Out of Reach: How Insecurity Prevents Humanitarian Aid from Accessing the Neediest

In a small number of crisis-affected countries, humanitarian organizations work amid active conflict and under direct threat of violence. This insecurity, reflected in rising aid worker casualty rates, significantly constrains humanitarian operations and hinders the ability of people in emergencies to access vital aid. Extensive fieldbased research in Afghanistan, southern Somalia, South Sudan and Syria measured humanitarian coverage (aid presence relative to the level of need) in each context to determine how this coverage is affected by insecurity. Results show that humanitarian operations are highly determined by security conditions, more than any other factor. As a result, coverage is uneven relative to need and appears politically skewed in favor of areas under control of Western-supported conflict parties. Additionally, humanitarian coverage in these war zones is even lower than it outwardly appears, as aid organizations tend to remain in the country (even after suffering attacks) but reduce and contract their field presence, adopting new, often suboptimal, means of programming.1


Introduction
International humanitarian law specifically proscribes violence against humanitarian organizations 2 and accords protected status to their facilities and activities during armed conflict. Despite these rules of war, aid personnel and operations frequently come under attack in conflict settings, used as proxy targets, revenue sources, and convenient tools for terror or propaganda purposes.
Over the past decade, major attacks on aid workers (killings, kidnappings and serious injuries from deliberate violence) have increased in both absolute and relative terms (Humanitarian Outcomes 2015a). The casualty toll has consistently been driven by a small number of highly violent contexts (less than ten per cent of the humanitarian emergency caseload each year), which together account for over 60 per cent of all such attacks. For the past five years, these were Afghanistan, Syria, South Sudan and Somalia (Figure 1).
Each of these contexts represent protracted civil conflicts involving armed stability non-state actors and asymmetric warfare tactics with serious impacts on local populations, over whom the warring parties vie for control. And in all of them, aid workers have been subject to both direct and collateral violence that has claimed more than 800 victims since 2011, of whom 277 lost their lives.
To measure the impact of insecurity on the humanitarian response requires determining the size and shape of the humanitarian footprint in each country. Unlike military deployments, however, the humanitarian response to an emergency is not unified and readily measurable. Rather, it comprises the loosely coordinated efforts of often hundreds of autonomous organizations taking largely independent decisions on where and how to operate. While the UN humanitarian office provides 'Who does What, Where' maps in many countries, these '3Ws' 3 maps typically do not give a sense of the magnitude of each agency's presence and activities or the extent to which they are covering people's needs.
Information on aid presence is even more difficult to derive in highly insecure settings. Aid organizations 4 are often reluctant to share information on their presence and activities due to a combination of security and reputational concerns. They may face conflicting pressures to keep their specific locations and activities quiet for the security of their staff and programs on the one hand, and on the other hand to exaggerate the extent of their presence for funding and public relations purposes, demonstrating to donors and the general public that they are capable of going where needed. With the actual size of the humanitarian footprint largely unknown, how it may shrink and/or reconfigure in situations of heightened insecurity, and what this means for the affected population, has never been clear.
There is also very little by way of prior research into this problem. Although there is considerable humanitarian literature devoted to the issue of 'humanitarian access' (UN OCHA 2010), including the constraints created by insecurity (Egeland, Harmer & Stoddard 2011;Steets, Reichhold & Sagmeister 2012), humanitarian negotiations with armed actors (Jackson 2014;Maurer 2014) and the impacts of sanctions and counter-terror regimes (Burniske,  Likewise, a good deal has been written about the why of violence against aid workers. The literature on the ethical challenges of humanitarian action in conflicts and the inability of humanitarians to be truly neutral parties (e.g. Terry 2002;Barnett 2011;Slim 2015) underpins this discourse. Some analysts have also shone a light on the internal polices and behaviors of aid organizations that can generate risk or lead to counterproductive security measures in these contexts (Egeland, Harmer & Stoddard 2011;Duffield 2012;Fast 2014). The goal of this research, however, was to provide the missing empirical evidence base to the humanitarian security discussion (the what and how of insecurity's effect on aid operations) in the hope of better informing and improving the immensely difficult task of aiding populations in war zones.

Methodology
The research consisted of quantitative data gathering on humanitarian presence at the field level over an 18-month period, key informant interviews, global demographic and humanitarian financing data compilation and analysis and surveys of the affected populations.
The study identified and recruited fieldbased researchers for each of the cases to collect data on the number of humanitarian organizations, activities and personnel at the subnational level (by province/region and where possible by district) for 2014 and prior years of the conflict. This was done by first collecting existing compiled data from the UN Humanitarian Coordination Office (OCHA), followed by direct inquiries to each of the agencies at country-level headquarters. The field-based researchers then cross checked these numbers by means of local researcher networks at the provincial and district levels.
In all, 273 semi-structured interviews were conducted with humanitarian practitioners in all four contexts and at the headquarters level. The interview questions were designed to elicit information on current and past activities and operational presence levels, perceptions of insecurity and the decisionmaking processes around initiating and changing field programs.
The study also ran remote surveys of local populations, which served as additional pieces of evidence to triangulate humanitarian presence information, as well as to glean residents' perspectives on security in their area and the barriers to accessing humanitarian assistance. Remote, mobile phone surveys, using 'interactive voice response' (IVR) technology, were used in Afghanistan, Somalia and South Sudan. In Syria, the threat of surveillance of mobile phone communications represented an unacceptable risk for respondents, and instead the study collaborated with a regionally based research partner to undertake in-person household surveying.
The quantitative analysis compared humanitarian presence data to insecurity levels as measured by incidents of major violence ( Table 1). This entailed analyzing and comparing data on:

Humanitarian presence
The dependent variables for analysis were the data collected by the field researchers on the numbers of organizations, projects and personnel at the subnational level in each of the four countries. The problem of missing presence data (in particular, spotty personnel data at the subnational level) was addressed through the application of two different estimation techniques for the purpose (Osborne 2013;Little & Rubin 2014). First, we ran the primary regression analyses discarding those observations and then we applied multiple imputation procedures to address the potential shortcomings of this approach.

Affected population and people in need
Subnational population data for the countries in our sample were drawn from different sources, selected for best reliability. For Afghanistan, we used data from NATO's Civil-Military Fusion Center, which has subnational population data for the period 2010-13. Using this data, we predicted values for missing years using a simple ordinary least squares (OLS) regression. For regions in Somalia, we took United Nations Development Programme estimates. For South Sudan, we used state-level data from the National Bureau of Statistics South Sudan for the year 2011 and used the population growth rate to extrapolate for subsequent years.
Estimates of the number of people in need of aid (PIN) at the national level were taken from the UN's coordinated Humanitarian Response Plan documents. Disaggregated PIN estimates were not available at the subnational level for most countries, so we used this national PIN figure to calculate the proportion of the population in need nationally and then calculate corresponding subnational figures for the four countries by applying the ratio to the subnational population. For Syria, the PIN numbers by governorate were available in UN documents (UN OCHA 2014a and 2014b) and these were found to correspond nearly exactly with the results of our formula for estimating.
The independent variables were the number of major attacks occurring in the subnational regions in each country. Two datasets provided data on these variables: the Aid Worker Security Database (AWSD), which tracks killings, kidnappings and serious injuries of aid workers by violence from 1997 to the present; and the Global Terrorism Database (GTD), which collects broader numbers of violent incidents and targets and, as such, provides an indicator of generalized violence and instability in an area irrespective of the humanitarian response. 5 For the analysis of the relationship of aid presence to insecurity, however, we use the GTD data for a broader indicator of violence in the environment. Summary statistics for the entire dataset (all countries) are displayed below.

Limitations
The study met with several -largely expected -challenges to amassing humanitarian presence data for each of the four contexts. Even with field-level researchers in each field context systematically inquiring agency by agency, the availability of presence data was limited. Reasons for this included organizations' differing perceptions of security risk, weaknesses in record keeping and concerns about public image. Written assurances of data anonymization and confidentiality protocols were not able to overcome some organizations' sensitivities around sharing operational information. This was particularly the case in Syria, were only 63 per cent of the known humanitarian organizations agreed to share their information and of those only a few did so at the requested level of granularity (i.e. staff and projects by district).
Finally, reputational concerns appeared to cause a reluctance to be transparent about how limited some reported operational presence actually was. The result was a lack of reliable data on humanitarian presence prior to 2011 (not needed for South Sudan and Syria but sought for the longer-running  crises of Afghanistan and Somalia), and only partial availability of staffing numbers from 2012-14 for all countries. Means of addressing these limitations in the quantitative analysis are described in the next section.
Survey findings were treated with modesty in respect to their significance. The remote survey method cannot guarantee randomness given the less than universal mobile phone ownership and coverage in these countries. Nevertheless, the target number of respondents for each country was 267, which would be the sample size required for the national populations at a 95 per cent confidence level with a confidence interval of six. In addition, despite the use of a female voice on the interactive recording, designed to increase the willingness of women to participate, the gender balance skewed heavily male (overall roughly 70:30 per cent, male: female). This likely not only reflects the reluctance of female respondents but also the gender imbalance of cell phone ownership in the countries.
Additional methodology details can be found at www.SAVEresearch.net.

Insecure countries attract a markedly smaller pool of humanitarian responders
Global data show that humanitarian organizations respond in smaller numbers to insecure emergency contexts, compared to more stable settings. Considerably fewer humanitarian organizations responded to highly violent, conflict-driven emergencies, irrespective of funding available and the needs of the population. Globally, the countries with the highest number of aid worker attacks had the lowest number of aid organizations responding per USD 100 million in funding and vice versa. On average, countries with no aid worker attacks had more than four times the number of organizations engaged in the response (Figure 2).
What this suggests in practice is a relatively small group of humanitarian actors operate in the highest risk locations. We tested this mathematically by using the presence data gathered in the most dangerous provinces within Afghanistan, Somalia and South Sudan and generating a ranked index of 'most present' organizations in each country and across the board. 6 For each country, the higher the index, the more widespread the organization's presence relative to the average for the regions in question. Summing them up across the countries lets us know which organizations have the most widespread presence in high-risk environments generally. This revealed that certain specific international non-governmental organizations (NGOs) tend to be among the major operators across all of the high-insecurity settings (Figure 3). They represent a subset of the largest (and some midsize) international humanitarian organizations, and their efforts are joined by a different constellation of national organizations in each context that often can achieve better access to certain areas than can their international counterparts.

Security conditions are the strongest determinant of aid presence
Insecurity dictates where aid agencies operate within high-risk countries, resulting in unequal coverage of needs. The quantitative data on organizational presence and activities, reaffirmed by interviews with practitioners, showed humanitarian operations clustered in more secure areas within these countries, irrespective of the relative level of need of the local populations. (The exception was capital cities, where aid organizations had their headquarters, despite high numbers of attacks on aid workers in those areas.) "Moreover, a path dependency was observed where security-related decisions in programming led to 'access inertia' among agencies. In other words, once they had contracted their presence, they had stronger incentives to remain in their comfort zone than to try to expand their geographic and programmatic reach."

Statistical analysis
Using the data on humanitarian presence we gathered in the four countries, we investigated whether it was possible to show a statistically significant relationship between insecurity and the level of aid coverage in a given area. The hypothesis was that these two variables would be negatively correlated (i.e. that an area where aid workers have been attacked would have a smaller humanitarian response). And indeed, plotting aid worker attack rates (number of AWSD-recorded incidents/number of aid personnel) against coverage levels (number of aid personnel/100,000 people in need), shows a pronounced negative relationship (Figure 4).
This would appear to show a very clear negative relationship between high insecurity and low humanitarian coverage, that is, the higher the degree of insecurity, the lower the level of humanitarian coverage of people in need. However, this simple bivariate relationship cannot be taken as proof, for reasons that are not immediately apparent. First, regions with larger aid worker populations may be likely to experience greater numbers of attacks in general simply because the aid presence offers potential targets. Second, as the number of aid workers in a region increases, the attack rate falls by construction, holding the numbers of attacks constant. This realization called for a deeper approach to the analysis for results to have sufficient rigor.
To get around the circularity problem of using aid worker attacks as our measure of insecurity, we instead used the data on a broader range of violent incidents and combat from the Global Terrorism Database (GTD) 7 as our independent variable. To make the coverage figures comparable across all four cases, we set different baselines for each country using fixed effects. Then, using OLS regressions, we attempted to measure the statistical impact of violent incidents on the level of humanitarian coverage at the subnational level. Results from that analysis were inconclusive, however (i.e. they did not pass the test for statistical significance). 8 The inconclusive results illustrate the difficulty inherent in performing statistical analysis on observational data (as opposed to a controlled experiment) due to confounding factors in the strategic and operational environment. These results certainly do not prove that insecurity does not have an impact on humanitarian coverage, but rather that it is obscured in formal models by the existence of other factors and drivers, such as the strategic behavior of both aid groups and armed actors. First, humanitarian organizations are likely to operate in areas that have been affected by conflict, due to the resultant needs of civilians. Second, militant groups are likely to increase the 'supply' of violence in areas where humanitarian actors gravitate, given that they can be attractive targets.
Finally, the humanitarians' perception of risk also speaks to another confounding element, which is the 'stickiness' of securitydriven operational decisions. Aid organizations tend to remain in locations and programming modalities where they feel comfortable and have strong disincentives to expand into the unfamiliar. In part this is because there is inherently greater risk involved in being new to an area where you have not yet had time to forge acceptance with local populations and conflict actors. But there is also an observed behavior in organizations (Stoddard, Haver & Czwarno 2016), reconfirmed by security professionals interviewed for this study, in which they will more readily raise the assessed risk level in response to new incidents than lower it in their absence. (How much time must pass since the last attack before one can confidently say the risk has decreased?) The elevated vigilance results in what looks very much like a path dependency, and at times inertia, of many agencies working in longterm insecure contexts.
Notwithstanding the lack of definitive regression results, the effects of insecurity on humanitarian coverage and operations can still be quantitatively (and qualitatively) observed. The interviews of humanitarian practitioners repeatedly confirmed that security concerns, more than any other factor, determined where, when and what sort of programming aid organizations implement. This is manifested both directly, as in Afghanistan where the direct targeting of aid workers prevented agencies from expanding to new provinces, and indirectly, as in South Sudan where ambient security concerns (i.e. the fear of combat ensuing/recurring) prevented investment in the facilities and logistics capacity necessary to maintain a sustained aid presence in field locations.

Insecurity limits technical complexity and targeting of aid activities
Humanitarian activities run the gamut from the simplest distributions of relief items to highly complex and technical programs. Getting data at the necessary level of granularity to differentiate between the types and complexity of different humanitarian projects is extremely difficult. However, global statistics on humanitarian funding do break down activities by project within broad sectors (food, water, heath, etc.) and we can use these figures to see patterns in types of programming between low and high insecurity countries.
We compared activities and funding per sector during 2011-14 between the emergency-affected countries with the highest insecurity (our four focus countries, plus Pakistan, Sudan, CAR, DRC and Yemenwhich had aid worker attacks numbering in the double digits) and the lowest insecurity (Algeria, Bangladesh, Burkina Faso, Egypt, Liberia, Malawi, Sierra Leone, Uganda, Thailand, and Zimbabwe -all of which had zero attacks reported against aid workers during the period). This showed that the more insecure countries had higher relative percentages of funding going to food and shelter/non-food items (NFI) distributions than the secure ones. Insecure settings also saw lower relative percentages of funding going to the health sector, which typically requires more technically complex programming requiring skilled personnel and sustained presence (Figure 5).
That protection activities have a higher proportion of total funding in insecure environments than secure ones is not surprising, but given the incidence of violence against civilians in civil conflict settings, one would expect the difference to be much greater than the figures show. This can be explained by insecurity limiting the ability of humanitarians to engage in more technical and presence-reliant activities. The findings for the water and sanitation (WASH) sector are less easily explained. On the one hand, some of this programming can be highly technical, requiring experts and access, which would lead one to expect it to be proportionally lower in insecure contexts than secure ones. However, in many cases the WASH sector category includes very basic distributions of 'hygiene kits' that require no more technical capacity than the distribution of food or shelter items. This, we suspect, explains the higher proportion of WASH funding in insecure environments. In addition, in some highly urban conflicts such as in Syria, the International Committee of the Red Cross (ICRC) and other agencies have prioritized support to rehabilitate the main supply lines, working with local authorities.
Aid personnel working in the four contexts reported that they were less able to specifically target the more vulnerable groups as is normal good practice. A less stable presence on the ground often means relying on intermediaries for distribution planning and beneficiary lists, reducing the ability to prioritize the neediest among the local population. Cultural injunctions, against certain programming for women and girls in areas of Afghanistan, for instance, were often cited as posing too high a security risk for organizations to challenge. In large areas of South Sudan and parts of Syria, a large portion of programming was limited to mobile deliveries as opposed to static programming with a sustained organizational presence.
In some cases, medical programming, particularly hospital services and trauma care, were seen to garner greater acceptance in the subfield regions considered the most dangerous. The reason for this is simply that they are needed and desired by all warring parties in the conflict, as well as the local population, which may explain why International NGOs (INGOs) with proven capacity to run this type of technical medical programming, even though a small minority within the global INGO community, are among the organizations ranked highest on the presence index. However, this demonstrably has not rendered them immune to targeted attacks and collateral violence.

Affected-populations perceive a declining aid presence, not meeting priority needs
The results of the surveys conducted for this study (by remote telecoms in Afghanistan, South Central Somalia and South Sudan and household canvassing in Syria) both supported the quantitative findings on humanitarian coverage, and revealed that perceptions of insecurity and access challenges diverge between humanitarians and the people they serve. For many questions, the results varied both between countries and locations surveyed within countries (discussed in the country-specific findings in the next section), but a few overall themes emerged as well.
Afghans and South Sudanese reported seeing fewer aid organizations working in their area in the past three years than had been there previously. In South Sudan, the perceived decline was starker, reflecting the cessation of many development-oriented aid projects that had been running before the outbreak of violence and evacuation of personnel. This decline in aid presence was also the dominant perception among Syrian respondents in all areas except Aleppo, which had seen the number of aid operations increase sharply in conjunction with stepped up cross-border operations from Turkey. 9 South Central Somalia respondents, conversely, reported seeing an increase in the number of aid organizations, likely reflecting the 'redeployment' of UN agencies and some INGOs in response to the famine that was declared in mid-2011, after years of very low presence.
Food aid was cited most often as the most urgent need by affected people in all countries except South Sudan, where more people reported protection as their first priority, with food coming second. 10 Children's education, though often not considered a basic humanitarian provision, ranked high as a priority need, particularly in Syria and Afghanistan.
In Somalia, respondents were much more unanimous in terms of the type of aid most urgently needed. For a large majority (85 per cent) it was food, and for the remainder it was water/sanitation assistance.
In Syria, where in-person surveys allowed a more nuanced breakdown of the question of needs, people were asked about the greatest needs for different groups. Respondents cited food as the greatest need for women, education for children, and cash for men. Numerous respondents cited an excess of food aid baskets and a desire for a greater diversity of aid, or preferably cash, so they would not need to sell portions of their food to meet other needs.
In general, far fewer recipients reported receiving cash assistance than food and hygiene items, health care and water/sanitation assistance. When Syrian aid recipients were asked if the aid they received addressed their most urgent needs, a significant proportion of them (43 per cent) answered 'no'. Majorities in all four countries agreed on only two points. The first was also the most surprising and counter-intuitive: respondents were mostly of the opinion that working in their location was not dangerous for aid organizations, in direct contradiction to aid organizations' perceptions (and what aid worker casualty figures support). Only in two subnational areas -Helmand, Afghanistan, and Aleppo, Syriadid majorities agree that aid groups faced specific danger. This does not mean that most respondents failed to acknowledge violence and insecurity in their region; in fact, they were seen as posing general impediments to people's ability to access aid. What this does suggest is that the people surveyed did not perceive a specific, direct threat against humanitarian workers and organizations.
The dramatic difference in perception might also reflect the greater distance and difficulty of forging meaningful dialogue between humanitarian organizations and local communities in insecure settings. This relates directly to the second point on which findings in all four countries were uniform: large majorities (65-94 per cent) reported that locals in their area had not been consulted by any aid organizations about their opinion of the aid being delivered.
Despite not perceiving aid organizations to be at risk, affected people in some countries nevertheless cited insecurity generally as a major impediment to accessing aid (Figure 6). Afghans and South Sudanese cited insecurity as the number one obstacle to aid in their areas. For Somalis it was the second largest impediment, after corruption. In Syria, most people felt that their biggest problem in terms of accessing humanitarian aid was simply that not enough of it was coming in.

Adaptive programming paradigms
The threat profile, political environment and donor interests that humanitarian actors must navigate are specific to each context and have resulted in different modes of adaptive programming. While there are exceptions in every case, it is possible to discern a general pattern or operational model for the humanitarian response that has emerged in each place (Figure 7).
In each of the operational models, humanitarian organizations have found a way to continue programming but each also entails significant downsides and results in uneven coverage of humanitarian need across each country, as discussed below.

Country-specific findings Afghanistan: Contracting presence and access inertia
In Afghanistan, humanitarian needs are most severe in the south and southeast of the country (UN OCHA 2013), where fighting with a resurgent Taliban has displaced tens of thousands, exacerbated pre-existing malnutrition and disrupted immunization programs. In addition, the southeast also hosts large refugee populations from Iran and Pakistan. Despite gains made in some development sectors during the post-Taliban period, humanitarian needs in many places have not abated, and in others have increased, due to intensifying conflict and newly occurring natural disasters and refugee crises.
The most common form of major attack against aid workers is kidnapping, typically resolved by the intervention of community leaders with the safe release of victims after a few days. Although this now-commonplace practice makes the lethality of aid worker attacks overall lower than in other places, Afghanistan has also had a relatively high number of ' complex' attacks employing sophisticated weaponry and explosives that are highly lethal.
The data gathered in Afghanistan show that, countrywide, the humanitarian organizational presence has not changed significantly since 2006. Their numbers have fluctuated around 160 operational 11 organizations, including roughly equal numbers of international and national NGOs (70-75 each, depending on the year), 6-9 UN agencies and two Red Cross/Crescent movement entities (and not counting a variety of governmental and commercial entities, working mostly on development and economic infrastructure projects). The overall numbers of operational organizations present held fairly stable despite a three-fold funding surge in response to heightened humanitarian needs from 2006-08.
Despite some 2,000 national NGOs officially registered with the government, practitioners on the ground report that less than 150 have the capacity to deliver humanitarian assistance, and only 70-80 have been operational in any given year.
Although a handful of large national NGOs have broad coverage across the country, the majority work only in one province (out of 34 provinces) and in just 1-4 districts per province (where the average number of districts per province is 12). The Afghan National Red Crescent Society works in nearly all provinces, and a majority of respondents in the affected population survey confirmed they were more present in those areas than either international or local aid organizations. However, their scope of action is limited to their own resources and those provided by the International Red Cross/Crescent movement. They were not used as an alternative channel for international donors and other organizations to extend the humanitarian presence.
The reduction in presence over the study period can be seen most clearly at the district level. Even though many organizations could continue to claim presence in the same number of provinces, the number of districts they were active in decreased by over 40 per cent between 2012 and 2014 (Figure 8).
To deal with insecurity, most INGOs in Afghanistan employ localization -a means of maximizing community acceptance -as their primary coping strategy. This means hiring all staff from the immediate vicinity of the project and reducing or completely eliminating the presence of non-local personnel, vehicles and organizational branding so that the work can blend into local community life. Those aid organizations that have localized their programming admit that this approach can undermine the technical quality of projects, since the local pool of potential hires often lacks the technical skills needed for certain activities. It can also delay implementation due to the need for supplemental staff training. Localized programming entails extensive communications, before and during project implementation, with local leaders, including state officials, tribal elders, local commanders or religious leaders. Careful and prolonged negotiation can cement acceptance and help provide security guarantees, but at the same time naturally hinders flexible programming and rapid emergency response. One INGO reported that it could only respond to floods in northern Afghanistan after a full month of negotiation, and one UN agency described months of negotiation, still ongoing, to gain secure access in Helmand province.
It also often means in practice that local power holders have veto power over aid activities and modalities. Gender issues tend to become the sticking point in this regard. Organizations in many locations reported they could not work with, or target, women and girls directly. Instead, they had to rely on male intermediaries from the community to see to it that community women also benefitted from organizational programming, albeit in a way that the organization could not directly assess or verify.
The localization model has enabled a stable population of humanitarian organizations to continue working in Afghanistan, but this has in some ways obscured the fact that the humanitarian presence has thinned out at the subfield (district) level generally and is concentrated in safer provinces, not those where the humanitarian need is greatest. The violence and insecurity that has spurred the needs in Helmand and other areas in the south has also prevented a concomitant increase in humanitarian presence, as agencies choose to remain in their relative comfort zones. The affectedpopulation surveys also bear this out. Of the provinces surveyed, Kandahar in the south had the lowest percentage of respondents who had seen aid in their area in the past three years.
Given the casualties continuing to occur in Afghanistan, it is hard to fault the majority of organizations who rely on this means of programming, but the downsides are apparent. In addition to the constraints on rapid response and quality of programming mentioned above, the prevailing access inertia among agencies has resulted in a clear imbalance of humanitarian coverage across the country, with aid agency presence still concentrated in the north and in provincial capitals (Map 1).
Access limitations have not been offset by identifying or helping to build up local partners at any scale or other potential modes of remote programming. Nor have humanitarian donor governments and the UN been successful in attracting significantly more organizations and activities to the south with 'pull funding' allocations for the purpose.

South Central Somalia: Remote management and compliance risk aversion
Like the armed opposition groups in Afghanistan, the militant Islamist group Al Shabaab in Somalia has targeted international aid entities, but has focused its violence primarily on Somalia National Government and AMISOM (African Union Mission to Somalia) forces, as well as UN agencies, due to the political role played by the UN in the country.
In 2011 and 2012, Al Shabaab also explicitly barred 19 international aid organizations, including UN agencies and some major INGOs, from South Central Somalia and has demanded payments from aid organizations in exchange for access to famine-affected areas under its control.
Despite aid workers' perceptions of Al Shabaab as the primary security threat, most violent incidents affecting NGOs are reported to be the result of inter-clan hostilities, often revolving around competition over the resources and employment of the aid 'industry', with different groups seeking a bigger piece of the action. Based on an analysis of security incidents, the main risks for aid organizations seem to have to do with their selection of suppliers, contractors, staff and beneficiaries. (This squares with the affected population survey in which respondents overwhelmingly saw corruption as the main barrier to their access to aid.) Humanitarian presence in Al Shabaabheld areas is hindered by more than direct, physical security threats. Because Al Shabaab has been designated as a terrorist organization, aid organizations must contend with serious legal and financial risks if they run afoul of the anti-diversion regulatory framework of the US and other governments. This has implications for aid organizations because even though both Al Shabaab's and the national government's local authorities regularly seek to extort payments and bribes from aid organizations, the counter-terror legal risks apply only regarding the former.
The result has been that humanitarian presence in South Central Somalia has been increasingly constrained by aid organizations seeking to mitigate their physical, fiduciary and legal risks within this extremely complex and forbidding environment (Map 2). As in Afghanistan, most of the aid organizations that have been attacked in Somalia have retained a programming presence in country. Only one pulled out completely on the grounds of physical insecurity. Others have modified and limited their approaches while maintaining some operations. Often this entails, first, limiting their activity to more critical (life-saving) projects and, second, reducing senior staff exposure.
Over the past decade, the main way international aid organizations sought to reduce their exposure was via partnerships with local actors. Recent legal and financial/fiduciary issues have caused activities to cease in specific locations, when some partners prove to be fraud or diversion risks. And although they Map 1: Insecurity and humanitarian coverage in Afghanistan, 2014. remain the majority of organizations doing direct program implementation in Somalia, local NGOs have seen their direct access to international funding decline steeply, likely as a result of fiduciary concerns. The number of allocations to national NGOs from the UN-managed Common Humanitarian Fund has dropped 77 per cent -from 142 in 2011 to 33 in 2014 -while the number of UN allocations (subsequently sub-contracted to NGOs) has increased 300 per cent -from 33 in 2011 to 99 in 2014.
As in the other settings, access constraints and risk considerations have influenced the types of projects implemented. Interviewees in this context, as in Syria, are of the opinion that health facilities tend to be the types of projects most accepted by local non-state armed actors as they benefit all community members and conflict parties equally, with little opportunity for diversion. In contrast, the practice of identifying and targeting the most vulnerable members of a community for aid has always been difficult in the Somali context, as it runs counter to cultural norms of equity and can exacerbate competition between local clans.
However, when asked to identify the most urgent need, an overwhelming majority of the Somali affected-population survey respondents answered 'food'.
Map 2: South Central Somalia NGO (international and national) presence per region, 2014.
Risk-driven delays in responding to rapidonset crises and newly emergent needs were also reported in Somalia. At the same time, the clustering of most aid organizations in Mogadishu and other urban areas seemingly contradicts the logic of avoiding the frontlines of the conflict, as this is where much of the violence is playing out. The small number of international staff working inside Somalia are largely confined to bunkerized living conditions and have little to no contact with the people their programming is serving. High staff turnover and low field-level institutional memory and contextual expertise are the natural results.
Donor governments have reinforced the presence trends in Somalia by de facto encouraging aid in areas controlled by the government. The thematic approach to funding, in particular, has shifted the focus of aid intervention to urban areas and reduced funds for rural projects, where Al Shabaab's presence is strong. In 2009, the US Office for Foreign Assets Control (OFAC) refused to issue a general waiver for humanitarian aid to areas under Al Shabaab's control. Consequently, most aid organizations that have redeployed in South Central Somalia since the 2011 emergency have done so in areas controlled by the government.
Humanitarian coverage of needs in Somalia is thus unbalanced and appears politicized. In terms of organizational presence alone, the data show clearly the over-representation in the Banadir region (where Mogadishu is located) and weak coverage in areas where Al Shabaab is strongest, such as Middle Juba and Lower Shabelle (Map 3).

South Sudan
The civil crisis that began in December 2013 in the newly independent South Sudan has caused massive displacement and a major humanitarian crisis. During early days of the violence, tens of thousands of South Sudanese fled to bases of the UN Mission in South Sudan (UNMISS), which, in an unprecedented action by UN peacekeeping forces, took them in and established Protection of Civilians (PoC) sites that continue to shelter people at the time of writing.
Although representing a small portion of the affected population, the PoCs were the focus of much of the initial humanitarian response, as it was easier and safer to provide aid to these few locations with consistent, secure access. Elsewhere in the field, battle lines were constantly moving and towns were taken and retaken by different sides. Humanitarian field facilities and assets existing before the conflict were largely looted or destroyed in the process.
As battle lines solidified several months into the conflict, humanitarian organizations began attempting to push the aid response out to the 'hard-to-reach' areas. And because the population in need was spread across large areas in the Greater Upper Nile region, with limited road access, air-drops and mobile responses became the predominant operational approach.
Aid organizations cited physical access, logistical constraints and scattered and displaced populations as the main challenges to South Sudan's humanitarian response. Unlike many other contexts, national staff have been at greater risk than international staff for direct conflict-related violence, due to the ethnic dimensions of the civil war. However, violent crime has recently increased in densely populated places such as Juba and the Maban county refugee camps, affecting international staff as well. Numerous carjackings, violent robberies and sexual assaults suggested an atmosphere of growing impunity.
For reasons of ease and safety of access, humanitarian organizations avoided establishing a sizable presence outside of the PoCs. Overall humanitarian field presence in the Greater Upper Nile region declined considerably in the two years since the start of the crisis, with a 12 per cent decrease in operational organizations and a 36 per cent decrease in humanitarian projects. Similar to the access inertia observed in Afghanistan, many organizations chose to remain operational in PoCs only, even though the 75,000 inhabitants of PoC sites accounted for less than ten per cent of the displaced and at-risk population. Only a few large organizations with independent funding were prepared to respond outside of the PoCs, which required a significant financial and logistical investment in a location to ensure fundamental needs for staff, such as accommodation structures, security, food, water, vehicles and evacuation routes. For most of those organizations, the active combat conditions and lack of pre-existing logistical infrastructure drove a wholesale shift in operational modalities from in situ programming in field locations to mobile deliveries (often referred to in South Sudan by the shorthand 'rapid response'). The nearly complete reliance on air transport has meant that humanitarian delivery in South Sudan has been more both more expensive, and more scattershot, than in any other protracted humanitarian crisis.
Aid workers interviewed in South Sudan viewed insecurity more in terms of the logistical constraints emerging from largescale instability related to the conflict. Interviewees were most concerned with military movement, as this presents a threat to both staff and assets. Similarly, a majority of the South Sudanese people sampled in the survey ranked insecurity as the most significant barrier to receiving aid, but they did not perceive aid organizations to be in specific danger of violence, implying that it was generalized insecurity (active conflict conditions) that was the hindrance. Moreover, more survey respondents perceived risks to receiving aid rather than providing it, perhaps implying the need for beneficiaries to cross lines or expose themselves to opposition groups to collect the aid.
Although South Sudan hosts larger numbers of aid organizations than the other contexts studied, it is important to note the difficulty in sustainably meeting the needs of a far-flung population with mobile deliveries. As the conflict entered its third year in 2014, a few organizations such as the ICRC and MSF (Doctors Without Borders) endeavored to maintain static operations, but the humanitarian presence was still largely focused in the three UNMISS PoC sites and in field locations perceived to be a safe distance from frontlines (Map 4).

Syria
The Syrian conflict, a multi-party civil war with regional and international dimensions, presents the most challenging political and security environment for humanitarian response in recent memory. By March 2015, the fifth year of the conflict, areas of control had solidified, divided between the Assad government mainly in the south and west; Kurdish forces in the north; increasingly small portions held by various rebel groups, including the Free Syrian Army; and a consolidation of the Islamic State (IS) territory across much of the west and central portions of Syria, centered in Raqqa and spanning the Syria-Iraq border.
The massive humanitarian needs in Syria stem mainly from conflict-related displacement, affecting several million people, and public infrastructure damage.
The initial humanitarian response in late 2011 was limited to a small number of actors. It expanded over the next two years then plateaued in 2014 at 54 international humanitarian organizations (UN, Red Cross/Crescent Movement and INGOs) and roughly 175 Syrian NGOs and diaspora organizations. From the beginning, the humanitarian response to needs inside Syria has been bifurcated between the aid efforts of organizations officially sanctioned by the government of Syria Humanitarian operations face insecurity stemming directly from the conflict, (bombardment, ground fighting and crossfire) as well as threats arising from the multiplication and fragmentation of armed groups, which includes interference with aid deliveries, attempted diversions and kidnapping. IS threats to NGOs in 2013, particularly kidnapping, prompted them to curtail all cross-border movements of international staff from Turkey. Although the overall number of security incidents involving aid operations has been higher in western areas controlled by various armed opposition groups, the severity of the incidents (i.e. killings and kidnappings) and the overall threat levels are higher in areas controlled by IS.
The difficulties of the security context and increasing pressure from IS as it consolidated its control in 2013 and 2014 severely limited options for aid operations in Syria.
Organizations implementing cross-border aid from Turkey were forced to rely increasingly on national staff and/or national partner organizations, as the risks to international staff traveling across the border became too great. At least two major INGOs ceased direct cross-border implementation completely and began to work only through local partners. They also increasingly emphasized the monitoring and evaluation of programs, as diversions by IS and violations of counter-terror legislation became more prominent risks after IS increasingly interfered with aid operations. This also led to a significant decrease of the humanitarian presence in Raqqa and Deir Ez-Zor governorates, with several INGOs and diaspora NGOs withdrawing most of their staff and suspending their activities, leaving only around five still operational there at the end of 2014.
For INGOs that were able to remain operational in IS areas, their programming sector appeared to play a major role. Those involved in health care (particularly hospital and trauma care services), as well as those doing WASH programming, managed to maintain their presence. According to practitioners interviewed from these agencies, better access for these programs can be explained by the greater acceptance of health (particularly life-saving) programs and the fact that more complex service-type programs are more difficult to maintain than aid commodity (or cash) distributions, giving organizations less to fear about IS potentially enriching itself with their aid.
Across the whole of Syria, however, the need to reach large numbers of people using limited distributional capacity has caused most aid programs to focus on simple humanitarian distributions. Non-food items (NFI) distributions was the single most common activity undertaken by aid organizations in 2014 (Figure 9).
Only a small portion of programming has been in the form of cash and vouchers as opposed to aid commodities. Fear of possible diversion and tight controls of cross-border cash transfers from Turkey seem to be largely why this modality has not been more utilized.
The household survey of affected people in Syria revealed that the population perceived that their needs were increasing, but the number of aid providers was going down. The exception was Aleppo, where most respondents reported an increase in aid organizations. This squares with interview and data evidence that, as cross-border operations were being squeezed out of IS areas, organizations were concentrating increasingly in Aleppo and Idlib. Like the other contexts studied, most affected people do not perceive a significant direct threat to aid workers (despite casualty figures to the contrary). Only in Aleppo did a majority agree that (international) aid groups faced specific danger.
The IS threats to, and explicit expulsions of, international aid organizations from areas under their control significantly reduced the cross-border humanitarian coverage from Turkey. A UN official observed that most of the humanitarian programming in Syria was concentrated in a 40-kilometer zone along the border from Idlib to Aleppo.
Meanwhile, IS-controlled areas in Raqqa and Deir-ez-Zor have the fewest relief operations. Government and held territories show the highest ratio of aid activities to affected population (Map 5).

Conclusions
These results suggest that humanitarian response in these high-insecurity contexts is both more durable and more limited in scope and reach than it might appear to policy makers and the general public. Certain humanitarian organizations (far fewer than needs demand) have been able to remain operational in countries undergoing active conflict, despite the high risk of targeted violence. But they have done so at the cost of the core humanitarian principle of impartiality, i.e. prioritizing those most in need. Without diminishing the achievements of humanitarians who work in dangerous places at great personal risk, it is important to recognize that aid organizations have incentives to appear more present than they actually are, which can obscure the reality that widespread needs are going unmet.
The constraining effects of insecurity on humanitarian operational presence and coverage of needs are considerable. While it should come as no surprise that insecurity makes accessing affected populations and meeting their needs more difficult, these findings illuminate the paucity of humanitarian coverage where it is often obscured, albeit by well-meaning humanitarian actors.
Another uncomfortable but inescapable conclusion of the research is that humanitarian coverage is not only uneven within and across contexts, but it is also proportionally lower in areas under control of militants in opposition to the government and to the Western powers that provide most of the humanitarian funding (e.g. areas controlled by IS in Syria and by Al Shabaab in South Central Somalia). The implications of this for the core humanitarian principles of impartiality, neutrality and humanity are fairly stark.
The conclusions also suggest a few potential areas for action: 1. Aid actors must increase operational transparency for a more accurate picture of coverage.
Reputational and financial concerns clearly create the tendency among some organizations to overstate their presence and territorialize service areas even when they are meeting just a fraction of the need. Apart from misrepresentation, agencies' general reluctance to fully disclose operational information (which this research study experienced first-hand) has resulted in a much weaker situational understanding of aid operations in arguably the most critical contexts. To avoid these tendencies and to present a clear picture of the scope and scale of the humanitarian response -and its gaps -the humanitarian community requires common measures of presence and coverage. Ideally, coverage would be measured not by the calculation of humanitarian presence over people in need, as used for the purposes of this quantitative analysis, but rather by the percentage of people in need being reached and served by the humanitarian response.
For this to happen, more work needs to be done on developing a common methodology for calculating the number of people in need from among the affected population. Likewise, more robust information-management systems need to be developed for mapping operational activity, as well as methodologies for tracking and reporting on the specific modality of rapid response deliveries and the populations they reach.
Greater transparency as to which actors are operating in these most difficult settings could provide the opportunity to deliver aid in a more effective and coordinated manner, gaining efficiencies. This is more of a normative challenge than a methodological one. It requires the organizational relationships that enable information to flow freely yet securely, in a way that benefits all parties in the process. Designing the system would not be difficult but, to work, it will require a critical mass of stakeholders to participate fully and consistently.
A related measure that would enhance both transparency and accountability would be humanitarian actors jointly investing in systematic, independently conducted remote surveys of affected populations. This would enhance knowledge of underserved areas, priority needs and issues of importance to local populations.
2. After identifying coverage gaps, the humanitarian system must prioritize finding means to fill them.
Collectively, humanitarian actors have met access constraints, if not with complacency then with a decided lack of urgency in finding means to reach the people in need that remain unassisted by the overall humanitarian response. This is not born of neglect or incompetence, but rather of the fundamentally fragmented nature of humanitarian response. Each organization being too small to cover more than a fraction of the people in need, each focuses on operating neutrally and impartially within the area where it has decided to be present. Facing at times formidable obstacles and threats, each organization does what it can, where it can, to the best of its ability. However, on the macro scale this amounts to partial and inequitable coverage for the country as a whole, as many like-minded agencies tend to cluster in the same places. The reality of sparse humanitarian coverage warrants a more strategic overview and stronger leadership. In addition to advocating for disaster-affected governments and nonstate armed actors to protect and aid civilians in areas they control, the various parts of the humanitarian system have responsibilities to find proactive and innovative means for reaching people in areas too risky for humanitarian organizations to operate. This should start with identifying the humanitarian actors who are already present and assessing what more they can absorb and implement to serve greater numbers of people. Second, when the limits of that potential capacity are reached, aid actors could make aggressive and concerted efforts to identify or help organize additional local/ national entities or mechanisms (e.g. community-based, commercial, religious, other) that could potentially deliver materials and services, even if they are not ideal humanitarian partners for political or other reasons.
3. Donor governments must accept responsibility for correcting aid coverage imbalances.
Core humanitarian principles are threatenedand there are attendant security risks -when donor funding strategies discourage programming in opposition-held territories. Although the problem is not universal, aid presence in many countries appears partial and politicized.
Map 5: Insecurity and humanitarian coverage in Syria, 2014.
Donors should encourage agencies to extend their presence and to devise solutions for presence gaps, and should remove the obstacles and disincentives to their doing so. Blanket humanitarian waivers and financial/legal exemptions for aid providers should be the norm when there are high levels of need. For their part, individual aid organizations must be frank about their own presence, limitations and capacities, and speak out forcefully when they know that needs are not being met.