The last decade has generated a robust study of the role of external interventions into civil wars. This literature builds on a rather small but influential foundation that at minimum pointed in the direction for a more systematic explorations for the conditions that lead to effective interventions. After a decade or more of research it seems appropriate to take stock of where we have gone, what our results tell us, and how we might further advance this important theoretical and policy issue. In this review I summarize and evaluate the large N, broadly cross-national studies on the conditions that lead to external interventions, and the conditions that lead to their success. I follow this with a discussion of how best to advance our understanding and move the research process forward.
Civil wars continue to provide a vexing problem for policy makers to grapple with. We know that on average there will be about two civil war onsets in each year, and this pattern has been reasonably robust over the past 60 years. From a policy perspective it is important to understand why groups take up arms against their states; theory and empirical evidence has contributed to our ability to understand these conditions1 and to forecast them2. A body of scholarship over the past ten years has attempted to weigh in on the related question of how external actors influence the course of civil wars. The implications of our understanding of the impact of external interventions have important political consequences, as former US Secretary of State Madeleine Albright’s recent NY Times editorial highlights.3
From both a theoretical and policy perspective understanding the conditions under which interventions work is important. Russia recently intervened in the renewed violence between Georgia and the breakaway region of South Ossetia. To what end? Albright lamented that nobody intervened in the Zimbabwean unrest, yet we do not have a sense of the consequences if there had been an intervention. From a broader theoretical perspective, interventions are international policies that sometimes violate international norms, sometimes support the continued oppression of people, and sometimes can bring armed violence to an end and provide for peace and stability. Where scholarship falls short is in articulating adequately where “sometimes” is “often”, or when facilitating peaceful settlements is more likely.
The evidence would suggest that interventions into civil wars remain a common feature of world politics, in spite of the former US Secretary’s complaint that leaders no longer have the stomach for such actions. According to her NYT editorial, state sovereignty rules the international system and the old days of interventions are behind us. The evidence will not bear out her argument, but it is important at this juncture to step back and evaluate what the past decade of systematic, broadly cross-national, research into the conditions for and consequences of external interventions has taught us, and why she might have it so wrong.
While conventional wisdom puts the beginning of outside interventions at the end of the Cold War, empirical evidence suggests that intervention policies have changed little since the end of WWII.4 And in spite of some early efforts to conceptually develop the ideas behind our understanding of interventions, the systematic study of outside interventions into internal conflicts did not take off until 1996.5 Two central questions of theoretical and policy relevance drive most of the research on interventions: 1) when to state leaders choose to intervene in internal conflicts; and 2) under what conditions do interventions work to achieve specific outcomes?
This article will provide a survey of what we know about interventions as a result of contemporary, largely cross-national, scholarship. What I leave out might be as important as that which I include. Two types of literature stand out: 1) the literature on UN Peacekeeping Operations, and 2) that adopting a more qualitative approach, often using single cases to describe specific interventions. The omission is not without logic, even though the insights from some of this literature could be fruitfully developed in studies of unilateral interventions.6 For example, one argument that is more fully developed in the peacekeeping literature involves the idea that peacekeepers are sent to the hardest cases. In effect, there is a process by which the peacekeeping data we observe are generated that would appear to be non-random.7 I will pick up this point more fully below, but suffice it to say at this juncture that logic points to a different selection process in unilateral, state-based interventions. My survey will emphasize the role of state interventions into ongoing civil wars, and will serve as a compliment to the recent survey on peacekeeping by Fortna and Howard.8 From this survey I will point to limitations and opportunities that derive from this collective body of research, areas where we seem to know something and those for which we remain largely in the dark. I end with a prescription for future research.
Such a survey is useful for a number of reasons, not the least of which is that over a decade of scholarship has addressed issues of when interventions take place and how interventions play out when they do take place. It is easy at a point like this – where studies accumulate but progress becomes increasingly halting – to step back and reflect on what we know and what we don’t, and to identify some of the pitfalls that have been skirted or avoided, pitfalls that must be overcome to ensure future progress. This should be the standard on which this review stands. I will work from a summary of the early literature on interventions toward a comprehensive overview of what contemporary scholarship contributes to our theoretical understanding of the role of external actors in the outcome of internal conflicts.
Three points of reference will motivate this review: 1) the outcomes from interventions, 2) the conditions under which states, or more directly, who intervenes and where, and 3) future directions that might help advance our understanding of the role of intervention in the dynamics of civil war. There has always been debate about the goals of interventions, whether they are to achieve geopolitical objectives or to advance humanitarian considerations such as bringing peace to the country. At this juncture we can only speak to the empirical patterns and leave research into the relationship between potential goals and specific outcomes for the future. At the core one could view decisions about who intervenes and into which conflicts from the perspective of realpolitik behaviors versus liberal concerns with cooperation and stability. How we think about outcomes and actors should influence the future direction of research efforts.
The Conceptual Beginnings
Roseneau9, Duner10, and Rasler11 provided some of the early foundations for the contemporary, and more systematic, study of external interventions. Roseneau’s early work posed the problem of how we can best understand the idea of an intervention as distinct from more “normal” forms of inter-state interactions. Everything is an intervention unless you set some parameters to identify those that are targeted at influencing an ongoing – or maybe yet to be started – conflict. His criterion involved a standard that focused on authority-targeted and convention breaking policies. The former requires that any intervention attempt to influence the authority structures, and we know from that one of the core concepts of civil wars is that they involve multiple centers of authority within one geographical entity.12 Targeting authority structures, whether it is to overthrow the ruling elite or to support the status quo, seems to be precisely the type of question that motivates policymakers. The convention breaking criterion reminded us that the interventions of interest are distinct from regular interactions. Recurring aid to a country at civil war does not necessarily constitute an intervention policy, even though cutting off that aid might. Roseneau never translated his conceptual arguments into empirical insights, but his definitional discussions were sufficiently compelling to form the foundation for numerous studies to follow.13
In a search for systematic analysis of the implementation and outcomes from external interventions into civil wars, Duner14 and Rasler15 stand out as early attempts to identify patterns in interstate behaviors. Rasler’s was a single-case longitudinal study of the Syrian intervention into the Lebanese civil war. Even though her study was not broadly cross-national, the cross-temporal variations in Syrian policy provided her with sufficient variation to draw interesting inferences. The implication of her analysis is that external military interventions can control internal violence if the intervener is willing to commit a sizeable military force. Early and reasonable weak Syrian interventions had little effect on the civil war, but a massive show of force helped bring the fighting to a halt. And while the argument is compelling, there are questions of how generalizable it may be, particularly given the US experience in Vietnam and what we are currently observing in Iraq and Afghanistan. At a minimum these are the types of questions that have immediate contemporary policy application, and reflect in many respects the types of intervention strategies at the heart of Madeleine Albright’s recent lamenting over the end of interventions.
Duner took a different early track to understand military interventions, focusing instead on the sequence by which interventions unfold. As he framed it, he was interested in the “anatomy of interventions”. The results from his sample of cases suggest that interventions tend to be grouped, sequenced, and escalatory. To put his results into the more contemporary framework adopted today, interventions tend to be part of a strategy that takes place in a strategic environment involving interveners and targets and that the sequence follows an identifiable logic.16 For at least a decade very little by way of systematic cross-national research followed up on these early efforts. Book-length treatments of interventions in specific civil wars provided most of the analysis17, and policy prescriptive discussions could be found in foreign policy journals18. The end of the Cold War seemed to move attention from the US—Soviet standoff into other questions of war and peace, and, importantly, interventions. Contemporary efforts to understand the determinants and consequences of external interventions build from these early foundations.
Contemporary Systematic Research
There are two clear dimensions to contemporary research on interventions that together form parts of the same puzzle, but in practical terms tend to be engaged separately. The first has a focus on the results from intervention efforts: do they work or not?; do civil wars last longer or are they bloodier with interventions than without? The second theme tends to focus on who is doing the intervening and which countries tend to be targets. For the most part these two dimensions are intellectually linked, and understanding outcomes requires some understanding of the countries, regions, and possibly types of conflicts involved. I will tie these two queries together at the end of this section.
Outcomes from Interventions
In 1996, as the intellectual opportunities that arose from the end of the Cold War began to take root, four articles took on the task of trying to understand the role that outside actors can play in determining the outcome of internal conflicts.19 The world had recently witnessed the genocide in Rwanda and the bloody breakup of Yugoslavia. In the Rwandan case international action was paltry at best and the call for the partition of Rwanda was heard in policy circles and editorials. Chaim Kauffman argued that partitions and population transfers were the only solution to conflicts rooted in such ethnic hatreds, and that to make them successful they would have to be enforced through military intervention. His article relied more on argument than evidence, but it certainly presented arguments that were testable. Sambanis subjected this argument to empirical verification and found that, in essence, partitions are not the answer.20
Patrick Regan21 asked the question of what types of interventions do work and under what conditions, with “work” being defined in terms of stopping the fighting. Using a new data set on military and economic interventions he tested whether different forms of intervention contributed to the end of the conflict. His results suggested that interventions that include a mix of military and economic levers would be more effective than either type alone, but that in general, conflicts with external interventions are less likely to end than those without. The latter inference may have been a result of a selection process whereby the most violent and longest running civil wars are the ones that attract interventions, but this qualification did not hold up to evidence.22 The success or failure of intervention policies, however, is not always defined in terms of conflict abatement, so the results and the theoretical argument required further refinement.
Mason and Fett23 used Correlates of War data to test arguments about the outcome of interventions. The COW definition of an intervention required the deployment of troops into the conflict zone, effectively setting the bar for what we consider an intervention quite high. They used an expected utility model of the outcome of civil wars to develop expectations, and they anticipated that external interventions would influence the outcomes of the war. Their results, however, did not support this expectation, and in fact in their models external interventions had no discernible impact on war outcomes. If the logic, that external interventions would buttress the recipient and make victory more likely, did not hold up to evidence, the problem was to be found in the crudeness of the indicators of interventions. So much of the logic in world politics revolves around the idea that power matters most, and when this does not stand up to evidence, at least in the case of civil wars, we have to look elsewhere. One of the results of their research was to call for more refined data on interventions, which up to that point consisted of data on interventions only when troops were inserted into a conflict zone.24 Mason and Fett’s work, however, pointed to the potential theoretical importance of relative power and in doing so highlighted the difficulty of making these relative judgments when one participate is not a state actor. We know how to measure the ‘power’ of a state but non-state actors posed more difficult problems; the resolve of rebel groups might temper a more standard measure of relative power. I will return to this point later.
Carment and James25 contributed to this early work by presenting a game theoretic treatment of the conditions or processes by which third parties intervene in civil wars. Their work helped lay the foundation for a number of later studies that develop strategic models of intervention.26 These strategic interaction models turn out to be quite important in our contemporary thinking about when and why states intervene, pointing to clearer theoretical refinements on when to expect interventions and when they are more likely to be effective.
At the core scholars and policy makers want to know what works and when. The trickiest part has been how to conceptualize the notion of what is meant by “works”, but this has stymied research for well over a generation.27 The most common outcomes that were explored involved either win, lose, or negotiations on the one hand, or the expected duration of the civil war on the other. Individual studies have picked up outcomes such as democratic transitions28, and genocides,29 but the core body of research has focused on how or when a civil war ends. Research into the expected duration of civil wars, given interventions, has been the most enduring in this genre of research, most of which comes to a very similar conclusion: interventions into ongoing civil wars are more likely to extend the length of the conflict.30 Regardless of the data that are used or the empirical specification, each of these studies finds that on average external interventions increase the expected duration of a civil war. The exception to this overarching result is by Collier, Hoeffler and Soderman, where they find that under certain conditions external interventions that support the rebels can decrease expected durations.31 They used Regan’s (2002) data but their sample was restricted to 1960, and they operationalize interventions as an increasing function of time. It is the latter change to the research designs of the other studies that presents both an innovation and a liability, though much more of an innovation. Effectively, by inflating the contribution of interventions for each unit of time that a conflict continues after the intervention, the intervention becomes more effective with each successive period. The innovative part of this is that they, for the first time, conceptualize interventions as something other than static behaviors, but the assumption of an inflator rather than a deflator, I think, is troubling. And since this is the only study to find results supporting an effect from interventions that decrease durations, the assumption deserves further review.
The question of “what happens to an intervention” or “when do interventions end” are important, though difficult ones to answer. In general the only interventions that have both a start and an end date are those in which troops from the intervening country are placed into the conflict zone. The Correlates of War requires this for its definition of an intervention, and studies by Balch-Lyndsay and colleagues32 adopt the COW definition and data. Interventions which are more refined in character, such as military hardware, intelligence, or monetary aid do not have a clear end date because the end point is when the form of aid or the piece of equipment becomes lost or obsolete. Since Collier et al. use Regan’s data, which includes various forms of aid as well as troops, it is virtually impossible to identify an end date for each intervention. By the functional form of their counting of interventions they make the assumption that a cache of weapons today becomes more effective over time, and as a result find that interventions on behalf of the rebels, 1960-1999 can shorten expected durations. The guns, however, may be lost, captured, or broken and in practice leaves the initial intervention degraded over the course of time. Their result demonstrating that interventions supporting the rebels decrease expected duration stands on its own, but it points to the importance of model specification. Regan and Aydin33 adopt the Collier et al. approach to thinking about the temporal impact of interventions and developed a deflator – rather than an inflator -- for each individual intervention. Using data that models a declining impact of interventions over time supports the general body of literature’s conclusion that interventions serve to extend the time that combatants fight.
One of the insights from recent scholarship is that civil war participants hold expectations about potential interventions, and that these prior expectations can influence the course of the war.34 Put differently, under some conditions rebels or governments plan on interventions and these expectations play into their strategies for prosecuting the war. For example, Akcinaroglu and Radziszewski frame this in terms of the effect of interstate rivalries on the duration of civil wars. If the state experiencing a civil war is involved in an interstate rivalry, the rebels fighting that state may expect to receive support from the rival country. In their model the geopolitical relationship among actors is an important determinant of the willingness and ability of rebels to press their case by force of arms, and their results point to the existence of a rivalry increasing the probability of observing an intervention supporting the rebels. However, the interventions themselves, as demonstrated by others, extend the duration of the conflict. Even with this more nuanced theoretical specification, their results support the general trend in duration models.
An important aspect of research into the effect of interventions on civil war duration is how robust the “negative” result is to model specification, research design, and data sources. It seems to matter little whether the requirement is troops on the ground or the provision of aid or arms, interventions generally extend durations, and countering – or as Balch-Lyndsay and Enterline35 frame it, balanced – interventions seem particularly prone to extending the duration of a civil war.
An alternative approach to understanding the outcome of interventions is to examine the influence of external support on the probability that we observe negotiations or victory by one side. It seems that many in the policy arena would view this as an implicit (if not explicit) outcome of interest, and that this more closely mirrors the questions asked by those who are responsible for implementing intervention policy. There is some convergence in this genre of research, even though not all studies come to the same conclusion. Mason, Weingarten and Fett36 use an expected utility framework to develop hypotheses posting that biased interventions, those that support one side in a conflict, will lead to an increased probability of that side winning and a decreased probability of a negotiated outcome. That is, victory should come to those who generate external support. Their evidence however, suggests that there is no significant relationship between biased support and the probability of victory, at least in the short term.
Rowland and Carment37 also use a game theoretic framework to develop the expectation that biased interventions are more likely to lead to an escalation of a civil war than its termination, in effect the exact opposite prediction from the Mason et al model. And although their evidence relies on a case study of the conflict in Kosovo, the implications seem to be quite convergent with the evidence presented by Mason and his colleagues.38 Moreover, this result garners some support from the work by Balch-Lyndsay, Enterline and Joyce in which they use a competing risk framework and find that biased interventions increase the time until the onset of negotiations, but a balanced, or offsetting, interventions can decrease the time until we observe negotiations. As the dictates of social science would have it, Gent articulates a formal theoretical model positing that biased interventions will decrease the time until we observe either rebel victory or negotiations.39 His model makes no prediction about the effect of an intervention supporting the government, and his results support this hypothesis that runs counter to the study by Balch-Lyndsay and colleagues. The most promising explanation for the differing results between Balch-Lyndsay et al, Mason and his colleagues, and Gent is their use of different data sets, which in turn require dramatically different standards for observing an intervention. One interpretation might be that interventions consisting of troops involved in the conflict decrease the prospects for an early victory or negotiations, but less intrusive interventions have a quite dramatic impact on the time it takes for either outcome to prevail. It would be possible to test for this differing outcome by selecting only Regan’s interventions that reach the level of the use of force by the intervener. If those results hold up then we would be on firmer ground thinking about the effect of differing magnitudes of intervention on negotiations or victory. To my knowledge this has not been done.
Cetinyan develops a formal model to explore the conditions under which external interventions will influence the course of civil wars, asking a somewhat broader question about the effect of interventions on probable civil war onsets.40 His model and subsequent analysis using the Minorities at Risk data demonstrate that external support does not affect the likelihood of the onset of a civil war (one of the only studies to explore this question), but it does influence the terms of a settlement, should one occur. Gershenson also develops a formal model of the role of external interventions, conceptualized in terms of sanctions, on the likelihood that the state engages rebels in negotiations.41 His analysis points to a double-edged sword, where weak sanctions against the state can hurt the rebel cause but strong sanctions can compel the state to engage rebel demands. Together these studies point toward a possible link between external interventions and the bargaining power of the rebels vis-à-vis the government they are fighting. This type of power-based argument would be consistent with a general understanding of the role of power and bargaining in world politics42, but subsequent research has not yet taken this line of enquiry further. It should. One of the primary difficulties is in deriving measures of rebel power when the sources of their power might be rooted in factors other than soldiers in arms. For example, rebel sanctuary across an international border43, ethnic affinities44, or the commitment of the rebels45.
These analyses rely on different model specifications and one of three data sets, each of which sets a different standard for conceptualizing an intervention. And since the bulk of the broadly cross-national studies also rely on similar conceptions of the expected outcomes of interventions, the results are comparable. In general, interventions do nothing to decrease the duration of a civil war46, they appear to be associated with more violence47, and it is somewhat unclear as to whether support for rebels will increase their chance for victory or decrease the time until negotiations start48.
Given that negotiations – at least as an outcome if not a form of intervention by way of mediation – should be so central to the conflict management aspects of interventions, it is remarkable how little systematic research there is on this topic. As described above, a few studies examine the role of military and/or economic interventions and the prospects for observing negotiations between the warring parties, but only a few recent articles examine the impact of externally driven negotiations, that is, mediation, on the outcome of the civil war49. If studies of bargaining, power, and resolve in interstate conflict have taught us anything it is that an expected payoff from a given choice is important for understanding outcomes, at least given the relative military capabilities of each side.50 Conceptualizing interventions as part of a strategic decision process should include the option of offering or engaging in mediation. Results adopting this frame of reference demonstrate that mediation is the only type of external intervention that is associated with shorter conflict durations, and the effect of mediation on duration is substantively huge compared to other forms of intervention.51 In effect, interventions that manipulate information about possible outcomes seem to have a considerably larger impact than those that only manipulate relative capabilities A likely explanation for why there is not a body of literature on the role of mediation in civil war outcomes is that data have not been widely available. Regan, Frank, and Aydin52 have addressed this problem by introducing a data set on external diplomatic interventions into civil wars that builds on the data used by Regan and Aydin (2006), while Greig and Regan53 test models of the offer to mediate and the conditions associated with its acceptance by the warring parties. Their results suggest that common interests drive the offer to mediate but conditions associated with the conflict are most influential in the acceptance of an offer.
As with the relative dearth of research on diplomatic forms of interventions, there are a number of other outcomes that could form the basis for understanding the impact of external interventions on internal conflict. Genocide prevention and democratic transitions are two outcomes for which interventions would seem to be tailor made. The outcry over the lack of action during the Rwandan genocide was grounded in an implicit assumption that an intervention could have worked to at least slow the tide of the murders. Krain provides possibly the only broadly cross-national analysis of the impact of interventions on genocides.54 His results suggest that if interventions were to target the perpetrators of the genocide, the severity of the violence could be reduced. But, he cautions, that neutral interventions into genocidal situations are largely ineffective. If Krain’s study were to hold up in further analysis, then the results could have important policy implications. Clearly the humanitarian motivation behind interventions has been the subject of considerable attention, but too few have addressed these questions in a broadly cross-national framework55, nor with a focus on what might be the gravest of humanitarian crises, genocide.
Bueno de Mesquita and Downs56 tested the idea that external interventions facilitate the transition to democracy. Working from expectations derived from a selectorate model they find that external interventions do not promote democratic transitions, or at best interventions contribute to a “facile” form of transition. In effect, interventions tend to restrict rather than facilitate democratic transitions in the targeted state. This general question has a precursor in work by Peceny.57 Building on a different body of literature Peceny shows that US interventions can promote democratic transitions, but only if they marry military interventions to the active promotion of elections within the targeted country. As with Krain’s research on genocide and intervention, two systematic studies of the effect of intervention on democratic transition are hardly enough to draw policy relevant inferences, particularly when the results tend to diverge in critical areas. However, the idea of promoting transitions toward democracy provides an integral link to Walter’s work on peace agreements that resolve civil wars.58 Her analysis demonstrates that fully implemented agreements require external guarantees in the form of interventions, and her notion of success requires a democratic transition, at least in the form of an election. One view of this is that you cannot generate peace until you remove the incompatibility, and elections provide one such mechanism for doing so. In the systematic and cross-national genre of research, this has been woefully underdeveloped.
Who Intervenes and Where
A critical part of any effort to model the outcome of interventions into civil wars is developing an understanding of who intervene and where. In the interstate war literature this generally shows up as models designed to account for selection processes, and in the realm of civil wars this is plainly evident in observations about UN interventions.59 At the level of intuition the process by which a United Nations intervention can be organized, approved and implemented requires that the conflict be large and destabilizing enough to bring it to the attention of the world community, and effectively, get the five permanent members of the Security Council to sign on to the intervention. Put differently, the hardest cases are most likely to be the ones that attract the attention of the UN, and these are presumably the most difficult cases in which an intervention can be successful.
The empirical evidence would suggest that many of the unilateral interventions are carried out by the major powers60, and conventional wisdom would posit that most of these are tied to the geopolitics of the Cold War. Recent research has begun to frame the decision to intervene in terms of strategic interactions, with implications for whether a civil war starts and the potential outcomes if an intervention takes place61. Few studies, however, link models of interventions to models of the outcome of intervention.62 Who intervenes is wound up in issues that transcend the geopolitics of the Cold War and is often conceptualized in terms of trans-border support for ethnic kin or insurgencies. Since many of the civil wars are not in neighboring countries to the world’s great powers, something else is pushing states to intervene. Khosla’s research suggest that neighbors are one of the biggest sources of interventions, and this result is consistent with the theoretical arguments presented by Davis and Moore.63 Carment et al.64 articulate a model of interventions which suggests that the twin factors of group affinity and cleavage drive intervention policies, and we know from others that cross-border ethnic kin can influence both65.
Empirical results support the notion that neighbors and major powers are the primary interveners, as are conditions of humanitarian crises, and that the decision to intervene is strategic.66 Gent argues and provides evidence to support his argument that the preferences of possible interveners structures who actually intervenes. Convergent preferences lead to free riding by allies, and as preferences diverge the incentive to intervene increases, presumably to the point that completely divergent preferences results in competing interventions. This would be consistent with evidence that major powers during the Cold War account for a considerable portion of the observed interventions. Regan’s results hold up under a selection model, and Thyne provides evidence that interventions extend expected durations when controlling for the selection into interventions.67 Although not framed explicitly as a selection model per se, Akcinaroglu and Radziszewski adopt a strategic framework involving the role of interstate rivalries to understand civil war durations, effectively using rivalries as a selection criterion.68 Their results suggest that interventions prolong civil wars, even when controlling for this strategic selection process. A game theoretic treatment by Rowland and Carment provides some of the theoretic logic behind these strategic calculations.69 Greig and Regan focus exclusively on intervention patterns with regard to mediation, with mediation being a form of diplomatic interventions.70 But any mediation that is offered must be accepted by the warring parties, so part of the puzzle revolves around which parties actually accept offers of mediation. Since mediation is voluntary and both (or all) parties must agree to participate, this provides for us a window on the role of intervention as part of a conflict management process. Presumably a party would not agree to participate in mediation if they did not conceive of an outcome that was potentially obtainable through negotiation that was preferred to their best expected outcome as a result of war. In this sense the observation of mediation provides at least a tacit understanding that there is a potential pathway out of war, even if find it is difficult. The results of their analysis demonstrate that prior relationships between the country at war and the party offering mediation has a considerable influence on who offers, but that it is not solely the strength of that relationship that determines when an offer is accepted. Conditions associated with the conflict and the reputation of the potential mediator are the dominant predictors of acceptance.
The idea that there are multiple pathways to intervention as well as multiple forms that an intervention can take leads directly to the concept that intervention policies reflect a substitutable set of options available to any potential actor. Why one state chooses to use military troops versus a less intrusive form such as a cache of weapons, or economic aid versus diplomatic negotiations remains elusive. But that such policy options are substitutable holds intuitive appeal when considered as a foreign policy choice.71 The concept of substitution has been used to understand the conditions under which the US chose one of four strategies with regard to intervening in a civil war: do nothing, military, economic, or diplomatic intervention.72 The strongest predictor of a change in US intervention policy was the amount of attention the conflict generated in the national media. Political party difference, trade, alliance, or other interveners were considerably weaker predictors of US policy than a measure of how visible the war is to the US public. These results only hold for the United States but are at least suggestive of earlier results that humanitarian events drive intervention policy. It might be naïve to think that geopolitics are unimportant but maybe equally simplistic to ignore the role of local politics.73 In effect this is a line of research that requires considerably more attention to understand how these mechanisms drive policy choices.
In the end we have a sense of who intervenes into which types of conflicts, but the few efforts to model outcomes as part of a structural process have not significantly changed the results derived from non-selection based research. One interpretation is that the strategic calculations of potential interveners are already worked into the models that account for outcomes. There is enough systematic evidence, however, to begin to tie things together in ways that can facilitate movement in the study of interventions; that is where I go next.