Michael G. Findley
Abstract
Does foreign aid build peace? The answer is of paramount importance for policy makers and practitioners, given that the world’s poor are growing increasingly concentrated in conflict-affected countries. Scholars have also demonstrated keen interest, primarily examining the relationship between foreign aid and civil wars. This review takes stock of the existing literature through a survey of key theoretical arguments connecting aid to the onset, dynamics, and recurrence of civil wars. It then articulates a key challenge posed by undertheorization of aid allocation, which is largely nonrandom, making the causal effects difficult to infer. I identify five areas in need of greater attention: microfoundational theoretical assumptions about aid flows; aid in the context of other foreign policy options; explicit articulation of other factors that may mediate or moderate aid’s effects; levels of observation and aggregation; and measurement.
Keywords
foreign aid, peacebuilding, civil wars, aid effectiveness, aid allocation, international development
INTRODUCTION
Since World War II, most countries have experienced violence, and in many countries, violence has exceeded the traditional threshold of 1,000 battle-related deaths to qualify as civil war (Pettersson & Wallensteen 2015). Civil war poses a fundamental problem against successful economic and human development (Collier et al. 2003). Most countries that have avoided civil war are making great strides toward sustainable development, whereas most countries plagued by war are not—and some are in fact reversing progress, often to a significant extent (World Bank 2011). A key concern is that poverty itself may trigger violence, which may in turn increase poverty, thus creating a persistent vicious cycle. World Bank President Jim Yong Kim recently articulated this syndrome, saying, “By 2030, 50% of the global poor will live in areas affected by conflict and fragility” (Kim 2017).
To mitigate the risk and impact of war while servicing their own foreign policy interests, wealthier states and international organizations frequently make use of foreign aid—public assistance with a grant or concessional component administered to promote development. Since World War II, donors have given a total of more than US $7 trillion in foreign economic assistance to developing countries (Tierney et al. 2011). The international community’s development assistance efforts are currently at unprecedented high levels. The Millennium Development Goals, adopted in 2000, encouraged donors to devote more aid to developing regions, building on a commitment first made in 1970 to give 0.7% of their gross domestic product. While many donors have yet to meet this commitment, overall aid spending has been slowly increasing since 2000.
Donors have made conflict-affected countries and regions a high priority in recent years (Collier et al. 2003, Ellison 2016, World Bank 2011). In their policies and guidelines, donors appear to believe that their efforts could be a tool used to mitigate civil war and achieve peace (DAC 2012, OECD 2016b). “Peace, Justice and Strong Institutions,” a key item (#16) of the Sustainable Development Goals, increases donors’ emphasis on reducing violence and promoting peace.
Scholarship on the relationship between aid and civil war is also increasing, partly through collaborations with practitioners. Within the scholarly literature, multiple causal processes connecting aid to civil war have been advanced, with little consensus about aid’s effects. A central goal of this article is to review the arguments and evidence reflected in studies that explore different theoretical claims. I organize the article around aid’s effects on the onset, dynamics, and recurrence of civil war.
The literature has not been sufficiently disciplined in its arguments. In many studies on aid and civil war, analyses of one stage of conflict processes are loosely applied to other stages when the applicable theoretical logics could actually be quite distinct. In particular, theories about the onset and termination of civil war may differ fundamentally from theories about the intensity of war. The implication is that macro- versus micro-level goals may shape aid differently, and we should be attuned to that possibility in future work.1
Studies of aid and civil war onset emphasize predation, especially rent seeking and looting, commitment and information problems, as well as indirect effects of aid designed to win the hearts and minds of the people. Scholarship considering whether aid affects civil war dynamics investigates similarly diverse factors, including predation, moral hazard, winning hearts and minds, opportunity costs, and international presence. Finally, scholarship on aid’s effects in the postwar phase also considers factors such as predation and winning hearts and minds, but it focuses on disaggregating aid by sector, examining the dynamics of economic recovery, understanding micro-level interventions, and addressing reconciliation challenges. The body of this article unpacks aid’s effects at each stage of conflict processes.2
While my primary focus is on how aid affects civil wars, I also devote some attention to how and why aid is allocated. This discussion highlights a consensus that aid is given in strategic, political ways—and is therefore nonrandom. This raises critical questions about how best to causally identify models considering the effects of aid on civil wars. Indeed, my section on allocation should raise serious concerns about the literature that is reviewed, highlighting the necessity of devoting much greater attention to the politics of aid allocation specifically in the context of wars and peace processes.
Following the discussion of allocation, I present ideas for the scholarly agenda. On the whole, I contend that aid’s positive or negative effects are likely to be relatively small in comparison to other foreign policy options and financial flows, that the effects are difficult to detect at country-year levels and are more likely to be found at subnational and micro levels, and that these subnational and micro-level effects are probably not large direct effects but make a difference as moderators or are conditioned by other variables. I now unpack these factors through the examination of five specific scholarly approaches.
First, microfoundational theoretical assumptions are not well examined, especially the process by which aid is allocated and flows through to the point at which it could affect recipient actors—the government, potential rebels, or ordinary citizens. In the midst of much-studied bargaining between governments and rebels, donors and governments also bargain with each other. All the bargaining is subject to failures, which have not been sufficiently theorized in this context. Moreover, questions about the fungibility and lootability of aid, which have important implications for the relationship between aid and war, are not settled. Indeed, scant attention has been devoted to validating frequently invoked theoretical assumptions concerning what governments and rebels know about the dynamics of aid, including spatial–temporal questions such as where aid is specifically located, when aid funding is actually disbursed, and when aid reaches its intended sites.
Second, foreign aid is just one tool within a suite of foreign policy options to address civil war. Decisions to allocate foreign aid, as opposed to engaging in other foreign policy activities, result from a selection process that likely has consequences for aid’s effects on civil war, but is often undertheorized—if not ignored—in studies of those effects. In addition, while states may directly employ foreign policy tools, they may also work to facilitate other mechanisms that alter the available resources that can affect civil war dynamics. At the very least, future work needs to better theorize, and perhaps explicitly integrate, other financial flows—foreign direct investment, remittances, sovereign lending, private foundation grants, debt relief, military assistance, arms shipments, or even domestic government fiscal transfers—to understand the constellation of financial flows that complement or substitute for aid and potentially affect civil wars.3 Analyses should separate these financial flows from aid flows, as warranted to distinguish the independent or codependent effects of aid on civil wars.
Third, studies regarding the aid–conflict nexus need to more fully engage the myriad factors that may moderate or mediate aid’s effects on civil war, in order to make strong causal inferences rather than merely detect correlations. Theoretical work is crucial because existing arguments about the direct effects of aid do not provide competing testable predictions. Understanding moderators or mediators may help disentangle the relationship. Some of the most significant theoretical advances in the past few years pertain to the moderating effects of territorial control and multi-donor spatial concentration of aid, as well as mediating factors such as aid-induced improvements in economic conditions and sharing of information about rebels.
Fourth, scientific ideals of cumulation and integration of knowledge have been complicated by the varied levels of theoretical and empirical aggregation across studies. For example, some micro-level studies employ incredibly refined aid information but do so for a unique type of aid (e.g., military-allocated or counterinsurgency aid) and a unique type of conflict environment (e.g., US counterinsurgency efforts). These studies are methodologically rigorous, but inference beyond these few sectoral cases may be problematic. Moreover, the drivers of variation in local contexts may differ from those at higher levels of aggregation. On the flip side, other studies examine more common types of aid across regions or countries but struggle to overcome methodological challenges associated with pooling data, such as accounting for important heterogeneity of countries and regions, as well as the aid itself.
Fifth, issues of measurement require concerted consideration. Over the past decade, measurement of aid has improved dramatically, in no small part thanks to AidData.4 Still, some of the data scholars would like to analyze are not systematically available and may not be for some time. Among the important missing elements are specific details on aid amounts (especially actual disbursements, which vary considerably from commitments), the precise timing of expenditures, and activity-level (within-project) details, particularly in multi-sector projects. Also, subnational analyses depend on high-resolution, location-specific information, which unfortunately cannot be obtained from most project records. Donors have little incentive to disaggregate in project planning documents. Only in exceptional cases, such as the United States’ aid to Iraq, do donors systematically and rigorously track aid disbursements.5
CONCEPTS AND PATTERNS
Foreign aid most commonly refers to official development assistance (ODA). ODA is public assistance—government bilateral or multilateral—that is administered to promote economic development and welfare and is concessional, meaning that it carries at least a 25% grant element (OECD 2017).6 This assistance is the most common policy lever developed countries use to improve economic development in other countries. Development here refers to “long-term, high rates of economic growth…accompanied by a wider economic transformation that benefits the poor and shares prosperity broadly” (DFID 2014, p. 3). While economic development is not the only outcome of interest for development assistance, historically it has been one of the most important. ODA is distinct from humanitarian aid, which provides immediate medical, food, and other assistance in emergency situations, including conflict-affected areas and refugee crises. ODA is also distinct from military aid, which is provided directly by militaries (or to militaries), though some such aid seeks to boost development. Another separate category is counterinsurgency aid, which is allocated by counterinsurgents directly, rather than through traditional development assistance channels.7 Finally, ODA is different from other official flows that (a) are insufficiently concessional to qualify as ODA and/or (b) reflect specific representational or commercial interests, rather than having a broad development intent (Tierney et al. 2011). These aid flows may overlap but often have distinct purposes. This article addresses each to some extent, explicitly disaggregating where appropriate.
ODA is the result of a long, complex process that unfolds at the intersection of donors’ national interests, recipients’ national interests, bureaucratic processes surrounding the allocation of aid, and the complex delegation chain of partners responsible for implementing aid. Yet, most studies of aid and conflict theorize about initial donor commitments (a consequence of the data, which until recently has been largely limited to highly aggregated commitment flows) and simply aggregate those commitments to consider their effects. By construction, therefore, most studies ignore donor-specific dynamics, including loan and grant negotiations over targeting and allocation. Taking commitment data as a proxy for the amount and location of aid within a country, researchers focus on how the existence of that aid affects governments and dissident groups who may use violence, including civil war.
Civil war broadly refers to the joint production of violence between a government and one or more opposition groups in which the number of battle-related deaths exceeds a stated threshold. Moreover, civil war violence is intended to replace a government or form a new political entity altogether, and it could also force capitulations on a particular policy issue or set of issues. Most scholars use thresholds of 1,000 deaths to define a civil war and 25 to characterize minor civil violence (Allansson et al. 2017). The conceptual and operational dimensions of civil war have been explored at length elsewhere. A key point is that civil war represents a form of political violence; the political nature of the acts distinguishes civil war violence from ordinary crime (LaFree & Dugan 2004). Different forms of political violence are likely more similar than they are different. Indeed, civil wars are composed of a variety of tactics, including the frequent use of terrorism, and are regularly intermixed with campaigns of nonviolent resistance.
Substantial foreign aid is allocated to countries that are affected by civil violence, although even more aid goes to countries that are not (Figure 1). Aid to both types of settings has been increasing for a couple of decades. Interestingly, although the number of civil wars peaked in the 1990s and then began to decline, aid appears to have stayed at higher levels and even climbed in the early 2000s. Section 1 of the Supplemental Appendix provides additional trend analysis comparing aid patterns in conflict and non-conflict settings, as well as top recipients of aid.
A relevant question stems from these empirical trends as well as the theoretical debates discussed below: Are these increasingly large amounts of aid, in fact, related to the occurrence of violence? Examining this relationship is critical because development organizations assume that development aid mitigates violence by redressing the underlying conditions, including poverty and poor governance. For example, James D. Wolfensohn (2001), former President of the World Bank, commented after 9/11 that terrorism was caused by poverty, and thus addressing terrorism required more foreign aid to alleviate poverty. If the conventional wisdom is correct, as articulated by many donors and scholars, then why do we often see a positive correlation between increased aid flows and increased prevalence of civil war?
DOES AID LEAD TO CIVIL WAR?
A vast literature considers whether foreign aid improves economic conditions, such as GDP growth. Some scholars contend that aid is effective and should be encouraged (Sachs 2006), whereas others argue that aid is ineffective and should be discouraged (Deaton 2013, Easterly 2006, Moyo 2009). Some have noted just how brittle the aid–growth results are (Hansen & Tarp 2001, Roodman 2007), raising questions about our ability to study aid effectiveness rigorously, especially absent better theory. In recent years, a growing literature contends that aid’s effects are conditional on a variety of factors, including governance quality, preexisting level of poverty, inequality, and ethnic heterogeneity (Burnside & Dollar 2000, Collier 2007, Clemens et al. 2012), or more simply that aid is not significantly correlated with growth (Doucouliagos & Paldam 2009).
Existing research on whether foreign aid affects civil war often conflates different periods of conflict for which the logics may not be the same. Accordingly, I separate aid’s effects with respect to civil war onset, civil war dynamics, and postwar recurrence, while also delineating the theorized direct, conditional, and indirect relationships. As is evident in each of these sections, extant arguments link aid both to more violence and to more peace. A key task is to catalogue theoretical approaches in order to identify testable implications that could help reconcile or discriminate among the theories.
Civil War Onset
Civil wars occur for myriad reasons, but arguably all are “rooted in endemic competition for resources across groups, with bargained solutions occasionally breaking down” (Blattman & Miguel 2010, p. 17). Aid could affect the probability of civil war onset in a number of ways. Table 1 enumerates the key theoretical arguments. Section 2 of the Supplemental Appendix includes citations alongside each of the predictions.
Table 1
Theoretical predictions for civil war onset
Direct effects.
Some of the earliest work on aid and civil war advances predation arguments, specifically citing the prize of state control. Elites in government have incentives to defend the state and their control over its resources; rebels have incentives to capture the state to acquire power and access to resources. Aid enters the equation as another set of resources. So long as the costs of violence are not sufficiently high, governments and rebels may choose to fight over those resources (Arcand & Chauvet 2001, Azam 1995, Grossman 1992).
A corollary of this argument also focuses on predation but shifts the focus to potential direct rebel control of the aid resources. Aid often bypasses the government and becomes susceptible to capture by various actors. In other words, aid may be distributed in a manner such that potential rebels can loot it to fund rebellion (Addison et al. 2002). Although rebels are the primary actors who would loot aid, arguably some weak, fragmented governments could resort to looting to fund violence, as did the Rwandan government during the lead-up to the 1994 genocide (Uvin 1998).
A third perspective suggests that earlier predation arguments overlook outcomes short of war that benefit all parties. Because war tends to be very costly, there is always some bargain that both sides should prefer, and aid could be part of the bargain. War still occurs, however, because bargaining breaks down. Prominent explanations for failures of bargaining include credible-commitment problems and incomplete information in the presence of incentives for combatants to misrepresent their strength and preferences (Fearon 1995). One reason agreements lack credibility is that foreign aid flows are highly volatile, making the macroeconomy difficult to manage (Bulř & Hamann 2008). Nielsen et al. (2011) contend that large, sudden decreases in aid (negative aid shocks)—whether coordinated by the donor community or not—upset the domestic balance of power among elites and citizens, making rebels more powerful relative to a government with less aid. When negotiating an end to conflict, the government must commit to future resource transfers to appease the rebels, but those promises are not credible because the shift in aid may prove temporary, paving the way for war to erupt once again. Dal Bó & Powell (2009) suggest a closely related alternative mechanism: In bad times (for example, when negative aid shocks occur), governments may offer assistance to opposition groups, but the opposition fears it is being “low-balled” and therefore rejects the deal in favor of violence. The general logic of these different scenarios is that bargains can avert war, and aid might directly contribute to bargaining success or failure depending on the circumstances.
Conditional effects.
Arguments about aid’s direct effects may apply only under certain circumstances. For example, predation arguments make most sense in contexts characterized by weak state institutions. Besley & Persson (2011) contend that while positive aid shocks amplify resources of interest to governments and oppositions, which may motivate them to invest in violence, countries with strong institutions can provide the proper checks and balances against such predation. Thus, the onset of civil war should be confined to a subset of countries with weak institutions that experience sudden, substantial increases in aid inflows.
Qualifying the argument about the deleterious effects of negative aid shocks, the appearance and activity of donors outside of the OECD Development Assistance Committee may have altered the landscape of aid.8 Arguably, these new donors behave differently from the traditional OECD-DAC donor organizations and may provide stability against the ebbs and flows of standard development aid. Evidence suggests that China, for example, may not follow the same rules as the OECD-DAC. Therefore, one could imagine that China provides a consistent presence, counteracting sanctioning initiated by traditional donors for corruption, repression, and violence. Theorizing about this possibility, Strange et al. (2017) contend that the presence of Chinese aid ameliorates the violence-inducing effects of negative aid shocks. Separately, others have suggested that the proliferation of donors in the international community, sometimes referred to as donor fragmentation, should reduce the risk of exposure to negative aid shocks, inasmuch as the actions of specific donors are less likely to dominate. In turn, donor fragmentation should reduce the risk of civil war outbreak (Gutting & Steinwand 2017). Finally, Arab donors appear to be increasingly active in a number of conflict-prone countries following the Arab Spring uprisings, though academic studies have yet to investigate the patterns closely.
Indirect effects.
Foreign aid is traditionally thought mainly to affect economic conditions; it is meant to improve the quality of life for citizens of a recipient country. Indeed, this is the key micro-level logic linking development to peacebuilding. Aid may have an indirect effect on civil war because improvement in economic conditions has been associated with fewer grievances and higher opportunity costs to violence, making war less desirable (Collier & Hoeffler 2004b). Thus, even prior to war onset, aid may help governments secure the hearts and minds of people so that rebellion never gets traction.
A related logic is that aid’s effects on the economy may work via the government. Specifically, if continually provided, aid relaxes government budget constraints, which should have a positive impact on growth if expenditures are appropriate and effective (Collier & Hoeffler 2002, Miguel et al. 2004). More immediately, aid could be used to buy off rebels and/or distributed among civilians so as to thwart support for potential rebellion (Azam & Delacroix 2006, Azam & Mesnard 2003). This argument relies on strong assumptions about the fungibility of aid, a topic to which I return later.
Aid may also work through government channels to decrease the likelihood of civil war by enabling the government to deter opposition groups. Rather than creating a more enticing prize for potential rebels, a government bolstered by aid may signal its strength to defeat any rebellion, strength that discourages opposition groups from initiating such activity (Collier 2011, Collier & Hoeffler 2002). Aid may boost state capacity to monitor and track rebels in the periphery, limiting rebel mobilization.
The timing of aid is pertinent. For example, assistance to a government may generate greater stability during democratic transitions or economic downturns, when civil wars are particularly likely to occur. Targeted aid at these moments may make political commitments between government and society more credible—and therefore less likely to break down into war (Savun & Tirone 2011, 2012).
Despite normative arguments in support of aid, studies demonstrate that aid allocation is not comprehensive, uniform, or fair. Aid is often politicized in ways that can exacerbate disparities and grievances along social cleavages, which in turn can result in rebellion. Jablonski (2014), for example, demonstrates that politicians often allocate aid to their coethnics, meaning that recipient political dynamics shape aid. By some accounts, the disparities and grievances may motivate civil violence (Esman & Herring 2003). Thus, aid may inadvertently cause war due to the politicized disbursement process.
Civil War Dynamics
During ongoing civil wars, development needs often go unmet, and humanitarian consequences can be dire. Unable to fulfill the basic needs of their populations, resource-constrained governments often demand assistance from international sources. At times, international actors offer unsolicited humanitarian assistance. During wars, foreign aid is intended to fulfill both development and humanitarian purposes and often is designed to address conflict dynamics, specifically conflict intensity. The key theoretical arguments linking foreign aid to the dynamics of civil war are summarized in Table 2. Section 2 of the Supplemental Appendix includes citations alongside each of the predictions.
Table 2
Theoretical predictions for civil war dynamics
Direct effects.
The earliest literature on the direct relationship between aid and civil war dynamics emphasized narratives of individual conflicts and identified devastating negative consequences of humanitarian aid, mostly stemming from looting for purposes of financing ongoing civil war (Anderson 1999, Bryer & Cairns 1997, Cooley & Ron 2002, De Waal 1997, Gourevitch 1999, Maren 2009, Polman 2010, Terry 2002). One estimate suggested that 20–80% of food aid shipments to Somalia during the 1990s were looted, stolen, or confiscated (Barnett 2011). An earlier study found that the stolen aid was traded for arms in Ethiopia (Perlez 1992). In the Biafran war, the fact that humanitarian aid had to be shipped on the rebel leader’s planes gave rebels influence over its distribution and allowed more capture and looting, which likely prolonged the war (Barnett 2011).
Humanitarian aid flows may also incentivize insurgent violence by threatening rebel authority. Rebels are likely to perceive aid intended to provide relief to citizens as a tool to consolidate government or counterinsurgent control in recipient areas. This challenge to rebel authority may motivate military or civilian violence (Blouin & Pallage 2008, Wood & Molfino 2016, Wood & Sullivan 2015). Similarly, Narang (2015) contends that humanitarian aid prolongs civil wars by altering the relative valuations of each side’s strength, thus affecting the set of bargains entertained by (or acceptable to) the government and rebels. Aid generates uncertainty about the distribution of capabilities, which makes negotiated bargains less likely to occur, increasing the chances that war continues.
Another negative consequence of aid is that local actors begin to depend on it, generating a moral hazard. If combatants resolve wars or terrorist campaigns, humanitarian aid may no longer be needed. Thus, there are incentives to prolong war. Beneficiaries of humanitarian aid therefore position themselves, often through the continuation of violence, to display need so that aid continues (Bapat 2011).
Finally, international actors may provide cause for continued violence. Intentionally or not, they impose structure and practices that could be the target of government or rebel strategies. For example, humanitarian intervention (closely tied to humanitarian aid) in Bosnia possibly led to the deaths of 20,000 people (Woodward 1995). In addition, it is increasingly common for aid workers with international organizations to become the target of attacks (Narang & Stanton 2016).
Conditional effects.
Development and counterinsurgency assistance can be expected to follow a different logic than humanitarian assistance. Development assistance is not typically deployed in conflict-affected areas of a country, instead being confined to more stable areas. This concentration raises questions about conditional effects. Counterinsurgency assistance may be used in conflict-affected areas but may also require certain conditions in order to be effective. In both instances, the logic is that governments or counterinsurgents want to win the hearts and minds of the people so that they do not join an insurgency, or so that insurgents defect and return to the civilian population.
The impact of development assistance on patterns of violence may depend on the extent to which aid is valuable and accessible. Arguably, when development aid takes the form of discrete projects and is spatially concentrated, making capture worthwhile—and is further from the reach of the government, making capture sustainable—rebels incur less risk in targeting aid and therefore use greater violence in targeting and using aid resources (Findley et al. 2011, Strandow et al. 2016). When aid is less concentrated or closer to areas of government control, rebel incentives to appropriate it should be severely diminished because the aid is less valuable or more difficult to capture.
Following a related logic, Sexton (2016) contends that counterinsurgency aid should reduce violence only when disbursed in areas already secured by counterinsurgent forces. In areas not under counterinsurgent control, insurgents worry that aid may encourage civilians to support counterinsurgents or the government. Insurgents therefore strategically resist, in violent ways, to prevent a shift in allegiance (Beath et al. 2016, Crost et al. 2014, Findley & Young 2007). Where insurgents have influence, they may seek to forestall the support for the government that aid would confer. As such, aid that is supposed to win hearts and minds may work only when there is already some semblance of control by those disbursing aid.
Indirect effects.
A recent line of reasoning contends that security is endogenous to aid, meaning that aid itself may generate the security necessary to bring about aid’s success. Berman et al. (2013) argue that smaller, well-conceived aid projects motivate citizens to share information with governments and counterinsurgents. Information sharing then allows security forces more easily to protect the aid, and therefore the aid can have a peace-promoting effect at a local level.
Traditional development programs may likewise incentivize civilians to support the government or counterinsurgent forces, often through sharing information about rebels. Such interactions can allow governments to engage with rebels better by improving government capabilities or territorial control. This dynamic could also lead insurgents to fear future support for the government and thus engender more violence against the government or counterinsurgents. Indiscriminate violence against civilians may also become more prevalent, especially when governments can no longer easily identify insurgent defectors and suspect whole villages of supporting or collaborating with the insurgents (Weintraub 2016).
Aid may also indirectly reduce violence when it raises the opportunity costs of recruitment and otherwise reduces insurgent influence (Crost et al. 2016), a logic broadly consistent with that of Berman et al. (2011). In this light, aid allocated during war that increases opportunity costs may act similar to aid granted in a prewar phase.
Post–Civil War Recurrence
War has substantial negative economic and social consequences (Collier et al. 2003, Ghobarah et al. 2003, World Bank 2011). When wars end, the task of rebuilding countries is enormous, and the international community commonly banks on the promise of foreign assistance. Scholars have advanced various theoretical logics connecting foreign aid to postwar recurrence. These are summarized in Table 3, and Section 2 of the Supplemental Appendix includes citations alongside each of the predictions. This work is complemented by that of policy-making professionals and practitioners who advance various strategies to build governance and deepen community relations following war. Yet few existing claims have been tested empirically.
Table 3
Theoretical predictions for post–civil war recurrence
Direct effects.
A central argument in the civil war literature holds that postwar international intervention is necessary to prevent the recurrence of war. The theoretical underpinnings are largely derived from bargaining theories of war, with particular attention devoted to military security guarantees (Walter 2002) that alleviate credible-commitment problems. Foreign aid is a key component of international intervention in postwar areas. Sometimes referred to as peacebuilding assistance, most foreign aid in this context is aimed toward peace implementation activities, which are designed to “directly contribute to postwar stabilization and strengthen the capacity of governments, national/local institutions and transitional or other relevant authorities,” according to the United Nations Peacebuilding Fund (http://www.unpbf.org). More broadly, peacebuilding refers to an amalgam of activities in which external actors attempt to prevent the recurrence of war. In general, scholars have contended that the more robust or multidimensional the peacebuilding approach, the more likely its success (Doyle & Sambanis 2006). This effect appears even stronger if other local conditions are favorable.
Humanitarian aid flows into postwar environments with the goal of addressing displacement and other immediate concerns, but in ways that may affect the balance of capabilities of conflict actors. Narang (2014) contends that humanitarian assistance favors the loser in a war and as such can empower the loser to seek to undermine the peace. Moreover, humanitarian assistance remains vulnerable to looting in the postwar period. As an example, following the civil war and genocide in Rwanda aid was looted within refugee camps and likely used to carry out residual postwar violence (Gourevitch 1999).
Conditional effects.
Most postwar aid from traditional donors is designed to build peace in context-specific ways. Drawing on the concept of aid as unearned income or a nontax revenue, Girod (2012) contends that aid is successful in the postwar environment only when recipients have incentives to pursue the development goals articulated by donors. If recipients are strategically important to donors or have natural resource rents, recipients can afford to waste aid resources, confident that assistance is relatively secure or alternative sources of financial support are available.
Because war may affect society and the economy in disparate ways, some have argued that aid’s effects in the postwar period depend on the particular sector to which aid is granted (Chauvet et al. 2010, Flores & Nooruddin 2009b, Donaubauer et al. 2016). The observational, quantitative literature is just beginning to make sector-specific arguments and disaggregate empirical tests accordingly. Limited investigations contend that aid to the transport and urban development sectors is more successful than education projects in postwar settings (Duponchel et al. 2010). More research is needed to fully unpack sectoral effects.
A separate set of studies evaluates the effects of specific development interventions. Each study concerns an individual aid project. The typical design builds a randomized controlled trial into the project from the outset. Whereas most aid projects cannot be evaluated using methods that assume randomization (and frankly are difficult to evaluate even with methods that relax this assumption), this recent wave of randomized approaches has led to the customization of select projects to allow for rigorous evaluation. Thus, while the projects may be intended as specific instances of a broader class of aid programs designed to prevent postwar recurrence, they may be sufficiently stylized that they are not representative—or only representative of other projects that have also been customized to allow for randomized evaluation. Among these interventions, community-driven reconstruction programs were popular for a time. Although the balance of the evidence indicates that these programs are rarely successful, in some exceptional instances they contributed to positive social cohesion and reconciliation outcomes (Cilliers et al. 2016, Fearon et al. 2009). Other donors have introduced public education programming that attempts to change potentially problematic aspects of political culture contributing to the risk of war recurrence, with mixed effects (Paluck & Green 2009). Some programs focus on drivers of recidivism, including cognitive behavioral therapy designed to improve a host of individual-specific perceptions and behaviors, along with cash transfers to address employment and income concerns. One study found that this combination substantially reduced crime and violence in a vulnerable population (Blattman et al. 2016). Numerous micro-level interventions have occurred or are occurring. In time, evidence from studies of these interventions may clarify the conditions under which foreign aid builds peace, has no effect, or promotes war recurrence.
Indirect effects.
By addressing economic challenges, aid should improve outcomes so that governments, rebel groups, and citizens have fewer incentives to return to war. Any indirect effects of aid on war through economic improvements necessarily take time. Indeed, aid in the postconflict period may take some time to absorb, but it should eventually propel economic growth in ways that are important for achieving peace (Addison & McGillivray 2004, Collier & Hoeffler 2004a, Donaubauer et al. 2016, Flores & Nooruddin 2009a). The prevailing evidence supports the idea that aid is effective in postwar environments, which has prompted calls for greater international investment (Quinn et al. 2007).
Despite much optimism and empirical support for this argument, specific mechanisms or conditions are neither well theorized nor rigorously tested (Adam et al. 2008, Elbadawi et al. 2008). The postwar period is complex. A mix of policies—high aid, low taxation, independent public service delivery, and low inflation—is likely required to help countries experience a greater peace dividend (Hoeffler 2012).
Not all work points to aid uniformly promoting peace. Suhrke et al. (2005) criticize prominent studies of recovery, arguing that they rely on too few observations, misspecify relationships, and are insufficiently guided by theory. Moreover, in contrast to the results of Collier & Hoeffler (2004a), others have found different timing effects: Aid distributed within the first three years after conflict may reduce time to recovery, whereas aid distributed seven years or longer after the end of conflict increases time to recovery (Flores & Nooruddin 2009a). Other emerging work finds only modest differences in postwar recurrence when comparing countries with high and low levels of aid (Nunnenkamp 2016).
THE CHALLENGE OF AID ALLOCATION
Studies of aid’s effects on civil war largely neglect aid allocation dynamics, which may complicate knowledge about aid. Aid allocation has been conventionally explained by comparing donors’ strategic interests with the needs of recipients. A large body of literature has generally concluded that donors’ strategic interests or political economy considerations are key drivers of foreign aid allocation (Alesina & Dollar 2000, Bueno de Mesquita & Smith 2009, Kuziemko & Werker 2006, McKinlay & Little 1977, Morrison 2012). In the classic political economy model, foreign donors strategically provide aid, which is assumed to be fungible and therefore under the control of politicians, to recipient governments in exchange for policy concessions. An emerging subnational literature on aid allocation appears to tell a similar story (Briggs 2017, Hodler & Raschky 2014). The political economy model is supported by considerable evidence. Donors also appear to be selective in their assessment of recipient characteristics, favoring countries with robust institutions and rule of law (Dollar & Levin 2006). Several studies contend that donors exhibit both motivations, pursuing strategic interests while meeting recipient needs (Claessens et al. 2009, Hoeffler & Outram 2011, Thiele et al. 2007). In any case, nonrandom allocation of foreign assistance appears to be the norm (Flores & Nooruddin 2009b).
Little is known about donor allocation behavior during war and peace processes. In a cross-national examination of aid allocation, Balla & Reinhardt (2008) claim that all donors condition their aid allocation on conflict; many allocate more aid to countries bordering conflict countries (Collier & Hoeffler 2004a). Other scholars contend that characteristics of the conflicts matter, e.g., wars fought over control of the government (as opposed to over territory) attract more postwar aid. The severity of the conflict, as gauged by the number of battle-related deaths, does not appear to affect aid allocation in broad terms (Kang & Meernik 2004). Areas with modest levels of conflict may actually attract aid, whereas those with the severe conflict discourage aid (Bezerra & Braithwaite 2016, Lis 2014). The pattern reflects a paradox: Donors increasingly call for aid to be allocated to fragile and conflict-affected areas (Wild et al. 2017), but less aid may go to the areas with higher violence due to security concerns (Fast 2014). This disparity affects the access and effectiveness of aid projects as well as the safety of aid providers. Moreover, donors may not even understand local contexts sufficiently well to address needs appropriately (Autesserre 2014).
For purposes of evaluating whether aid effectively builds peace, understanding allocation dynamics is important for at least two reasons. First, if aid is given for geopolitical reasons, then there is clear selection bias in who gets aid and how it could matter within countries. Second, if aid is fungible, as much of the allocation literature has assumed, then aid is not altogether different from other nontax revenues, an observation with implications for aid’s effects on conflict.
Dreher et al. (2016) contend that aid given by donors with clear strategic interests may be less effective than aid given otherwise. Strategic selectivity in allocation to conflict-affected areas likely influences its effectiveness and disrupts our ability to assess effectiveness. Indeed, because aid allocation is clearly nonrandom, the causal identification challenge is acute. Many other factors could correlate with aid allocation and thus confound any inferences made about the effects of aid.
Unfortunately, little research offers a clear causal identification strategy. Thus, we need to be cautious about what inferences we draw. One study found that once nonrandom selection of allocation is controlled for, there may be no relationship between World Bank aid and the recurrence of conflict (Flores & Nooruddin 2009b). Moreover, humanitarian aid is typically given to the most difficult places. A correlation analysis that fails to assess causality could therefore erroneously suggest that humanitarian aid prolongs war.
Much closer attention should be paid to foreign donor allocation behavior in conflict-affected areas. Particularly, we need to understand better (a) why aid is chosen and not some other foreign policy tool, (b) why a specific target country is chosen over another, (c) why specific locations or recipients within a target country are chosen over others, and (d) why specific types (for examples, sectors) of aid are chosen over others. Unpacking these decisions will require input from qualitative and mixed-method studies that can probe the allocation process with greater depth and specificity.
Some recent studies have pursued rigorous causal identification via research designs that rely on natural experiments or are experimental or quasi-experimental. Among these designs are controlled trials that randomize the implementation of projects (e.g., community-driven reconstruction) across a given set of units (e.g., villages). Other studies employ methodological techniques to achieve causal identification. One such technique is instrumental variable regression, which identifies plausible sources of exogenous variation (e.g., random rotation onto donor governing councils). Another technique is matching, which uses statistical algorithms to find control units that closely resemble units that receive aid, thereby facilitating causal comparisons (Carnegie & Marinov 2017, Crost et al. 2016, Flores & Nooruddin 2009b, Nielsen et al. 2011, Sexton 2016, Weintraub 2016, Wood & Sullivan 2015).
The causal identification challenge posed by nonrandom aid allocation stems not only from exogenous sources of variation but also from underlying theoretical assumptions. For example, if aid is allocated for geopolitical strategic reasons and used primarily to buy political influence, then politicians in the recipient country need relatively unfettered access to the aid resources, which is the foundation of the fungibility assumption. Once politicians exchange policy support, there is little or no accountability in terms of the development effectiveness of aid funds. In such a case, aid may not be altogether different from other nontax revenues, such as oil, and could perhaps be even worse (Djankov et al. 2008, Morrison 2012). Aid could be a special case of the resource curse and potentially carry with it many of the ill effects associated with resources such as oil (Ross 2004).
The fungibility assumption, and the political economy model more generally, have come into question. One reason is that aid may not be nearly as fungible as some scholars have suggested (Van de Walle & Mu 2007). Indeed, Altincekic & Bearce (2014) note that one of the key studies cited on fungibility (Feyzioglu et al. 1998) shows that aid is fungible only in a single sector and does not consider fungibility an intrinsic, universal trait. Furthermore, others have questioned the argument that aid is fully capturable by elites, demonstrating that donors have significant control over aid resources and appear to exercise it (Bermeo 2016, Dietrich 2013, Findley et al. 2017).
If aid allocation is nonrandom, and if aid is not purely fungible, then many of the assumptions, arguments, and inferences in the aid and conflict literature require further consideration. Indeed, these issues may be most fundamental if we are to reach accurate conclusions about the effects of aid on civil wars.
DIRECTIONS FOR FUTURE RESEARCH
Five topics should be elevated on the scholarly agenda: exploring the microfoundations of the process by which aid flows to recipient-country actors; considering the full range of foreign policy options and financial flows; theoretically innovating in the analysis of mediators and moderators of aid’s effects; more carefully engaging different levels of observation and aggregation; and improving measurement.
Microfoundations
The turn to subnational studies of the relationship between aid and conflict has yielded fruitful micro-level insights. Yet the field still needs to devote greater attention to developing and evaluating microfoundational theoretical assumptions. Indeed, most work on aid and conflict provides only a superficial treatment of the complex aid flow process, from donor-country preferences and resulting commitments through final disbursements to various recipient-country actors and locations. Conventional practice is to assume that aid is disbursed as committed and that recipients comply fully, then create an aggregated aid variable that is plugged into a country-year (or perhaps administrative unit–year) statistical model as if it were a single, homogeneous treatment administered consistently and reliably.9 An examination of the significant difference between commitments and disbursements (though this duality is itself an oversimplification) would underscore the need for more careful consideration. The conventional focus ignores important functional concerns such as how donor–government interactions influence aid flows. Even in micro-level interventions that consider potentially more tractable sets of activities, implementers vary in the allocation and delivery of aid services. The simplifying assumptions in most of the literature belie the extreme complexity and heterogeneity in the process of aid flows. Several key considerations follow.
In a provocative new book, Swedlund (2017) chronicles the fads and fashions in foreign aid delivery, highlighting the constantly shifting bargaining process. Her central contention is that most deals reached between donors and governments are not credible over the long run and thus eventually break down, resulting in ongoing revisions in aid flows. Both donors and recipient governments regularly negotiate and renegotiate the terms of aid arrangements, which helps explain the disconnect between commitments and disbursements. In the context of aid and conflict, this flux in arrangements has important indirect implications for potential rebels, given the potential for aid to affect the distribution of power. Moreover, donor organizational culture may be a powerful determinant of patterns of aid (Autesserre 2014, Campbell 2018), but its influence can be obscure to donors themselves, not to mention governments or rebels. Ultimately, aid is neither fully fungible, and thus subject to full control by recipients, nor strictly programmed, and thus subject to full control by donors. Theories emphasizing predation, state capacity, private information (within a bargaining approach), and strategic importance need to engage more seriously with the dynamic, negotiated nature of aid allocation.
Most studies of aid and conflict make assumptions about government and opposition preferences, information, and strategies, each of which warrant validation. Studies emphasizing wartime bargaining dynamics, specifically shifts in power through mechanisms such as aid shocks, assume at some level that rebels are aware of aid increases or decreases (or the possibility thereof). The aid delivery process may be opaque, however, raising the question of what the rebels actually understand.
A big obstacle is scholars’ lack of access to first-hand perspectives of opposition leaders. An ongoing study is seeking to address this gap by interviewing high-level ex-commanders of the Lord’s Resistance Army in Uganda, specifically probing what these rebel leaders understand and do not understand about aid dynamics (Findley & Backer 2017). The results reveal highly strategic behavior in the midst of substantial information shortfalls. In particular, leaders describe their decision-making calculus about how and when to loot different types of aid; their decisions were designed to induce uncertainty on the part of the government or aid workers, as well as within the local community. Thus, the interviews confirm theoretical approaches emphasizing rational, strategic behavior, but also reveal nuances of looting target choice, with important implications for theories that emphasize bargaining, predation, and winning hearts and minds.
The timescale on which aid is allocated and disbursed also needs greater attention if we are to construct appropriate claims that aid can influence government and/or opposition behavior. Timing is relevant to a variety of theoretical approaches, especially bargaining models. It is also germane to sorting out possible endogeneity stemming from anticipation effects: Governments or rebels may form expectations about future aid and try to anticipate its allocation, although this would necessitate the formation of expectations about specific characteristics of aid. Should aid change conflict propensities one or more years from the time of commitment? If donors attempt to condition aid commitments on future conflict, how far in advance are they trying to anticipate? Some studies assume that donors can anticipate future conflict and adjust aid accordingly (De Ree & Nillesen 2009, p. 5), whereas others cast doubt on this possibility (Nielsen et al. 2011). Which view is correct? In large measure, the broader aid effectiveness literature does not make clear arguments about the lag time between commitment and disbursement, let alone impact. The situation is no less muddy in the context of studies of aid and conflict. For most theoretical and empirical approaches, attention to time horizons of aid is needed to specify models properly and evaluate aid’s effects.
Foreign Policy Substitutability
Foreign policies of states consist of a suite of activities and aid is potentially one. States may provide military assistance, offer debt relief, send arms shipments, make loans, set up punitive sanctions, engage in trade, etc.—all in addition to providing foreign aid. States may also enable other financing, including foreign direct investment, remittances, and support from private foundations. Meanwhile, governments that receive aid also independently collect revenues and make (or withhold) resource transfers to their populations. A growing trend in the international political economy literature is to consider how different financial flows compete with or reinforce each other (Ahlquist 2006, Singer 2010, Wellhausen 2015). With few exceptions (Flores & Nooruddin 2011), scholars have not applied such analysis to the civil war context, especially when contemplating the influence of aid.
In absolute terms, foreign aid comprises a relatively small share of overall financial flows, despite its prominence in scholarly and public debates. Financial flows are notoriously difficult to track and likely prone to severe measurement error. That said, efforts to gauge these flows are worth comparative examination. According to the AidData Portal (http://www.aiddata.org, accessed April 18, 2017), states and international organizations since 1946 have given more than $7 trillion in international aid. Foreign direct investment inflows total $28.2 trillion, remittance inflows total $7.6 trillion, and US private foundation aid totals $12.2 billion. Sovereign lending also provides an increasing source of funds to developing countries, and debt rates are higher for conflict-affected countries relative to non-conflict affected countries (World Bank 2017). Moreover, many states give substantial military assistance abroad, though amounts are difficult to track globally because few incentives exist to report military assistance fully and accurately. Estimates suggest that the United States alone provided $848 billion in military assistance from 1946 through 2012 (USAID 2014). Arms transfers can be equally difficult to quantify but are no less consequential as a foreign policy tool. External financial flows aside, governments of developing countries extract and expend resources in support of various development activities. Reliable figures for government budgets/spending are not easy to estimate broadly, but from initiatives to publish government budgets/expenditures for developing countries (World Bank 2016), it is clear that governments engage in many development activities similar in function to international development. Section 3 of the Supplemental Appendix discusses other economic foreign policy options.
The emphasis on foreign aid is likely because this funding is easier to track than other financial flows. Some literature addresses other financial flows and foreign policy activities in isolation. To the extent that they correlate with aid and influence civil war, both of which are plausible, there may be significant challenges to inference about the effects of aid on civil war without proper theorization and rigorous causal identification. Inference is made more difficult by the likelihood of consequential selection effects related to the complex decision to use a foreign policy strategy, such as aid. A further layer of complexity is the distinction between allocation of aid and allocation of domestic government resources within the recipient country, especially since transfers of those resources are endogenous to the distribution of capabilities and will surely be affected by changes to that distribution.
Moderators and Mediators
Some of the most important insights into the connection between aid and civil war have come from directly theorizing and testing moderators and mediators through either relevant statistical approaches or more intricate measurement strategies. The effects of aid on civil war may hold in certain conditions but not others, and scholars have proposed some moderating contextual factors. In particular, aid is more effective at producing peace to the extent that it is distributed in areas of uncontested military control, securely controlled by government, and given at moderate levels for targeted activities (Berman et al. 2013, Sexton 2016). Aid can be more dangerous to the extent that it is spatially concentrated or diffused such that the government cannot provide sufficient security (Strandow et al. 2016), or when aid is fungible such that the government can use it to fight the war (Findley et al. 2011).
Arguably, aid influences some intermediate factor that, in turn, affects civil war. A prominent thesis is that aid helps win the hearts and minds of the people (Beath et al. 2016), which may make them more supportive of the government and more willing to share information with counterinsurgents or the government (Berman et al. 2011, Weintraub 2016). Aid may also work by altering domestic political constraints. In Colombia, for example, foreign aid provided resources that enabled the government to expand the war to FARC-controlled regions without needing to cut back on support for other domestic priorities. Other moderating factors, including regime type, may affect how nontax revenues affect the incentives for civil war (Paine 2016), and these factors could apply to foreign aid. And although the bargaining theory of war has largely dominated most explanations, over the past decade there has been renewed interest in ethnicity and civil war (Cederman 2011); issues of ethnicity may shape aid allocation and effectiveness dynamics, a topic that has not received much attention (Jablonski 2014).
The broader aid literature has already moved toward an examination of moderators and mediators. Indeed, much of the conventional wisdom on aid effectiveness can be traced directly to Burnside & Dollar (2000), who theorize about good governance’s moderation of the effects of foreign aid on economic growth. To chart new theoretical and empirical territory, the aid and conflict literature needs to take moderators and mediators much more seriously—probing, for example, what it means for aid to be secure, outside of US counterinsurgency aid cases. This question is central to the Aiding Resilience project led by Paul Huth and David Backer, with funding from the US Department of Defense’s Minerva Initiative. In addition, examining micro-level sectoral aid dynamics will likely be important for studying whether aid repairs economies after war and potentially insulates them against the recurrence of violence.
Observation and Aggregation
The relationship between aid and conflict has been studied at multiple levels. Most work originally focused on the cross-national level, examining aggregate patterns. In recent years, scholars have argued that both aid and conflict are fundamentally micro-level phenomena (Findley et al. 2011), and there has been a move toward subnational analyses (Parks et al. 2016). Using observational data, some studies analyze subnational dynamics across multiple countries (Bezerra & Braithwaite 2016, Wood & Sullivan 2015)10 while others examine subnational processes within single countries (Berman et al. 2011, 2013; Iyengar et al. 2011; Sexton 2016). Much of this single-country work considers a unique type of aid: counterinsurgency aid administered by the US military, mostly through the Commander’s Emergency Response Program. Key ideas emerging from these studies are (a) that aid projects should be smaller and more secure, improving delivery to the population that can, in turn, provide better information to counterinsurgents or the government, and (b) that aid may decrease violence only when allocated to uncontested areas. These studies advance the literature by utilizing data that are likely much more reliable than traditional development assistance in ways that potentially allow better causal identification. A critical follow-up question is whether counterinsurgency aid is fundamentally different from broader development assistance.
Another set of studies considers the role of specific aid interventions that cover a more limited set of participants in more confined, subnational areas (Beath et al. 2016; Crost et al. 2014, 2016; Khanna & Zimmermann 2017). These studies have found that community-driven development programs can increase violence, that conditional cash transfers can decrease violence, and that cognitive behavioral therapy coupled with cash transfers can reduce crime and violence. These studies may carry the highest internal validity but are perhaps least comparable with respect to earlier studies and offer the least external validity.
Which level is most appropriate, or can we learn from each one? The discipline’s turn toward better causal identification may privilege micro-level randomized program evaluations or other quasi-experimental methods, an emphasis that is sorely needed. The task of generalization is not straightforward, however, and caution is required in making broader inferences from such micro-level studies (Cartwright & Hardie 2012). Examples abound of studies using careful causal identification strategies but obtaining different results in different areas. Of note, community-driven development program success may vary by location and implementer (Avdeenko & Gilligan 2015; Casey et al. 2012; Fearon et al. 2009, 2015; Humphreys et al. 2015; King & Samii 2014). The proposal to implement small, secure, and informed development may not generalize (Berman et al. 2013, Sexton 2016), though the effects may depend on other contextual factors such as whether areas are under control of the counterinsurgency. Micro-level studies need to be repeated in multiple places with explicit attention to contextual factors that may lead to heterogeneous treatment effects. Strong causal identification strategies are not the exclusive domain of randomized interventions, of course, and studies at other levels have attempted to identify strategies (Carnegie & Marinov 2017, Nielsen et al. 2011).
Whereas micro-level interventions with strong identification strategies offer much appeal, the appropriate level of observation may be better served by following theoretical considerations. There is a substantial mismatch between micro-level program interventions and theories about government and rebel incentives to capture aid, or respond to shifts in aid. Indeed, empirical testing of such theories may require higher levels of aggregation, relying on the theoretical logic and accepting less “rigorous” identification strategies. Other theories, such as those emphasizing opportunity costs, lend themselves to tailored studies of micro-level interventions (Blattman et al. 2016).
The appropriate level of aggregation in studies may depend on the types of aid and process by which aid is allocated. For example, budget aid may be better for testing aggregate theories about state capture, whereas sectoral project-level aid may be more suitable for testing theories about localized aid. To date, few studies have disaggregated by aid sector (Duponchel et al. 2010, Nunnenkamp 2016), channel of delivery (Dietrich 2013), or other appropriate attributes.
Measurement
Despite progress in collecting and coding data on foreign aid, much of the data scholars would like to analyze are not systematically available. The International Aid Transparency Initiative (IATI) is making a valiant effort to develop and facilitate comprehensive aid reporting, but it has experienced mixed success across the donor community. Even if all donors sign on, the IATI standards will need to widen and deepen over time. These processes will be complicated, to say the least.
Official development assistance (ODA) was traditionally tracked by the OECD Creditor Reporting System, which included only traditional donors, defined as those belonging to the OECD. In the early 2000s, the AidData project began to add multilateral donors, which were not historically tracked by the OECD, and then to track development finance more generally. AidData has clearly advanced the tracking of development finance by improving data on sectoral aid, subnational locations, and nontraditional donors.
Yet measurement issues are pernicious, and they are not trivial. Perhaps most important, subnational analyses depend on fine-grained, location- and activity-specific information, which unfortunately cannot be disaggregated from most project records. Despite the promising turn toward subnational foreign aid dynamics, development donors still do not as a matter of course subnationally disaggregate aid projects by location, and project records do not allow scholars to do this disaggregation. Consequently, scholars cannot easily examine how subnational variation in aid to specific locations correlates with subnational patterns of violence. IATI is pressing for more subnational reporting, but donor compliance is a challenge.
CONCLUSION
Aid and civil wars are connected in consequential ways. Scholars and practitioners, using theoretical arguments and field-level data, have reached this conclusion. The empirical regularity alone is sufficiently compelling to establish it. But how is aid expected to affect civil war, and how would we go about identifying that effect?11
This article has not settled these questions because there is no simple argument that foreign aid should consistently ameliorate or exacerbate civil war. At each stage of a civil war, aid has been theorized to increase and decrease violence, potentially conditional on other factors. Numerous theoretical connections have been advanced, with little attention to sorting out which theories have more credible empirical support. A key next step is for scholars to develop testable implications that explicitly allow the integration of complementary theories or discrimination among more and less useful theories. For example, arguments for indirect effects, highlighting (a) winning hearts and minds and (b) redistribution, likely complement each other, but we have little understanding of how. Either one may reduce the likelihood of civil war. If redistribution is part of a hearts-and-minds strategy, does the likelihood decrease even further? Many dimensions of this relationship could be explored to refine our understanding. The set of testable implications is too large to enumerate here, and the next generation of studies should directly engage promising possibilities.
The scientific implications of this discussion are perhaps discouraging. Theoretical arguments that make direct, conditional, and indirect connections between aid and civil war rest on a challenging set of microfoundational assumptions and do not consider effectively the nonrandom allocation of aid. With the empirical complexities of data, measurement, and identification challenges, making accurate inferences feels dishearteningly difficult. The inferential challenges (from macro-level correlational studies) and problems of generalizability (from better-identified micro-level studies) are, of course, ubiquitous to all domains of scientific inquiry. Science is cumulative, and in the study of foreign aid and civil war, as in other topics, newer studies will inevitably use better methods and more precise data to challenge and refine existing studies. The collective efforts of the research and policy communities, if wisely executed, should lead to cumulative and integrative knowledge capable of benefitting developing societies.
The normative implications of this discussion are challenging to reconcile. Aid probably saves lives, but it probably contributes to suffering as well. Is aid a net positive or net negative? This welfare question is complicated, if not impossible, to analyze. Policy makers need answers, yet will continue making aid allocation decisions even without those answers. Scholars have not weighed in substantially as part of these deliberations about policy and programming but are increasingly equipped to do so. Policy makers, for their part, could consult the accumulating scientific research base, whether or not academics initiate engagement. Meanwhile, much hangs in the balance throughout the world.
disclosure statement
The author is not aware of any affiliations, memberships, funding, or financial holdings that might be perceived as affecting the objectivity of this review.
acknowledgments
I thank David Backer, Paul Bezerra, Andrew Boutton, Alex Braithwaite, Susanna Campbell, Michael Denly, Tom Flores, Desha Girod, Adam Harris, Paul Huth, Neil Narang, Irfan Nooruddin, Brad Parks, Peter Rudloff, Burcu Savun, Renard Sexton, Jacob Shapiro, Daniel Strandow, Haley Swedlund, Catherine Weaver, Michael Weintraub, Rachel Wellhausen, Reed Wood, Austin Wright, and Joseph Young for excellent comments, as well as Kyosuke Kikuta, Elizabeth Teare, and Megan Thomas for research assistance.
