Identifying Pathways to Peace: How International Support Can Help Prevent Conflict Recurrence

This article provides new evidence on how the international community can effectively foster peace after civil war. It adopts a disaggregated, yet comprehensive view on international peacebuilding efforts, expanding the current literature’s narrow focus on peacekeeping or aggregated aid flows. We distinguish five areas of peacebuilding support (peacekeeping, non-military security support, support for politics and governance, for socio-economic development and for societal conflict transformation) and analyse which types or combinations are particularly effective and if this varies between different country contexts. Applying configurational analysis (QCA) to all 36 peace episodes after civil wars between 1990-2014, we find that 1) peacekeeping is only one important component of effective post-conflict support, 2) the largest share of peaceful cases is explained by support for politics and governance, 3) only combined international efforts across all types of support can address difficult contexts and 4) countries neglected by the international community are prone to experience conflict recurrence. The insights gained by the QCA are complemented with three illustrative case studies, based on 90 qualitative interviews. This article provides new insights on how different types of post-conflict engagement can increase a country’s chance for building peace after civil war. It thereby contributes to the academic debate on the effectiveness of international peacebuilding and at the same time can help inform policy makers engaged in post-conflict situations.


Introduction
Recurring violence continues to haunt many countries that have experienced civil war. As a consequence, pressing questions remain about how international support can help countries exit from the "conflict trap" (Collier et al. 2003). In particular, the debate about how best to design policy interventions in the face of conflict and fragility is far from concluded.
The current peacebuilding literature is in many regards ambiguous about the role specific elements of peace support play. Arguments have been made that peacebuilding without sufficient attention to security is likely to fail (Holm and Eide 2000), yet that extensive security support not embedded in domestic institutions may also be ill-guided (Ryan 2009). For some, economic support is key to success (Collier et al. 2003), while others claim that too much economic support impacts on a country in a similarly detrimental way as natural resource rents unless it comes with a broader strategy of political reforms (Bräutigam and Knack 2004). Political, and especially democratic, reforms, in turn, have been criticized to sow the seeds of unrest rather than help stabilise a country (Mansfield and Snyder 2008). In sum, the effect of most types of external peacebuilding support appears to depend on the combination with other elements of support or with certain context factors. Such combinations, however, are still underexplored.
Against this background, this article takes a comprehensive look at external support for post-conflict countries. It seeks to identify if patterns of international engagement exist that are systematically associated with sustained peace after civil war. For this we distinguish five main areas of post-conflict peacebuilding 1 : support for politics and governance; for socio-economic development; for societal conflict transformation; peacekeeping; and non-military support for security. Qualitative Comparative Analysis (QCA) helps us analyse patterns in 36 cases of international peacebuilding support. Finally, the findings are discussed drawing on three cases: Cambodia, Indonesia and Liberia. We use 90 interviews with domestic and international experts to shed light on causal mechanisms at play in the pathways identified by the configurational analysis and to better understand contextual factors determining each pathway.
Four findings stand out: first, peacekeeping is only one important component of effective postconflict support. Second, although democratization might sometimes have conflict-inducing effects, international engagement aimed at building political institutions, democracy and governance can in fact contribute to peace after civil war. Third, only a combination of all types of support explains sustained peace also in cases with a higher risk of conflict recurrence. Fourth, and more generally, we find that international support clearly matters: no case without any substantial external support avoided civil war recurrence.
We contribute to the literature on peacebuilding in two regards. First, compared to previous research our analysis is more comprehensive because we take the whole range of peacebuilding activities into account in an encompassing, yet disaggregated approach. Second, by applying a configurational analysis we are able to identify whether different types and combinations of peacebuilding support are particularly beneficial and whether this varies between different country contexts. 1 Following former UN Secretary General Boutros Boutros-Ghali (1992, para 21), we refer to peacebuilding support as external activities aimed at preventing the recurrence of violence in post-conflict societies.

Peacebuilding after civil war
Although international actors regularly become active in post-conflict countries in multiple ways -by sending peacekeeping troops, providing ODA or engaging in mediation, for example -systematic knowledge about international impact on post-conflict peace remains limited (Findley 2018;Hoeffler 2014). Instead, cross-country studies that analyse what factors influence post-conflict peace mostly focus on domestic rather than international aspects (Collier, Hoeffler, and Söderbom 2008;Walter 2004;Quinn, Mason, and Gurses 2007;Flores and Nooruddin 2012;Toft 2009). While these studies do sometimes include aid (measured as ODA flows) to account for international dynamics, they do not analyse them in detail and come to diverging conclusions (Collier, Hoeffler, and Söderbom 2008;Collier and Hoeffler 2004a;Flores and Nooruddin 2009). Scholars that look at international engagement more closely tend to focus on specific types of activities and disagree whether international actors harm or strengthen peace.
On the one hand, the existing literature provides convincing evidence that international support can make an important difference. The most consistent evidence concerns the military dimension of postconflict support: peacekeeping troops can effectively reduce the risk of conflict recurrence (Doyle and Sambanis 2000;Fortna 2004;Collier, Hoeffler, and Söderbom 2008;Mason et al. 2011;Doyle and Sambanis 2006). Regarding another important aspect of peacebuilding, Savun and Tirone (2011) show that democracy support can reduce democratising countries' risk of experiencing internal strife; and evidence from comparative case studies indicates that even in post-conflict situations, democracy support can effectively foster peace (Mross 2019). Furthermore, Matanock (2017) and Smidt (2016) show that international election monitoring, an important component of support for politics and governance in post-conflict countries, can stabilise elections. Next to specific types of support, Fiedler et al. (2020) show that how peacebuilding aid is provided matters for its effectiveness.
On the other hand, some scholars are highly critical of international peace support. There is an extensive and vibrant -albeit mainly theoretical -debate amongst critical peacebuilding scholars on the appropriateness and effectiveness of, as well as inconsistencies and problematic power-relations embedded in, international interventions to support peace (Richmond and Mac Ginty 2013;Autesserre 2014;Chandler 2017). A key debate in this strand of literature has critically engaged with the "liberal peacebuilding" paradigm, which pushes post-conflict countries to democratise and hold elections early on (Paris 2004). This is said to endanger peace and several authors empirically show that democratisation more generally and elections in particular can be associated with renewed conflict (Paris 2004;Mansfield and Snyder 2002;Cederman, Hug, and Krebs 2010). Furthermore, some anecdotal and statistical evidence exists that humanitarian aid as well as aid shocks can under specific circumstances fuel conflict onset (Nielsen et al. 2011;Anderson 1999).
The empirical literature so far either analyses international aid in post-conflict contexts as one abstract aggregate figure or focuses on specific subtypes, leaving an important research gap unaddressed: it fails to systematically examine the relationships between and across the different sectors. One of the most influential studies in the peacekeeping literature suggests that multidimensional peacekeeping, which combines troop deployment with developmental elements of peacebuilding support, has so far proven to be the most effective approach in promoting peace (Doyle and Sambanis 2000). This suggests that broader strategies than just troops are needed if the international community wants to impact on peace in post-conflict countries. However, research has not analysed what broader strategies of combined support should look like. In this vain, Hoeffler (2014, 75) concluded in a literature review that "our knowledge about the optimal policy mix of economic, diplomatic and military interventions is still limited". This paper aims to fill this gap by not only disaggregating ODA and studying different areas of engagement alongside peacekeeping troops -security, politics and governance, socio-economic development and societal conflict transformation -but also analysing their interplay.
Theoretical framework -how can peacebuilding support contribute to peace?
Building peace in a society that experienced civil war is ultimately a domestic process. That said, peacebuilding can be influenced by external actors. Although some authors argue that countries should rather be left to their own fate, even at the risk of state failure (Herbst 2004;Weinstein 2005), or that external interventions have often been ineffective if not harmful (Paris 2004), convincing evidence suggests that peace can indeed be furthered from outside by setting incentives and supporting domestic actors and institutions (Zürcher et al. 2013;Fiedler 2018;Fortna 2004). How and why such support can theoretically help to build peace, and under which conditions it may be less likely to succeed, can be understood best when we disaggregate peacebuilding support and look more closely into the logic of impact of the different types of international engagement. To do so, we conceptually distinguish five areas of international peacebuilding: peacekeeping, security ODA, politics and governance, socio-economic development and societal conflict transformation (Barnett et al. 2007;Smith 2004). This section first discusses why and how we expect that international support in the different areas affects sustainable peace and second why combinations might be particularly effective.
Supporting the reestablishment of security through the deployment of peacekeeping forces is perhaps the most obvious pillar of international efforts to prevent the recurrence of civil war. Empirical research has consistently shown UN peacekeeping to be "remarkably effective at bringing peace" (Walter, Howard, and Fortna 2020). Why exactly this is the case is less clear. In her seminal 2008 book, Fortna proposed four main mechanisms: 1) changing the cost-benefit calculations of warfare, 2) reducing the credible commitment problem, 3) preventing an accidental return to the battlefield, and 4) averting political exclusion. However, the evidence on which mechanisms can best explain why and how peacekeeping after civil war works is still inconclusive (Walter, Howard, and Fortna 2020).
Besides direct provision of security by military means, non-military security support ("security ODA") can also strenghten peace in post-conflict contexts. The main logic behind this type of support is to re-establish and consolidate the state's monopoly of violence pursuing the twin goal of providing effective public security and strengthening state legitimacy (Toft 2009;Krebs and Licklider 2015;Ansorg and Gordon 2019). Activities such as disarmament, demobilisation and reintegration (DDR) programmes for former rebels, small arms control or humanitarian demining aim to create a sense of security and allow a return to normal life. This should revitalise social and economic activities and strengthen trust in the post-conflict regime. Moreover, security sector reforms (SSR) that focus on professionalising the state security forces and establishing democratic control over them are supposed to foster state legitimacy and provide 'institutional guarantees of non-recurrence' by preventing authoritarian abuse of the security forces. Support for politics and governance is another central component of international peacebuilding. Here the underlying logic is that functioning and legitimate state institutions are key for overcoming conflict. Theoretically, functioning democratic institutions are supposed to provide non-violent channels to express, and deal with, competing interests and grievances (Hegre et al. 2001;Call 2012). Engagement in this area typically includes assistance to elections, constitution-writing, technical state capacity, the rule of law, human rights and civil society, which is supposed to increase the capacity, independence and legitimacy of state institutions so that they can better serve as mechanisms of conflict management.
We conceptualize the fourth area of support as "societal conflict transformation" (SCT). Such aid, aimed to mend the social fabric, is based on the assumption that it is crucial to actively deal with the violent past of a civil war in order to build a peaceful future. By helping to reveal and actively deal with atrocities committed during the civil war and their legacy in society, such support is supposed to build trust and generate legitimacy for the new order (Skaar, Malca, and Eide 2015;Loyle and Appel 2017;Samset, Petersen, and Wang 2007). Activities for SCT are supposed to strengthen peace based on various theoretical mechanisms: revealing the truth (uncovering atrocities committed during the conflict), providing justice (holding perpetrators accountable), fostering reconciliation (healing wounds of the past and overcoming grievances within society) and victim restitution (compensating individuals for harms and losses) (Skaar, Malca, and Eide 2015).
Finally, socio-economic development is another area of engagement that can theoretically strenghten peace, in this case by improving the living conditions of the population, which should alleviate grievances, give the population a stake in peace and hence reduce the risk of arms being taken up again (Del Castillo 2008). Activities to this end include, for example, the reconstruction of infrastructure and the provision of basic services. Moreover, service delivery and more long-term policies for economic development are meant to help restore confidence in state institutions (Del Castillo 2008).
Against this background, we theorise that international support in each of the areas could contribute to strengthinging peace. However, this does not mean neglecting potentially negative effects of international engagment. While there is little empirical evidence so far that international support fuels conflict recurrence, individual analyses of ineffectual post-conflict aid are common and negative effects theoretically possible. The critical peacebuilding debate for example contends that donors' push for elections in post-conflict countries had destabilising effects (Paris 2004;Chandler 2017). Transitional justice, a core component of what donors support through societal conflict transformation, was long marked by a (albeit mainly theoretical) debate on whether some types of transitional justice might spark rather than reduce conflict (Snyder and Vinjamuri 2003). Regarding a third core component of international peacebuilding, Autesserre (2014), for example, argues for the DRC that attempts to establish state security backfired and led to greater insecurity instead. Still, most empirical evidence does not clearly suggest negative effects but rather points toward some positive (e.g. peacekeeping) and some inconclusive (e.g. ODA) findings.
While international engagement in each of the five areas of peacebuilding might be important and in some contexts one type alone might effectively foster peace, it is also clear that important interdependencies exist between them. Support for SCT and for politics and governance, for example, is closely connected because addressing war crimes and strengthening the rule of law more generally can reinforce each other. In a similar vain, the success of DDR programmes, one main component of security ODA, strongly depends on socio-economic development because former combatants need jobs. Likewise, reparation programs, a cornerstone of SCT, are more easily funded from a state budget when the economy grows.
Combining different types of engagement can also be important when support in one area might reduce risks entailed in processes targeted by another type of support and thus ensure that the net benefits for peace prevail. For example, to avoid conflict-inducing effects of 'dealing with the past', it might be important to combine support for SCT with efforts to create institutional 'guarantees of non-recurrence' by supporting SSR and DDR or strenghening democratic processes. Furthermore, linking support for socio-economic development with governance reforms might be advisable to avoid the "resource curse" effect of aid referred to by Bräutigam and Knack (2004). Similarly, combining support for politics and governance with security-related support might be worthwhile. Peacekeeping troops accompanying electoral support, for example, can help decrease the risk of destabilizing elections by acting as security guarantors. Similarly, demobilization support can also reduce the ability of warring parties to return to warfare if unsatisfied with election results. Indeed, evidence shows that the presence of peacekeeing troops or DDR processes starting before the first post-conflict elections significantly reduces the risk of civil war recurrence (Brancati and Snyder 2012;Joshi, Melander, and Quinn 2017).
In sum, we contribute to the debate on determinants of post-conflict peace by taking a more nuanced, systematic look at the combined effect of international engagement. Naturally, many domestic factors, such as economic development or characteristics of the previous conflict, influence whether or not conflict recurs -these lie at the heart of empirical research on post-conflict societies (Walter 2004;Quinn, Mason, and Gurses 2007;Collier, Hoeffler, and Söderbom 2008). Domestic factors are crucial and the effect of external peacebuilding is contingent upon them. For example, focusing international efforts on mainly one type of support might work in contexts with a relatively lower risk of recurrence, whereas more difficult contexts might call for international engagement in several areas of peacebuilding. However, research has so far not analysed whether particular combinations of support can help sustain peace and what role the post-conflict context plays here. To answer these questions the following analysis takes all five areas of peacebuilding into account and systematically investigates their individual and combined effect on peace after civil war.

Research design, method and empirical strategy
To identify effective international strategies of peace support, we use Qualitative Comparative Analysis (Ragin 2008(Ragin , 1987Schneider and Wagemann 2012). QCA is ideally suited for our purpose because it focuses on the effect of combinations of conditions; it allows identifying diverging paths to the same outcome, with each path characterised by a specific configuration of conditions; and its set-theoretic logic is in line with our research interest: we aim to better understand if substantial amounts of certain types and combinations of external support are sufficient to sustain peace. An additional advantage for our analysis is the fact that QCA works well with relatively small numbers of cases, such as the 36 post-conflict episodes in our population. For recent high-quality applications of this method see (Mello 2020;Ide et al. 2020).
The logic of QCA is based on set memberships. In its dichotomous ('crisp-set') version, QCA variables (the conditions and the outcome) will adopt the values of 1 (member) or 0 (non-member). We use the fuzzy-set version, which indicates the degree of each case's membership in a set. The calibration process transforms observational data into fuzzy-sets ranging from 0 to 1, for which defining the cross-over point at 0.5 is key, which separates members of a set from non-members. A central element of QCA is the truth table. Each row in the truth table represents one possible combination of the conditions, and cases are assigned to the row that represents them best. Based on the empirical data, the truth table also indicates if the configuration represented in a row can be considered sufficient for the selected outcome. 2 Based on the data at hand, algorithm-based logical minimisation reduces the combinations (rows) considered sufficient for the outcome to the simplest possible solution term without violating any of the statements made by them.
Different types of solutions can be reached depending on the treatment of logical remainders (theoretically possible but empirically not observed combinations). Since we have strong theoretical assumptions, we interpret the intermediate solution, which is not purely guided by parsimony (as the parsimonious solution) but takes theoretical knowledge into account in the minimization process where empirical data is lacking. By conducting the Enhanced Standard Analysis (ESA) we ensure that the minimization does not rely on untenable assumptions. 3 Two main parameters of fit inform the interpretation of QCA results: scores of consistency and of coverage. 4 The consistency score indicates to which degree the empirical data is in line with (or deviates from) a result. The coverage score provides information on the share of the outcome that is explained by the results. 5 Both scores run between 0 (worst) and 1 (perfect).
The benefits of QCA come at a cost. As the method takes into account all theoretically possible combinations of the factors under examination, each condition added to the analysis increases the number of possible combinations exponentially. In reality, not all combinations may be matched by empirical cases (the situation of 'limited diversity'). Since limited diversity becomes problematic when too many potential combinations are not matched with data, it is only possible to include a limited number of conditions (typically four to six) into one analysis, depending on the number of cases at hand. As Mello (2021, 27) points out, the balance to strike in QCA is to "keep the number of conditions small, while allowing for enough complexity to investigate various configurations of relevant conditions". Hence, it is crucial to carefully select the conditions that are included. 2 In this process, the raw consistency threshold set by the researcher specifies the level of inconsistency allowed in considering a combination of conditions (i.e. truth table row) sufficient for the outcome, which can be guided by a gap between rows with higher and lower consistency scores (see Schneider and Wagemann 2012). 3 We do this by testing for simultaneous subset relations, contradictory simplifying assumptions or necessary conditions. See Schneider and Wagemann (2012) on the Enhanced Standard Analysis and different solutions types. 4 We also consider the Proportional Reduction in Inconsistency (PRI) score, which indicates how far a condition might be considered simultaneously sufficient for the outcome and its negation (simultaneous subset relations).
As stated above, the main result of a QCA computation is the identification of alternative "pathways" that lead to the outcome under investigation and are each characterised by a specific configuration of conditions. While an identified configuration suggests a causal relationship, it does not specify the causal mechanism at play (why does a certain path lead to an outcome of interest?) nor can it identify common characteristics of cases covered by the same path outside the limited number of conditions entered into the analysis. For this reason, we employ two additional strategies. First, case studies serve to support causal interpretations (see Goertz 2017, 59;Schneider and Rohlfing 2016). While a full process tracing to demonstrate causality is beyond the scope of this paper, we use qualitative insights from typical cases to check the plausibility of the causal relationships suggested by the different paths. Second, our interpretation of results takes into account a number of additional context factors that might help explain why a certain configuration of peacebuilding support works for the cases covered by that particular path.

Defining population and outcome
We identify post-conflict peace episodes using the UCDP/PRIO dataset 6 on armed, state-based conflict. We consider a country involved in several armed conflicts as one system in conflict, which is why our analysis takes countries that emerged from conflict -rather than conflict dyads that endedas units of analysis.
Because we are interested in the effects of international support after civil war, we only include cases where violence was particularly severe. In a slight deviation from a common practice in civil war research, we posit that all conflicts combined must have led to at least 1,000 battle deaths in any two consecutive years during the conflict period. This threshold broadens the number of cases eligible for inclusion and, inter alia, allows including cases that barely missed the 1,000 threshold in one calendar year. Finally, we consider a civil war "over" when a country has experienced a minimum of one calendar year without any armed conflict on its territory. Our period of observation includes all peace episodes that began in 1990 or later, because it was only in the 1990s that international peacebuilding emerged as a regular practice. Based on the above criteria, we arrive at a list of 36 peace episodes in 28 different countries. 7 Consistent with our definition of post-conflict episodes, we define peace as the absence of major armed conflict after civil war. In order to explain 'sustained peace', we add the notion of longevity and consider peace as sustained if peace lasted at least for five years. Clearly, the intensity of renewed violence also matters. Therefore, we create a fuzzy-set that ranges from entirely peaceful to full recurrence of civil war, considering post-conflict countries that experienced renewed minor conflict (below 100 battle deaths per year) still as relatively peaceful. 8 6 Version. 4-2016. 7 See Appendix 1 . 8 We use the direct method of calibration. See Appendix 1.

Measuring international support
We conceptualize four of our conditions, namely substantial security ODA, substantial international support for politics and governance, for socio-economic development, and for societal conflict transformation, as official development assistance given to these purposes. We approximate substantial support in these areas through financial contributions provided by OECD/DAC donors and multilateral organizations using the AidData database and its purpose codes. Support for politics and governance is well captured by AidData's purpose code 151 "Government and Civil Society", which includes topics such as elections, civil society, legal and judicial development or government administration. Developmental activities connected to security (security ODA), including DDR and SSR efforts, are measured using AidData's purpose code 152 "Conflict prevention and resolution, peace and security". The indicator for support for socio-economic development covers topics such as education, health, social services and infrastructure.
In the case of societal conflict transformation (SCT) additional coding efforts were required as no code exists to capture this area of engagement. We compiled the first dataset of projects in this area 9 , using project information provided by AidData in a two-step coding process: First, we ran an automatized pre-selection based on a list of search terms that capture one or more of the central elements of SCT -Truth, Justice, Victim Restitution and Reconciliation. Second, we identified SCT projects through hand-coding based on project descriptions.
We use ODA commitments for these four conditions because the quality of data on disbursements is poor, especially in the earlier periods. This comes with the caveat that we thereby capture all planned activities rather than only those that were implemented. Nevertheless, commitments disclose what developmental activities donors and recipient countries jointly agreed upon and constitute a good approximation of international engagement across the different issue areas. Since the same volume of aid is likely to have a different effect depending on the size of a country, ODA commitments are measured per capita.
Our fifth condition -peacekeeping support -is measured through the number of military personnel deployed for activities relating to the internal conflict. 10 In all conditions we measure support in the first up to five post-conflict years. For cases that experienced recurrence after less than five years, only the peaceful years were taken to calculate the average amount of support provided. Since no strong theoretical basis exists for when support should be considered substantial, we used evident gaps between the amounts received (ODA or peacekeeping troops) by the countries, combined with case knowledge, to differentiate between substantial and non-substantial support. 11 Table 1 summarizes the conditions capturing international engagement, data sources as well as the calibration thresholds chosen. 9 Muck and Wiebelhaus-Brahm (2016) focus on trials and truth commissions only. 10 based on data from SIPRI, the International Military Intervention Dataset and mission homepages. 11 Evident gaps were identified based on tables displaying the amount of support the different countries received per capita in the first five post-conflict years (see Appendix 1). The 0.5 anchor was set where a relatively big gap exists in the data at a level of support that can plausibly differentiate cases that are 'more in than out' of the set of substantial support from those that are 'more out than in'. We then plotted the calibration pattern across all cases to verify that

Measuring a country's predisposition for conflict
Approximating a case's 'predisposition' for conflict recurrence is important for two reasons: 1) to take into account local and international factors besides peacebuilding support known to influence civil war recurrence and 2) to account for the fact that cases with a particularly high risk might require a different engagement, or defy any efforts by the international community.
We measure a country's predisposition for conflict based on six factors the quantitative literature has relatively consistently identified to increase a country's risk of civil war onset or recurrence (Collier and Hoeffler 2004b;Doyle and Sambanis 2006;Quinn, Mason, and Gurses 2007;Collier, Hoeffler, and Söderbom 2008;Walter 2004) 12 : the presence of multiple factions in the previous war, a less intense previous war, a short previous war, low-economic development, natural resource dependency and conflicts in neighbouring countries. To control for a particularly challenging context for international engagement, we consider cases that exhibit four or more of these factors to be at a high risk of civil war recurrence. For more information on the rationale behind each factor, the concrete measurement and data sources used for each variable see Appendix 4.
the distribution is not distorted. Details on the calibration for all cases and conditions as well as scatter plots of each fuzzy-set against its base variable are rendered in Appendices 1 and 2.

Findings
Analysing 36 episodes of countries coming out of major civil wars after 1990, we use fuzzy-set QCA to search for patterns of international support that explain peace. 13 Figure 1 summarizes the essence of our findings: In contexts with a comparatively low risk of recurrence (~PRED), either peacekeeping (PK) or non-military security support (SC) or support for politics and governance (PG) are key to sustain peace, whereas high-risk contexts (PRED) require a comprehensive effort encompassing all five areas of peacebuilding. In turn, all cases that did not receive any substantial support experienced conflict recurrence. In the following, we present these findings in more detail. We start with an analysis of necessity -as it is good practice in QCA, even if we are primarily interested in the relationship of sufficiency. However, none of our conditions or their combinations can be considered necessary for peace or conflict recurrence (see Appendix 6). For the analysis of sufficiency, we set the threshold for inclusion of truth table rows (see Table 2) as sufficient at 0.75, based on a gap in the data. 14 Following our theoretical framework, we include two directional expectations to derive the intermediate solution: We expect that a high predisposition is negative for peace, and peacekeeping troops are conducive to peace.  Our analysis reveals four alternative paths to peace (see table 3). The first three paths describe countries with a low predisposition for conflict recurrence. In the first, we see a presence of PK and absence of SC as well as SCT, in the second the presence of SC and absence of SE and SCT and in the third the presence of PG and absence of SCT. The last path, which also covers countries with a high predisposition, displays a combination of all types of international support.
The solution is consistent (with an overall consistency score of 0.86) -as are the individual paths. The solution explains 14 of the 16 cases that remained peaceful. 15 There are no contradictory cases, that is, cases which are represented by a path but do not share the outcome. 16 15 Azerbaijan and Nepal are not explained. 16 Also called 'deviant cases for consistency in kind'. Yet, two cases -Tajikistan and Peru -deviate in degree: although representing the respective paths relatively well, they did not remain entirely peaceful. Solution ~PRED*PK*~SC*~SCT + ~PRED*SC*~SE*~SCT + ~PRED*PG*~SCT + PK*SC*SE*PG*SCT => PEACE 0.86 0.63 Note: Full circles mean a condition was present in all cases, empty circles (as well as ~ in the solution term) denote it was absent. Empty cells indicate that a particular condition is not a necessary part of the path because the condition was present in some of the cases covered by this path but absent in others. PRED: High conflict predisposition, PK: Peacekeeping, SC: Security ODA, SE: Socio-economic development, PG: Politics and governance, SCT: Societal Conflict Transformation. For countries with multiple peace episodes, a number attached to the country name denotes the start year of the respective episode. Cases in bold are "uniquely covered cases", i.e., they are explained by one path only.
The focus of our analysis is on explaining sustained peace. Examining those cases that did not remain peaceful, however, helps address the notion that international support might also be harmful. Table  4 provides the intermediate solution for conflict recurrence. The results back our general finding that international support can foster peace.  The first two paths cover a large (overlapping) number of cases, and both display different combinations of absences, suggesting that the absence of donor support contributes to conflict recurrence. Strikingly, six of these eleven cases did not receive any substantial support across the five areas of engagement and all of them experienced civil war recurrence. 17 The pattern is confirmed by four recurrence cases not included in the solution paths: Yemen, Iraq, Uganda, and Ethiopia. Characterized by a lower predisposition, they, too lacked substantial international support in all five areas. Overall, the results therefore lead us to conclude that post-civil war countries neglected by the international community can hardly escape the conflict trap.
The latter three paths to recurrence are notable by the presence of some forms of support (PG, SCT, SE) in combination with the absence of others (PK, SC, again SE). From the mere QCA solution it is not possible to discern if and to what extent these configurations represent causal relationships. Theoretically, the support provided in the cases covered by paths R3 -R5 could have contributed to peace breaking down. In the context of all our other findings, however, a different reading is more plausible. These paths cover only cases with a high predisposition of recurrence. Keeping in mind from our previous analysis that in high-predisposition cases lasting peace can only be found where encompassing support was provided, it seems more appropriate to conclude that, if anything, it is the absence of additional forms of support rather than the presence of some elements that contributed to recurrence. This interpretation is also in line with paths R1 and R2, where absences are the defining features. Nonetheless, future research engaging in thorough case study analyses of recurrence episodes included in paths R3 -R5 could possibly shed additional light on the role international support, or the lack thereof, played for the breakdown of post-war peace.
Another notable pattern stands out when bringing the two analyses together: three paths to peace include the absence of support to SCT whilst two paths to recurrence entail the presence of SCT support. This is interesting from a theoretical perspective because there is a critical debate on the potentially conflict-inducing effects of dealing with the past through transitional justice, one main element of this type of support. However, extensive robustness checks show that these absences of SCT in the pathways to peace are the least robust components of the analysis, calling into question the weight they should receive in the interpretation. Furthermore, the case studies on the respective paths discussed in the next section did not reveal that the absence of SCT was important in Cambodia or Indonesia to explain why peace prevailed. Further qualitative research would be needed to analyse whether the presence of SCT combined with PG support as described in paths R3 and R4 did contribute to recurrence or whether it was rather, as we argue above, the absence of a more encompassing approach that can explain this result. Overall, we therefore do not interpret these results as demonstrating that supporting SCT endangers peace.
The results hold against a wide range of robustness tests (presented in Appendix 9). Going beyond those robustness tests recommended as standards of good practice for QCA (Schneider and Wagemann 2012;Skaaning 2011), we perform five types of robustness tests by altering 1) calibration & consistency thresholds, 2) case selection, 3) operationalization of the outcome (e.g. using a qualitative measure), and 4) model specifications (e.g. changing the period of analysis or frequency threshold). The methodological triangulation presented in the next section, including plausibility checks with three qualitative case studies, adds another layer of robustness to our findings.
Running the QCA across 86 specifications demonstrates that the results are highly robust. For QCA results to be considered robust, a solution should be in a subset or superset relation to the main solution formula, with variations in consistency and coverage that do not call for substantively different interpretations (Schneider and Wagemann 2012;Skaaning 2011). In almost all tests, this is the case, with the majority even yielding solutions that are identical (with minor variations in consistency and coverage) to the standard model, thus strongly increasing the confidence in the results. Looking at the paths individually, path 1 passes this test in all specifications, path 4 in all except one: When using an alternative, qualitative measure of peace (HIIK), the path appears slightly adapted: ~PRED*PG*SE*SC*PK, strengthening our interpretation that all conditions are jointly needed to overcome a high predisposition. Path 3 is equal or in a subset/superset relationship in 84 of the tests. In the remaining two, ~SCT is replaced by ~SE. Path 2 is also highly robust across the vast majority of the tests. In those specifications where the path is not equal to the standard model or in a superset/subset relation, ~SCT is replaced with ~PG, indicating that the absences included in the solution are the least robust components of the solution. Generally, the large number of robustness tests significantly increase our confidence in the results, and none of the tests contradicts our substantive interpretation.
In order to further assess their plausibility and inform our interpretation, we complement our QCA results with illustrative insights from three typical cases (representing all paths except the first, that confirms the well-researched importance of peacekeeping). 18 The cases serve to conduct a 'plausibility check' of the causal mechanism suggested by the different paths, which should also give us guidance on how to interpret the absences of support included in the pathways to peace. This step is based on 90 qualitative interviews conducted in 2017 in Cambodia, Indonesia and Liberia with representatives of domestic civil society, public institutions, the media as well as bilateral and multilateral donor agencies. 19 Moreover, we aim to better understand if cases covered by the same pathway share similar contextual factors that might explain the existence of different pathways. Table 5 summarizes the additional background data compiled for this purpose. 20 The first of our four paths, "Protecting peace", is characterised by substantial peacekeeping in cases without a high predisposition of civil war recurrence and in the absence of substantial security ODA or SCT support. This path confirms previous research finding a positive effect of peacekeeping 21 but, surprisingly, covers only two cases -Mozambique and Tajikistan. In both countries, one party to the conflict used its initial grip on power to consolidate its position once international peacekeeping had contributed to stabilising the situation. Although major outbreaks of renewed violence were avoided 18 For more details on the selection see Appendix 7. For selecting cases based on QCA see Schneider and Rohlfing (2013). 19 More detailed information is presented in Appendix 9.
20 Data on the type of conflict come from Vogt et al. (2015), on practiced power-sharing in the first five post-conflict years from Ottmann and Vüllers (2015), on democracy from V-Dem (Coppedge et al. 2016) and on conflict ending from Kreutz (2010). 21 Therefore, we did not conduct an in-depth case study on this path. in both countries, both have seen an increasing exclusion of political opposition, including former warring factions, which has raised doubts about the longevity of the country's post-civil war peace (ICG 2016). International peacekeeping, it seems, was useful to reduce the immediate risk of conflict recurrence. Issues with the exclusion of political opponents, however, have remained and contributed to periodic instability.
Another security-related path to peace in our results is characterised by a substantial provision of nonmilitary security support, which is why we refer to it as "Securing peace". Cases covered by this path do not display a "high predisposition" and received neither substantial amounts of support for SCT nor in the field of socio-economic development.
Cambodia illustrates the pattern of "Securing peace". 22 Despite an internationally sponsored peace agreement in 1991, the civil war between the Cambodian government and the Khmer Rouge came to a final end only in 1998 when most remaining rebel fighters accepted an offer to demobilise and return into society. International support for Cambodia post-1998 reflected the political circumstances of a "victors' peace": Having won the civil war, the government was able to dictate the terms of peace and be selective in opening the doors for external support. International non-military support for security was welcomed because three decades of warfare had left the country with a legacy of millions of anti-personnel mines and unexploded ordnance. As the latter was largely the consequence of US bombing during the 1960s and 1970s, many Western donors engaged heavily in humanitarian demining after the end of the conflict.
Interviewees agreed that peace in Cambodia lasted due to several factors, in which international support did not play the most crucial role: A strong ruling party managed to provide basic security and portray dissent as a threat to the country's stability. The Hun Sen government introduced a cooptation scheme of former Khmer Rouge and other oppositional fighters, dubbed "win-win policy".
In the words of one senior government official, this was "a great success for integration". 23 Although the country had experienced some of the gravest mass atrocities in modern history, the government did not consider societal conflict transformation a priority. Quite to the contrary, an open dealing with the past would have threatened to reach into the highest levels of the new power elite and was thus openly opposed. Despite international pressure, a special tribunal was not established until 2003, after which it took another three years to come into operation and eventually tried and convicted only a handful of former Khmer Rouge leaders. Clearly, the absence of SCT support in the early post-civil war years was dictated less by donor preferences than by a lack of opportunity granted by the ruling government. Meanwhile, steady economic growth provided jobs for an impoverished population and the resources needed by the government to secure a buy-in from the new power elite. 24 External assistance in the field of humanitarian demining contributed by helping foster a sense of increasing security in everyday life.
Comparing Cambodia to the other cases covered by the "Securing peace" path (Angola after 2002 and Sri Lanka after 2009), it appears that this pattern might represent a typical scenario of authoritarian consolidation by a ruling party. In Angola, the ruling MPLA established peace in 2002 after having defeated the UNITA rebels. Since then, "the MPLA government has successfully recast itself as the party of stability and peace, and embarked on an ambitious reconstruction drive financed by growing oil revenues and oil-backed credit lines" (BTI 2018, 3) that have helped to establish a neopatrimonial type of rule. In Sri Lanka violent conflict was ended at the hands of President Rajapaksa through a heavy military campaign against the insurgent Tamil Tigers. Their defeat set the scene for an increasingly authoritarian and exclusionary rule that lasted until 2015.
In sum, these comparative observations suggest two different interpretations. On the one hand, proponents of the victor's peace theory could argue that the absence of a large-scale truth and reconciliation policy helped avoid new divisions in society at a crucial point in time and "let time heal". On the other, the cases might also indicate that the absence of SCT support in the "Securing peace" path was dictated more by necessity than by design: International support went where it was allowed to do so, helping to secure a peace that was already won by the government. However, as mentioned before, the robustness tests point towards the first explanation: although the presence of SC in cases without PRED is highly robust, the absence of SCT is less so. Either way, questions need to be asked as to the longevity of a peace built upon the increasing exclusion of political opponents.

Pathway 3: Institutionalizing peace
We find one path to peace, "Institutionalizing peace", which does not include any type of security support. Instead, support for politics and governance provides an alternative pathway to sustain peace in cases with a low predisposition for conflict and in absence of support for SCT. This path explains the largest share of cases, demonstrating the important role support for politics and governance can play in post-conflict situations.
Indonesia is a case in point. 25 The country has struggled with various internal conflicts -the two most prominent civil wars being Timor-Leste, a region which gained independence after violent struggle in 2002, and Aceh, which fought for independence for almost 30 years but eventually remained part of the country through a negotiated settlement facilitated by international actors in 2005. Since then Indonesia has managed to remain relatively stable. 26 Interviewees agreed that this is mainly due to Indonesia's decentralization process, which was part of the reform packages adopted after Indonesia's transition toward democracy in 1998 and highly supported by the international community. 27 The decentralization process has been described as "the most important factor" 28 to explain peace and "an absolute success" 29 regarding its effect on stability in the country. Decentralization has increased possibilities for participation at the sub-national level, resulted in resources being controlled more directly by the local governments as well as reducing the "distance between the rule and the rulers". 30 This has reduced grievances and improved relations with the national government, although corruption remains a considerable problem. The peace agreement granted Aceh the right to establish its own local parties. It is exactly this possibility to participate locally, as well as gaining greater power over fiscal resources that interview partners mentioned as the most crucial factors why Aceh has remained peaceful. 31 As one interviewee put it, the local elections in Aceh brought "the separatists into the society by giving them a ship, i.e. a political party, where they can run in the system." 32 This allowed "institutionalized communication, instead of rioting and protesting" 33 between the local and national level in Indonesia. The international community, in turn, strongly supported decentralization, both at the national level and in Aceh. 34 Regarding the absence of SCT in the "Institutionalizing peace" path one could hypothesize that the future-oriented focus on institutional solutions marked by a division of power was only possible in the context of a trade-off that entailed not actively dealing with the violent past in return for granting autonomy. However, the qualitative evidence does not indicate that supporting SCT would have led to renewed animosities. To the contrary, many interviewees judged it as highly problematic for the peace process that Indonesia has not had a process of dealing with the past. 35 Furthermore, no larger scale reconciliation or dialogue projects were implemented because the conflict mainly took place with the central government, making peacebuilding efforts between societal groups less necessary. 36 The absence of SCT should therefore not be seen as a necessary element of this pathway to peace but instead be better explained by the conflict structure as well as the government's unwillingness to address a delicate topic. This supports our QCA findings that the absence of SCT does not constitute a robust necessary component of the PG pathway to peace.
Looking at the other cases covered by this path reveals that almost all of them -namely Peru, Indonesia, Nicaragua, Guatemala, El Salvador and Serbia (including Kosovo) -embarked on a democratization process either shortly before or after the conflict ended. The only exception is Cambodia, a multiply covered case, 37 which however, is well captured by the path "Securing peace". As in Indonesia, all of these democratization processes were accompanied by a devolution of government through an internationally supported decentralization process. While not all countries successfully transformed into decentralized democracies, they did not slide back into more authoritarian systems either and remained peaceful. These results are in line with Savun & Tirone's (2011) finding that democracy support helps avert the negative effects democratization can have on peace, but specifically confirms it for post-conflict cases. Overall, our findings speak to the debate 30 Domestic academic, 22-11-2017. 31 Jakarta: Domestic bilateral cooperation representatives, 22-11-2017;Domestic INGO representative, 23-11-2017; international representative from bilateral cooperation, 14-11-17;Banda Aceh: Policy Expert, 27-11-2017. 32 Domestic INGO representative, 23.11-2017 Domestic bilateral cooperation representatives 22-11-17.
35 Interview s with civil society representative s in Banda Aceh and Jakarta, 29-11-2017, 27-11-2017. 36 Member of former rebel movement, Banda Aceh, 27-11-2017 37 Multiply covered means that according to the underlying data, two paths explain the case equally well, so that both logics could be at play.
whether supporting democracy in a post-conflict context can help strengthen peace, or risks triggering renewed violence. Although democratization might sometimes have conflict-inducing effects, international engagement aimed at building political institutions and democracy can in fact contribute to peace after civil war.

Pathway 4: The encompassing approach
The fourth path identified by the QCA is the 'Encompassing approach'. With combined support across all areas of engagement, this is the only path to peace that applies to post-conflict countries irrespective of their predisposition. It suggests that external support that embeds security-related efforts in support for politics, socio-economic development as well as societal reconciliation can contribute to peace and is even able to overcome a negative predisposition. The robust UN peacekeeping mission played an important role, serving as "a stabilizing force" 40 preventing an escalation of violence at various instances. Its broad mandate reached beyond the military component and exemplifies the interaction between the different areas of engagement in the encompassing approach. As one interviewee stated: "The presence of the UN troops set the stage for disarmament", which the mission actively assisted, as well as the subsequent conduct of democratic elections. 41 Many interviewees agreed that international support for the democratic development of Liberia was important for peace. With this support, three elections were successfully held, leading to a peaceful change of government in 2018. "Those elections helped to stabilize the country's governance structure and they helped to get legitimacy to its political leadership.
[…] Substantial support to the elections, that has been part of the vast support to the country's democracy". 42 To address the huge capacity constraints, substantial efforts by the UN and other international donors to train and equip government institutions helped to rebuild the largely defunct state and facilitated a normalization of the situation. 43 Next to electoral support and strengthening government capacity, "there was lots of 38 The following discussion is based on 40 interviews conducted in Monrovia in 2017. In addition, humanitarian assistance and quick impact projects helped to alleviate urgent socioeconomic needs until basic infrastructure -roads, schools, hospitals -was rebuilt and the provision of social services improved. 46 This helped to create a peace dividend for the population. It strengthened one factor that many interviewees stressed as key to sustain peace: the war-tiredness of the population. "I think that is the first and important thing to recognize, that people have chosen to remain peaceful. Second is the support from the UN, and also from the international community". 47 Although interviewees also voiced criticism about problems related to reform processes and persisting challenges to the Liberian peace process, overall they painted a clear picture of positive developments that were strongly facilitated by the substantial international engagement.
In sum, the path shows that it is important to embed international support for security in broader developmental activities combined with efforts to strengthen political institutions and reconcile the war-torn society. The three cases explained by the 'Encompassing approach' -Bosnia, Liberia, Sierra Leone -share a particularly complex conflict context, where a number of factors coincided. Their civil wars involved several fighting factions, were characterized by regional conflict dynamics, ethnic polarization and large-scale atrocities against civilians. They left the countries with a devastated economy and a weak or even dysfunctional state. All three cases are small states, in which comprehensive peace agreements mediated by international actors ended the civil war, making comprehensive international engagement easier. The results confirm previous research that multidimensional peacekeeping is particularly effective when moving beyond a mere security focus and including broader developmental and political activities.

Summary observations
Two important questions remain. Firstly, could it be that donors choose to engage in easier contexts only and therefore "insufficient" support in high predisposition cases is a consequence of high risk, not a cause of peace breaking down? Neither our data nor existing evidence supports this notion. As mentioned, four of the recurrence cases actually display a low predisposition for conflict and therefore could have been considered "easy" cases by the international community but were not chosen for support. Moreover, almost one third of all cases with a high predisposition did nonetheless receive substantial engagement in at least one area -hence, a clear pattern of international actors "cherry picking" easy contexts does not emerge. Nor is it supported by the literature on peacekeeping operations and post-conflict aid. Fortna (2008, 173) specifically addresses this question and provides strong evidence to the contrary. UN peacekeeping missions are in fact "much more likely to deploy when the danger of war recurring is particularly high". Similarly, Walter (2010, 24) posits that after civil wars, "aid tends to be offered to countries that are particularly susceptible to renewed civil war".
Second, one could raise the general question whether the patterns of international support found to be associated with peace in the analysis represent a conscious, strategic decision by international actors or rather the country context that allowed only certain types of support. Clearly, the identified patterns of support can partially be explained by the specific environment: all countries in the "Institutionalizing peace" path were experiencing a democratic opening, hence making substantial donor engagement in the area of politics and governance possible. The countries in the path "Securing peace" exhibit the exact opposite context, characterized by authoritarian tendencies where donor engagement had to be limited to specific areas. The cases represented by the "Encompassing approach" in turn are marked by particularly devastating civil wars creating high needs for international engagement while at the same time these small, politically weak countries clearly enabled substantial donor engagement in all five areas. However, none of the paths can entirely be explained by country contexts and additional robustness checks confirm that while the mentioned factors represent interesting context, they do not bias our analyses. 48 Overall, the case studies as well as the additional domestic factors taken into account indicate that the configurations of international engagement identified by the QCA should be understood against the background of country contexts that permit certain types of engagement while discouraging others.

Conclusion
How can international actors effectively support peace after civil war? This paper is the first to unpack the broad category of peacebuilding and look beyond peacekeeping or general aid flows to identify which components or combinations of international support can contribute to sustained peace. Complementing a QCA of 36 post-conflict episodes with additional context information and illustrative case studies reveals four key insights: First, our results confirm that peacekeeping can make a difference, yet also demonstrate that it is but one important component of effective post-conflict support. Second, a focus on support for politics and governance can effectively strengthen peace. Third, only a combination of all types of support explains sustained peace in cases with a high risk of conflict recurrence. Finally, and more generally, the analysis demonstrates that international support matters. All ten cases that did not receive substantial support in any of the areas of engagement experienced recurrence. This includes cases with both a high and a low predisposition for renewed conflict. Hence, a general neglect of a country emerging from civil war by the international community seems an almost certain pathway to renewed violence.
The analysis consistently explains why post-conflict countries remained peaceful by taking a closer look at the type and combinations of international support. By doing so, it unveiled new avenues for further research. First of all, the analysis does not cover cases that suffer from protracted conflict, like Somalia or Afghanistan, but ending ongoing conflict is likely to require different kinds of engagement than preventing renewed violence. Furthermore, it might be valuable to further disaggregate specific areas of engagement as these cover a wide variety of activities. For example, disaggregating support for politics and governance, which explains a large share of cases, might yield valuable insights into how exactly it contributes to peace. Finally, taking a closer look at differences between support provided by different types of international actors, such as bilateral or multilateral donors, might also yield interesting insights.
The findings have clear policy implications. International actors have supported post-conflict countries for more than two decades. Our analysis shows that such international support can indeed make a difference. This analysis also demonstrates that different country contexts condition what types of support can be provided. Given the controversial debate about detrimental effects of democratization on peace, it is noteworthy that international support for politics and governance in democratizing post-conflict contexts can in fact stabilize peace after civil war.

Funding
We gratefully acknowledge that the research towards this article was funded by the Federal Ministry for Economic Cooperation and Development (BMZ) under the Supporting Sustainable Peace project grant.

Data
The replication data for this article are available on the ISQ Dataverse, at https://dataverse.harvard.edu/dataverse/isq.