Abstract
A prominent explanation of the rising attacks on aid workers is that aid workers in proximity to peacekeepers or soldiers of an intervention force are seen to be associated with them. This article leverages high-quality, disaggregated data collected by the UN about attacks on humanitarians in Darfur between January 2008 and March 2009 to examine this “blurring of the lines” explantion. We argue that rather than a blurred lines logic, spatial clustering of peacekeepers and humanitarians in a given area explains why aid workers are often attacked in proximity to peacekeepers. Consistent with this argument, we find that when controlling for humanitarian activity, the number of peacekeepers in an area is no longer significantly associated with attacks on aid workers. This study has important implications for current discussions on the relationship between humanitarians and peacekeepers and the concept of humanitarian space.
1 Introduction1
The emergence of new, disaggregated datasets focused on violence against civilians has facilitated analysis of the dynamics and strategic logic of belligerents in armed conflict.2 Likewise, analyses of the effects of peacekeeping interventions have shed light on how these interventions affect attacks on civilians.3 An emerging body of research examines attacks on peacekeepers,4 the obstruction of peacekeepers,5 and attacks on other third parties such as aid workers and human rights defenders.6 Less understood, however, is how peacekeeping interventions affect other types of interveners. Several studies suggest that the presence of peacekeepers increases the level of risk for humanitarian and aid workers.7 A common explanation for this finding is that aid workers in proximity to peacekeepers or soldiers of an intervention force are seen to be associated, which makes these aid workers relatively likely to be attacked.8
Most of the quantitative studies cited above that have examined the links between peacekeeping presence and attacks on aid workers are cross-national in nature and therefore unable to unpack the proposed causal mechanisms. This raises other questions: What accounts for the association between the presence of peacekeepers and attacks on aid workers? How does the presence of peacekeepers affect aid workers? Are aid workers attacked because belligerents view them as associated with the kinetic or politicized actions of peacekeepers?
We argue that previous studies that find an association between number of peacekeepers and attacks on humanitarians fail to control for the extent of aid operations in proximity to peacekeeping interventions. In this article, we draw on a rich and detailed dataset of violence and insecurity in Darfur from 2008–2009 to examine the micro-level dynamics of attacks against aid workers. Crucially, we try to account for the number of aid operations that cluster around peacekeeping bases. We find that when controlling for aid activities in a given area, the significant effect of the number of peacekeepers in that same area on attacks on aid workers disappears. Thus, instead of explanations relying on the negative ramifications of political or securitised associations on the safety of aid workers, humanitarians in close proximity to peacekeepers might simply be more likely to be attacked because violence, peacekeepers, and humanitarians are all highly clustered in space. In other words, the causal mechanism at play is physical proximity rather than other associations. Given the limited nature of our dataset (temporal and location), however, we are reluctant to extrapolate these findings to other contexts. Instead, this finding illustrates the crucial importance of mixed methods in unpacking the micro-level dynamics of attacks on aid workers.
This article makes at least three essential contributions to the literature. First, it adds to the emerging quantitative literature that looks at the micro-level to explain violence against aid workers. Despite a growing literature focused on the drivers of attacks against aid workers, many of these theoretical discussions lack strong empirical evidence, illustrating a persistent and pernicious gap between theory and existing evidence.9 For example, in an important contribution to the literature on attacks against aid workers, Narang and Stanton analyse variation in violence against aid workers across Afghanistan’s 34 provinces between 2008 and 2012. Narang and Stanton, however, do not look at the impact of the number of international soldiers in proximity to aid workers on violence.10 To our knowledge, we conduct the first study that examines the effects of associational logics using a micro-level research design, making this study distinct in design and location.
Second, we attempt to deal with what Fast refers to as the ‘denominator issue’ when studying violence against aid workers: high or low levels of violence against aid workers may simply reflect many or few humanitarian personnel providing assistance or protection in a given area, so the presence of aid workers within a given context needs to be accounted for in research design.11 Third, and perhaps most importantly, our study emphasizes the need to examine attacks against aid workers in a holistic manner: that is, to not only examine violence against aid workers, but also to examine incidents that affect aid delivery. While highly publicized and tragic examples of extreme violence against aid workers, such as killing and sexual violence are important, in terms of impact on aid delivery it is also essential to consider this alongside more mundane forms of obstruction which may cumulatively prove a greater impediment to humanitarian work. Though extreme forms of violence are, of course, abhorrent, our data illustrate the more numerous obstructive types of events affect aid work, such as hijackings or theft, to potentially much greater cumulative effect. The complexity of these attacks, combined with our findings, point to the importance of mixed methods in understanding causes and consequences.
This article proceeds as follows. The first section reviews the literature on attacks on aid workers, after which we present our argument that rather than being the result of associational logics, the relationship between peacekeeping presence and violence against aid workers is a spurious relationship that results from spatial clustering. Next, we present our research design. The subsequent section summarises the main findings and the results of robustness checks. The final section concludes with implications for future research.
2 Explaining Violence against Aid
It has become something of an accepted truth of our times that aid workers are increasingly seen as targets when operating in spaces affected by violence and conflict,12 though some have sought to challenge and provide more nuance to this claim.13 The literature on ‘new wars’ provides one explanation for increased attacks on aid workers.14 Rather than pointing to changing dynamics in international interventions, this explanation points to changing conflict dynamics among belligerents. Others have pointed out that aid workers are perhaps less careful and therefore more likely to be attacked in the proximity of peacekeepers.15
However, the most prominent explanation of an increase in attacks on aid workers is the role that the politicized or securitized aid has played in compromising humanitarian principles,16 and by extension, is responsible for increased violence against aid workers.17 This trend is commonly presented as a consequence of a lack of respect for either humanitarian space or for international law. Since many aid workers operate in proximity of military personnel (and in some cases cooperate with them), they are seen to be associated to political, security, or military agendas.18 We explore the literature on this explanation of attacks on aid workers below.
2.1 Crowding Humanitarianism Space? Associations between Aid and Peacekeeping
The concept of humanitarian space has received much attention.19 Much of this literature suggests that humanitarian space is ‘shrinking’ or ‘under attack’.20 Four main reasons are typically cited as the reasons for this: military involvement in humanitarianism; the use of aid as a foreign policy tool; the rise of multi-mandate organisations; and ‘new’, political forms of humanitarianism.21 All four of these explanations rely on claims about the principled nature of humanitarian action – that humanitarian space has previously been respected because of its adherence to humanitarian principles of humanity, neutrality, impartiality and independence – 22 and that the erosion of humanitarian space and, in turn, humanitarian principles, is simultaneously responsible for attacks on humanitarian action.
The beginning of the erosion of humanitarian space is typically traced to the rise in interventionism following the end of the Cold War. The so-called ‘New World Order’, heralded by the first Gulf War, offered military solutions to conflict and suffering around the world. In Conor Foley’s words, ‘humanitarianism went to war’.23 The problem of military involvement in humanitarian action, for those who see it as threatening humanitarian space, only intensified following September 11th 2001. In the context of the War on Terror, governments focused their foreign aid budgets on providing assistance to ‘failed’ or ‘fragile’ states because they believed these states, and the destabilising conflicts within them, posed a threat to Western security by providing ‘breeding grounds’ for terrorism. This obviously securitized and politicized aid, the claim goes, undermines principled humanitarian action and international law.24
The invasions of, first, Afghanistan in 2001, and Iraq in 2003 by the US and its allies are seen as particularly problematic, as commentators have noted: ‘[o]ne of the legacies of the Afghan adventure is the blurring of lines between humanitarian and military operations”.25 The specific charge is that military involvement in humanitarianism blurs the lines between the military and humanitarian actors, which makes them appear associated. This logic became more visible following comments from high profile individuals such as former US Secretary of State Colin Powell calling ngo s ‘force multipliers’ in Iraq26 or US Special Envoy Richard Holbrooke attributing intelligence information in Afghanistan to aid agencies.27 As a result, “aid agency staff, both international and local, are increasingly seen as legitimate targets by combatants … and this prevents them for fulfilling their humanitarian mandate”.28
There is no such thing as impartial humanitarian assistance or cmo [civil-military operations] in coin. Whenever someone is helped, someone else is hurt, not least the insurgents. So civil and humanitarian assistance personnel often become targets. Protecting them is a matter not only of providing a close-in defense, but also of creating a secure environment by co-opting local beneficiaries of aid and their leaders.29
Peacekeeping missions increasingly rely upon coin strategies.30 The UN Security Council has mandated peacekeepers in the Democratic Republic of Congo (drc) to ‘neutralise’ and ‘disarm’ certain armed groups and has mandated peacekeepers to use all necessary means to ‘stabilize’ the Central African Republic (car) and Mali.31 These types of mandates not only alter the way peacekeepers operate, but also the relationship between peacekeepers and humanitarians, often through the sharing of data and information. For instance, UN peacekeepers and humanitarian organizations often share information, which undermines humanitarian neutrality if the information is used to guide UN military operations.32 ngo s and UN agencies, such as the United Nations Office for the Coordination of Humanitarian Affairs (ocha), use and contribute to common operational datasets (cod s), to which UN peace operations also have access. Given ngo and UN agency physical presence in armed conflicts, these data are valuable to UN peacekeeping operations. Information analysts in the UN peacekeeping operation in Mali have developed “targeting packs” on groups and individuals who are considered a threat to the mission. They then shared these targeting packs with French intelligence supporting counter-terrorism operation in Mali.33 This not only blurred the lines between the UN peacekeeping activities and the French counter-terrorism operation in Mali, but also indirectly the lines distinguishing humanitarian operations from UN peacekeeping action. According to Karlsrud, for this reason humanitarians increasingly “insist on their neutrality and are wary about sharing more sensitive information with peace operations, especially those with robust mandates”.34 In short, this explanation for attacks, centered on the compromised humanitarian space narrative, follows a logic that holds that aid workers in proximity of peacekeepers or soldiers of an intervention force are seen to be associated, which increases the risks for aid workers.35
What are the consequences of an increased operational association between peacekeepers and humanitarians? ngo s have long claimed that they are attacked because humanitarian space has been compromised by the shared operational space and interaction between aid and military activities.36 In the 1990s, aid workers in Kosovo were arrested and charged with spying, all a result of the maps and sundry items that could be used for intelligence gathering.37 This was a particular issue in the post-9/11 wars in Afghanistan and Iraq. For example, calls to reclaim humanitarian space and renew the separation between the military and humanitarian actors grew stronger in the aftermath of the bombing of the United Nations headquarters in Iraq in August 2003.38 Numerous other examples illustrate the squeeze on humanitarian space in Iraq and Afghanistan. Military forces in both Iraq and Afghanistan provided food and other assistance, often while wearing civilian clothing but still while carrying weapons. Humanitarians objected, pointing out that such acts made it difficult to distinguish military forces from civilian aid workers. The yellow food packets dropped in Afghanistan as part of the humanitarian effort also, disturbingly, resembled the cluster bombs the US military also dropped to target Taliban and al-Qaeda forces. After human rights and other groups objected, the US military changed the colour of the cluster bombs.39
More recently, humanitarians have expressed similar concerns about associational logics, or a blurring of these lines, in the drc. For instance, several ngo s issued a complaint to the UN in July 2014 opposing the use of drones by the UN for humanitarian information gathering. Since UN peacekeepers were using drones,40 ngo representatives believed that the use of drones for humanitarian purposes would blur the lines between the military and humanitarian actors in the conflict. As one ngo representative put it, “If data gathered during a flight with a humanitarian objective informs combat operations or is used for military intelligence, there is a clear compromise of neutrality”.41
Some quantitative evidence does indeed exist to support the claim that the presence of peacekeepers in a given country makes attacks on aid workers more likely. In a study on attacks on aid workers – comprising 408 major incidents of violence against aid workers between 1997 and 2005, Stoddard and colleagues examined the impact of contextual circumstances on the likelihood that aid workers were targeted, including, among others, the intensity of the conflict, the presence of UN peacekeeping forces, and the presence of global terrorist movement cells. Of all the contextual variables studied, they found only the presence of UN peacekeeping forces significantly influenced the likelihood that aid workers were attacked. Where UN peacekeepers were present, they reported “a slight but statistically significant increase in the number of international staff victims”.42
In a more recent study using the same dataset, Hoelscher and colleagues found a correlation between the presence of peacekeepers in a country, and took their analysis a step further. Recognizing that the presence of a peacekeeping mission often occurs simultaneously to armed conflict, they examined the budget of peacekeeping missions and proposed that “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”.43 Their findings confirmed their hypothesis. They did not, however, point out that the budget of a peacekeeping mission is also highly correlated with the severity of a conflict.
Two issues arise, however, in testing the associational or blurred lines argument. The first concerns the question of intent: without explicitly asking those attacking aid workers, determining intent – and the role of such associations in motivating such attacks – is virtually impossible, particularly in cross-national or even national comparisons over time. The second, which Hoelscher and colleagues highlighted in their analysis, is the correlation among armed conflict, peacekeeping, and aid worker presence. We turn to this issue next.
2.2 The Denominator Issue and the Spatial Clustering of Aid Workers and Peacekeepers
In this article, given the challenges of investigating intent, we instead explore associational logics and the blurred lines hypothesis by examining proximity. We argue that aid workers are likely to be attacked in close proximity to peacekeepers because the locations of peacekeepers, aid workers, armed violence, and attacks on aid workers are not randomly distributed. Peacekeepers are likely to be deployed where armed fighting is taking place,44 and areas in which violence takes place often require humanitarian aid. Humanitarian organisations are therefore likely to be present in close proximity to peacekeepers for protective and logistical reasons, in that both have mandates to protect civilians and they are present in larger towns or cities, which are easier to supply. Moreover, the UN’s presence in situations of armed conflict often includes both peacekeeping forces and humanitarian agencies. For example, the UN in the drc not only consists of the peacekeeping mission (monusco), but also 21 programs, funds, and specialized agencies working on development and delivering humanitarian assistance across the drc.45
In addition, the extent of violence in a given area positively influences whether the UN establishes a peacekeeping base,46 while areas with a higher conflict intensity also needs more humanitarian aid. Hoelscher et al. found, in turn, that the presence and intensity of armed conflicts increased the risk of attacks on humanitarian workers. Compared to different types of violent attacks, abductions were especially likely in more intense armed conflict.47 In other words, the more violence that takes place in a given area, the more humanitarians and peacekeepers will operate in this area, and those higher levels of armed fighting are associated with more attacks on humanitarians. Indeed, in the first published study that looks at sub-national determinants of violence against aid workers, Narang and Stanton found that armed violence, the number of aid workers operating in an area, and attacks on aid workers were all highly clustered in space.48
Hence, when assessing whether attacks on aid workers are correlated with the presence of peacekeepers, it is crucial to account for the number of aid workers operating in this area, what Fast refers to as the “denominator issue”.49 In an analysis of violence against aid workers, Barnett made a similar point with regard to the variation in violence against aid workers in different countries: “A high number of deaths in a country may merely reflect the fact that there were many humanitarian personnel in those countries, and conversely the low number of deaths may be because there were very few humanitarian personnel present”.50 Specific attacks on aid workers resulting in death or injury represent only a fraction of the range of attacks affecting aid workers, agencies, and aid delivery in a conflict zone. Aid agencies receive threats against programmes and personnel, aid workers are harassed and stopped at checkpoints and sometimes even abducted, and vehicles, equipment or medical supplies are stolen from compounds, warehouses, and medical facilities. None of these necessarily result in death or injury to aid workers. Yet all of these examples affect the delivery of assistance and provision of protection, which accounts for the presence of aid workers in the first place.51 All too often these types of events are excluded from analyses that focus only on aid workers.
Hypothesis 1: High numbers of peacekeepers do not increase the likelihood of attacks, defined as incidents in which aid workers are killed or injured as a result of the use of force.
Hypothesis 2: High numbers of peacekeepers do not increase the likelihood of an incident that affects aid delivery, including incidents of hijacking, abduction, or theft.
3 Data and Methodology
To investigate the hypotheses above, we conduct a statistical analysis on the correlates of attacks on aid workers in Darfur, as well as incidents that affected aid delivery but did not necessarily result in death or injury of aid workers, such as hijacking, abduction, or theft. We draw on a unique dataset generated by the Joint Missions Analysis Centre (jmac) between 3 January 2008 and 6 April 2009.52 A crucial aspect of any jmac unit within a peacekeeping mission is to provide information that helps peacekeepers in fulfilling the mandate of the mission.53 Since unamid’s deployment to Darfur in 2007 expressly mandated providing a secure environment for humanitarian assistance,54 the jmac dataset is well-suited to study attacks on aid workers and incidents that affect aid delivery. To check the comprehensiveness of our data, we compared the jmac incident data – relating to attacks on aid workers, hijacking, abduction, and theft – to two other datasets that record data on these incidents. This comparison, which we include in the appendix, illustrates the comprehensiveness of the jmac data.
“On 21 February, unamid was informed that two local staff members of an ingo, [name withheld], were killed and four injured in an attack by Arab militia. The incident took place in the Sand Valley between Karamje and Dengia”.55
“On 11 March, unamid forces were advised by unamid Police that six unidentified armed men had attacked the [name withheld] clinic in Saraf Umra. Three international staff members and two national staff members were abducted. The national staff members were subsequently released. […] The kidnappers are reportedly demanding a ransom of two million Sudanese Pounds. […] [Name withheld] is going to evacuate staff members from Saraf Umra today”.56
This observation also exemplifies why we have included abductions in our aid delivery category, since it illustrates the disruption on aid delivery (for the clinic and the subsequent evacuation of aid workers) even as it does not mention any harm to aid workers.
“On 21 March, a group of unidentified armed men robbed a former ingo ([name withheld]) warehouse at the Al Salaam idp Camp. The bandits made off with all the contents of the warehouse. GoS police and the Humanitarian Aid Commission (hac) are investigating”.57
The variable that records whether aid workers were violently attacked and the variable that records attacks affecting aid delivery are not mutually exclusive. For instance, a hijacking in which humanitarians are killed would fall within both categories because the event affects both aid workers and aid delivery (e.g. the vehicle is no longer available to support programming). However, as we discuss further below, these categories barely overlap. Out of the 175 events recorded in our dataset, only 11 classified as hijacking, theft, or abduction resulted in death or injury.58
On the basis of the locations of the incidents included in the in the original jmac dataset, we have assigned location coordinates for the incidents included in the dataset, allowing analysis examining temporal and spatial variation. We use the jmac data to construct our variables of interest in a dataset for which we use the prio Grid data as the unit of analysis. The prio Grid divides Darfur in 213 grid cells of 0.5 x 0.5 degrees (approximately 55 x 55 km at the equator).59 An advantage of using grid cells is that, in contrast to localities or conflict zones, they are not endogenous to a range of factors that might also be correlated with attacks on aid workers, such as armed fighting for example. Moreover, taking grid cells as opposed to single events as the unit of analysis allows us to look at the impact of the number of peacekeepers and aid workers operating in a given area. We draw on data from Fjelde et al. to determine how many peacekeepers were deployed in grid cells in Darfur.60
In order to account for the level of humanitarian activity in a grid cell, we use the ocha-compiled “Sudan: Who is Doing What Where (3W)” document from August 2008.61 The 3W document is a core humanitarian coordination dataset used to identify gaps, avoid duplication of efforts, and plan for future humanitarian response. Obviously the number of operations is not a perfect proxy for the number of aid workers.62 Logically, however, the number of aid workers will be higher in locations with more humanitarian operations, which provides us with the best grid-level estimate for the number of aid workers.
It follows from the 3W data that 78 different humanitarian organisations had a combined total of 1352 operations running across Darfur. Of these, unicef, the World Food Program, and the International Committee of the Red Cross ran the most operations. We list all the different humanitarian organizations and the number of operations they ran in the appendix. The 3W data reports where these operations occurred. We geo-coded these locations, making it possible to determine whether and how much humanitarian activity takes place in a given grid cell. The total of 1352 operations running in Darfur were conducted at 150 unique locations throughout 57 grid cells in Darfur. The 3W data on humanitarian activity in Darfur thus allows us to create an approximation of humanitarian activity in a grid cell.
However, since we use a snapshot of operations from August 2008, the 3W data does not vary from month to month. Monthly 3W data simply do not exist for the time period for our study. Nevertheless, we are confident that this lack of variation is unlikely to greatly affect our findings, since the number of aid workers deployed in Darfur did not vary significantly over the time horizon for this study.63 According to the United Nations Mission in Sudan (unmis) around 14,700 aid workers were active in all of Darfur as of January 2008, 17,100 aid workers as of 1 April 2008, 16,366 aid workers as of 1 October 2008, 17,700 as of December 2008, and 16,250 as of early March 2009.64
Our dataset covers 15 months, the period between January 2008 and March 2009, giving us 3195 (213 * 15) observations in the first instance. However, in many of the gird cells no humanitarian activity took place. For instance, most of the grid cells in northern Darfur are located in the desert where few people live, in which little fighting occurred, and where no aid activity or peacekeepers were located. Observing attacks on aid workers and peacekeepers is unlikely in these areas. We therefore exclude all grids cells from our dataset in which no humanitarian activity takes place. Including these grids would create many “empty cells” with no meaningful observations, but that still generate units and thus potentially bias the results.65 Aid workers were active in a total of 57 grid cells. Hence, we draw on 855 (57 * 15) observations.66
Figure 1 shows the spatial distribution of attacks on aid workers and on aid delivery involving hijacking, abduction, and theft. The figure displays only those grid cells in which peacekeepers were present and humanitarian activity occurred. Darker shades of the grid cells indicate more humanitarian activity. The blue triangles reflect the locations of the different UN peacekeeping bases. The black dots designate the locations of attacks on aid workers or aid delivery. Aid workers were directly attacked (resulting in death or injury) 27 times in 26 grid cell months in Darfur between January 2008 and March 2009. Aid delivery was affected 155 times in 113 grid cell months. Only seven of the 155 incidents of hijacking, abduction, or theft were also classified as a violent attack on aid workers that resulted in death or injury.
The spatial distribution of attacks on aid workers in Darfur, April 2008–March 2009
Citation: Journal of International Peacekeeping 25, 4 (2022) ; 10.1163/18754112-25040003
Only two attacks took place outside of the grid cells in which the 3W file records any humanitarian operations. This is shown in figure 1 by the two black dots that are plotted on the map but do not fall within a grid cell. The incidents that did not occur in the grid cells make up 1.1 percent of the total of 175 incidents in which aid workers were attacked or aid delivery affected by abduction, hijacking, or theft. Both incidents involved the hijacking of a car traveling from one town to another.67 This gives us further confidence that the 3W file accurately describes humanitarian activities in Darfur.
The spatial clustering of aid workers/delivery and peacekeepers is clear: all peacekeepers deployed are located in an area with at least some humanitarian activity, and most are located in areas with a relatively high degree of humanitarian activity. Figure 1 thus visually illustrates why it is crucial to deal with the denominator issue when studying the association between peacekeepers presence and attacks on aid workers/delivery. The presence of peacekeeping bases is highly spatially correlated with locations where aid workers or delivery are attacked, yet there is simply also a high level of humanitarian activity in these areas.
In addition to including variables on the number of peacekeepers and humanitarian operations, we incorporate variables about the number of armed clashes and attacks on civilians in a given area, for which we also draw on the jmac data. Furthermore, we include the population size of a given grid cell in our model, taking the logged version of the population size of the grid taken from the Gridded Population of the World version 3.68 Since it has previously been observed that armed clashes in Darfur have taken place mostly in mountainous areas,69 we include a variable that measures the proportion of mountainous terrain within the grid cell based on data taken from a high-resolution mountain raster developed for unep’s Mountain Watch Report.70 Controlling for these potential confounders helps to mitigate the problem of omitted variable bias as much as possible, but we acknowledge that endogeneity could still possibly bias our findings. The empirical analysis in this paper should be seen within the correlational realm and not the causational one.
Since the dependent variables – the occurrence of an attacks on aid workers/delivery in a given grid cell in a given month – are binary variables, we use a logit model. We use robust standard errors, clustered on the grid cell level, in order to allow for the clustering of attacks. In addition, to control for temporal dependence, the models include a variable that measures the number of months since the last observation of an attack on aid workers or delivery in that grid cell.
4 Findings
Model 1 in Table 1 shows the association between the number of peacekeepers and attacks on aid workers in a given grid cell, without controlling for the degree of humanitarian activity in that same grid cell. This association is found to be positive and statistically significant at the 1% level. Crucially, however, this model does not control for humanitarian activity in a given area. Model 1 therefore may suffer from omitted variable bias; this finding by itself provides no evidence of the associational logic.
Indeed, Model 2 shows that the impact of the number of peacekeepers in a given area is no longer statistically significant when controlling for the effect of humanitarian activity. This suggests that, contrary to the logic of associational targeting, peacekeeping presence is not significantly correlated with attacks on aid workers. Instead, it seems that aid workers cluster in space near the locations of peacekeepers – and that the number of aid operations in a given area increases the likelihood that aid workers are attacked. Similarly, the significant effect of armed clashes in a given grid cell at the 5% level in Model 1 disappears in Model 2. This suggests that aid workers are disproportionally likely to operate in areas that experience high levels of armed clashes. In other words, rather than being a direct effect of armed fighting, aid workers are more likely to be attacked in areas afflicted by armed fighting because they operate in these areas in greater numbers. Thus, more fighting, more aid operations, and more attacks.
The two graphs on top in Figure 2, below, show the marginal impact of the number of peacekeeping troops on the likelihood of an attack on aid workers. Comparing these graphs shows that in addition to this effect turning insignificant, the substantive impact of peacekeeping troops is greatly reduced when controlling for humanitarian activity. The left graph is based on Model 1, in which the effect of peacekeeping troops seems strong. This model predicts that the likelihood of an attack on aid workers in grid cells with humanitarian activity but no peacekeepers is 2.2 percent. This increases to 9.5 percent in grid cells with around 900 peacekeepers deployed. By contrast, in Model 2, on the right, we control for humanitarian activity. This model predicts that the probability of an attack on aid workers is 2.6 percent in areas in which no peacekeepers are based, compared to 4.3 percent in areas in which 900 peacekeepers are based. However, we cannot be confident about the predicted probabilities based on Model 2, because the effect of peacekeepers is found to be statistically insignificant in Model 2.
Predictive margins with 95 percent confidence intervals on the likelihood of attacks on aid workers in Darfur, January 2008–March 2009
Citation: Journal of International Peacekeeping 25, 4 (2022) ; 10.1163/18754112-25040003
Models 3 and 4 replicate Models 1 and 2 but focused on the effect of the explanatory variables on the likelihood of attacks on aid delivery, comprising hijacking, abduction, and theft. Similar to Models 1 and 2, we find that the number of peacekeepers in a given grid cell has a positive and statistically significant effect in Model 3, which becomes insignificant when controlling for the level of humanitarian activity in Model 4.
The two bottom graphs in Figure 2 show how the substantive impact of peacekeeping troops on the probability of hijacking, abduction, or theft is greatly reduced when controlling for humanitarian activity. Model 3 suggests the probability of an incident affecting aid delivery in given grid cell-month increases from 10.5 percent in areas where no peacekeepers are deployed to 38.7 percent in areas where 900 peacekeepers are deployed. Based on Model 4, where we control for humanitarian activity, this probability grows only from 12.2 percent to 18.6 percent.
In short, we do not find a significant correlation between the number of peacekeepers on the one hand and incidents involving attacks on aid workers (H1) or aid delivery (abductions, hijackings, and theft – H2) on the other hand. This provides support for both hypothesis 1 and 2. Moreover, the stark difference between the number of attacks on aid workers (27 times in 26 grid cell months in Darfur between January 2008 and March 2009) and on aid delivery, in the form of hijacking, abduction, or theft (155 times in 113 grid cell months over the same period) points to the necessity of examining a more holistic range of attacks, which vary in terms of the scale and type of violence used as well as in their immediate impact.
5 Conclusion
In the conclusion of their cross-national study on violence against aid workers, Hoelscher and colleagues emphasise the need “to unpack the micro-level undercurrents of these attacks”.71 This article contributes to an emerging literature that turns to the micro or sub-national level to explain attacks on aid workers.72 Based on data on violence against aid workers and aid delivery in Darfur between January 2008 and March 2009, we are unable to support the associational logic to explain attacks, which holds that aid workers are more at risk when operating in the same terrain as military forces, including peacekeepers, because they are seen as associated with their political or military agendas. Instead, we find that the correlation between peacekeeping presence and violence against aid workers and aid delivery is spurious because peacekeepers and aid workers cluster in space. Although this is in principle a ‘null-finding’ because we do not find a significant association between the number of peacekeepers and attacks on aid workers or aid delivery, it still is important because it contrasts with established wisdom, although further investigation would be welcome.73
Prior to this study, the links between the presence of peacekeepers and attacks on aid workers have solely been examined based on country-level or cross-country data, and solely in terms of aid worker death, injury, and kidnapping. In addition, the ‘denominator’ issue regarding the number of aid workers operating in a given country or area has been insufficiently taken into account in previous research.74 This is an understandable, if serious, omission from the literature, since the high number of attacks on aid workers in countries and sub-national districts where peacekeepers are present might simply reflect the possibility that armed actors have greater opportunity or incentives to attack aid workers in areas where more of them are present. The reason for not controlling for the number of aid workers in cross-national studies likely results from the difficulty – even impossibility – of obtaining reliable cross-national data on the number of aid workers across countries. Using a proxy for the number of aid workers might only work with sub-national data. This points to the need for caution if generalizing from cross-national data to specific national contexts. The weakness with our study and that of Narang and Stanton,75 however, is precisely this: that these findings are not necessarily generalizable beyond the country on which these sub-national analyses focus. To move beyond this problem, a promising avenue for future research would therefore be to identify a proxy for the number of aid workers that travels well across different countries, or perhaps even across sub-national units of different countries.
A further reflection on the generalizability of our study is in order. One possible reason why, all else being equal, we do not find a significant association between the number of peacekeepers in an area and attacks on humanitarians could be that humanitarians in Darfur have worked to protect humanitarian space and limit their associations with peacekeepers. For instance, Mills observes that ngo s in Darfur have generally been wary about asking peacekeepers for protection because this could create the perception of taking sides.76 Anecdotal evidence suggests that humanitarians in other countries where UN peacekeepers are deployed have also taken great care to avoid associations with peacekeeping forces.77 It might be harder for aid workers to maintain this humanitarian space in non-peacekeeping contexts. Given the strong association in the literature on blurred lines with the cases of Iraq and Afghanistan, both of which involved military forces and not peacekeeping missions, the logic may still hold in those cases. Further research is necessary to explore the explanations we posit here. Accordingly, our study illustrates the importance of studying a diversity of contexts and their unique dynamics, and the need to generate micro-level datasets for other countries. Much of the literature about violence against aid workers focuses on the most extreme contexts for this type of violence, for which Iraq and Afghanistan have been tragically illustrative. Another possible avenue for future research is to analyse the associational or blurred lines logic that compares countries in which a peacekeeping operation is deployed on the one hand, and countries in which a multilateral intervention force is deployed, like in Iraq and Afghanistan, on the other.
Finally, if the logic of association and blurred lines cannot fully account for violence against aid workers and aid delivery, how might we explain such attacks? Our analysis suggests a need to look beyond explanations that focus primarily on one cause, such as blurred lines, to explain such attacks. Instead, explanations of violence against aid workers and aid delivery must account for broader contextual dynamics, and therefore require qualitative research to complement quantitative analysis and to dig deeper into time, place, or culture. These contextual dynamics might include attitudes towards aid providers, their effectiveness and relationships with local populations,78 or to account for the strategic logics of violence.79 They may also include the broader political economy of a conflict and the role that aid plays within it.80 These explanations must examine questions of agency, look toward the ways that aid organisations themselves generate vulnerability to risk, and scrutinize the more specific vulnerabilities of individual aid workers.81 This requires us to move beyond a focus on the most violent incidents and to think about more ‘everyday’ forms of aid obstruction which may or may not be targeted at specific aid workers. This, in turn, pushes for what many have consistently called for – going beyond the numbers to more closely (and qualitatively) examine the specificities of context and individual events to shed light on the dynamics of violence against aid in situations of conflict.82
Equal authorship is implied.
For example: Stathis N. Kalyvas, The Logic of Violence in Civil Wars (Cambridge: Cambridge University Press, 2006); Reed Wood, ‘Rebel capability and strategic violence against civilians’, Journal of Peace Research, vol. 47, 2010, pp. 601–614; Kristine Eck and Lisa Hultman, ‘One-Sided Violence Against Civilians in War: Insights from New Fatality Data’, Journal of Peace Research, vol. 44, 2007, pp. 233–246; Reed Wood and Jacob Kathman, ‘Too Much of a Bad Thing? Civilian Victimization and Bargaining in Civil War’, British Journal of Political Science, vol. 44, 2013, pp. 685–706; Amanda Murdie and Craig S. Stapley, ‘Why Target the “Good Guys”? The Determinants of Terrorism Against ngo s’, International Interactions, vol. 40, 2014, pp. 79–102; Alex De Waal, Chad Hazlett, Christian Davenport, Joshua Kennedy, ‘The epidemiology of lethal violence in Darfur: Using micro-data to explore complex patterns of ongoing armed conflict’, Social Science & Medicine, vol. 120, 2014, pp. 368–377.
Jacob Kathman and Reed Wood, ‘Stopping the Killing During the “Peace”: Peacekeeping and the Severity of Postconflict Civilian Victimization’, Foreign Policy Analysis, vol. 12, 2016, pp. 149–169; Vincenzo Bove and Andrea Ruggeri, ‘Kinds of Blue: Diversity in UN Peacekeeping Missions and Civilian Protection’. British Journal of Political Science, vol. 46, no. 3, 2016, pp. 681–700; Lisa Hultman, Jacob Kathman, and Megan Shannon, ‘United Nations Peacekeeping and Civilian Protection in Civil War’, American Journal of Political Science, vol. 57, 2013, pp. 875–891.
Nynke Salverda, ‘Blue helmets as targets: A quantitative analysis of rebel violence against peacekeepers, 1989–2002’, Journal of Peace Research, vol. 50, no. 6, 2013, pp. 707–720; Hanne Fjelde, Lisa Hultman, Sara Lindberg Bromley, ‘Offsetting Losses: Bargaining Power and Rebel Attacks on Peacekeepers’, International Studies Quarterly, vol. 60, no. 4, 2016, pp. 611–623.
Allard Duursma, ‘Obstruction and intimidation of peacekeepers: How armed actors undermine civilian protection efforts’, Journal of Peace Research, vol. 56, no. 2, 2019, pp. 234–248.
Larissa Fast, Aid in Danger: The Perils and Promise of Humanitarianism (Philadelphia: University of Pennsylvania Press, 2014); Abby Stoddard, Necessary Risks: Professional Humanitarianism and Violence against Aid Workers (Cham, Switzerland: Palgrave Macmillan); Karen Bennett, Danna Ingleton, Alice Nah, and James Savage, ‘Critical Perspectives on the Security and Protection of Human Rights Defenders’, The International Journal of Human Rights, vol. 19, no. 7, 2015, pp. 883–895; Rohini Haar, Róisín Read, Larissa Fast, Karl Blanchet, Stephanie Rinaldi, Bertrand Taithe, Christina Wille, and Leonard Rubenstein, ‘Violence against healthcare in conflict: a systematic review of the literature and agenda for future research’, Conflict and Health, vol. 15, no. 1, 2022, pp. 1–18.
Abby Stoddard, Adele Harmer, and Katherine Haver, Providing Aid in Insecure Environments: Trends in Policy and Operations (hpg Report, London: odi, 2006); Kristian Hoelscher, Jason Miklian, and Håvard Mokleiv Nygård, ‘Conflict, Peacekeeping, and Humanitarian Security: Understanding Violent Attacks Against Aid Workers’, International Peacekeeping, vol. 24, 2017, pp. 538–565.
Mark Duffield, Joanna Macrae, and Devon Curtis, ‘Editorial: Politics and Humanitarian Aid’, Disasters, vol. 25, no. 4, 2001, pp. 269–274; David Mitchell, ‘Blurred Lines? Provincial Reconstruction Teams and ngo Insecurity in Afghanistan, 2010–2011’, Stability: International Journal of Security and Development, vol. 4, no. 1, 2015, pp. 1–18.
Fast, Aid in Danger.
Neil Narang and Jessica Stanton, ‘A Strategic Logic of Attacking Aid Workers: Evidence from Violence in Afghanistan’, International Studies Quarterly, vol. 61, 2017, pp. 38–51.
Fast, Aid in Danger, p. 369; Larissa Fast, ‘Mind the gap: Documenting and explaining violence against aid workers’, European Journal of International Relations, vol. 16, 2010, pp. 365–389.
Karl Mattli and Jörg Gasser, ‘A Neutral, Impartial and Independent Approach: Key to icrc’s Acceptance in Iraq’ International Review of the Red Cross, vol. 90, no. 869, 2008, pp. 153–68; Sarah Kenyon Lischer, ‘Military Intervention and the Humanitarian Force Multiplier’, Global Governance, vol. 13, 2007, pp. 99–118; anso and care International, ngo Insecurity in Afghanistan (Kabul: anso and care International, May 2005); United Nations Security Council, Aid Operations under Increasing Threat as State, Non-State Combatants Ignore International Law, Humanitarian Affairs Chief Warns Security Council, sc/13760, (1 April 2019), https://www.un.org/press/en/2019/sc13760.doc.htm.
Fiona Terry, ‘The International Committee of the Red Cross in Afghanistan: Reasserting the Neutrality of Humanitarian Action’, International Review of the Red Cross, vol. 93, no. 881, 2011, pp. 173–88; Fast, Aid in Danger; Michaël Neuman and Fabrice Weissman, Saving Lives and Staying Alive: Humanitarian Security in the Age of Risk Management (London: Hurst, 2016); Narang and Stanton, ‘A Strategic Logic of Attacking Aid Workers’.
Mary Kaldor, New and Old Wars (Palo Alto: Stanford University Press, 2007).
Stoddard et al., Providing Aid in Insecure Environments, p. 19; Neuman and Weissman, Saving Lives and Staying Alive.
Nicolas De Torrenté, ‘Humanitarian Action under Attack: Reflections on the Iraq War’, Harvard Human Rights Journal, vol. 17, 2004, pp. 1–30.
Lischer, ‘Military Intervention and the Humanitarian Force Multiplier’; Francis Kofi Abiew, ‘Humanitarian Action under Fire: Reflections on the Role of ngo s in Conflict and Post-Conflict Situations’, International Peacekeeping, vol. 19, no. 2, 2012, pp. 203–216; Jason-Louis Carmichael and Mohammad Karamouzian, ‘Deadly Professions: Violent Attacks against Aid-Workers and the Health Implications for Local Populations’, International Journal of Health Policy Management, vol. 2, no. 2, 2014, pp. 65–67.
Mitchell, ‘Blurred Lines?; Duffield et al., Editorial: Politics and Humanitarian Aid’.
For example see, Erik Abild, ‘Creating humanitarian space: a case study of Somalia’, Refugee Survey Quarterly, vol. 29, 2010, pp. 67–102; Sarah Collinson and Samir Elhawary, Humanitarian space: a review of trends and issues, hpg Report 32 (London: odi, 2012); D. Robert DeChaine, ‘Humanitarian Space and the Social Imaginary: Médecins Sans Frontières/Doctors Without Borders and the Rhetoric of Global Community’, Journal of Communication Inquiry, vol. 26, 2002, pp. 354–369; Karen Guttieri, ‘Humanitarian space in insecure environments: a shifting paradigm’, Strategic Insights, vol. 4, no. 11, 2005; Dorothea Hilhorst and Bram J. Jansen, ‘Humanitarian space as arena: A perspective on the everyday politics of aid’, Development and Change, vol. 41, 2010, pp. 1117–1139; Lara Olson, ‘Fighting for Humanitarian Space: ngo s in Afghanistan’, Journal of Military and Strategic Studies, vol. 9, 2006; Róisín Shannon, ‘Playing with principles in an era of securitized aid negotiating humanitarian space in post-9/11 Afghanistan’, Progress in Development Studies, vol. 9: 2009, pp. 15–36.
For example, see: Greg Hansen, Iraq: More Challenges Ahead for a Fractured Humanitarian Enterprise (Medford, MA: Feinstein International Center, 2008); Cynthia Brassard-Boudreau and Don Hubert, ‘Shrinking Humanitarian Space? Trends and Prospects on Security and Access’, The Journal of Humanitarian Assistance, vol. 24, 2010.
For example, see: Joanna Macrae, ‘The Death of Humanitarianism? An Anatomy of the Attack’, Refugee Survey Quarterly, vol. 17, no. 1 1998, pp. 309–317; Claude Bruderlein and Pierre Gassmann, ‘Managing security risks in hazardous missions: The challenges of securing United Nations access to vulnerable groups’, Harvard Human Rights Journal, vol. 19, pp. 63–93.
For example, see: Ed Schenkenberg van Mierop, ‘Coming clean on neutrality and independence: The need to assess the application of humanitarian principles’, International Review of the Red Cross, vol. 97, no. 897/898, 2016, 295–318; Daniel Thürer, ‘Dunant’s pyramid: thoughts on the humanitarian space’, International Review of the Red Cross, vol. 89, no. 865, 2007, pp. 47–61.
Conor Foley, The Thin Blue Line: How Humanitarianism Went to War (New York: Verso, 2008).
Jamie A. Williamson, ‘Using Humanitarian Aid to Win Hearts and Minds: A Costly Failure? International Review of the Red Cross, vol. 93, no. 884, 2011, pp. 1035–1061; Antonio Donini, Larissa Fast, Greg Hansen, Simon Harris, Larry Minear, Tasneem Mowjee, and Andrew Wilder, Humanitarian Agenda 2015: Final Report. The State of the Humanitarian Enterprise (Boston, MA: Feinstein International Center, Tufts University, 2008).
James Denslow, ‘Reclaiming the humanitarian space’, The Guardian, 7 August 2010. See also: De Torrenté, ‘Humanitarian Action under Attack’; Olson, ‘Fighting for Humanitarian Space’; Mattli and Gasser, ‘A Neutral, Impartial and Independent Approach’; Marcos Ferreiro, ‘Blurring of Lines in Complex Emergencies: Consequences for the Humanitarian Community’, Journal of Humanitarian Assistance, 2012, pp. 1–28.
Colin Powell, Secretary of State Colin L. Powell Remarks to the National Foreign Policy Conference for Leaders of Non- Governmental Organizations, 26 October 2001.
Trend News Agency, ‘Envoy laments weak U.S. knowledge about Taliban’, Seattle Times, 7 April 2009.
Denslow, ‘Reclaiming the humanitarian space’.
David Petraeus and James Amos, Department of the US Army Counterinsurgency Manual (Washington, DC, 2006), A7.
Karsten Friis, ‘Peacekeeping and Counter-insurgency – Two of a Kind?, International Peacekeeping, vol. 17, no. 1, 2010, pp. 49–66; Bruno Charbonneau, Intervention in Mali: Building peace between peacekeeping and counterterrorism’, Journal of Contemporary African Studies, vol. 35, no. 4, 2017, pp. 415–431.
John Karlsrud, The UN at War: Peace Operations in a New Era (Cham: Springer International Publishing, 2017).
Allard Duursma, ‘Information Processing Challenges in Peacekeeping Operations: A Case Study on Peacekeeping Information Collection Efforts in Mali’, International Peacekeeping, vol. 25, no. 3, 2018, pp. 446–468.
Duursma, ‘Information Processing Challenges in Peacekeeping Operations’, p. 461.
Karlsrud, The UN at War, p. 74
Mitchell, ‘Blurred Lines?; Duffield et al., Editorial: Politics and Humanitarian Aid’.
For example: anso and care International, ngo Insecurity in Afghanistan.
David Rieff, A Bed for the Night: Humanitarianism in Crisis (New York: Simon & Shuster, 2002), pp. 221–222.
Kaldor, New and Old Wars, p. 136.
Edward Epstein, ‘Pentagon to change color of air-drop food packets / Afghans may mistake bombs for yellow rations’, Chronicle Washington Bureau, 3 November 2001.
John Karlsrud and Frederik Rosén, ‘In the Eye of the Beholder? UN and the Use of Drones to Protect Civilians’, Stability: International Journal of Security and Development, vol. 2, no. 2, 2013.
The New Humanitarian, ‘ngo s against monusco drones for humanitarian work’, 23 July 2014.
Stoddard et al., Providing Aid in Insecure Environments, p. 19. However, rather than stressing the blurred lines logic, Stoddard et al. argue that the presence of peacekeepers makes aid workers less careful and therefore more likely to be targeted: “The fact that internationals become more exposed to violence when a peacekeeping or peace-support force is present (while incidents against nationals decline) is likely to stem from a feeling among international agencies that the force provides a measure of ambient security that allows them to extend their operational presence, and gives their international staff greater freedom of movement than they would otherwise have”. Stoddard et al., Providing Aid in Insecure Environments, p. 19.
Hoelscher et al., ‘Conflict, Peacekeeping, and Humanitarian Security’, p. 554.
Andrea Ruggeri, Han Dorussen, and Theodora-Ismene Gizelis, ‘On the Frontline Every Day? Subnational Deployment of United Nations Peacekeepers’, British Journal of Political Science, vol. 48, no. 4, 2016, pp. 1005–1025.
United Nations, ‘United Nations DR Congo’, available at: https://drcongo.un.org/, accessed on 16 December 2020.
Ashly Adam Townsen and Bryce W. Reeder, ‘Where Do Peacekeepers Go When They Go? Explaining the Spatial Heterogeneity of Peacekeeping Deployments’, Journal of International Peacekeeping, vol. 18, no. 1, 2014, pp. 69–91; Ruggeri et al., ‘On the Frontline Every Day’.
Hoelscher et al., ‘Conflict, Peacekeeping, and Humanitarian Security’.
Narang and Stanton, ‘A Strategic Logic of Attacking Aid Workers’.
Fast,’ Mind the gap’, p. 369.
Katy Barnett, Security Report for Humanitarian Organizations, echo Security Review (Brussels: Directorate-General for Humanitarian Aid – echo, European Commission, 2004).
Fast, Aid in Danger.
The data have been provided by the African Union High-Level Panel on Darfur (aupd) with assistance from Alex de Waal who acted as an adviser to the aupd. See: De Waal et al., ‘The epidemiology of lethal violence in Darfur’.
Allard Duursma, ‘Counting Deaths While Keeping Peace: An Assessment of the jmac’s Field Information and Analysis Capacity in Darfur’, International Peacekeeping, vol. 24, no. 5, pp. 823–847.
UN Security Council Resolution 1769.
Observation 2341 in the jmac dataset.
Observation 2590 in the jmac dataset.
Observation 2732 in the jmac dataset.
Note, however, that both events happen subsequently in some grid-cell months, so there is overlap in that sense.
Andreas Forø Tollefsen, Håvard Strand, and Halvard Buhaug, ‘prio-grid: A Unified Spatial Data Structure’, Journal of Peace Research, vol. 49, no. 2, 2012, pp. 363–374.
Hanne Fjelde, Lisa Hultman and Desirée Nilsson, ‘Protection through Presence: How Peacekeepers Reduce Violence against Civilians’, International Organization, vol. 73, no. 1, 2019, pp. 103–131. Fjelde et al. have coded the locations of peacekeeping bases on the basis of several deployment maps published in this period by unamid.
We obtained the August 2008 3W file from ocha Sudan, which is included in our replication dataset.
While they represent the best available approximation of formal, international humanitarian operations in a given context, the 3W data often exclude local humanitarian actors or agencies (for a summary of the different types of 3W data and their strengths and limitations, see: David Megginson, ‘The rough guide to aid-activity data on hdx’, The Centre for Humanitarian Data, 6 May 2020. Available at: https://centre.humdata.org/the-rough-guide-to-aid-activity-data-on-hdx/, accessed 27 August 2020.
Email correspondence with a Sudanese humanitarian working for a local ngo, which has offices in Khartoum and Darfur, on 9 March 2017; Telephone interview with unicef staff member on 13 June 2017; Telephone interview with a former ocha staff member on 14 June 2017.
UN Mission in Sudan. Sudan: Darfur Humanitarian Profile No. 30 – Situation as of 1 January 2008; UN Mission in Sudan. Sudan: Darfur Humanitarian Profile No. 31 – Situation as of 01 Apr 2008; UN Mission in Sudan. Sudan: Darfur Humanitarian Profile No. 33 – Situation as of 1 October 2008. Available at: http://reliefweb.int/report/sudan/sudan-darfur-humanitarian-profile-no-30-situation-01-jan-2008; http://reliefweb.int/report/sudan/sudan-darfur-humanitarian-profile-no-31-situation-01-apr-2008; https://www.ecoi.net/file_upload/1504_1226486578_darfur-humanitarian-needs-profile-no-33.pdf. For the figure of March 2009, we draw on UN-ocha (2009) Sudan: ocha Situation Report No. 1, 08 Mar 2009 – Expulsion of key ngo s from Darfur. http://reliefweb.int/sites/reliefweb.int/files/resources/ECA64643771CB0B885257574007C81C5-Full_Report.pdf. We report the number of humanitarians in Darfur as of early March 2009, because in response to the icc indictment against President Omar al-Bashir, around 7,302 national staff and 308 international staff were expelled from Sudan. See: Allard Duursma and Tanja Müller, ‘The icc indictment against Al-Bashir and its repercussions for peacekeeping and humanitarian operations in Darfur’, Third World Quarterly, vol. 40, no. 5, 2019, pp. 890–907; Overseas Development Institute, ‘Where to Now? Agency Expulsions in Sudan: Consequences and Next Steps’, alnap Working Paper, 2009. Yet, these (now former) aid workers were still present in Darfur in March 2009.
Karsten Donnay, Disaggregating Civil Conflict: Theoretical, Methodological and Empirical Contributions (eth Zurich, 2014), p. 25.
In order to determine whether dropping the grid cells in which no humanitarians operated influences our findings, we ran the models using the full dataset of 3195 observations. This, however, did not affect our findings.
Previous research suggests travel between places is one of the most dangerous activities for humanitarians. Stoddard et al., Providing Aid in Insecure Environments; Fast, Aid in Danger.
Center for International Earth Science Information Network (ciesin) and Centro Internacional de Agricultura Tropical (ciat), Gridded Population of the World, Version 3 (gpw v3): Population Count Grid (New York: Palisades, 2005).
Julie Flint and Alex de Waal, Darfur: A New History of a Long War (London: Zed Books, 2008); Johan Brosché and Allard Duursma, ‘Hurdles to peace: a level-of-analysis approach to resolving Sudan’s civil wars’, Third World Quarterly, vol. 39, no. 3, 2018, pp. 560–576.
S. Blyth, B. Groombridge, I. Lysenko, et al., ‘Mountain Watch: environmental change & sustainable development in mountains’, unep-wcmc Biodiversity Series 12, 2002.
Hoelscher et al., ‘Conflict, Peacekeeping, and Humanitarian Security’. See also: Fast, Aid in Danger.
Narang and Stanton, ‘A Strategic Logic of Attacking Aid Workers’.
Mitchell, ‘Blurred Lines?; Duffield et al., Editorial: Politics and Humanitarian Aid’.
For a notable exception, see Narang and Stanton, 2017.
Narang and Stanton, ‘A Strategic Logic of Attacking Aid Workers’.
Kurt Mills, ‘Constructing humanitarian space in Darfur’, The International Journal of Human Rights, vol. 17, no. 5, 2013, pp. 605–618.
The New Humanitarian, ‘ngo s against monusco drones for humanitarian work’.
Fast, Aid in Danger.
Kalyvas, The Logic of Violence in Civil Wars.
For example: Reed Wood and Christopher Sullivan, ‘Doing harm by doing good? The negative externalities of humanitarian aid provision during civil conflict’, Journal of Politics, vol. 77, no. 3, 2015, pp. 736–748; Michael Findley, ‘Does Foreign Aid Build Peace?’ Annual Review of Political Science, vol. 21, 2018, pp. 359–384.
Fast, Aid in Danger; eisf, Managing the Security of Aid Workers with Diverse Profiles (European Interagency Security Forum, 2018).
Fast, Aid in Danger; Neuman and Weissman, Saving Lives and Staying Alive; Silke Roth, ‘Aid work as edgework – Voluntary risk-taking and security in humanitarian assistance, development and human rights work’, Journal of Risk Research, vol. 18, no. 2, 2015, pp. 139–155.
