannon, Civilian Protection. 65 Hultman, Kathman, and Shannon, Beyond Keeping Peace. 66 Hultman, Kathman, and Shannon, United Nations Peacekeeping Dynamics. 67 Hegre, Hultman, and Nygård, Simulating the Effect.

548 K. HOELSCHER ET AL. 4. Data and empirical strategy To examine our hypotheses, we use time-series data across a global sample of countries between 1997 and 2014, with the country-year as unit of analysis. Our baseline model is a negative binomial regression with standard errors clustered on the country level. 68 This estimates the count of events, here aid worker attacks, when events are over-dispersed compared to what the Poisson distribution would predict. 69 We fit a negative binomial regression where this over-dispersion is modelled by assuming that each observation follows the Poisson distribution, but in addition, a variable v i is added to the individual effects where e vi is gamma distributed with mean 0 and variance a. 70 Let i be index country-years, then our model is given by (the Offset parameter is simply a variable whose coefficient is fixed at 1, inclusion of this is necessary since we are comparing counts over different times): 4.1. Dependent variables y i Poisson(m i ), m i = exp (X i b + Offset i + v i ), ( e vi Gamma 1 ) a, a. Our main dependent variables measure lethal and non-lethal attacks on humanitarian workers between 1997 and 2014 using incident data drawn from the AWSD. 71 The AWSD uses systematic media filtering and information directly provided by aid agencies to compile counts and descriptions of global attacks on humanitarian aid workers, cross-checking figures with regional and field-level consortiums. 72 Our analyses primarily use combined lethal and non-lethal attacks on aid workers as the main dependent variable, yet also disaggregate analyses for aid workers killed, wounded, and kidnapped. 4.2. Independent variables Our key independent variables address factors related to (i) conflict dynamics, (ii) humanitarian operations, and (iii) other relevant political, economic and demographic factors. 68 As the data contain a large number of zero values, we also fit a zero-inflated negative binomial, as well as a Poisson regression and a simple OLS with country fixed and random effects to test robustness. Reported in supplementary files (appendix), these specifications yield substantively the same results. 69 Formally, this means that the variance of the counts does not equal the mean of the counts. 70 Hilbe, Negative Binomial Regression. 71 Humanitarian Outcomes, Aid Worker Security Database. 72 For full description of the data and methodology, see Humanitarian Outcomes, Aid Worker Security Database.

INTERNATIONAL PEACEKEEPING 549 4.2.1. Conflict dynamics To measure conflict presence (Hypothesis 1a), we use a dummy based on whether the UCDP/PRIO Armed Conflict Database (ACD) 73 registers an internal armed conflict in the country. To measure conflict intensity (Hypothesis 1b) we use dummies for high-intensity conflict and low-intensity conflict, and a measure of the total count of battle deaths in a given year, all from UCDP. For conflict type (Hypothesis 2), we use the UCDP/PRIO ACD measure of whether the conflict is territorial (challengers want either succession or regional autonomy); or governmental (challengers want to change the composition or political system of government). To test whether aid workers are attacked more frequently where civilians are targeted (Hypothesis 3), we include a measure of the number of deaths from one-sided violence in the country. 74 4.2.2. Humanitarian operations For Hypothesis 4 assessing foreign military intervention, we use a dummy indicating the deployment and presence of a NATO or US military ground mission in agivenyear. 75 We choose this as it is to our knowledge the best proxy measure to operationalize 76 situations where humanitarian and military personnel and operations are merged in both perception and reality. 77 We use NATO mission data from NATO; 78 while data for US ground missions is from Grossman. 79 In Hypothesis 5 we also measure the effect of presence (5a), size (5b), and mandate (5c) of UN PKOs. Here we use a dummy measuring the presence of a PKO and a measure of the total PKO budget, both derived from; 80 and dummy variables indicating whether the PKO mandate is traditional or transformational. 81 73 Gleditsch et al., Armed Conflict ; Pettersson and Wallensteen, Armed Conflicts. 74 Eck and Hultman, One-Sided Violence. 75 Data in Gaibulloev et al., Personnel Contributions, provide more detailed estimates of troop numbers in UN and non-un missions, yet was inadequate for the purposes of this article due to insufficient data coverage. 76 Further alternative measures to operationalize this variable may include where and how many US military personnel are deployed overseas. However, the US military generally has a small number of personnel in over 130 countries around the world, limiting the theoretical fit and inferential power of such a variable. 77 While drawing on the blurred lines argument relating foreign military presence as a motivating effect for attacks on aid workers, our proxy variable is not sufficient to test or infer this type of motivation. We instead simply assess the relationship between aid worker attacks and the presence of foreign military forces and stop short of claiming evidence for the motivations for these attacks. 78 NATO, Operations and Missions. 79 Grossman, U.S. Military Interventions. 80 Hegre, Hultman, and Nygård, Simulating the Effect. Kathman, Personnel Commitments, also provides data on number of peacekeepers deployed, yet a global sample of countries is unavailable. We instead use PKO budget and mandate data which are routinely used as proxies for troop size. 81 Doyle and Sambanis, Making War.

550 K. HOELSCHER ET AL. 4.2.3. Control variables We control for several additional factors. Generally, strong democratic and autocratic states are more able than weak or transitional states to ensure a modicum of territorial security and legitimacy, 82 and we use we use the absolute and squared polity score based on the Polity IV index 83 to measure political institutional consolidation. Economic conditions including low economic development and high inequality may play a role in how humanitarian workers are targeted, thus we control for GDP per capita 84 and income inequality. 85 We also include the log total national population since countries with larger populations generally see more conflict 86 and/or might attract more aid workers. To further control for other aspects of state capacity not captured by our political or economic independent variables, we include variables measuring the number of years since last regime change and the number of years the country has been in peace. Finally, to account for inertia effects of aid worker attacks and potential autocorrelation in the data, we include a lagged dependent in all estimations reported below. 5. Results and discussion Table 1 reports the first set of results for explanatory variables related to conflict dynamics, estimating their effect on the total count of both lethal and non-lethal aid worker attacks. Results strongly support hypotheses 1a and 1b that the presence and severity of conflict is related to greater numbers of attacks on humanitarian workers. The estimates in Column 1 for the effect of internal armed conflict are large and clearly different from zero. This holds for both minor conflicts (those incurring between 25 and 999 battlerelated deaths per year) and major conflicts (those incurring over 1000 battle deaths per year). Both in this estimation and all subsequent estimations we find, unsurprisingly, that aid workers are much more likely to be attacked in countries experiencing conflict than in peaceful countries. Somewhat surprisingly though, the difference between minor and major conflicts is not that large regarding the effects on the number of expected attacks. An average country with a minor conflict is likely to see seven aid worker attacks annually; while a similar country with a major armed conflict is only likely to see 14 attacks. 87 82 Hegre et al. 2001, Democratic Civil Peace ; Fox and Hoelscher, Political Order. 83 Marshall and Jaggers, Polity IV. The Polity IV index measures a country s placement on a 21-point scale between full autocracy ( 10) and full democracy (10). 84 World Bank, World Development Indicators. 85 Solt, Income Inequality Database. 86 Hegre and Sambanis, Sensitivity Analysis. 87 Proportionally this increase is, of course, substantial.

INTERNATIONAL PEACEKEEPING 551 The relationship also holds when measuring conflict intensity using battlerelated deaths (Column 2). Figure 3 illustrates the effect of battle deaths on the expected number of aid worker attacks. It simulates the expected count of aid worker attacks for an average country in conflict as log battle deaths increases, 88 showing a strong and significant effect of conflict intensity. An increase in battle deaths from around 400 deaths per year (a medium-intensity conflict) to 2000 deaths a year (a high-intensity conflict) roughly doubles the expected amount of aid worker attacks. For aid organizations the lesson here is clear, the more violent the situation they deploy to, the greater the risk they face. Column 3 analyses Hypothesis 2 to ascertain whether aid worker attacks are conditioned by aims of non-state conflict actors. Whether a conflict is fought over government or territory is only defined for countries in conflict, and consequently, we only include these states in this analysis. We find no evidence indicating that rebel motivation significantly influences aid worker attacks. Conflicts where rebels are seeking secession or regional autonomy (and where rebels often have a strong regional presence) do not show more attacks on aid workers than in conflicts motivated to change the government or the policies of the state. We find no evidence indicating that countries experiencing one-sided violence, i.e. where either the government or insurgents are actively targeting civilians, have higher rates of attacks on aid workers (Hypothesis 3). Column 4 reports that the effect of one-sided violence on aid worker attacks is essentially zero. We find no evidence that aid workers are at increased risk of attack where civilians are targeted, offering support to work by Narang 89 and others suggesting that there are different motivations to attack aid workers and civilians. As periods of one-sided violence are situations where civilian populations are especially vulnerable and in need of humanitarian assistance, that these situations do not appear to be comparatively more dangerous to aid workers may encourage the international community to provide greater support. Table 1 grouped all attacks on aid workers, both lethal and non-lethal, together, yet certain conflict dynamics may encourage different types of attacks on aid workers. Delving deeper, Table 2 distinguishes between counts of aid workers killed (Column 1), wounded (Column 2), and kidnapped (Column 3). Regarding the effect of conflict intensity, we find little or no difference between whether aid workers are killed or wounded, but that kidnappings occur at a much lower rate in minor armed conflicts than major armed conflicts. We also observe that economic development affects 88 King, Tomz, and Wittenberg, Improving Interpretation. An average country in conflict has a (log) population of 10; 11 years since last regime change, and 15 attacks at t 1. We use this country profile for all simulations below. 89 Narang, Biting the Hand.

552 K. HOELSCHER ET AL. Table 1. Negative binomial regression, Conflict Dynamics, 1997 2014. (1) (2) (3) (4) Conflict BRD Territory One-sided Minor conflict 1.170*** Dropped (0.287) Major conflict 2.094*** 0.943*** (0.375) (0.200) ln (Battle deaths) 0.146** (0.048) Territorial 0.100 (0.360) One-sided violence 0.000988 (0.001) ln(population) 0.0123 0.00338 0.00749 0.00713 (0.030) (0.027) (0.038) (0.027) ln(gdp capita) 0.0123 0.0124* 0.0119 0.0129* (0.006) (0.006) (0.007) (0.006) ln(time in peace) 0.145 0.208* 0.0532 0.206* (0.104) (0.101) (0.174) (0.105) Time since regime change 0.579*** 0.565*** 0.218 0.489** (0.149) (0.144) (0.160) (0.149) Polity 2 0.335** 0.429** 0.133 0.571*** (0.116) (0.138) (0.111) (0.118) Polity^2 0.00686 0.00844 0.0245* 0.00853 (0.005) (0.005) (0.013) (0.006) Aid worker attacks (t 1) 0.126*** 0.127*** 0.0754*** 0.137*** (0.036) (0.034) (0.013) (0.031) _cons 1.076 0.822 1.518 0.568 (1.168) (1.122) (1.727) (1.162) Year dummies Yes Yes Yes Yes lnalpha _cons 1.675*** 1.785*** 1.187*** 1.782*** (0.166) (0.165) (0.172) (0.163) AIC 2994.6 3038.1 1589.6 3036.9 ll 1470.3 1493.1 783.8 1492.4 N 2641 2641 408 2641 Note: Country clustered standard errors in parentheses. *p <.05, **p <.01, ***p <.001. lethal and non-lethal attacks differently. While less developed countries see more killed and wounded aid workers, this is not the case for kidnappings where the effect of log GDP per capita is indistinguishable from zero. Table 3 looks at how the presence of international military forces (Hypothesis 4) and the dynamics of UNPKOs (Hypothesis 5) affect violence against aid workers. Hypothesis 4 tests whether the presence of NATO or US forces is related to greater risk, drawing upon the blurred lines argument which posits that due to perceptions that humanitarian and military agendas are conflated, attacks on aid workers will occur more frequently in countries where foreign militaries are present. Results of the effect of US military or NATO deployment on aid worker attacks are reported in Column 1. We find no evidence indicating that countries where such forces are present have more attacks on aid workers, with the estimated effect failing to even approach significance, and unable to be reliably distinguished from

INTERNATIONAL PEACEKEEPING 553 Table 2. Negative binomial regression, Conflict Dynamics, aid worker attacks disaggregated, 1997 2014. (1) (2) (3) Killed Wounded Kidnapped Minor conflict 1.311*** 1.081** 0.877* (0.303) (0.352) (0.425) Major conflict 2.710*** 1.862*** 2.392*** (0.385) (0.434) (0.621) Polity 2 0.0458 0.0182 0.0440 (0.030) (0.033) (0.038) Polity^2 0.00898 0.00892 0.0194* (0.007) (0.007) (0.008) ln(population) 0.0602 0.239* 0.150 (0.131) (0.111) (0.125) ln(gdp capita) 0.595*** 0.643*** 0.427 (0.150) (0.142) (0.254) ln(time in peace) 0.406* 0.263* 0.329* (0.171) (0.117) (0.153) Time since regime change 0.00835 0.000785 0.00729 (0.006) (0.008) (0.010) Aid worker killed (t 1) 0.250*** (0.057) Aid worker wounded (t 1) 0.205** (0.073) Aid worker kidnapped (t 1) 0.285** (0.104) _cons 1.140 0.0388 0.333 (1.131) (1.520) (1.762) Year dummies Yes Yes No lnalpha _cons 1.340*** 1.700*** 2.635*** (0.194) (0.211) (0.391) AIC 1783.6 1920.3 1318.9 ll 864.8 933.1 648.4 N 2641 2641 2641 Note: Country clustered standard errors in parentheses. *p <.05, **p <.01, ***p <.001. zero. This supports Mitchell, 90 who challenges the idea that risks are increased for humanitarian actors operating in countries with an international military force present. To further examine the role of military presence we look at PKOs in Hypothesis 5, and estimate the effects of peacekeeping presence, budgets, and mandates. Results for the presence of UNPKOs (5a) are shown in Column 2. We find that countries with UNPKOs present see more attacks on aid workers. This is not surprising. The coarse dummy measure essentially only distinguishes between countries that currently are in or have recently been in conflict and those in peace, and given this these results align with our findings in Hypothesis 1. We unpack this further in Column 3, which reports results for the budget of the PKO force (5b). For this, we find a positive 90 Mitchell, Blurred Lines.

554 K. HOELSCHER ET AL. Table 3. Negative binomial regression, Humanitarian operations, 1997 2014. (1) (2) (3) (4) NATO/US Presence Budget Mandate Conflict 1.058*** 1.083*** 1.100*** 1.066*** (0.203) (0.199) (0.203) (0.195) NATO/US 0.266 (0.353) PKO 0.958** (0.337) ln(pko budget) 0.143* (0.058) PKO transformational 0.375 (0.429) PKO traditional 1.354** (0.476) Polity 2 0.0133 0.0336 0.0293 0.0413 (0.030) (0.031) (0.031) (0.033) Polity^2 0.0121 0.0174** 0.0151* 0.0197*** (0.006) (0.006) (0.006) (0.005) ln(population) 0.148 0.210* 0.198 0.218* (0.102) (0.105) (0.104) (0.106) ln(gdp capita) 0.586*** 0.470** 0.455** 0.503** (0.143) (0.175) (0.173) (0.185) ln(time in peace) 0.340** 0.210 0.252* 0.201 (0.123) (0.112) (0.121) (0.110) Time since regime change 0.00702 0.00588 0.00654 0.00617 (0.005) (0.005) (0.005) (0.005) Aid worker attacks (t 1) 0.125*** 0.127** 0.125** 0.130*** (0.035) (0.040) (0.040) (0.039) _cons 1.086 0.777 0.827 0.999 (1.126) (1.139) (1.133) (1.178) Year dummies Yes No No No lnalpha _cons 1.673*** 1.728*** 1.735*** 1.724*** (0.165) (0.175) (0.177) (0.172) aic 2994.2 2998.5 3008.0 2995.2 ll 1470.1 1488.2 1493.0 1485.6 N 2641 2641 2641 2641 Note: Country clustered standard errors in parentheses. *p <.05, **p <.01, ***p <.001. result that is clearly distinguishable from 0. In other words, the larger the budget of the peacekeeping force which we assume is highly correlated with the size of the force the more attacks against aid workers we expect to see. These results could be the effect of a selection bias, in that UNPKOs occur in more violent areas where the risk of aid worker attacks would likely be higher. There is a limit to the extent to which we can address these types of endogenous effects in the present paper, but as a first attempt we re-run the analysis in the last two columns using genetic matching. 91 Results of the matching analysis, and a more detailed description of the procedure, 91 Sekhon, Matching Package for R. See also Gilligan and Sergenti, Do UN Interventions, for an example of matching analysis applied to UN interventions.

INTERNATIONAL PEACEKEEPING 555 Figure 4. Expected number of aid worker attacks as (log) PKO budget increases (from Table 3). are reported in appendix Table A2. As expected, the results of PKO variables is weaker after matching, a clear sign that there is a selection effect. Nonetheless, we still find a positive and significant effect of both the PKO budget and mandate variable. This effect is substantively interesting, and is simulated in Figure 4. It shows that an increase in UNPKO budgets from USD $7 million, the budget of a small observer force like the 1988 1991 United Nations Iran Iraq Military Observer Group, to USD $400 million, the budget of a force the size of United Nations Interim Force in Lebanon, doubles the expected count of aid worker attacks. At first glance this might appear disheartening deploying PKO forces is associated with more attacks on aid workers. 92 This finding, however, is mediated by the result for the type of PKO mandate (Hypothesis 6c) reported in Column 4. Here PKOs with traditional mandates are indeed associated with more aid worker attacks; whereas there is no effect for PKOs with transformational mandates. This result supports recent research showing that only PKOs with transformational mandates are effective at reducing conflict. 93 In contrast, traditional peacekeeping forces mostly tasked with observing the terms of truce or peace agreements, or policing a buffer zone and assisting in negotiating a 92 Conversely, the UN may selectively deploy larger PKOs to situations where aid workers are more likely to be attacked. We cannot rule out such a relationship using our approach here. If anything, however, this contradicts the more general argument that PKOs are only deployed to the easy cases, yet a recent study by Vivalt, Peacekeepers Help, that tries to tackle this endogeneity finds no support for such an argument. 93 Hegre, Hultman, and Nygård, Simulating the Effect.

556 K. HOELSCHER ET AL. peace agreement are not equipped to use lethal force to protect themselves and civilians, and seem less capable of creating environments safe for aid workers. Nevertheless, we find little clear support for Hypothesis 5, as the presence of PKOs, either traditional or transformational, and regardless of budget, does not reduce aid worker attacks. In addition to analyses reported here, in supplementary files (appendix) we report robustness tests (appendix Table A1) that check (1) whether the results are sensitive to modelling excess zeros in the data, or (2) sensitive to assuming a Poisson instead of a Negative binomial distribution. We find our results are not sensitive to these modelling choices. Second, it could be that aid worker attacks are associated with a particular set of countries or that we have omitted time-constant country-level variables. We therefore re-estimated all our models with OLS regressions with country fixed effects and tested a subset of our result with country random effects. Last, given the high number of aid worker attacks occurring in Syria and Afghanistan, it could be that these particular cases drive our results. We therefore re-estimated without these two countries. Our results are robust to this, and suggest that our findings are not driven by the presence of absence of potentially influential outliers. 6. Limitations to the analysis This study is among the first to analyse attacks against aid workers. As such, we recognize that the present analysis has a number of limitations that future research should address. First is the quality of the data. Despite the AWSD providing a robust data source, refinements and improvements may include: the quality of coverage and extent of reporting of attacks; unpacking what constitutes an incident ; and important in determining risk rates accurately defining and determining the number of workers in the field. 94 While the data landscape can be improved, we note that if a definition is consistent over time, and a database populated with a consistent definition is used, then it does strengthen our ability to draw inferences. However, we are still left with limitations connected to the highly aggregated nature of available data, and are unable to perform potentially important sub-national analysis. Further, existing data are ill-equipped to examine certain questions including: how agency deployment methods influence the number of attacks per aid worker; which types of aid workers under which conditions are most at risk; gender disparities; the forms, roles, and activities undertaken 94 For example, Fast, Mind the Gap.

INTERNATIONAL PEACEKEEPING 557 by various organizations; and, to an extent, the question of if the data itself are comprehensive enough to register all attacks. 95 Regarding the spatial unit of analysis, a large literature has documented the local-level dynamics of conflict. 96 In conducting a country-year level analysis, we aggregate what are likely often local phenomena up to the country level. This potentially masks large within-country variation in the location and causes of aid worker attacks, which may be important in countries such as the Democratic Republic of Congo and Syria. Therefore, as availability of geo-located data on where aid worker attacks occur is limited, it is possible that attacks we associate with ongoing conflicts may occur far from actual conflict zones. Alternatively, in certain countries aid workers may stay clear of active conflict zones, reducing their exposure and risk and thus depressing the number of reported attacks in the country overall. Moreover, data may also include attacks possibly unrelated to field activities done by aid workers. Humanitarian security may also be affected by decisions regarding the characteristics of aid workers deployed. As noted, there are concerns that national staff are placed in more precarious situations than international counterparts. Yet given their local knowledge and profile, local workers may be a safer alternative to visible expatriate staff in dangerous settings. Such dynamics could influence the results reported above, but unfortunately, we lack adequate data to either test or control for this using the empirical strategy here. We also recognize that the most dangerous environments for aid workers may fall outside the bounds of this study as we look at where aid workers are, not where they have left or where situations are too risky to enter. Getting answers might necessitate in-depth case studies of large international humanitarian organizations. This would enable an examination of precisely the factors that influence how organizations make decisions about how, why, who, and where to deploy. Combining this with case studies of conflict dynamics in relevantcountriescouldallowamappingofhowconflictdynamicsinteractwith various aspects of humanitarian organizations operations. Further, different humanitarian organizations may have fundamentally different ways of dealing with the risk of attacks. Ideally, comparative research would examine varied security protocols and operational responses to begin to answer these questions. 7. Conclusions In this article, we have considered humanitarian risk and integration to inform an examination of the factors associated with violence against aid 95 In addition to those discussed here, certain limitations may relate to specific aspects of the data. For example, kidnappings are under-reported for various reasons, including reputational issues for INGOs and not setting a precedent. Killings, conversely, are nearly always reported. Economic attacks (such as kidnapping or extortion) might therefore be under-represented in the data. 96 For example, Buhaug and Gates, Geography of Civil War.