ArticlePDF Available

What Do We (Not) Know About Development Aid and Violence? A Systematic Review


Abstract and Figures

The paper presents findings from the first-ever systematic review of the causal impact of development aid on violence in countries affected by civil war. The review identifies 19 studies: Fourteen within-country studies from Afghanistan, Iraq, Colombia, Philippines, and India, and five cross-national studies. These studies investigate the impact of six aid types: Community-driven development, conditional cash transfers, public employment scheme, humanitarian aid, infrastructure, and aid provided by military commanders in Iraq and Afghanistan. The evidence for a violence-dampening effect of aid in conflict zones is not strong. Aid in conflict zones is more likely to exacerbate violence than to dampen violence. A violence-dampening effect of aid appears to be conditional on a relatively secure environment for aid projects to be implemented. A violence-increasing effect occurs when aid is misappropriated by violent actors, or when violent actors sabotage aid projects in order to disrupt the cooperation between the local population and the government.
Content may be subject to copyright.
What Do We (Not) Know About Development Aid and
Violence? A Systematic Review
Christoph Zürcher
Source Information
World Development
October 2017Volume, 98(IssueComplete)Page, p.506To-522
This is the final draft. Access the published version here:
doi: 10.1016/j.worlddev.2017.05.013
What Do We (Not) Know about Development Aid and
Violence? A Systematic Review
The paper presents findings from the first-ever systematic review of the causal impact of
development aid on violence in countries affected by civil war. The review identifies 19 studies:
Fourteen within-country studies from Afghanistan, Iraq, Colombia, Philippines and India, and
five cross-national studies. These studies investigate the impact of six aid types: Community
driven development, conditional cash transfers, public employment scheme, humanitarian aid,
infrastructure and aid provided by military commanders in Iraq and Afghanistan. The evidence
for a violence-dampening effect of aid in conflict zones is not strong. Aid in conflict zones is
more likely to exacerbate violence than to dampen violence. A violence-dampening effect of aid
appears to be conditional on a relatively secure environment for aid projects to be implemented.
A violence-increasing effect occurs when aid is misappropriated by violent actors, or when
violent actors sabotage aid projects in order to disrupt the cooperation between the local
population and the government.
Keywords: development aid; violence; insurgency; COIN; systematic review
The past fifteen years witnessed an unprecedented securitization of foreign aid. In the wake of the
9/11 terror attacks, war-torn and fragile states came to be seen as a global security threat (Patrick,
2011; WDR, 2011). Foreign aid, alongside with diplomacy and defense, was turned into a major
strategic tool for addressing this threat. The lengthy asymmetrical wars in Iraq and Afghanistan
only reinforced the expectation that development assistance could be used as an effective means
to stabilize war-torn countries. Multi-lateral and bi-lateral donors increasingly saw aid as an
important instrument for addressing development and security issues simultaneously (Brown and
Grävingholt 2014) and the military lavishly spent aid as “monetary ammunition” in
counterinsurgencies in order to win mind and hearts.
But can development aid really reduce violence? The debate about the effects of aid on conflict
is not a new one. Development economists have argued, on mainly theoretical grounds, that aid
should help to lower the risks of war. The literature on civil war has long claimed that low levels
of economic development and low growth rates tend to increase the risk for war. Growth spurred
by aid should therefore reduce the risk for war (Collier and Hoeffler 2002). Unfortunately,
whether aid actually leads to economic growth remains hotly debated (cf. Doucouliagos and
Paldam 2009). Another strand of the literature linked aid to reduced risk of war by suggesting
that aid might be fungible, allowing recipient governments to boost their military spending. This
in turn should deter potential rebels (Collier and Hoeffler 2007, Ree and Nilesen 2009). Other
scholars were more skeptical and believed that aid might actually increase the propensity for war.
A prominent argument was that foreign aid could increase the spoils to be won from rebellion
(Azam 1995; Grossman 1991). As a result, seizing the state may become an attractive option for
rebels, which would make civil war more likely (Arcand & Chauvet, 2001; Grossman, 1992).
This argument tied in with the sizeable qualitative “do-no-harm” literature which warned that aid
could be misappropriated by local violent actors and used for sustaining violence (for example,
Anderson, 1999; Uvin, 1998). Scholars have argued that aid can be misused by rebels for
financing their war (Bradbury and Kleinmann, 2010; Goodhand, 2002; De Waal, 1997; Easterly,
2001), that it can alleviate the pressure of local actors to provide basic services to their
constituencies, freeing up resources that can be invested in violence (Polman, 2010; Duffield
1994; Luttwak 1999), that it can fuel conflict by increasing corruption (Goodhand, 2005; 2006),
or provide perverse incentives to private security firms to fuel conflict (Aikins, 2010; Wilder and
Gordon, 2009).
This scholarly debate has been renewed and intensified over the last decade. As billions of aid
dollars are being spent in the hope that foreign aid can buy stability, a new wave of scholarship
emerged, committed to rigorous tests and evidence-based policies. Newly available disaggregated
sectoral and subnational aid data and subnational security event data opened up new avenues for
causal inference and led to theoretical innovation. This paper takes stock of the available
evidence on the impacts of aid on violence. It is based on a systematic review covering the years
20012016 (cut-off date: November 1, 2016). A systematic review differs from a traditional
literature review in important ways. It is primarily a stock-taking exercise, designed to identify all
available evidence on a given topic.1 Systematic reviews depend largely on what studies are
available, how they were carried out (the quality of the tests) and the outcomes that were
measured. The most important criteria for including a study in a systematic review is that the
study has a clearly defined identification strategy which allows inferring causal mechanisms.
This is a high threshold which will often reduce the number of included studies, but relaxing that
threshold would lead to the inclusion of studies which do not provide robust causal evidence.
Furthermore, a systematic review requires a transparent search strategy based on a search
protocol and transparent criteria for inclusion and exclusion, which are a priori defined in order to
minimize any selection bias. Systematic reviews are thus different from traditional reviews,
where authors are at liberty to include and exclude studies based on, for example, theoretical
preferences or anticipated findings.
Whether or not a study meets the inclusion criteria is determined by reliable and replicable
coding procedure. For this review, three researchers independently assessed the studies. Only
studies that met the inclusion criteria according to all three researchers were included in the final
Four inclusion and exclusion criteria were used:
1) The independent variable is development aid, or a closely related concept, such as foreign aid,
foreign assistance, humanitarian aid, etc. Military aid was excluded.
2) The dependent variable is violence, or a closely related concept such as armed conflict, civil
war, insurgency etc. Also included were the opposite of these concepts, such as security, stability,
counterinsurgency, etc.
3) Only published studies were included. Working papers and grey literature were not included.
1 I follow the standard definition for systematic reviews; for example, see the Campbell
Collaboration, “What is a systematic review” ( Also, Waddington
et al. (2012). Similar definitions are offered by DFID (2012) and Petticrew & Roberts (2006).
4) Finally and most importantly, only studies with a clear and transparent identification strategy
allowing for causal inference were included. The minimum threshold for this criterion is that the
methodological set-up of the studies allows assessing the counterfactual: what would have
happened without the intervention. Such a criterion does not a priori exclude qualitative studies.
Careful process-tracing or structured comparison allow for discussing the counterfactual.
Nevertheless, all included studies turned out to be quantitative studies with an experimental or
quasi-experimental design.
The following steps were carried out to identify studies to be included. The researchers had
previously identified ten seminal papers that needed to be included in the review. Search terms
based on concepts found in these studies were selected and tested in preliminary searches
conducted in EconLit. This helped determine appropriate keywords that would yield relevant
After further validation by all researchers, a final search strategy was devised that included the
two core concepts of this review: development aid and violence. For each of these concepts,
keywords were identified along with relevant subject terms found in the database’s unique
thesaurus, when appropriate. Searches were executed by a research librarian in the following
electronic databases: PAIS International (ProQuest), EconLit (ProQuest), International Political
Science Abstracts (EBSCO), Worldwide Political Science Abstracts (ProQuest) and Web of
Science (Social Sciences Citation Index). Searches were limited to articles published in English
between 2001 and November 2016. Results were then exported to a bibliographic management
tool and duplicates were removed
Upon completion of the database searches, three researchers screened the identified articles to
exclude those which did not meet the criteria for inclusion. Eventually, 107 full text studies were
read by all three researchers. 88 were excluded because they did not meet all inclusion criteria.
By far the most frequent reason for exclusion was that the study was descriptive in nature without
a clear causal identification strategy. We also excluded studies that make a formal argument but
do not provide an empirical application (for example Child and Scoones 2015 and Scoones
2013), and studies that used the “wrong “ independent variable (for example, transnational
terrorism, in Young and Findley 2011, Azam and Delacroix 2006, Azam and Thelen. 2008). We
also excluded four studies which used ODA as their independent variable (Collier and Hoeffler
2007, Nielsen 2011, Ree and Nillesen 2009, Tahir 2015), because this high level of aggregation
masks important differences between aid sectors and made it impossible to infer causal processes
(cf. Findley et al. 2011,Young & Findley 2011). Unpublished studies and working papers were
also excluded (for example Arcand and Labonne,2010, van Weezel, Stijn, 2015, Iyengar et al.
2011, Böhnke et al., 2015), but their content was assessed in order to ensure that there was no
publication bias in the included studies.
Figure 1 details the screening process.
The final sample consisted of 19 studies, which can be considered a normal sample size for
systematic reviews in the field of development.2
2 For example, the median of included studies in all 26 systematic reviews which the
International Initiative for Impact Evaluation (3ie) has conducted between 2009 and 2017 is 20.
The average is 27.8, and the range is 5 to 93. 14 out of 27 studies are based on a sample smaller
than 19. Note that for this calculation, I exclude one outlier which is based on 420 studies on
educational outcomes in middle-income countries. Without excluding this outlier, the average is
Out of 19 included studies, 18 are published in peer-reviewed journals, and one is a report
published by the World Bank. 14 studies focus on the sub-national level: Six on Afghanistan,
three on India, two on Iraq, two on the Philippines and one on Colombia. Five studies are cross-
national. Two of these are global in reach and focus on all civil wars after 1945, respectively after
1989. One focuses on all wars (civil and international) in non-OECD countries after 1971; two
focus on 22 sub-Saharan African states between 1989 and 2008. With the exception of one study
(Nunn & Qian 2014) which also includes international wars, all studies investigate the impact of
aid in civil wars. This should give us confidence that differences in outcomes are not driven by
conceptually different forms of violence.
The studies investigate the effects of six different aid types: community driven development
programs (7 instances), commander emergency response program (6), humanitarian aid (5),
employment programs (3), conditional cash transfer programs (2), and large-scale infrastructure
programs (1). Note that one study can cover more than one aid type.
This paper makes several important contributions.
It is the first systematic review on the topic and therefore the first paper to map and summarize
the available evidence base for assessing the impact of aid in conflict zones. With 19 studies, the
evidence base is surprisingly small, given the massive sums which donors spend on aid in
conflict zones. Even more surprising is the fact that development organizations have so far made
no serious efforts to rigorously evaluate their efforts. Out of 19 studies, only one was conducted
by a development actor, while four were financed by the military, underlining once more the
securitization of aid.
Secondly, the study shows that the evidence for a violence-dampening effect of aid in conflict
zones is not strong. On the aggregate, aid in conflict zones is more likely to exacerbate violence
than to dampen violence. This is especially true for “font-line aid” which is spent in highly
insecure regions, with strong presence of anti-government forces. Most worrisome is that
40 and the median is 23. One important reason for what may seem a small number of studies is
that rigorous tests require rich, longitudinal data which is not often available from poor, war-torn
humanitarian aid is consistently associated with more, not less violence. But also aid given by the
military and community driven development projects implemented in insecure region rarely
dampen violence.
Thirdly, the review offers a stock-taking of the possible causal paths that lead to more or less
violence. The reviewed studies identify three causal paths that reduce violence and two that
increase violence. The causal mechanisms which are most often thought to account for a
violence-increasing effect of aid are “predation” and “sabotage”. Predation refers to aid being
misappropriated and reinvested in financing violence. Sabotage refers to a strategic response of
insurgents to the possibility that aid can mobilize support for the government. In order to counter
that threat, insurgents sabotage aid projects and deter collaboration between the local population
and the government. Importantly, the evidence shows that all six aid types which the studies
review can trigger predation or sabotage. This means that there is no aid type which is immune to
doing harm.
Fourthly, there is strong evidence that a violence-reducing or violence-increasing effect of aid is
not primarily caused by how much aid is given, or by what type of aid is given, but rather by the
conflict environment in which aid is implemented. Aid injected in insecure regions appears to
increase violence. By pointing out the importance of the local conflict environment, this review
moves the discussion away from whether aid works towards under what conditions aid may
work. An important implication for scholars is that future research must engage in a much more
systematic way with the scope conditions for aid effectiveness, especially with the sub-national
security environment. As long as our understanding of scope conditions and causal processes is
limited, it would wise to not allocate aid projects in highly insecure regions, since the available
evidence strongly suggests that this will do more harm than good.
The remainder of this paper is organized in four sections. The first section describes the six types
of aid which the reviewed studies investigate. The second section summarizes the evidence of the
impact of aid on violence. The third section discusses the causal mechanisms which the reviewed
studies propose. The last sections discusses the results and offers conclusions, including
suggestions for future research.
1. Adams, Greg. 2015. “Honing the proper edge: CERP and the two-sided potential of
military-led development in Afghanistan”. The Economics of Peace and Security
Journal, v. 10, n. 2: 53-61
2. Beath, Andrew, Fotini Christia, and Ruben Enikolopov. 2012. “Winning Hearts and
Minds through Development: Evidence from a Field Experiment in Afghanistan“. Policy
Research Working Paper 6129. (The World Bank)
3. Berman, Eli, Jacob N. Shapiro, and Joseph H. Felter. 2011. “Can Hearts and Minds Be
Bought? The Economics of Counterinsurgency in Iraq.” Journal of Political Economy
119, no. 4: 766819.
4. Berman, Eli, Joseph Felter, Jacob N. Shapiro, and Erin Troland. 2013. Modest, Secure
and Informed: Successful Development in Conflict Zones”. American Economic
Review: Papers & Proceedings,103(3): 512517
5. Böhnke, Jan Rasmus, and Christoph Zürcher. 2013. “Aid, Minds and Hearts: The Impact
of Aid in Conflict Zones.” Conflict Management and Peace Science 30, 5: 41132.
6. Child, Travers B. 2014. “Hearts and Minds Cannot Be Bought: Ineffective
Reconstruction in Afghanistan.” Economics of Peace and Security Journal 9, 2: 4349.
7. Chou, Tiffany. 2012. “Does Development Assistance Reduce Violence? Evidence from
Afghanistan.” Economics of Peace and Security Journal 7, no. 2: 513.
8. Crost, Benjamin, Joseph Felter, and Patrick Johnston. 2014. “Aid under Fire:
Development Projects and Civil Conflict.” The American Economic Review 104, no. 6:
9. Crost, Benjamin, Joseph H. Felter, Patrick B. Johnston. 2016. “Conditional Cash
Transfers, Civil Conflict and Insurgent Influence: Experimental Evidence from the
Philippines.” Journal of Development Economics, v. 118 (Jan. 2016): 171-182.
10. Dasgupta, Aditya and Gawande, Kishore and Kapur, Devesh. 2016. (When) Do Anti-
Poverty Programs Reduce Violence? India's Rural Employment Guarantee and Maoist
Conflict (April 22, 2016, forthcoming in International Organization).
11. Hoelscher, Kristian, Jason Miklian, and Krishna Chaitanya Vadlamannati. 2012.
“Hearts and Mines: A District-Level Analysis of the Maoist Conflict in India.”
International Area Studies Review 15, 2: 14160.
12. Khanna, Gaurav, and Laura Zimmermann. 2014. “Fighting Maoist Violence with
Promises: Evidence from India’s Employment Guarantee Scheme.” Economics of Peace
and Security Journal 9, no. 1: 3036.
13. Narang, N. “Humanitarian Assistance and the Duration of Peace after Civil War.”
Journal of Politics 76, no. 2 (April 2014): 44660.
14. Narang, N. 2015. “Assisting Uncertainty: How Humanitarian Aid Can Inadvertently
Prolong Civil War.” International Studies Quarterly 59 (1): 18495.
15. Nunn, Nathan, and Nancy Qian. 2014. “US Food Aid and Civil Conflict.” American
Economic Review 104 (6): 163066.
16. Sexton, Renard. 2016 “Aid as a Tool against Insurgency: Evidence from Contested and
Controlled Territory in Afghanistan,American Political Science Review,110, 4: 731-
17. Weintraub, Michael. 2016. “Do all good things go together? Development Assistance
and Violence in Insurgency”. Journal of Politics,78,4: 989-1002
18. Wood, Reed M., and Emily Molfino. 2016. “Aiding Victims, Abetting Violence: The
Influence of Humanitarian Aid on Violence Patterns During Civil Conflict.” Journal of
Global Security Studies 1 (3): 186-203
19. Wood, Reed, and Christopher Sullivan. 2015. “Doing Harm by Doing Good? The
Negative Externalities of Humanitarian Aid Provision during Civil Conflict.” Journal of
Politics 77 (3).
The 19 studies included in this review investigate the impact of six different types of aid:
community driven development programs, commander emergency response program,
humanitarian aid, employment programs, conditional cash transfer programs and large-scale
infrastructure programs. The next section briefly describes the main characteristic of these aid
Community driven development (CDD)
CDD programs are classical development tools primarily aimed at poverty-reduction. The
defining feature of CDD is that it requires the participation and continuous involvement of the
local communities. Typically, communities first assess their needs and prioritize them in a
participatory way and then apply for a grant. CDD interventions are typically small infrastructure
such as irrigation, flood protection, rehabilitation of transport infrastructure and the like, or
capacity building measures. CDD programs have long been viewed as particularly relevant
development interventions in conflict zones (cf. World Bank 2013, World Development Report
2011) because the programs are flexible, small-scale and demand driven, thereby promising local
ownership and quick results. Because of the involvement of the communities, development actors
may get access to remote and insecure regions which would otherwise be off-limit. Often, CDD
interventions are also used as a means to reestablish a direct link between communities and the
subnational government, which in many cases is lacking after conflict.
Commander Emergency Reconstruction Program (CERP)
CERP is a program of the US military that was specifically designed as a tool for
counterinsurgency. It is “monetary ammunition” intended to improve security through non-lethal
means (Bodnar and Gwinn, 2010). The program was considered critical to supporting military
3 I exclude large-scale infrastructure from this discussion. Only one study (Berman et al. 2013)
tries to measure the effect of large-scale infrastructure built by the US Army Corps of Engineers
in Iraq. But since this study is mainly on CERP, it provides not enough information on the type
and scale of the infrastructure to warrant a discussion here.
commanders in the field (SIGIR, 2013). CERP funds are handed out by US commanders to fund
projects for local communities with the aim of garnering the population’s support and
cooperation where the military is active. Between 2003 and 2008, CERP spending amounted to
$3 billion in Iraq (Berman et al., 2011) and $2.2 billion in Afghanistan (Child, 2014).
CERP funded a wide range of projects, for example road construction, cash-for-work,
agricultural assistance, water projects, militia payoffs, healthcare, or rehabilitation of schools.
Since its main objective was to win over the local population, commanders funded projects which
they thought the communities wanted or needed most. CERP projects were rather small on
average (around $100k) and rather short (around 80 days, cf. Berman et al. 2011). However,
there were also large projects well over $100k, mainly for roads and bridges. In Afghanistan 60%
of the spending went onto road and bridges (Johnson et al, 2012).
Conditional Cash Transfer (CCT)
CCT programs are another staple of traditional development programming. CCT programs intend
to reduce poverty by providing grants to poor households, based on some number of conditions.
For example, households may have to ensure that their children attend school and receive a
variety of medical treatments in order to receive funding. The monetary value of transfers is
typically small, but can make a significant contribution to the income of the poorest households.
CCT programs primarily intend to protect the most vulnerable segments of society. Recently
CCT programs have been increasingly used in contexts of conflict and fragility, for example, in
Afghanistan, Colombia, the Philippines, Sudan, Somalia, Haiti, Liberia and Mozambique (for an
overview see Holmes 2009). While they are not primarily intended as a tool for conflict
transformation and stabilization, development actors hope that CCT can also increase social
capital and trust, strengthen the sense of sense of citizenship and improve the social contract
between communities and the government, all of which is particularly important in post-conflict
Employment Programs
Employment creation programs include a range of interventions, ranging from enabling macro-
level policy measures to stimulate employment growth to interventions aimed at promoting self-
employment and to direct employment creating (Holmes et al. 2013), the latter often in the form
of cash-for-work, food-for-work or temporal guaranteed employment. Employment programs are
primarily meant as a means to fight poverty and inequality. However, in contexts of fragility and
conflict, employment schemes are also seen as contributing to stability (2011 World
Development Report WDR). The United Nations Policy for Post-Conflict Employment Creation
(UN PCEI 2009 ) states that employment has a crucial role to play for growth, reintegration and
sustainable peace. Accordingly, donors have supported numerous direct employment creation
initiatives as part of post-conflict reconstruction in countries such as Iraq, Afghanistan, Nepal and
Liberia and elsewhere (cf. McCord and Chopra 2010, McCord and Slater 2009).
While authors have remarked that there is no robust evidence so far to support these claims
(Brück et al. 2015, Holmes et al. 2013), the theoretical arguments for a stabilizing effect of
employment are plausible. Employment may increase the state’s legitimacy, reduce poverty,
inequality and grievance, all of which are seen as drives of conflict, or raise the cost of violence
by drawing the population away from insurgent activity (GIZ 2013, Blatmann and Annan 2015).
Humanitarian Aid
Humanitarian aid is defined as assistance designed to save lives, alleviate suffering and maintain
and protect human dignity during and in the aftermath of emergencies. Humanitarian aid includes
relief and assistance, protection, support services, and material assistance like food and medical
supplies. Running refugee camps, which can include services such as education, training and
labor programs, also falls into the humanitarian aid category. The majority of humanitarian aid is
allocated in conflict zones. Much of it is distributed by NGO, because government actors may not
have access to conflict regions.
These six types of aid fall by and large in distinct categories, but there are some overlaps. This is
especially true for CDD programs and CERP, which both deliver a wide range of similar project
outputs. CERP and many CDD programs aid at improving local basic infrastructure and service
delivery and predominantly provide small infrastructure. There may also be some overlap
between CERP and humanitarian aid, as some CERP funding was spent on urgent humanitarian
needs (cf. Sexton 2016). However, there are important differences in the mode of delivery. CDD
require the structured and prolonged engagement with local communities. CERP projects were
commissioned by commanders and rarely based on structured engagement with the communities.
Furthermore, CDD projects and humanitarian aid is delivered by civilian organizations, in
collaboration with the local communities, whereas CERP projects were usually delivered by the
military. Also, CERP projects were implemented in the most insecure regions, because they were
primarily meant to improve the environment for the military in contested regions. CDD projects,
while also used in conflict-affected states, are usually only implemented in regions where NGOs
have secure access.
In theory, there could also be some overlaps between CDD, CERP and employment schemes.
Many CDD projects use local labor to build infrastructure, and a small fraction of CERP projects
involved cash-for-work. In both cases, the employment effects would be small, short term, and a
side effect of the infrastructure project. By contrast, the employment scheme in our sample is a
massive public employment scheme which guarantees 100 days of work per year for all members
of eligible communities.
The next section summarizes the impact of aid on violence. I group aid interventions in the
aforementioned six categories, but remain aware of the fact that there are some overlaps in
On the aggregate, out of the 24 interventions which the sample covers, only seven had a violence-
reducing effect. Three had a heterogeneous treatment effect, six had no effect at all, and nine had
a violence-increasing effect. When looking at the different aid types separately, the following
picture emerges: Humanitarian aid never had a violence-dampening effect: In all five cases
humanitarian aid actually increased violence. Of seven community driven development
programs, only one had an unqualified violence dampening effect. Another one had a violence-
dampening effect only in relatively secure regions, but not in insecure regions. Three had no
effect, and two increased violence. Likewise, for aid which is given by the military, only one out
of six studies finds an unqualified violence dampening effect (Berman et al. 2011). A subsequent
study by the same authors qualifies that result in important ways. Berman et al. (2013) find that
the violence dampening effect actually occurs only in interaction with increased troop presence.
Two studies find no effect, and three more found that CERP projects could also increase
violence, when they were large, or implemented in insecure regions. Out of two conditional cash
transfer programs, one dampened violence, but the other one increased violence. Finally, two out
of three studies on employment programs find a violence reducing effect. However, this finding
must be treated with caution because all three studies investigate the same program, hence
external validity may be limited.
In sum, the evidence for a violence-dampening effect of aid in conflict zones is not strong. No
effects, heterogeneous effects and violence increasing effects appear to occur much more often.
Table 1 summarizes the results. The next section offers a more detailed discussion of the
Table 1: Aid Types and Outcomes
Aid Programs and Source
USAID Community Stabilization
Program (CSP)
Berman et al. (2013)
National Solidarity Program
Beath et al. (2012)
reducing / only
in secure regions
USAID Community Action
Berman et al. (2013)
No effect
National Solidarity Program
Chou (2012)
No effect
USAID Governance Local
Community Development in
Chou (2012)
No effect
Asian Development Bank
Community-Driven Development
Project in the Philippines
Crost et al. (2016)
Multi-sectoral, community-level
Böhnke and Zürcher (2013)
insecurity of
Pantawid Pamilyang Pilipino
Program / Bridging Program for
the Filipino Family, Phillipines
Crost et al. (2014)
World Bank / Interamerican
Development Bank, Familias en
acion, Colombia
Weintraub (2016)
National Rural Employment
Guarantee Act (NREGA)
Dasgupta et al. (2016)
National Rural Employment
Guarantee Act (NREGA)
Khanna & Zimmermann (2014)
increasing in the
short run
National Rural Employment
Guarantee Act (NREGA)
Hoelscher et al. (2012)
Program of US Army Corps
Berman et al. (2013)
No effect
Commanders Emergency
Response Program (CERP)
Berman et al. (2011)
Commanders Emergency
Response Program (CERP)
Berman et al. (2013)
decreasing, but
only in
interaction with
troop strength
Commanders Emergency
Response Program (CERP)
Sexton (2016)
reducing in
regions under
increasing in
contested regions
Commanders Emergency
Response Program (CERP)
Chou (2012)
No effect
Commanders Emergency
Response Program (CERP)
Child (2014)
No effect
Commanders Emergency
Response Program (CERP)
Adams (2015)
increasing for
projects >
decreasing for
projects <
Food aid
(Nunn and Qian 2014)
Humanitarian aid (all)
Wood and Molfino (2016)
Humanitarian aid (all)
Wood and Sullivan (2015)
Humanitarian aid (all)
Narang (2015)
All civil war
1969 2008
Humanitarian aid (all)
Narang (2014)
Post civil
1989 - 1999
CERP is the only aid type which is primarily intended to reduce violence. Development
outcomes are a secondary objective. By far the most influential studies on CERP are Berman et
al. ( (2011, 2013). It is appropriate to say that these studies relaunched the discussion about the
impacts of aid on violence, not least by triggering three replication studies. Berman et al. (2011)
measure violence as the number of attacks against US and Iraqi government per district half-year.
Data come from a declassified version of “significant activity” (SIGACT) reports collected by
the US army. The studies employ a first-difference design where changes in violence are
regressed on changes in aid spending, controlling for previous levels of violence and troop
strength. Berman et al. (2011) find that smaller CERP projects (under $50k) reduce insurgent
violence. Berman et al. (2013) improve over their preceding study by adding a control for troop
presence, recognizing that their previous finding might have captured the effect of troop presence
rather than the effect of CERP. The measurement for troop presence is based on newspaper
reporting. The revised study finds again that smaller CERP projects (under $50k) reduce
insurgent violence, but only in interaction with larger numbers of troops (e.g. it is the interaction
term which reaches significance).
The authors explain the violence-reducing effect by an information-centric model of
counterinsurgency. The model assumes that local communities possess critical information on the
activities of insurgents. The prospect of rewards in the form of development aid acts as an
incentive for local communities to share this information with the government and its
international allies. As a result, the government’s counterinsurgency efforts become more
effective and security eventually increases.
The Berman et al. studies has been replicated three times in Afghanistan, where CERP was also
widely used by the U.S. military. Two studies (Chou 2012 and Child 2014) could not find an
effect of CERP spending on insurgent violence.
Adams (2015) also replicates the Berman et al. study in Afghanistan. He uses declassified
military data from SIGACT for measuring the depended variable which includes around 107,350
insurgent-initiated events and covers 32 months between 2011and 2013.The data is parsed to the
district level, and adjusted to a per capita basis. Effects are estimated with OLS, using lags
ranging from one to three months. The study finds that small CERP projects (<USD50,000) are
associated with a statistically significant reduction in violence but larger CERP projects actually
led to an increase in violence. Both effects hold for the three months lags only. It is possible that
Adams (2015) found an effect whereas Chou (2012) and Child (2014) did not because of slightly
different data sources and model specifications. Chou (2012) and Child (2014) use one-month
lags, and Child (2014) also used a different source for measuring violence (the Worldwide
Incidents Tracking System).
Adams (2015) also reports the results from a qualitative survey with nine Civil Affairs Officers
who were familiar with CERP in Afghanistan. The results suggest that the respondents did not
think that CERP was an effective tool for reducing violence, nor did they think that CERP funds
were given conditional, and in exchange for information, which is, as mentioned before, an
important element on the causal theory proposed by Berman et al. (2011).
Sexton (2016) provides another study on the effects of CERP in Afghanistan. He uses variation in
week-per-week CERP spending per district week in all Afghan districts (instead of levels of
CERP spending, as the pervious studies did). This measure is chosen because it is assumed that
this variation is quasi random (caused by the unpredictable bureaucracy), whereas CERP
spending per se is endogenous to violence. The study finds that CERP has a violence-reducing
effect in regions which are under the control of the government and its allies, but a violence-
increasing effect in regions which are contested or under insurgent control. Control is proxied be
the presence of absence of a FOB (a battalion level forward operating base). The author argues
that the violence-increasing effect is caused by the attempts of insurgents to sabotage aid
programs which might win over the population to the government. These attempts at sabotage
can only be carried out in districts which are not yet secured by the government. This is why
more aid creates more violence in non-secured districts, but can dampen violence in secured
districts. Sexton (2016) also rejects the information-centric model. He argues that if increased
CERP spending would buy actionable information, then increased spending should be associated
with increased COIN activity. However, he finds no evidence for this.
With six studies dedicated to CERP, it is among the best-researched aid type in our sample. The
overall evidence for a violence-reducing effect is not strong. One study finds a violence
decreasing effect (Berman et al. 2011). Two studies find no effect (Chou 2012 and Child 2014).
Three studies find qualified effects: Berman at al. (2013) find that CERP dampens violence but
only in conjunction with increased troop levels. Adams (2015) finds that CERP increases
violence when projects are >$50.000 and dampens violence when projects are < $50.000 and
Sexton (2016) finds that CERP increases violence in territories which are not under the control of
the counter-insurgents, but dampens violence where counterinsurgents have control.
One observations stands out: If CERP has a violence-dampening effect, it is most likely only in
regions which are relatively secure. In insecure regions beyond the control of the government or
the military, CERP is likely to increase violence.
Unlike CERP, CDD programs are not designed as a counterinsurgency tool. CDD programs are
classical development tools primarily aimed at poverty-reduction. Two studies in our sample are
devoted to testing the effects of specific CDD programs (Beath et al. 2012; Cost et al.). One study
(Böhnke & Zürcher 2013) investigates the impact of multi-sectoral, community level
development aid in rural areas of North East Afghanistan. Two studies, while primarily focusing
on the impact of CERP, also each include two CDD programs in their investigation (Berman et
al., 2013; Chou 2012).
Of these seven community driven development programs, only one had an unqualified violence
dampening effect. Another one had a violence-dampening effect only in relatively secure regions,
but not in insecure regions. Three had no effect, and two increased violence.
Beath et al. (2012) investigate the impact of the National Solidarity Program (NSP) in
Afghanistan. NSP is a nation-wide, community-driven development program in Afghanistan
which gave block grants to Afghan communities in order to implement projects selected by the
communities themselves. The average size of the block grants was around US $30,000. The
study’s identification strategy employs the fact that the World Bank, as the main donor,
administered a randomized experiment in order to measure the impacts of NSP. In each of the 10
districts, 50 villages were selected to be included in the study, 25 of which were then selected as
treatment villages using a matched-pair randomization procedure.
Results suggest that NSP improved villagers’ perceptions of security and reduced the number of
security incidents recorded by ISAF in the long run (15 to 30 months after projects were
selected). However, these positive effects were only observed in eight of ten districts. In two
eastern districts, where initial levels of violence were higher, no effect was found. The study also
estimated the effect of NSP on a number of other outcomes and finds that NSP is associated with
perceived welfare gains, improved attitudes toward government officials, NGOs and ISAF
soldiers. Again, these positive effects were not found for the two eastern districts.
According to the authors, one explanation for this is that the government’s attempts to improve
material wellbeing are likely to have a strong effect on attitudes toward the government in
regions where the population is primarily concerned with economic conditions rather than
security. When public goods are provided in these regions, community members are less likely to
join the insurgency. In regions with high levels of violence, however, security is likely to be the
primary concern, so that marginal improvements in economic outcomes will be insufficient to
change people’s attitudes toward the government. The authors argue that their results suggest that
development programs are more effective in preventing the spread of violence, rather than in
reducing the level of violence in already insecure regions.
Crost et al. (2016) investigate the effects of KALAHI-CIDSS, a CDD program implemented by
the Philippine government and funded through World Bank. Between 2003 and 2008, more than
4,000 villages in 184 municipalities received aid through KALAHI-CIDSS. Typical of CDD
programs, the objectives of KALAHI-CIDSS were to mobilize communities by giving grants
which could be used for small, local infrastructure or capacity building projects. Participating
communities received approximately US $6,000 per grant. Their dependent variable is causalities
of civil war, measured at the municipal level per month. The data come from original reports of
the Armed Forces of the Philippines (AFP) between 2002 and 2006. These data are similar to the
US military’s “Significant Activities” (SIGACTS) database. The data allows distinguishing
between government- and insurgent-initiated incidents, as well as between causalities suffered by
government forces, insurgents and civilians. Since eligibility of the program was restricted to the
poor household in the forth quartile only, the study exploits this “cut-off” by using a regression
discontinuity design that compares municipalities just below the cut-off (treatment) with
municipalities just above the treatment. The results indicate that the program led to increased
causalities over the entire three-year period. The effect is, however, small in actual casualties and
translates to less than 3 killed within a municipality of an average population size of around
30.000. The study does not intend to test one specific causal mechanism. However, the authors
suggest that the most likely causal mechanism linking the CDD program to increased violence is
sabotage. Since rebels benefit from anti-government sentiments, they may have an incentive to
sabotage programs, which may repair negative attitudes of local communities towards the
government. Hence they may seek to derail the possible positive effects of CDD programs.
Böhnke & Zürcher (2013) investigate the impact of multi-sectoral, community level development
aid in rural areas of North East Afghanistan on perceived fear of violent actors. Data come from
two surveys among 2000 respondents in North East Afghanistan, conducted in 2007 and 2009.
Their measurement for aid is based on respondents’ perceptions of how much aid their
communities received in various sectors. This is a strictly perception-based measure, but the
authors demonstrate that it is correlated with an objective measurement of aid (defined as the
number of projects in a given community). The results suggest that more (perceived) community
aid is associated with higher perceived fear of violent actors. The authors suggest that
communities which received relatively large amounts of aid felt threatened because they fear that
cooperation with international actors has made them a target for Taliban reprisal attacks.
Berman et al. (2013) find some evidence for a violence-reducing effect of a CDD program, the
USAID-funded Community Stabilization Program (CSP).4 It should be noted, however, that their
study is predominately interested in CERP, and testing for the effect of CDD programs is done en
passant. Chou (2012) also included two CDD programs (NSP, and USAID Local Community
Development in Afghanistan) in her evaluation of CERP in Afghanistan, but found no effect.
In sum, we find little evidence for a violence-dampening effect of CDD programs in conflict
zones. As with CERP project, CDD projects appear to have a violence-reducing effect only when
the environment is reasonably secure. Under more adverse conditions, however, CDD can
increase violence. This effect may be driven by attempts of insurgents to sabotage the
cooperative relations between local communities and the government, or because rebels violently
loot aid.
CCT programs are another staple of traditional development programming. CCT programs intend
to reduce poverty by providing grants to poor households, based on some number of conditions.
Two studies estimate the effect of CCT programs on violence and reach opposite conclusions.
Crost et al. (2016) estimate the effect of a nationwide CCT program which financed transfers to
approximately one million households in all regions of the Philippines. The study estimates the
effects of the program on the annual number of conflict incidents per village and on the level of
4 This finding is in contrast to a report on the same program by the office of the inspector general
which states that “we do not have a reasonable basis for asserting that CSP activities in the
community infrastructure and essential services component were contributing to the overall
improvements in security in Iraq” (Office of Inspector General 2008: 4). Moreover, the audit also
pointed out that “CSP projects are highly vulnerable to fraud and exploitation which may have in
fact occurred, with potential adverse consequences to Coalition personnel (Office of Inspector
General 2008: 4).
insurgent influence in the village. Both measures are based on data from the Philippine military.
The study exploits the fact that the program was designed by the World Bank as an experiment
where 130 villages were randomly divided into a treatment group and a control group.
Observations were aggregated to a one-year pretreatment period and a one-year post treatment
period. Results suggest that the CCTs reduced the number of incidents in treatment villages
within one year after treatment. Also, treated villages experienced a decrease in insurgent
influence compared to control villages.
The authors propose two possible explanations for the observed violence suppressing effect: The
first is the opportunity cost model, which implies that the program reduced conflict by making it
more costly for insurgents to recruit combatants in treated villages. The second is the
information-centric model, which implies that the program increased popular support for the
government which led to more cooperation and information sharing between the government and
the villages.
Weintraub (2016) tells a different story. This study investigates the effects of the nation-wide
CCT program, Familias en Acción in Colombia, rolled out in 2002, using a sample of 57 treated
and 65 untreated municipalities. Data are drawn from the Human Rights Observatory Database
compiled by the Presidency of Colombia. This dataset has municipal-level data on violent events,
including the type of armed action perpetrated by various violent, non-state actors. The effect of
the program is estimated with a difference-in-differences strategy. The study exploits the fact that
an earlier evaluation study of the program constructed a data set where treated municipalities
were matched with untreated (Attanasio, Meghir and Vera-Hernandez 2004). The results suggest
that the program led to more killings and indiscriminate violent incidents by the FARC. The
effect appeared to be especially accentuated in the poorest municipalities and in municipalities
where coca was cultivated.
The observed effect is attributed to the insurgent’s attempts to sabotage the increased cooperation
between the beneficiaries of the program and the government. Insurgents, facing the threat of
losing control, will penalize “collaborators” with violence. Further specifying the mechanism, the
study argues that poor communities who depend most on aid are more likely to become
collaborators, and therefore more likely to become a target of insurgent violence.
How can we explain that these two studies reach different conclusions? It is plausible that the
violence increasing effect of the CTT in Colombia is attributable to the same logic that we
observed for CERP and CDD: Development aid triggered a violent backlash by rebels who feared
that they would loose poplar backing when communities received aid from the government, and
the military forces in Colombia were not capable of preventing that backlash. In the Philippines,
by contrast, security forces could shield the communities from rebel reprisals. The difference in
outcome would thus be explained by a different security environment and different military
capabilities of the rebels.
Employment programs, often in the form of cash-for-work, are another widely used development
tool. Three studies in our sample investigate the effect of one such program, the National Rural
Employment Guarantee Act (NREGA) in India. NREGA is an employment development
program introduced by the Indian government in 2006 that guarantees at least 100 days of wage-
employment to every rural household. It is a vast public employment scheme reaching up to 47.9
million rural households annually, generating 210 million person-days of employment for the
rural poor to date. While the key objective of NREGA is poverty reduction, the Indian
government hopes that it will also contribute to reducing violence in the regions most affected by
Maoist insurgencies (Hoelscher et al. 2012).
Dasgupta et al. (2016) investigate whether districts which adopted NREGA experienced lower
levels of violence compared to districts which did not adopt NREGA. Results indicate that
NREGA caused a roughly 50 percent reduction in violent incidents and deaths. The study also
shows that the effect is largest in districts which experienced little rainfall, suggesting the
NREGA serves as a substitute for foregone agricultural wages. The authors take this as support
for the opportunity cost model: The wage labor which the program provided to the rural poor
made recruitment for Maoist insurgents more costly. One innovative contribution of this study is
to highlight the role played by state capacity in shaping these effects. The performance of the
program is highly contingent upon local administrative capabilities. The results suggest that
NREGA's violence-reducing effects concentrated in states and districts which implemented the
program effectively and therefore provided greater levels of employment provision under the
Hoelscher et al. (2012) also find a violence-reducing effect of NREGA. Using a cross-sectional
model for the entire period from 20042010, the study finds that the percentage of households
per district participating in NREGA is associated with less battle deaths, less violent incidents,
and fewer districts which record violent incidents. The authors attribute the observed effect to the
fact that the employment program for the rural poor increased the opportunity costs for the
By contrast, Khanna and Zimmermann (2014), using a difference-in-difference-design, find that
the program led to an increase in Maoist-related violence in the short run. This increase in
violence appears to be driven by police-initiated attacks rather than by Maoist-initiated attacks.
The authors argue that such empirical patterns are consistent with the information-centric model
which predicts that civilians are more willing to share information with the police when they are
a recipient of a development program, thereby allowing government troops to crack down more
efficiently on the insurgents.
At first glance, the results of Khanna and Zimmermann (2014) appear to contradict Hoelscher
(2012) and Dasgupta et al. (2016)5. However, the spike of violence that Khanna and
Zimmermann (2014) observe is short-time only. Eight months after the implementation of the
program violence decreases. This is consistent with the logic of the information-sharing model: If
the program indeed led to better cooperation between communities and security forces, then we
would expect a short time increase of counter-insurgency measures, followed by an reduction in
overall violence. According to the authors, the overall findings of the study quite plausibly
indicatethe possibility that the police was more successful in catching Maoists right after
5 One possible explanation, put forward by Dasgupta et al. (2016), is that these studies use
different data sources. While Khanna and Zimmermann (2014) used data based on English news
clips which may over-report violence in urban regions, Dasgupta et al. (2016) constructed the
data based on local language news clips which may provide better and more balanced coverage
of rural areas.
NREGA implementation, and that this may lead to a fall in overall Maoist-related activities s in
the longer run (Khanna and Zimmermann 2014:33).
In sum, there appears to be evidence that employment programs may indeed dampen violence,
either by increasing opportunity costs for rebels, or by increasing cooperation between
communities and security forces, which makes COIN more effective. Some caution is in order,
however, since the evidence stems from only one program, which raises doubts about external
validity. With this in mind, it is interesting to speculate about why employment programs appear
to be more effective in dampening violence than CERP, CDD, CTT or humanitarian aid. I offer
two possible answers. Firstly, it is possible that the Maoist insurgency in India is primarily
driven by economic deprivation, which would explain why creating economic opportunities can
reduce insurgent activities. By contrast, insurgencies in Iraq and Afghanistan are to a very large
extent driven by ideology and religion, which make economic opportunities much less effective.
Secondly, it is possible that NREGA was implemented in regions where the government was
relatively strong to begin with. NREAG requires that government officials collect community
level data and are present in order to administer the work program. This is only possible in
regions which are largely under government control. The level of control is therefor much higher
than in regions where CERP programs or humanitarian emergency programs are implemented. It
is possible that the benign effect of NREGA is conditioned on pre-existing government control.
The evidence on humanitarian aid is unequivocal: All five studies in our sample find that
humanitarian aid increases violence.
Nunn and Qian (2014) study the effect of US food aid on conflict in recipient countries. Their
sample consists of a panel of 125 non-OECD countries between 1971 and 2006. Study variables
are onset and duration of conflict. In order to counter endogeneity problems, the authors use an
instrument for food aid based on exogenous time variation in US wheat production, which is
primarily driven by changes in US weather conditions. Surplus wheat is bought by the
government at fixed prices and then shipped to developing countries as food aid. Thus, US wheat
production is positively correlated with US food aid shipments in the following year. The authors
construct the interaction of last year's US wheat production and the frequency that a country
receives any US food aid and use this as an instrument for the amount of food aid received by a
country in a given year. The study finds US food aid increased the duration of civil conflicts, but
had no effect on interstate conflicts or the onset of civil conflicts. The effect is most pronounced
in countries with a recent history of civil conflict. The study is not designed to uncover the causal
mechanisms, but the authors refer to the large do-no-harm literature which suggests that stolen
aid is frequently used to finance the war.
Narang (2014) investigates the effect of humanitarian bilateral and multilateral aid disbursement
on the duration of peace, using a panel dataset of civil conflicts between 1989 and 1999. A
duration models is employed to estimate the effect of aid on the risk of peace failing in a
particular year. He finds that post-conflict states treated with higher levels of humanitarian
assistance exhibit shorter spells of peace; however, this effect only occurs after conflicts that
ended with a decisive victory. For conflicts which ended in negotiated settlement or stalemates
no effect is found. The author argues that humanitarian aid is usually disproportionally given to
the losers of the war, and that the aid can help the losing side to reconstitute its war effort. In
other words, aid can support or even create a revisionist party with the incentive to change the
postwar settlement on the battlefield. It should be mentioned that this is a theoretical argument.
The study does not offer supporting evidence for the alleged causal mechanisms. Such a test
would have to show that recipients of humanitarian aid diverted aid for their war efforts, by
stealing or taxing the aid.
Narnag (2015) investigates whether humanitarian aid pro-longs civil wars, using a cross-national
panel data on humanitarian aid disbursed between 1969 and 2008. Effects are estimated with Cox
proportional hazards models. The study finds that increased levels of humanitarian assistance
lengthen civil wars, particularly those involving rebels on the outskirts of a state. The author
notes that these findings are compatible with a range of causal mechanisms: Misappropriated aid
could finance the insurgency; humanitarian aid could create protected spaces (such as refugee
camps) that shield combatants from costly attacks; fungible aid could free up resources for
violence; or local power-brokers could prolong the war in order to continue “taxing” the
incoming aid. On a more general level, the author suggests that aid may exacerbate information
failures: By making war less costly, humanitarian assistance can inadvertently prolong fighting
by slowing down the accrual of information that allows opponents to converge on more
congruent estimates of relative strength which would lead to negotiated settlements” (Narang
2015: 184).
Wood and Sullivan (2015) investigate whether humanitarian aid can encourage rebel violence
against civilians. The authors suggest two possible causal mechanisms: First, aid may encourage
predation, which may result in abuses against the local population. Second, aid may be perceived
by rebels as a challenge to their authority, because aid may increase cooperation between the
local population and the government. Rebels may use violence to sabotage that cooperation.
The depended variable is the number of attacks on civilian targets by insurgents. Spatially
disaggregated conflict event data come from the Uppsala Conflict Data Program's (UCDP)
Georeferenced Event data set, which is based on media reports. The independent variable is
project-level bi- and multilateral humanitarian aid commitments, lagged by one year. Data come
from the UCDP/AidData georeferenced data set. The unit of analysis is grid / year, whereas the
grid is based on the PRIO-Grid system. A cell is roughly 55 # 55 km at the equator. The data
represents 22 sub-Saharan African states between 1989 and 2008. Effects are estimated by cross-
sectional regression models. Supporting evidence comes from a matched sample allowing for a
difference-in-difference model. Results support the argument that humanitarian aid is associated
with increased rebel violence. The study does not test for whether the effect is caused by
predation or sabotage.
Finally, Wood and Molfino (2016) explore whether humanitarian aid increased violence between
the government and rebels. The alleged causal mechanism is that injecting humanitarian aid into
a locality increases the incentives for rebels to challenge the government for control over
territory in which aid accumulates, thus leading to an increased risk of violence. The unit of
analysis is first order administrative unit (i.e., districts, communes) / year. The independent
variable is humanitarian aid commitments per unit, and the dependent variable are battles
between rebels and the government. Data sources are identical with Wood and Sullivan (2015).
Effects are estimated with Poisson regression, and supplemented with propensity score matching,
allowing for difference-in-difference estimates. Results provide support for the assumption that
humanitarian aid increases the subsequent frequency of conflict between rebel and government
The previous sections took stock of the evidence base for the impact of aid on violence. As we
have seen, the results are not encouraging. More often than not, aid is statistically associated with
an increase in violence. But statistical correlations can only take one so far. This section now
offers on overview of the causal mechanisms at work. Specifying these mechanisms make
theories more complete and persuasive and can help to design adequate policy responses, as
different mechanisms suggest different policy interventions. In what follows, I identify all causal
mechanisms which the reviewed studies propose. I then briefly discuss the observable
implications for the causal mechanism and assess to what extent the discussed studies tested for
the presence or absence of these observable implications. I find that most studies fail to provide
strong evidence for the presence of their alleged mechanism (or rather, the presence of the
observable implications) and suggest ways of how future research could conduct much stronger
tests for the presence of specific causal mechanisms.
Hearts and minds is perhaps the most famous concept used to describe possible links between aid
and violence. Campaigns to win over the hearts and minds of the local population have been used
by French and British colonial administrators, by the US army in Vietnam , and most recently in
Iraq and Afghanistan by coalition forces. The assumption is that aid can help win civilians’
"hearts and minds" by providing public goods. If the goods and services are valuable to them,
communities will develop more positive attitudes towards the government and will be less likely
to support the insurgency, which eventually will dampen violence. There is some initial evidence
for this mechanism in our sample. Beath et al. (2012) show that a large CDD program was
associated with more positive attitudes towards the government and less reported security
incidents. Similarly, Böhnke & Zürcher (2013) show that aid led to more acceptance and more
legitimacy for the subnational government and for development actors. These findings tie in with
a large literature that shows that the ability to provide basic public services to the population can
increase legitimacy for the egovernment (See McLoughin 2015 for an overview).
However, it should be noted that even if aid leads to more positive attitudes, more legitimacy,
and more acceptance, this does not necessarily translate into less violence. Behavioral changes
that could reduce violence include that communities no longer support insurgents by providing
fighters, shelter, food or information; that communities increase internal policing, making it more
difficult for insurgents to recruit fighters, or that communities increase collaboration with the
government by providing information. Such behavioral changes may follow from attitudinal
changes, but don’t have to. In other words, the “hearts and minds” mechanism explains the initial
attitudinal changes, but not the subsequent behavioral changes which are necessary if violence is
to be reduced. As such, it is an underspecified causal model which can explain the first element
in a causal chain, but not the necessary subsequent ones.
The information-centric model is a more complete causal mechanisms. Thus far, Berman et al.
(2011, 2013) have provided its most complete specification. The model assumes that local
communities possess critical information on the activities of insurgents. The prospect of rewards
in the form of development aid acts as an incentive for local communities to share this
information with the government and its international allies. As a result, the government’s
counterinsurgency efforts become more effective and security eventually increases. Berman et al.
(2011, 2013) and Crost et al. (2016) attribute violence reduction to this mechanism. Weintraub
(2016) and Khanna and Zimmermann (2014) also assume that aid encouraged information
sharing, but speculate that this information led to more violence, because security forces
increased their counterinsurgency activities (Khanna and Zimmermann 2014) or because
insurgents used violence to punish those who shared information (Weintraub 2016).
None of these studies provide support for the presence of information sharing beyond the claim
that it is compatible with the data. One straightforward way of testing the presence of the
mechanism would be to conduct interviews with military commanders, asking them whether the
handing out of development projects bought them reliable information on which they could act.
Adams (2015) does this and finds no support for information sharing. Also, Sexton (2016) finds
that CERP spending did not result in increased COIN activities, as would have been the case if
CERP spending had led to “tips” on which the military could act.
Besides the lack of empirical support for the presence of the mechanism, there is also theoretical
concern. The model requires that aid is given conditionally: “The violence-reducing property of
service provision requires conditional provision: the community benefits from services only if the
government controls the territory. If the community benefited from services regardless of who
won, provision would not motivate information sharing” (Berman et al., 2013, p. 523). This
assumption is problematic because it is unlikely that such an ex-ante conditionality was widely
practiced, since this would mean that commanders would withhold all funds until control is
established. This defies both the urge of bureaucracies to spend allocated funds quickly and the
intended use of CERP, which is using aid precisely as a means to establish control.6 Many
qualitative accounts of CERP spending suggest that funds were given without much planning or
oversight, not to mention conditionality. As a matter of fact, military reconstruction aid in the
context of COIN has become for many development experts a prime example of wasteful
spending with poor oversight (Wilder and Fishstein, 2012; Williamson, 2011; Special Inspector,
2011; Stein, 2011; Suhrke, 2006; Committee on Foreign Relations, 2011; Wilton Park, 2010).
Finally, the “information-sharing model” might also oversimplify the dynamics between foreign
counterinsurgents and local communities. According to many testimonials, this dynamic is
characterized by an information asymmetry, where local communities use their informational
advantage in a much more strategic way than the model presumes . Local actors often release
6 Berman et al. (2011, footnote 11) refer to a survey among officers and officials with CERP
implementation authority in Afghanistan, conducted in October and November of 2010 in which
61 percent of the 210 respondents indicated that they would “halt implementation of a CERP
project if the local population increased its support for anti-government elements.” It is unclear,
however, how many respondents actually did halt a project. Another study, based on 44 semi-
structured interviews with Canadian civilian and military officials involved in aid projects (not
necessarily CERP projects) in fragile states, found not one instance of an aid project that was
actually withdrawn (Bourgoin et al., 2013).
biased or incomplete information in order to influence international actors in a way that favors
local interests. It is hard to see how under such circumstance development aid would consistently
buy reliable information. In the future, further qualitative work may reveal if and when a
violence-reducing effect is indeed caused by increased information sharing or whether additional
or alternative causal paths are at work.
The violence-suppressing effect of aid is often explained by an opportunity cost model. Economic
opportunities, it is argued, can provide employment for young men, which makes the recruitment
of fighters more expensive (Grossmann, 1991, 1999; Collier, 2000; Collier and Hoeffler, 2004).
The opportunity cost model is most closely associated with employment programs, often in the
form of cash-for-work. Crost et al. (2016), Dasgupta et al. (2016) and Hoelscher et al. (2012)
both attribute the observed violence-reducing effect of a large employment scheme to increased
opportunity costs. Additionally, Dasgupta et al. (2016) demonstrate that the violence-reducing
effect of the program tended to be stronger in regions with unusually little monsoon rain than in
regions with normal monsoon. This suggests that the employment scheme offered compensatory
income for farmers suffering from bad harvests caused by a lack of rain. Without the employment
scheme, farmers might have turned to insurgent activities which offers some additional income.
The opportunity cost model does not necessarily work only through employment schemes,
however. Every labor-intensive and rent-generating aid program can, theoretically, increase
opportunity costs for insurgency. For example, Crost et al. (2014) report a violence-reducing
effect of a CCT program in the Philippines. These cash transfers, they argue, boosted the local
economy and created higher incomes from peaceful activities, which in turn made joining the
rebellion less attractive.
The opportunity cost mechanism is a theoretically attractive preposition because of its simplicity.
Yet, it is evident that the model depends on one crucial precondition: The violence-dampening
effect will only take place if insurgents are first and foremost motivated by private economic
gains. We should not expect to see an effect on insurgencies which are predominately motivated
by ideology.
The logic of the sabotage mechanism posits that aid will lead to better relations and more
cooperation between the population and the government. More popular support for the
government could then lead to information-sharing, and it could help the government to establish
or deepen control over contested regions. Insurgents, keen to sabotage the cooperative relations
between the local population and the government, then use violence against “collaborators” to
deter further collaboration. Eight out of nine studies that report a positive correlation between aid
and violence attribute this to the sabotage model For example, Crost et al. (2014) show that a
CDD program caused an increase in casualties a result of insurgent-initiated attacks, which is
consistent with the assumption that a successful community-driven development can increase
support for the government which then leads to sabotage by the insurgency. Weintraub (2016)
proposes a two-pronged model that combines the information-centric and sabotage models:
Development aid buys information and, in reaction, insurgents target the population to sabotage
the information sharing which threatens the insurgentscontrol of territory. Sexton (2016) shows
that CERP funds in districts which are contested between rebels and government increase
violence, and argues that this is caused by insurgents attempts to sabotage cooperation between
governments and local communities. Wood Wood and Sullivan (2015) demonstrate that injecting
humanitarian aid in conflict zones leads to increased insurgent violence against civilians, which is
compatible with the sabotage model. While this is compatible with the sabotage mechanism, the
authors also note that it is equally compatible with the predation mechanisms, which assumes that
aid flows increase violence between warring parties as insurgents try to control territories where
aid is distributed. Indeed, sabotage and predation are not mutually exclusive. Both are important
objectives for insurgents. There is ample evidence from Afghanistan that the Taliban seek to
control and tax all aid in regions which they control, while at the same time deterring the local
population from collaborating with the Afghan government and those aid organization which are
deemed hostile. If sabotage would take place without predation, then we would see an uptick in
violence against development projects and against civilians, but not necessarily an uptick in
violence between the warring parties. Fine-grained event data might allow for such tests.
The theoretical foundations of the predation model dates back to the works of Hirshleifer (1991,
Grossman (1991) and Skaperda (1992), who all argued that aid exacerbates violence because it is
an additional resource which fuels conflict. Predation is also at the core the “do-no-harm”
literature which sees aid a lootable resource which can incite or prolong violent conflicts
(Anderson, 1999; Uvin, 1998). The logic of the predation model found much support in
qualitative single case studies. Scholars showed, for example, that stolen aid fueled rebel violence
(Bradbury and Kleinmann, 2010; Goodhand, 2002; De Waal, 1997;) or that fungible aid relived
local actors from the burden of delivering basic goods to their constituencies, freeing up
resources that were be invested in violence (Polman, 2010). There is considerable qualitative
evidence that predation of aid flows may have led to increased violence in Afghanistan (Wilder
and Fishstein, 2012; Wilton Park, 2010; Karell, 2015).
Wood and Molfino (2016), Wood and Sullivan (2015), Narang (2014, 2015) and Nunn and Qian
(2014) all attribute spikes in violence after aid injections to the predation mechanisms.
Furthermore the data of Sexton (2016) and Böhnke & Zürcher (2013) are both compatible with
the predation mechanism. As mentioned above, predation may be difficult to distinguish from
sabotage, as both can go hand in hand. If only predation, but not sabotage would take place, we
would expect to see an uptick of violence between warring parties, but not necessarily against
local civilian population or development actors.
The reviewed studies offer a useful spectrum of five causal mechanisms. But there may be
alternative mechanisms which are not covered by the sample. One such mechanism which is
conspicuously absent form the sample is that aid might dampen violence by addressing
grievances. The literature on civil wars has long ago identified group-level grievances, especially
the real or perceived lack of current and future political and economic opportunities, as one
source of violence (for example, Gurr, 2000). When one particular group, typically an ethnic
minority, holds economic and social grievances, well-targeted aid might enable redistributive
policies that can lessen inequalities, create solidarity links between population groups and
remedy grievances (Azam, 2001; Azam and Mesnard, 2003; Justino, 2007). Arcand at al. (2010)
provide some evidence for this mechanism by pointing out that aid provided for the grievance-
driven Moro Islamic Liberation Front in the Philippines created concrete improvements in access
to government services, a greater sense of inclusion in local decision-making, and a greater sense
of empowerment. None of the studies in the sample consider that aid might reduce grievances,
and the data on which the studies are based do not suggest that the observed violence dampening
effect ay be caused by reduced grievances. Nevertheless, future research might provide evidence
for grievance reducing effects of aid linked to violence reduction.
Another unmentioned causal mechanism which might explain how aid can lead to reduction in
violence, at least in the short run, is buying-off strongmen. There is abundant evidence that aid
and particularly local contracting has very often enriched local strong men in Iraq and
Afghanistan (for example, SIGAR, 2009, 2011). It is worthwhile speculating about the effect of
this on violence. Local strongmen control the means of violence. They are also well placed to tap
into aid flows, for example by rigging contracts, extorting rents from contractors, or selling
protection to contractors, all of which is paid for by the aid program. One of the best-documented
examples is the transportation sector in Afghanistan. The US government spends hundreds of
millions a year to private contractors on trucking services in Afghanistan. These contractors pay
large amounts to local warlords across Afghanistan in exchange for “protection” in the form of
supply convoys to support U.S. troops (Warlord, Inc., 2010). This protection racket has become a
major source of funding for violent entrepreneurs, and they have a vested interest in keeping the
funds coming. Local warlords have the capacity to police their constituency and they can offer
their militias for protection. They will offer these services as long as they can extort rents. Too
high levels of violence would harm their business, hence violent entrepreneurs have an interested
in keeping violence at bay. But prolonged loyalty requires prolonged bribing. If bribes are
deemed to be too small, violence will return.
None of the studies in our sample explicitly investigate whether the observed effects of aid could
be explained by a buying-off-the strongmen model, but, for example, the data of Berman et al.
(2011, 2013) and Sexton (2016) are compatible with such a causal model. If the violence-
dampening effect is indeed caused by bribing strongmen, then we would see a reoccurrence of
violence once the rents from aid projects stop flowing. None of the reviewed studies currently
offers a test for how long violence is sustained, partly because temporal data on violence may not
be fine-grained enough. Future research should, however, investigate whether is likely that aid
has been misappropriated by local strong men, and whether violence returns once rents stop
flowing. Both would suggest that the aid-dampening effect is caused by buying-off strongmen.
Table 2 summarizes the discussed mechanisms.
Table 2: Causal Mechanisms
Violence-reducing Mechanisms
Assumed in:
Hearts-and-Minds leads to less violence
Aid provides public goods to local communities. Local communities value
these goods, which makes it less likely that the local population supports or
joins the insurgency.
Beath et al. (2012)
Böhnke & Zürcher (2013)
Information sharing leads to less violence
Local communities often have private information on the insurgency. The
promise of aid can incentivise local communities to share this intelligence
with the government which will make counterinsurgency more effective and
eventually reduce violence.
Crost et al. (2016)
Berman et al. (2013)
Berman et al. (2011)
Child (2014)
Chou (2012)
Opportunity cost leads to less violence
Aid provides public goods (esp. more employment opportunities), which
increases the opportunity costs for the insurgency. Violence is reduced as a
Crost et al. (2016)
Dasgupta et al. (2016)
Hoelscher, et al. (2012)
Buying off strongmen leads to less violence
Local strongmen siphon off rents from aid programs and suppress violence as
long as rents are deemed adequate.
No example in the sample;
however, the observations of
Berman et al. (2011, 2013),
Dasgupta et al. (2016) are
compatible with this
Addressed grievances leads to less violence
Aid successfully addresses economic and political grievances, which were
drivers of violence. By addressing these grievances, violence will be reduced.
Not in the sample
Violence-increasing Mechanisms
Sabotage leads to more violence
Aid leads to more cooperation and more intelligence sharing between the
local population and government; insurgents respond by deterring the local
population from cooperating by applying selective or indiscriminate violence,
and by sabotaging the aid programs.
Weintraub (2016)
Crost et al. (2014)
Sexton (2016)
Khanna&Zimmermann (2014)
Crost et al. (2014)
Wood and Molfino (2016)
Wood and Sullivan (2015)
Böhnke & Zürcher (2013)
Predation leads to more violence
Aid is a lootable resource. More resources can enable insurgents to continue
fighting, or it can fuel competition between armed groups.
Wood and Molfino (2016)
Wood and Sullivan (2015)
Narang (2014, 2015)
Nunn and Qian (2014)
Also, compatible with Sexton
(2016) and Böhnke & Zürcher
The available evidence does not suggest that aid in conflict zones is an effective instrument to
dampen violence. Using a “one case-one-vote-count, we find that of the 24 cases which the
sample covers, only seven had a violence-reducing effect. Three had a heterogeneous treatment
effect, six had no effect at all, and nine had a violence-increasing effect.
This picture does not change when we use a “one-intervention-one votecount. Recall that some
of the 24 cases cover the same intervention. We can thus collapse these 24 cases into 13
observations: CERP in Afghanistan; CERP in Iraq; six different CDD programs, two different
CTT programs, one employment scheme; food aid; humanitarian aid. Out of these 13
interventions, only two unconditionally dampened violence; two had no effect, two decreased
violence, conditional on the security of the location, two had a heterogeneous effect, and five
increased violence. Again, the evidence of a violence-dampening effect is not strong. No effects,
heterogeneous effects and violence-increasing effects are far more frequent.
Importantly, looking only at those interventions which dampened violence, we see that this effect
appears to be conditional on other factors. Beath at al. (2012) provide evidence that a CDD
program in Afghanistan increased security, but only in regions which were relatively secure to
begin with. Berman et al. (2013) show that CERP in Iraq helped to dampen violence, but only in
interaction with increased troop strength. Without additional forces, no effect was found.
Likewise, Sexton (2016) shows that CERP in Afghanistan could dampen violence but only when
implemented in regions already controlled by the military. In contested regions, the same project
would increase violence. Adams (2015) finds that only small CERP projects in Afghanistan
dampened violence, while expansive ones increased violence. He does not have controls for how
secure a regions was, but it is possible that the size of the projects actually picks up the security
level of the region where the project was implemented. We know that a disproportionate share of
CERP funding in Afghanistan went to highly insecure regions, and that the most funded sector in
those regions was expensive road construction (Johnson et al. 2012). Since the more expensive
projects where disproportionately implemented in highly insecure regions, the association
between project size and increased violence may actually pick up the effect of a highly insecure
In sum, the evidence strongly suggests that aid in conflict zones will only have violence-
dampening effect when the aid is injected in regions which are already relatively stable. Aid
injected in insecure regions is likely to increase violence. This observation is reinforced when
we look at the impacts of various types of aid. Humanitarian aid, which the is most “front-line”
type of aid and is allocated disproportionally to regions which are most hit by violence and
insecurity, is consistently associated with in increase in violence. Likewise, CERP projects are
also “front-line” projects, and we have seen that CERP increased violence in insecure regions.
The only aid type in our sample that is mainly associated with a violence-dampening effect is a
large public employment scheme in India. Since we have evidence on only one such program,
the external validity of the evidence is limited. But it is nevertheless interesting to speculate about
the cause for the violence decreasing effect of this program. It is possible that the success of this
employment scheme is conditional on a pre-existing relatively stable security situation. The
implementation of such a large employment scheme requires the presence of capable state
administrationa condition which is rarely met in regions under control by the insurgents. Also,
recall that the violence dampening effect of the program is increased by strong state capacity. All
of this suggests that the positive effect of this employment scheme is conditional on a relatively
stable security situation and the presence of a responsive state administration conditions which
are usually not present in “front-line” regions.
In sum, there is strong evidence that a violence-dampening effect of aid is conditional on a
relatively benign security environment. Why would this be the case? The studies offer two
different theoretical answers. Berman et al. (2013). argue that aid is more valuable to the
community when the aid project is protected from “extortion, capture or destruction (p.513).
Hence, when troops are present to protect the aid project, its value to the community is increased
to a level when the community begins to trade information for aid. This in turn makes COIN
more effective, and violence is eventually reduced. A second answer is offered by Sexton (2016),
Wood and Sullivan (2015) and Wood and Molfino (2016). These studies argue that rebel groups
resist the implementation of aid projects because aid could potentially undermine their position.
This is because aid has the potential to win hearts and minds of the local population for the
government and increase cooperation between the government and the local population and
thereby helps the government to re-establish or deepen its control over a territory. Rebels will
therefore punish the local population for cooperating with the government and with aid
organizations, as long as they have the military capacity to do so. Injecting aid in regions which
are controlled by rebels will therefore likely lead to more, not less violence. Sexton (2016) finds
support for this assumption in a quasi-experimental design.7
The finding that the impact of aid is in important ways modified by the local security
environment challenges the widespread belief of many aid practitioners that the impact depends
first and foremost on the aid type, and that some types of aid are better suited for conflict settings
than others. In particular, participatory community driven development has often been heralded
as especially well-suited for conflict and fragile contexts (World Bank 2013, World Development
Report 2011). Such an expectation may be misplaced. No aid type in the sample is consistently
associated with a violence-dampening effect. Rather, all reviewed aid types can also lead to
more violence, if they are injected in insecure environments. This is because the causal path that
7 More evidence for the modifying effect of the local security environment comes from recent
survey data from Afghanistan. When asked whether respondents could “think of an instance, in
your community or in a neighboring community, when groups such as the Taliban used violence
in order to obstruct a development project?” 6.4 % said they could think of such an instance. In
districts that were mostly under government control, the number dropped to 4%. In contested
districts, the number reached 10% and in a district that was fully under the control of armed
groups, the number reached 15%. The survey was conducted as a booster sample of “The Asia
Foundation / Survey of the Afghan People 2017,on behalf of the German Federal Ministry of
Economic Cooperation and Development. See
links aid to violence runs through either predation or sabotage, and no type of aid is immune to
predation or sabotage.
With regard to predation, it is evident that some types of aid such as food, medication, fuel,
building materials and the like are easy to steal. Other types of aid may be more difficult to steal,
but they can always be “taxed”. There is much evidence that the Taliban tax aid organizations
and their local contractors in return for safety guarantees and permission to work. Furthermore,
the Taliban routinely extract “taxes” from villages that they control. In sum, a share of all aid
resources injected in a war economy will always flow to those who control the war economy,
which explains why no aid type is immune to predation.
Just like there is no type of aid which is immune to predation, there is also no type of aid which is
immune to sabotage. Every type of aid that is regarded by insurgents as a potentially effective
tool for improving relations between the local population and the government may be sabotaged.
Sabotage is a strategic response by insurgents. It is triggered by the promise of aid to increase
cooperation between the local population and the government. As we have seen, the studies in
our sample report that insurgents sabotaged all reviewed aid types, such as food aid (Naunn and
Qian 2014), humanitarian aid (Wood and Sullivan 2015), CCTs (Weintraub 2016), CDDs (Crost
et al. 2014) and CERP (Sexton 2016). This suggests that no aid type is immune to sabotage.
A systematic review is also a useful tool for identifying knowledge gaps. Four such gaps stand
out and should be addressed by future research. Firstly, the evidence base is surprisingly small.
19 studies may not be an unusual small sample for a systematic review, but given the billions of
aid dollars spent on stabilization and the recent push by development organizations for more
rigorous evaluations, one might have expected more studies. Especially surprising is that
development organizations have not yet made serious efforts at evaluating the impact of their
programs on stability. Only one study in our sample was initiated and conducted by a
development organization, whereas four studies were financed by the US military. More studies
would not only broaden the evidence base, but also address justified concerns about limited
external validity of many studies.
Secondly, the list of reviewed aid types is by no means exhaustive. The sample is not surprisingly
biased toward “front-line” aid, such as humanitarian aid, CDD and CERP. Future research should
investigate the effects of other aid sectors which are often prioritized in the early yeas of post-
conflict regions, such as interventions in health, energy and electrification, water and sanitation,
and transport networks. However, this review strongly suggests that the effects of aid are
moderated by the environment in which aid is implemented, hence we have so far little reason to
believe that other aid sectors would have a consistently “good” effect. But only more evidence
can tell.
Thirdly, we found that our understanding of causality is still limited. While the reviewed studies
propose a useful range of possible causal mechanisms, they do not usually provide strong tests
that actually prove the presence of the alleged mechanisms. At best, they show that the used data
is compatible with the theoreticized mechanism. Evidently, it is possible and desirable to deduce
for all causal mechanism a set of observable implications, which then would allow to conduct
better tests. In my discussion of the causal mechanisms I have suggested some observable
implications for each causal mechanism. For example, support for the information-centric model
could be found by asking COIN actors to what extent the delivery of aid has bought them
information on which they could successfully act. An observable implication of the opportunity
cost mechanisms should be a substantial increase in employment opportunities in a given
geographical space, which can be identified with surveys or even qualitative observations.
Support for the “buying-off strongmen” model could come from the observation that aid has been
taxed by local strong men, and whether violence returns once aid rents stop flowing. An
observable impaction from the sabotage model would be that aid triggers insurgent-initiated
violence against aid workers and their “collaborators”, but not necessarily violence between
government forces and insurgents These and other tests will be feasible once we have more fine-
grained data on violence, especially on whether violence was initiated by rebels or government,
for how long violence was sustained, and who the primary target was. There are many data
collection initiatives underway which will substantially improve the available data and will
enable such tests in the near future. These new more granular data should be accompanied by a
shift to multi-method approaches, where econometric approaches are combined with micro-level
qualitative approaches which are well suited to capture the local conflict dynamics.
Fourthly, the most important question that future research needs to tackle in a systematic way is
not so much whether aid can buy stability or not, but rather which environment is more or less
conducive to a violent-reducing effect of aid. As long as we lack a clear understanding of how
various environments shape aid effects, much aid will continue to be misspent or may even cause
harm. Only two studies in our sample make an attempt to measure the security environment at the
sub-national level. Berman et al. (2013) use a measure of troop strength based on news reports,
and Sexton (2016) uses the presence of a forward battalion as a proxy for control. While these
studies should be commended for trying to measure the security environment at the subnational
level, the reliability and validity of these measures may raise some concern. The sheer presence
of foreign troops may not necessarily proxy that control has been wrestled away from insurgents.
To what extent government or rebels are in control of a territory is presumably best assessed
based on local intelligence. The most pressing task for future research on the impact of aid in
conflict is to find a valid and reliable measurement for the local security environment and include
it as a control or modifier in the identification strategies. This will help to identify the conditions
under which aid may actually help.
Finally, the results of this review also have policy implications. As we have seen, aid in conflict
zones is more likely to increase violence than to dampen violence. A violence-dampening effect
is conditional on a relatively secure local environment, where violence-increasing strategic
responses to aid injections such as predation or sabotage are less likely. Allocating the bulk of the
aid to the most insecure regions, as it was practiced in Iraq and Afghanistan, is most likely a
strategic mistake which exacerbates insecurity. Aid in conflict zones should be allocated to
relatively secure, government-controlled districts. While aid is not an instrument for bringing
back insecure districts from violence, it may help to protect relatively stable district from slipping
into war. Doing so would require a policy change. It would also require that development actors
develop clear criteria for when a region is secure enough in order to allocate aid. Only a thorough
analysis of the local, district level political economy provides such information. Aid
organizations should invest in this type of local knowledge, and use it for their allocation
Adams, Greg. 2015. Honing the proper edge: CERP and the two-sided potential of military-led
development in Afghanistan. The Economics of Peace and Security Journal 10(2): 53-60.
Aikins, M. 2010. “Last stand in Kandahar. Can the military’s massive counterinsugency gamble
salvage the Afghan war?” The Walrus (December).
Anderson, M B. 1999. Do No Harm. Boulder, CO: Lynne Reinner.
Arcand Jean-Louis, Bah Adama and Julien Labonne. 2010. “Conflict, Ideology and Foreign Aid.”
CERDI, Etudes et Documents, E 2010.21
Attanasio, Orazio, Costas Meghir and Marcos Vera-Hernandez. 2004. Baseline Report on the
Evaluation of Familias en Acción. Technical report University College London.
Azam, J. 2001. “The Redistributive State and Conflicts in Africa,” Journal of Peace Research, 38,
Azam, J. P., and A. Delacroix. 2006. “Aid and the Delegated Fight against Terrorism.” Review of
Development Economics 10 (2): 330344
Azam, J. P., and V. Thelen. 2008. “The Roles of Foreign Aid and Education in the War on Terror.”
Public Choice 135 (3): 375397.
Azam, J., and A. Mesnard. 2003. “Civil War and the Social Contract,” Public Choice, 115, 455
Beath, Andrew, Fotini Christia, and Ruben Enikolopov. 2012. “Winning Hearts and Minds
Through Development: Evidence from a Field Experiment in Afghanistan“. Policy Research
Working Paper 6129. (The World Bank)
Berman, Eli, Jacob N. Shapiro, and Joseph H. Felter. 2011. “Can Hearts and Minds Be Bought?
The Economics of Counterinsurgency in Iraq.” Journal of Political Economy 119, no. 4: 766
Berman, Eli, Joseph Felter, Jacob N. Shapiro, and Erin Troland. 2013. “Modest, Secure and
Informed: Successful Development in Conflict Zones”. American Economic Review: Papers &
Proceedings, 103(3): 512517
Berman, Eli, Michael Callen, Joseph H. Felter, and Jacob N. Shapiro. 2011. “Do Working Men
Rebel? Insurgency and Unemployment in Afghanistan, Iraq, and the Philippines.” Journal of
Conflict Resolution 55 (4): 496528.
Bodnar, Seth, and Jeremy Gwinn. 2010. “‘Monetary Ammunition’ in a Counterinsurgency.”
Parameters 40 (3): 91102.
Böhnke, Jan Rasmus, and Christoph Zürcher. 2013. “Aid, Minds and Hearts: The Impact of Aid in
Conflict Zones.” Conflict Management and Peace Science 30, 5: 41132.
Böhnke, Jan, Jan Koehler, and Christoph Zürcher. 2015. “Assessing the Impact of Development
Cooperation in North East Afghanistan 20052013. Final Report”. German Federal Ministry of
Economic Cooperation and Development. Accessible at
Bourgoin, Gaëlle, Tim Brown, and Christoph Zürcher. 2013. “How Canadian Practitioners See
Their Engagement in Fragile States. A Research Note.” Canadian Journal of Development
Studies / Revue Canadienne D’études Du Développement 34 (4): 56984.
Bradbury M and Kleinmann M .2010. Winning Hearts and Minds? Examining the Relationship
Between Aid and Security in Kenya. Boston, MA: Feinstein International Center.
Brown, Stephen, and Jörn Grävingholt, eds. 2014. The Securitization of Foreign Aid. Basingstoke
Palgrave Palgrave Macmillan.
Brück, T. et al. 2015. “Theories of Change Linking Employment and Stability in Fragile and
Conflict-Affected Situations,”. Scoping Study for the Working Group on Employment for
Stability of the Dutch Government, Knowledge Platform Security and Rule of Law, The Hague.
Child, Travers B. 2014. “Hearts and Minds Cannot Be Bought: Ineffective Reconstruction in
Afghanistan.” Economics of Peace and Security Journal 9, 2: 4349.
Child, Travers B., and David Scoones. 2015. “Community Preferences, Insurgency, and the
Success of Reconstruction Spending.Defence and Peace Economics 0 (0): 119.
Chou, Tiffany. 2012. “Does Development Assistance Reduce Violence? Evidence from
Afghanistan.” Economics of Peace and Security Journal 7, no. 2: 513.
Collier, Paul and Anke Hoeffler.2004. “Aid, policy and growth in post-conflict societies”.
European Economic Review (48), 5, pp. 1125.
Collier, Paul, and Anke Hoeffler. 2007. “Unintended Consequences: Does Aid Promote Arms
Races?” Oxford Bulletin of Economics and Statistics 69 (1): 127.
Collier, Paul. 2000. “Rebellion as a Quasi-Criminal Activity.” The Journal of Conflict Resolution
44 (6): 83953.
Committee on Foreign Relations, United States Senate. 2011. “Evaluating U.S. Foreign Assistance
to Afghanistan.” Majority Staff Report. Washington, DC. June 8. http://www.
Crost, Benjamin, Joseph Felter, and Patrick Johnston. 2014. “Aid under Fire: Development
Projects and Civil Conflict.The American Economic Review 104, no. 6: 183356.
Crost, Benjamin, Joseph H. Felter, Patrick B. Johnston. 2016. “Conditional Cash Transfers, Civil
Conflict and Insurgent Influence: Experimental Evidence from the Philippines.” Journal of
Development Economics, v. 118 (Jan. 2016): 171-182
Dasgupta, Aditya and Gawande, Kishore and Kapur, Devesh, 2016. (When) Do Anti-Poverty
Programs Reduce Violence? India's Rural Employment Guarantee and Maoist Conflict (April
22, 2016). Available at
SSRN: or
(forthcoming in International Organization)
de Ree, Joppe, and Eleonora Nillesen. 2009. “Aiding Violence or Peace? The Impact of Foreign
Aid on the Risk of Civil Conflict in Sub-Saharan Africa.” Journal of Development Economics
88 (2): 30113.
De Waal, Alexander. 1997. Famine crimes: politics & the disaster relief industry in Africa. Indiana
University Press.
DFID. 2012. “Making systematic reviews work for international development research”. Briefing
paper no. 1, January 2012. Accessed on December 12, 2016, at
Doucouliagos, H., Paldam, M. 2009. “The aid effectiveness literature: the sad results of 40 years of
research”. Journal of Economic Surveys, 23(3), 433461.
Duffield, John. 1994. ‘‘Dangerous Precedents: Famine Relief in Somalia 19911993.’’ In The
Political Economy of Internal War: Asset Transfer, Complex Emergencies, and International
Aid, eds. J. Macrae and A. B. Zwi. London: Zed Books in association with Save the Children
Fund (UK).
Easterly, William. 2001. The Elusive Quest for Growth: Economists’ Adventures and
Misadventures in the Tropics. Cambridge, MA: MIT Press.
Findley, Michael G., Josh Powell, Daniel Strandow, and Jeff Tanner. 2011. “The Localized
Geography of Foreign Aid: A New Dataset and Application to Violent Armed Conflict.” World
Development, 39 (11): 19952009.
GHA 2013: Global Humanitarian Assistance, Annual Report 2013.
GIZ. 2015. “Employment Promotion in Contexts of Conflict, Fragility and Violence Opportunities
and Challenges for Peacebuilding”. Published by GIZ, Bonn and Eschborn, September 2015.
Goodhand J. 2002. “Aiding violence or building peace? The role of international aid in
Afghanistan”. Third World Quarterly 23(5): 837859.
Goodhand J. 2005. Aid, Conflict, and Peacebuilding in Sri Lanka, 20002005. Colombo: The Asia
Goodhand J. 2006. Aiding Peace?: The Role of NGOs in Armed Conflict. Boulder, CO: Lynne
Grossman, H. I .1999: “Kleptocracy and Revolutions,” Oxford Economic Papers, 51(2), 267283.
Grossman, H. I. 1991: “A General Equilibrium Model of Insurrections,” American Economic
Review, 81(4), 912921.
Grossman, H. I. 1992. “Foreign Aid and Insurrection,” Defence and Peace Economics, 3(4), 275-
Hirshleifer, Jack. 1991. “The paradox of power”. Economics and Politics 3:177-200.
Hoelscher, Kristian, Jason Miklian, and Krishna Chaitanya Vadlamannati. 2012. “Hearts and
Mines: A District-Level Analysis of the Maoist Conflict in India.” International Area Studies
Review 15, 2: 14160.
Holmes, R., 2009, ‘Cash Transfers in Post-Conflict Contexts', ODI Project Briefings, no. 32,
Overseas Development Institute, London
Holmes, R., McCord, A., Hagen-Zanker, J., with Gina Bergh and Franzisca Zanker. 2013.”What is
the evidence on the impact of employment creation on stability and poverty reduction in fragile
states A systematic review”. ODI, London, May 2013.
Iyengar, Radha, Jonathan Monten, and Matthew Hanson. 2011. “Building Peace: The Impact of
Aid on the Labor Market for Insurgents”. NBER Working Paper No. 17297, Issued in August
2011 (National Bureau of Economic Research)
Johnson, Gregory; Ramachandran, Vijaya; Walz, Julie. 2012.” CERP in Afghanistan: Refining
Military Capabilities in Development Activities”. PRISM Security Studies Journal. March
2012, Vol. 3 Issue 2: 81 98.
Justino, P. 2007. “Carrot or Stick? Redistributive Transfers versus Policing in Contexts of Civil
Unrest,” Households in Conflict Network Working Papers 3.
Karell, Daniel. 2015. “Aid, power, and grievances: Lessons for war and peace from rural
Afghanistan”. The Economics of Peace and Security Journal 10(2): 28-42
Keen, David, 1998. The Economic Functions of Violence in Civil Wars, Oxford: Oxford University
Khanna, Gaurav, and Laura Zimmermann. 2014. “Fighting Maoist Violence with Promises:
Evidence from India’s Employment Guarantee Scheme.” Economics of Peace and Security
Journal 9, no. 1: 3036.
Luttwak, Edward N. 1999. “Give War a Chance”. Foreign Affairs, 78 (4): 3644.
McCord A, Chopra D. 2010. “Review of public works programmes in the South Asia region”.
London: Overseas Development Institute. Unpublished.
McCord A, Slater R. 2009. “Overview of public works programmes in sub-Saharan Africa”.
London: Overseas Development Institute.
Nielsen, Richard A. 2011. “Foreign Aid Shocks as a Cause of Violent Armed Conflict.” American
Journal of Political Science 55 (2): 21932.
Nunn, Nathan, and Nancy Qian. 2014. “US Food Aid and Civil Conflict.” American Economic
Review 104 (6): 163066.
Office of Inspector General. 2008. Audit of USAID/Iraq´s Community Stabilization Program. Audit
Report No. E-267-08-001-P. Baghdad.
267-08-001-p.pdf. Accessed December 12, 2014
Patrick, Stewart. 2007. "Failed" States and Global Security: Empirical Questions and Policy
Dilemmas." International Studies Review 9, no. 4 (2007): 644-62.
Petticrew, M., & Roberts, H. (2006). Systematic reviews in the social sciences: A practical guide.
Oxford: Blackwell Wood, Reed M. and and Molfino, Emily. 2016. 2Aiding Victims, Abetting
Violence: The Influence of Humanitarian Aid on Violence Patterns During Civil Conflict”.
Journal of Global Security Studies, 2016, 118
Polman, Linda. 2010. “The Crisis Caravan: What’s Wrong with Humanitarian Aid”?. 1st U.S. ed..
New York: Metropolitan Books.
Raleigh, Clionadh, and Havard Hegre. 2005. ‘‘Introducing ACLED: An Armed Conflict Location
and Event Dataset.'' Paper presented to the conference on ‘‘Disaggregating the Study of Civil
War and Transnational Violence,'' University of California Institute of Global Conflict and
Cooperation, San Diego, CA, 78 March.
Ree, Joppe de, and Eleonora Nillesen. 2009. “Aiding Violence or Peace? The Impact of Foreign
Aid on the Risk of Civil Conflict in Sub-Saharan Africa.” Journal of Development Economics
88 (2): 30113.
Scoones, David. 2013. “Winning Hearts and Minds: Public Good Provision in the Shadow of
Insurgency.” Peace Economics, Peace Science and Public Policy 19 (1): 1731.
Sexton, Renard. 2016 “Aid as a Tool against Insurgency: Evidence from Contested and Controlled
Territory in Afghanistan,” (forthcoming in American Political Science Review)
SIGAR Special Inspector General for Afghanistan Reconstruction. 2011. “Quarterly Report to the
United States Congress.” January.
SIGAR 2009. Special Inspector General for Afghanistan Reconstruction. 2009. “Increased
Visibility, Monitoring and Planning Needed for Commanders Emergency Response Program in
Afghanistan”. SIGAR Audit 09 05.
SIGIR. 2013. Special Inspector General for Iraq Reconstruction. 2013. Lessons Learned on the
Department of Defense’s Commander’s Emergency Response Program in Iraq (SIGIR 13-005),
January 24, 2013.
Stein, Michelle M. “Lawmakers Question CERP Funds in Afghanistan.” Military Times. August 8. lawmakers-question-cerp-funds-in- afghanistan/.
Suhrke, Astri. 2006. “When More is Less: Aiding Statebuilding in Afghanistan.” FRIDE Working
Paper. September.
Tahir, Nadia. 2015. “Does Aid Cause Conflict in Pakistan?” Defence and Peace Economics 0 (0):
UN PCEIR 2009. Policy for Post-Conflict Employment Creation and Income Generation, and
Reintegration (PCEIR). Geneva: International Labour Organization.
Uvin P. 1998. Aiding Violence: The Development Enterprise in Rwanda. West Hartford, CT:
Kumarian Press
van Weezel, Stijn, 2015. “A Spatial Analysis of the Effect of Foreign Aid in Conflict Areas” (June
22, 2015, unpubished).
Waal De, Alexander. 1997. “Famine crimes: politics & the disaster relief industry in Africa.”
Indiana University Press.
Waddington, Hugh, Howard White, Birte Snilstveit, Jorge Garcia Hombrados, Martina Vojtkova,
Philip Davies, Ami Bhavsar, et al. 2012. “How to Do a Good Systematic Review of Effects in
International Development: A Tool Kit.” Journal of Development Effectiveness 4 (3): 35987.
Warlord, Inc. Extortion and Corruption Along the U.S: Supply Chain in Afghanistan. 2010. Report
of the Majority Staff, Subcommittee on National Security and Foreign Affairs.
Weintraub, Michael. 2016. “Do all good things go together? Development Assistance and Violence
in Insurgency” (forthcoming in Journal of Politics)
Wilder A and Gordon S .2009. “Money can’t buy America love”. Foreign Policy, 1 December
(online). Available at:
love?hidecomments=yes) (accessed 28 November 2010).
Wilder A and Gordon S .2009. “Money can’t buy America love”. Foreign Policy, 1 December
(online). Available at:
love?hidecomments=yes) (accessed 28 November 2010).
Wilder, Andrew, and Paul Fishstein. 2012. Winning Hearts and Minds? Examining the
Relationship between Aid and Security in Afghanistan. Feinstein International Center, Tufts
Williamson, Jamie A. 2011. “Using Humanitarian Aid to ‘win Hearts and Minds’: A Costly
Failure?.” International Review of the Red Cross 93 (884): 103561.
Wilton Park. 2010. “Winning ‘Hearts and Minds’ In Afghanistan: Assessing the Effectiveness of
Development Aid in COIN Operations,” Report on Wilton Park Conference 1022, held March
1114, 2010 (July 22, 2010).
Wood, Reed M. 2010. “Rebel Capability and Strategic Violence against Civilians.” Journal of
Peace Research 47 (5): 60114.
Wood, Reed M. 2012. “Armed Intervention and Civilian Victimization in Intrastate conflicts.
Journal of Peace Research 49 (5): 64760.
Wood, Reed M. 2014. “From Loss to Looting? Battlefield Costs and Rebel Incentives for
Violence.” International Organization 68 (04): 97999.
Wood, Reed M., Molfino, Emily. 2016. “Aiding Victims, Abetting Violence: The Influence of
Humanitarian Aid on Violence Patterns During Civil Conflict”. Journal of Global Security
Studies, 2016, 118.
World Bank. 2013. “Designing Community-Driven Development Operations in Fragile and
Conflict-Affected Situations: Lessons from a Stocktaking.” Social Development Department,
World Bank, Washington, DC.
World Development Report 2011: Conflict, Security, and Development. 2011. World Bank
Young, Joseph K., and Michael G. Findley. 2011. “Can Peace Be Purchased? A Sectoral-Level
Analysis of Aid’s Influence on Transnational Terrorism.” Public Choice 149 (3-4): 36581.
... Moreover, although previous findings can be reconciled, they are frequently contradictory. The same studies that find a negative association between social programs and property crime show a null (Chioda et al. 2016;Camacho and Mejía 2013;Rios Salgado and Llano Jaramillo 2021;Watson et al. 2020) or positive link between CCTs and armed violence (Weintraub 2016;Zürcher 2017). Therefore, it seems critical to theorize about how CCTs can generate negative externalities that may attenuate and even revert their violencereducing impacts. ...
... In addition to their individual-level effects, cash transfer programs can increase the overall value of localities and, in doing so, trigger political, criminal, and armed territorial competition (Cameron and Shah 2013;Casas 2018;Díaz-Cayeros et al. 2016;Zürcher 2017). Cameron and Shah (2013), for example, examined the impact of the distribution of cash fuel subsidies in Indonesia and found the deployment of the program to be associated with increases in political and ordinary violence, revealing efforts by clientelist and criminal actors to capture the benefits of the program. ...
... Cameron and Shah (2013), for example, examined the impact of the distribution of cash fuel subsidies in Indonesia and found the deployment of the program to be associated with increases in political and ordinary violence, revealing efforts by clientelist and criminal actors to capture the benefits of the program. Similarly, studies about civil war have examined the issue and found that the introduction of foreign and local aid exacerbates conflict between armed actors (Zürcher 2017). Wood and Molfino (2016) found that in sub-Saharan Africa the introduction of humanitarian aid led to an increase in the number of confrontations between rebel and government forces. ...
Because conditional cash transfer programs (CCTs) can address the deep roots of violence, many scholars and policymakers have assumed them to be an effective and innocuous tool to take on the issue. I argue that while CCTs may have positive economic effects, they can also trigger social discord, criminal predation, and political conflict and, in doing so, increase violence. To test this claim, I take advantage of the exogenous shock caused by the randomized expansion of Mexico’s flagship CCT, PROGRESA/Oportunidades. I find that that the experimental introduction of the program increased rather than decreased violence. Then, I analyze all the data compiled by LAPOP on the issue over the years. I find that, other things constant, Latin Americans are more exposed to violence and insecurity when they participate in CCTs than when they do not. These findings urge us to reconsider the effects of social programs on violence.
... Consequently, government and insurgent troops will seek to gain control of the disaster-affected territories, usually by force, hence driving conflict escalation. Similarly, rebels might attack these areas in order to sabotage aid projects that are intended to strengthen the social contract between the government and the population (Zürcher 2016 One of the most frequently discussed causal pathways between disasters and armed conflict is the emergence of a power vacuum or of state weakness (Brzoska 2018;Hollis 2018;Meierding 2013). Disaster can limit the access of state forces to certain areas (e.g., because roads are washed away), thus providing the rebels with an opportunity for military expansion and territorial or organizational consolidation. ...
Full-text available
A ground-breaking study on how natural disasters can escalate or defuse wars, insurgencies, and other strife. Armed conflict and natural disasters have plagued the twenty-first century. Not since the end of World War II has the number of armed conflicts been higher. At the same time, natural disasters have increased in frequency and intensity over the past two decades, their impacts worsened by climate change, urbanization, and persistent social and economic inequalities. Providing the first comprehensive analysis of the interplay between natural disasters and armed conflict, Catastrophes, Confrontations, and Constraints explores the extent to which disasters facilitate the escalation or abatement of armed conflicts—as well as the ways and contexts in which combatants exploit these catastrophes. Tobias Ide utilizes both qualitative insights and quantitative data to explain the link between disasters and the (de-)escalation of armed conflict and presents over thirty case studies of earthquakes, droughts, floods, and storms in Africa, the Middle East, Asia, and Latin America. He also examines the impact of COVID-19 on armed conflicts in Iraq, Afghanistan, Nigeria, and the Philippines.
... Most of these investigations operate at the country level [4][5][6][7][8][9][10]. Recently, given the emergence of highly granular data on project locations of development cooperation, a growing number of studies consider subnational variations when investigating this relationship [11]. So far, there is no clear consensus on the relationship. ...
Full-text available
We know little about the general geographic allocation of development projects in post-conflict regions, and specifically of gender-focused projects. In this study, we explore whether donor agencies prefer to work in “safe” places or dare to operate in conflict-affected zones. Using Colombia as a case study, we combine data on battle deaths from the UCDP Georeferenced Event Dataset (1994–2004) with georeferenced information on the location of development projects from the Colombia AIMS dataset (2006–2013) and manually geocode data for German-funded development projects (2012–2018) with gender as a significant objective. Using count models (N = 1120), we find a statistically strong and positive relationship: an increase in battle deaths increases the number of development projects (with and without gender-focus) in a municipality. Interaction models further reveal an amplification of this relationship for regions with a large proportion of female-headed households, as well as a high number of formally employed and literate women. A context-sensitive interpretation of our findings suggests that (1) development projects in general, and German-funded gender-focused projects in particular, dare to operate in post-conflict settings; (2) women may play an active role as community leaders and mobilizers to influence the allocation of development programs to certain regions.
As Official Development Assistance (ODA) tops 180 billion USD per year, there is a need to understand the mechanisms underlying aid effectiveness. Over the past decade we have seen some low- and middle-income countries become developed nations with record economic growth. Others remain in development purgatory, unable to provide their citizens with access to essential services. In an effort to improve aid effectiveness, the prescriptive nature of aid, where (typically) Western countries allocate funds based on perceived need or the strategic priorities of donors is being reconsidered in favour of locally-led development, whereby recipient governments and sometimes citizens are involved in the allocation and delivery of development aid. Meeting the preferences of donors (both governments and citizens) has been a longstanding priority for international development organisations and democratically governed societies. Understanding how these donor preferences relate to recipient preferences is a more recent consideration. This systematic review analysed 58 stated preference studies to summarise the evidence around donor and recipient preferences for aid and, to the extent possible, draw conclusions on where donor and recipient preferences diverge. While the different approaches, methods, and attributes specified by included studies led to difficulties drawing comparisons, we found that donors had a stronger preference than recipients for aid to the health sector, and that aid effectiveness could be more important to donors than recipients when deciding how to allocate aid. Importantly, our review identifies a paucity of literature assessing recipient perspectives for aid using stated preference methods. The dearth of studies conducted from the recipient perspective is perplexing after more than 30 years of 'alignment with recipient preferences', 'local ownership of aid', 'locally-led development' and 'decolonisation of aid'. Our work points to a need for further research describing preferences for aid across a consistent set of attributes in both donor and recipient populations.
Prevalent counterinsurgency theories posit that small development aid projects can help stabilize regions in conflict. A widely assumed mechanism runs through citizen attitudes, often called “winning hearts and minds,” where aid brings economic benefits and sways public perceptions, leading to more cooperation and, eventually, less violence. Following a preregistered research design, we test this claim using difference‐in‐differences, leveraging original survey data, and new geocoded information about infrastructure projects in northern Afghanistan. We find that aid improves perceived economic conditions but erodes attitudes toward government and improves perceptions of insurgents. These attitudinal effects do not translate into changes in violence or territorial control. Testing mechanisms, we find projects with robust local consultation have fewer negative attitudinal effects, as do health and education projects. These findings challenge the “hearts and minds” theory but complement the wider literature on legitimacy, suggesting that foreign aid can improve human development but rarely meaningfully brings political stabilization.
Though the principle of “Do No Harm” is widely accepted, conflict sensitivity is insufficiently implemented in practice. This is frequently perceived as a failure of knowledge or capacity, but more often speaks to deeper challenges and gaps in organisational cultures, individual mindsets, and larger barriers in the aid system. Interpeace supports organisations to adopt peace responsiveness – a holistic approach to transform the ability of actors to act in a conflict-sensitive manner and to deliberately contribute to peace. This paper focuses on the organisational, individual, and systems-wide changes required to implement conflict sensitivity and peace responsiveness in practice.
Full-text available
“Hearts and minds” theory contends development aid strengthens community support for counterinsurgents by providing jobs and public goods. Based on field interviews in Kabul, we develop an alternative theoretical framework emphasizing instead the ideological preferences of civilians. In our model, some aid projects are ideologically contentious while others are benign. Given a mix of foreign aid, each civilian supports either the counterinsurgents or rebels, depending on his/her idiosyncratic preferences. In this setting, greater provisions of aid can actually erode community support. Donors therefore calibrate the mix of foreign aid to appease population groups with relatively strong ideological sensibilities. Individual-level analysis based on unique Afghan data substantiates key features of our theory. Benign projects lead to favorable opinions of development, while contentious aid has the opposite effect. Moreover, favorable opinions of development are associated with stronger support for government and counterinsurgents, and weaker support for rebels.
Previous research has shown that nonviolent resistance (NVR) campaigns are beneficial for democratization. However, research to date has not considered whether nonviolent revolutions succeed long term in bringing about democratic consolidation. In this paper, I address this gap by analyzing the effect of NVR on democratic consolidation, using data on Huntington’s consolidation criteria of two peaceful turnovers of power. The results suggest that initiating a democratic transition through NVR is not necessarily beneficial for achieving the first peaceful turnover of power. However, given that a democratic regime achieved a successful first turnover of power, NVR substantially increases the probability of completing democratic consolidation through a second peaceful turnover of power.
Theory and extensive evidence connect poverty and underdevelopment to civil conflict, yet evidence on the impact of development programs on violence is surprisingly mixed. To break this impasse, we exploit a within-country policy experiment to examine the conditions under which anti-poverty programs reduce violence. The roll-out of India's National Rural Employment Guarantee Scheme caused a large long-run reduction in Maoist conflict violence, as measured with an original data set based on local language press sources. These pacifying effects were not uniform, however, but overwhelmingly concentrated in districts with sufficient pre-existing local state capacity to implement the program effectively. The results demonstrate the potential for anti-poverty programs to mitigate violent civil conflict by improving livelihoods, but also highlight the crucial role of state capacity in shaping these effects.
Findings in political science, economics, and security studies suggest that during civil war aid can be used to help establish control of contested areas and reduce insurgent violence by winning the “hearts and minds” of the population. These accounts typically ignore the strategic implications of aid distribution by progovernment forces, namely that rebel groups should resist the implementation of aid projects that would undermine their position. Using a new dataset of fine-grained and geolocated violence incidents in Afghanistan and random variation in the administration of some U.S. counterinsurgency aid, I show that insurgents strategically respond to counterinsurgency aid in contested districts by resisting through violent means. The results indicate that civilian aid only reduces insurgent violence when distributed in districts already controlled by progovernment forces; when allocated to contested districts civilian aid in fact causes a significant increase in insurgent violence. The results also indicate that the effect of counterinsurgency aid on violence varies by project type, and can be overwhelmed by macrolevel strategic changes in the conflict.
Annual allocations of bilateral and multilateral humanitarian assistance to conflict-affected states total billions of dollars each year. Humanitarian assistance plays a vital role in sustaining vulnerable populations. However, inflows of such aid may also exacerbate violence by both threatening insurgents and creating incentives for these groups to extend or deepen control over the areas in which aid is concentrated. Insurgent efforts to ameliorate threats and co-opt resources ultimately raise the risk of conflict between insurgent and counterinsurgent forces. We use recently constructed geo-located data on both aid commitments and conflict events in a sample of twenty sub-Saharan African countries during the post–Cold War era to evaluate the impact of aid on violence. Even after accounting for the non-random assignment of aid within conflict zones, we find that humanitarian aid increases the frequency of subsequent violent engagements between rebel and government forces in the areas in which aid is concentrated. Importantly, however, we find no evidence that other forms of foreign development aid exacerbate or prolong violence in the areas in which they are allocated.
International organizations and national governments deliver billions of dollars in development assistance each year to citizens in ongoing insurgencies. Existing work shows that development assistance shapes insurgent violence, yet underspecified mechanisms and a failure to test all competing explanations have hampered knowledge cumulation. This article proposes a new theory of insurgent territorial loss to argue that development assistance should increase overall violence by insurgents, and indiscriminate violence in particular. Anti-poverty subsidies incentivize information sharing with the government, increasing incumbent territorial control and forcing insurgents to rely on increased and more indiscriminate violence to recapture territory. A difference-in-differences identification strategy combined with matching tests all existing mechanisms against one another in the context of a conditional cash transfer program in Colombia. The empirics provide support for the territorial loss mechanism, while competing explanations do not find support. Development assistance frequently produces welfare gains yet may also lead to increased insurgent violence.
The persistence and brutality of contemporary civil wars have left many analysts puzzled. Traditional interpretations describe civil wars as simple confrontations between two sides, as explosions of mindless violence, or as disrupting apparently benevolent development processes within countries. These approaches do not fully take into account the rational economic calculations that drive many civil conflicts in the late twentieth century. This paper argues that, to understand violence in civil wars, we need to understand the economic dimensions underpinning it. Economic activities arising from war fall into seven categories: pillage; extorting protection money; controlling or monopolising trade; exploiting labour; gaining access to land, water and mineral resources; stealing aid supplies; and advantages for the military. If these short-term benefits suggest that there is more to civil wars than simply winning, so too does the prevalence and persistence of behaviour that is, in military terms, counter-productive. This can take two forms: cooperating with the ‘enemy’; and mounting attacks that increase, rather than reduce, political and military opposition. This paper describes two forms of economic violence: ‘top-down’, which is incited by political leaders and entrepreneurs; and ‘bottom-up’, where violence is actively embraced by ‘ordinary’ people, either civilians or low-ranking soldiers. Seven conditions can encourage top-down economic violence: a weak state; rebel movements that lack strong external finance or support; an undemocratic or ‘exclusive’ regime under threat; economic crisis; ethnic divisions that cut across class lines; the existence of valuable commodities; and prolonged conflict. Three conditions are particularly conducive to bottom-up violence: deep social and economic exclusion; the absence of a strong revolutionary organisation; and impunity for violent acts. To achieve more lasting solutions to civil conflicts, it needs to be acknowledged that violence can present economic opportunities. This paper concludes that outside intervention must take into account the political and economic interests of the violent. Intervention must provide realistic economic alternatives to violence, and must handle democratic transitions and the introduction of free markets with sensitivity. This is likely to mean strengthening and improving the institutions of the state, such as schools, social-security systems and establishing a more accountable police force and Army.
Using a newer and expanded dataset as well as a survey of practitioner perceptions, this article adds to a recent body of literature on reconstruction and violence in Afghanistan. Data are taken from military-led development projects by way of the United States military’s Commander’s Emergency Response Program (CERP) and, to measure violence, from U.S. military Significant Activity reports. The results suggest that, at great cost, large-budget CERP efforts (those in excess of USD50,000 per project) are associated with an increase in violence and thus counter-productive to military stability goals. In contrast, small projects (below USD50,000), which comprise a smaller proportion of total CERP allocations, are associated in statistically significant ways with reductions in violence. To explore why CERP projects may have these effects, the article also examines administrative modalities for CERP spending. The results suggest that timely, flexible expenditure of CERP funds are most effective at reducing violence. [JEL codes: D74, O53]