Abstract
Disarmament, Demobilization, and Reintegration (DDR) is a crucial component of transitions from conflict. DDR starts the demilitarization of politics, a critical first step in reducing the feasibility of political success through violence and preventing conflict recurrence. Existing research on DDR is hampered by the absence of cross-national data. This paper introduces DDR-40, a pioneering global dataset documenting 83 DDR programs with 57 yearly attributes, encompassing 407 program-years (1980–2020). DDR-40 captures detailed program-specific characteristics, including target groups, membership size, group type, cantonment period, budget, implementation scores for disarmament, demobilization, and reintegration, implementation bodies, and program outcomes. For the duration of every program, the dataset also codes the implementation of other peace agreement provisions, including commitments to amnesty, prisoner release, boundary demarcation, women’s rights, and children’s rights, as well as provisions for various institutional reforms, such as executive, legislative, electoral, constitutional, judicial, and military reform. The paper outlines the dataset structure, provides descriptive statistics, and shows how the data can be used to explain conflict recurrence during DDR.
Introduction
Following the Toncontín Peace Agreement of 1990, international organizations spent 67 million USD to disarm and demobilize the Contras in Nicaragua, marking a monumental effort to bring peace to a war-torn nation (World Bank 1993). 23,000 Contras were demobilized, and more than 15,000 small arms, along with heavy weaponry, were handed over to the United Nations (Cox 1996). However, despite these substantial efforts and investments, the peace was tenuous and fragile. The peace agreement provisions – promises of land, executive reform, development assistance, and resettlement – remained unfulfilled. The socio-economic grievances and political marginalization fueled a desperate response: the Contras rearmed after a successful disarmament and demobilization effort to form the Recontras. As in many other conflicts, rearmament was not just a cry for survival fueled by a security dilemma but a powerful statement of unresolved struggles for justice and reform.
Disarmament, Demobilization, and Reintegration (DDR) is the primary political effort toward peace in countries engulfed in or emerging from war (Berdal and Ucko 2009; Humphreys and Weinstein 2007; Muggah 2009). DDR has become a central component of the majority of large-scale peace operations, mostly conducted under the auspices of the United Nations or regional organizations (Campbell and Di Salvatore 2024; Di Salvatore et al. 2022; Matanock and Lichtenheld 2022). On the one hand, DDR starts the demilitarization of politics, a critical first step in peacebuilding to reduce the feasibility of political success through violence and prevent conflict recurrence (Joshi et al. 2017; Lyons 2005; Özerdem 2009). On the other hand, DDR has a direct impact on the balance of military power and can lead parties to reconsider or contest the original terms of the settlement (Ruggeri et al. 2017), thus preventing the full transition to peace once rebels disarm.
Recurrence of conflict is the most salient symptom of DDR failure, leading some countries with multiple failed programs to a conflict trap. United Nations missions whose mandates include expansive DDR tasks are associated with higher post-conflict homicide rates even as battle deaths fall, suggesting that poorly integrated ex-fighters may drift into organized crime rather than politics (Di Salvatore 2019). Rearmed non-state actors sometimes have more devastating effects on post-conflict nations than the initial well-organized rebellions with politically motivated, socially integrated groups. Joseph Kony’s infamous Lord’s Resistance Army in Uganda emerged out of the country’s first DDR program (1992–1996) with the Ugandan People’s Democratic Army (UPDA) rebels (Borzello 2007). Similarly, challenges of third party actors and national authorities to bring stability to complex cases, such as Iraq and Afghanistan, are closely linked to the DDR challenge in both countries (Berdal and Ucko 2009).
Given the high stakes involved in DDR, a substantial body of literature has explored the determinants of DDR success or failure through country-specific and regional case studies (Bangura 2023; Blattman and Annan 2016; Daly 2016; Gilligan et al. 2013; Humphreys and Weinstein 2006; Kaplan and Nussio 2018; McMullin 2013; Muggah 2009; Schulhofer-Wohl and Sambanis 2010; Sharif and Zuluaga 2026). Nevertheless, evaluations of the effectiveness and relevance of DDR programs are hampered by lack of cross-national data. It is impossible to answer questions, such as (a) why do armed forces remobilize after going through DDR, (b) do the number of combatants, the budget, group type, and other program features impact DDR failure, and (c) which concurrent peacebuilding efforts have the greatest impact on DDR outcomes?
To facilitate cross-national research on DDR, this paper presents DDR-40, a pioneering global dataset of the universe of DDR programs (1980–2020). The dataset identifies 83 programs and presents them as counting process data with 57 yearly covariates. Encompassing 407 program-years, the variables capture detailed program-specific characteristics, such as target groups, membership size, group type, cantonment period, budget, implementation scores for disarmament, demobilization, and reintegration, implementation bodies, and program outcomes. In addition, the data facilitate analysis of DDR success or failure within the broader peacebuilding context by coding the implementation of peace agreement provisions that reflect protection of rights and institutional reform. For every year of a DDR program, the dataset codes implementation scores for amnesty, prisoner release, boundary demarcation, women’s rights, children’s rights, and social/economic reform, as well as provisions for various institutional reforms, such as executive, legislative, electoral, constitutional, judicial, and military reform. The dataset in its current iteration does not contain any missing values.
For each of the 83 programs, DDR-40 provides a narrative justifying coding decisions and presenting findings, based on academic and non-academic sources; together, these narratives offer a qualitative record of four decades of DDR. The qualitative and quantitative data collection project responds to a widespread call by scholars and practitioners to take into account the political context in which DDR programs take place (Giustozzi 2012; Humphreys and Weinstein 2006; Muggah 2005; Munive 2013). In what follows, the paper first identifies the gaps in existing data before presenting the variables and descriptive statistics. Finally, it discusses regression results with fixed and random effects, as well as machine learning algorithms, to show how the data can be used to explain recurrence of conflict during DDR. The data and replication files are available on the Harvard Dataverse (Sharif 2026).
Gaps in Data on DDR Programs
Current quantitative research on peace processes primarily relies on three major datasets: The Uppsala Conflict Data Program Peace Agreements Dataset (UCDP-PA), the Peace Accord Matrix (PAM), and PA-X Peace Agreements Database (PA-X). 1 While these datasets are invaluable for studying peace processes, they exclude DDR programs that exist outside of peace agreements. Moreover, UCDP-PA and PA-X only record the presence of DDR provisions in peace agreements without tracking their implementation. Although PAM includes yearly implementation data, it is limited to thirty-five comprehensive peace agreements (1989–2012), seven of which lack DDR provisions. 2 Several analytical reports have compiled lists of DDR programs (Banholzer 2014; Caramés and Sanz 2009; Schulhofer-Wohl and Sambanis 2010; Sharif 2018), but they do not allow for systematic cross-national comparisons.
DDR can be a national or international initiative that takes place during or after conflict. When conflict ends in government or rebel victory, the DDR program will not be associated with a peace agreement. As a result, datasets focused solely on peace agreements overlook DDR programs that take place in these contexts. Similarly, DDR can be initiated after the signing of a peace agreement that lacks specific provisions for disarmament, demobilization, or reintegration of either the military or rebel forces. The DDR program in Zimbabwe is an example, which was not part of the peace agreement, but it played a major role in ensuring peace between the Zimbabwe African National Union (ZANU) and the Zimbabwe African People’s Union (ZAPU).
Coverage of DDR-40 programs in UCDP Peace Agreements (UCDP-PA) and Peace Accord Matrix (PAM) datasets
The DDR-40 Dataset
Under UN Integrated DDR Standards, Disarmament, Demobilization, and Reintegration (DDR) involves: (i) collecting and disposing of combatants’ weapons, (ii) formally releasing them from military command, and (iii) supporting their sustainable social and economic reintegration (United Nations 2006). DDR-40 codes programs that include at least two of these components: Disarmament: collection, documentation, control, and disposal of small arms, light/heavy weapons, ammunition, and explosives held by ex-combatants or their command structures. Demobilization: formal discharge of combatants from their armed organization, typically verified at cantonment or assembly sites. Reintegration: post-discharge assistance (cash, training, placement, psychosocial support) aimed at enabling a sustainable civilian livelihood.
DDR-40 provides counting process data on all DDR programs from 1980 to 2020, covering 83 programs and 407 program-years. Each row of the dataset has a unique program ID-year, identifying the country, the specific program, and the year in which the program was in progress. For instance,
Each program is coded from its official start year, defined as the calendar year in which the first recognized DDR activity occurs, such as the opening of cantonment sites or the signing of a financing agreement, through its official end year, when the lead agency formally closes the program, holds a closing ceremony, or issues a final completion report. DDR programs are bounded, project-style interventions with clear start and end dates, so the number of rows mirrors their actual duration. Programs still active in 2020 are coded through that year and flagged as ongoing. The series is not truncated when some factions rearm; rearmament is recorded for the relevant year and coding continues until documented closure, reflecting the fact that DDR activities often proceed despite partial remobilization.
The dataset includes only operational DDR programs, defined as initiatives that have begun collecting weapons, formally discharging combatants, and providing reintegration assistance. Aspirational plans are excluded. For example, South Sudan mandated a new DDR process in 2015, but the agreement collapsed before any combatants were processed; the 2018 plan also stalled at pilot workshops. Policies limited to single-pillar disarmament are likewise excluded, as in Ayatollah Khomeini’s post-revolution order to confiscate civilian weapons in Iran, which involved no organized demobilization or reintegration. The data begin in 1980 with the first international DDR effort in Namibia and include seven ongoing programs in Colombia, Mali, Somalia, Cameroon, the Philippines, the Democratic Republic of the Congo, and the Central African Republic.
To avoid conflating distinct efforts within a protracted conflict, each DDR program is treated as a separate case and recorded for every year it remains active. Successive DDR programs with the same conflicting parties, such as Angola’s three programs in 1991, 1995, and 2002 with the National Union for the Total Independence of Angola (UNITA), receive separate program IDs and are coded separately. Figure 1 plots all programs by start year and the log of target combatants, with points colored by continent. The three largest caseloads and most recent programs in each region are annotated, illustrating the late-1990s–2000s surge in DDR in Africa and the variation in scale from Colombia’s 2020 paramilitary demobilization to Ethiopia’s 1991 mega-program. Global distribution of DDR programs by start year and log target number of combatants in each program. Colors denote continent; labeled points mark the three largest and three most recent cases within each region.
Data collection started in June 2016 and ended in August 2024. To compile the data, I analyzed both academic and non-academic sources in nine languages, encompassing the official languages of most countries represented in the dataset: English, French, German, Dutch, Spanish, Portuguese, Persian, Arabic, and Urdu. The non-academic sources included texts of peace and ceasefire agreements, documents from the UN Secretary-General, U.S. Department of State country assessments, publications by the Organization for International Migration (IOM), truth commission findings, World Bank publications, reports from human rights organizations, and news articles from LexisNexis. When necessary, I consulted officials at the UN Headquarters, the World Bank, and DDR offices in the relevant countries to address gaps in the publicly available information. The Appendix provides a detailed account of these sources and how they were used.
To fill the gap in published works, particularly for DDR programs predating the advent of the Internet, I attended meetings of former and current DDR Section Chiefs at the UN Headquarters in New York and obtained confidential documents related to DDR execution, program challenges, and budget management. These documents were instrumental in addressing the shortcomings in publicly available records. The dataset in its current iteration contains no missing values. To ensure inter-coder reliability, a team of graduate students at the University of Notre Dame independently re-coded seventy percent of the data. The re-coding results were then compared with the original coding. In line with the ethical standards of desk research (Hoover Green and Cohen 2021), any discrepancies were resolved by assessing the reliability of sources, consulting additional materials, and engaging in discussions about coding strategies.
Variables Capturing Program Features
The first set of variables captures core features of each DDR program, beginning with the name of the target group and the number of combatants involved, as larger cohorts can pose greater reintegration challenges (McMullin 2013). It also includes a binary variable indicating whether the DDR process involves multiple armed groups, which may make DDR more complex. An additional categorical indicator distinguishes whether the group undergoing DDR is part of the state military or a non-state actor and, if the latter, whether it is organizationally integrated or fragmented (Staniland 2014). 3 This allows the dataset to capture the cohesion dynamics that may shape compliance with the DDR process (Sharif 2022; Sharif and Zuluaga 2026).
Poorly executed DDR programs can become a catalyst for future conflicts. Partial disarmament, for instance, can lead to distribution of weapons across the country and into neighboring states (Colletta et al. 1996; Knight and Özerdem 2004; Muggah 2006). The dataset includes three critical variables capturing the implementation of the three elements of DDR. The first, disarmament, constitutes collection, documentation, control, and disposal of weapons from combatants, as well as the development of arms management programs. Following PAM (Joshi et al. 2015), the dataset codes the level of disarmament implementation annually on a scale of 0–3, indicating whether implementation was absent (0), minimal (1), intermediate (2), or full (3). Minimal implementation (1) is characterized by the registration of weapons with minimal actual turnover, while full implementation (3) indicates that the vast majority of weapons have been surrendered by combatants.
Demobilization, the second variable, refers to the formal discharge of combatants from armed groups. This process typically occurs in stages, beginning with the processing of individuals in temporary centers and potentially involving cantonment sites. The dataset tracks the degree of demobilization achieved each year, from minimal, where only a small number of combatants are discharged, to full implementation, where nearly all eligible participants have been demobilized. The third variable, reintegration, addresses the process by which ex-combatants are transitioned back into civilian life. This may include compensation packages, job placement, and vocational training. The reintegration variable captures the extent to which these programs have been successfully implemented, with levels of implementation ranging from null (0) to minimal (1), where few combatants are reintegrated, to medium implementation (2), to full implementation (3), where the majority have been successfully integrated into society.
Other variables capture operational aspects of DDR programs, such as the cantonment period, which refers to the number of months combatants are housed in designated camps during the demobilization process, while acknowledging that effectiveness of cantonment is closely linked to the availability of resources, the quality of services provided at these sites, and the durability of wartime bonds during DDR (Knight and Özerdem 2004; Sharif 2023). The dataset also records the origin of the DDR program, whether it was established as part of a peace agreement, a national initiative, or an international effort, which may impact the program’s legitimacy and potential challenges. When established by a peace agreement, the accord’s name and date are also noted. Some DDR programs are executed by national governments with no international support, which may have consequences on DDR effectiveness (Humphreys and Weinstein 2007). The dataset identifies the actors responsible for executing the DDR program, distinguishing between national bodies, international organizations, or joint efforts. The budget, in millions of USD, captures total funding for disarmament, cantonment, reintegration, logistics, administration, and security. 4
Descriptive statistics for DDR program features
Peace Provisions and Implementation Variables
DDR programs typically unfold alongside other peacebuilding and state consolidation efforts, creating a strong, interdependent relationship between DDR and peace agreement implementation (Berdal 1996). Peace agreements with DDR provisions are more likely to maintain peace (Matanock 2017), just as DDR is rarely successful without robust peacebuilding (Özerdem 2002). To capture this symbiosis, DDR-40 codes the implementation of key peace agreement provisions; i.e., goal-oriented reforms or stipulations costly to one or both conflict actors and falling within discrete policy domains (Joshi et al. 2015). Following PAM benchmarks, implementation is recorded on a 0–3 ordinal scale: 0 for no action, 1 for minimal, 2 for intermediate, and 3 for full implementation. Appendix Section 3 contains detailed explanations of coding rules.
Provisions cover both rights protections and institutional reforms. 5 The first provision is “Amnesty,” which concerns the legal forgiveness of crimes committed during the conflict, with implementation measured by the degree to which it is granted by the government to the target groups. A provision for “Prisoner Release” commits the government to the release of political prisoners, with implementation tracked by the extent of progress in freeing detained individuals. Next, “Truth and Reconciliation” mechanisms aim to investigate human rights violations and foster reconciliation, with implementation gauged by the operational status of these bodies. “Women’s Rights” provisions include measures for the civil, political, and economic rights of women, including specific DDR measures for female soldiers, with implementation measured by progress in realizing these rights. “Children’s Rights” provisions target the protection and reintegration of child soldiers, with implementation assessed based on improvements in legal and institutional frameworks for children’s rights.
“Economic and Social Development” aims to improve economic and social conditions, particularly for marginalized populations, with implementation tracked by the development and operationalization of relevant programs. The “Dispute Resolution” provision establishes committees to resolve conflicts arising during the peace process, with implementation measured by the effectiveness and activity of these committees. “Boundary Demarcation” involves changes to internal political boundaries, with implementation evaluated by the progress of these boundary adjustments. “Natural Resource Management” addresses the equitable use and management of natural resources, with implementation measured by the adoption and enforcement of new resource management laws. The “Decentralization/Federalism” provision commits the government to transfer power to regional or local authorities, with implementation measured by the establishment and operationalization of decentralized institutions.
Among provisions aimed at reforming state institutions, “Constitutional Reform” creates changes to the nation’s constitution, with implementation tracked by the adoption and enforcement of constitutional changes. “Electoral/Political Party Reform” focuses on altering the electoral system and political party regulations, including the creation of parties by rebels, with implementation assessed by the extent to which new electoral laws and political party reforms are enacted. “Executive Reform” deals with changes to the executive branch, such as altering powers or selection processes, with implementation measured by the degree to which these reforms are realized. The “Judiciary Reform” provision aims to reform the judicial branch, often by altering the process of selecting judges or the structure of the judiciary, with implementation measured by the enactment of these reforms. The promise of “Military Reform” typically involves integrating rebel forces into the national military or restructuring the armed forces, with implementation assessed by the degree of military integration and reform. Finally, “Legislative Reform” involves transformations in the legislative branch, including power-sharing arrangements and reforms in legislative procedures, with implementation measured by the progress in enacting these reforms.
Descriptive statistics for peace accord implementation variables
Final–year implementation scores.
DDR Outcomes
Evaluation of DDR programs suffers from a level-of-analysis problem. DDR operates on the individual level but aims at preventing countries or regions from returning to armed conflict. Evaluations of DDR are often conducted across the two levels. On the individual level, evaluation criteria can be based on the number of combatants disarmed and demobilized, the weapons collected, the quality of reintegration programs offered to ex-combatants, and the extent to which they are economically, socially, and politically reintegrated (Garibay 2006; Humphreys and Weinstein 2007; Özerdem 2002). On the national level, DDR can be evaluated as having failed or not failed, depending on whether combatants rearm and re-engage in conflict (Muggah 2009).
Defining rearmament as failure of DDR corresponds roughly to the notion of negative peace, i.e., the absence of violence (Galtung 1969). In the peacebuilding literature, negative peace as an outcome has come to represent a reductive notion of peace, because sustaining peace is almost impossible without addressing drivers of violence and structural causes of it (Darby and Mac Ginty 2008). Understanding peacebuilding as a process and not a goal (Özerdem 2009), it would be almost impossible to devise outcome measures for a long-term process that is meant to bring about “quality peace:” overhaul institutions of post-conflict states, improve their economic conditions, attend to issues of justice and reconciliation, and ensure violence does not re-emerge among former conflicting parties or new ones (Joshi and Wallensteen 2018). 6
To allow cross-national comparative analysis, DDR-40 includes outcome variables that indicate the recurrence of conflict during or after DDR, using data from the UCDP/PRIO Armed Conflict Dataset version 24.1 and the UCDP One-sided Violence Dataset version 24.1 (Davies et al. 2024). The variable “Rearmed during DDR” captures whether a group that was part of the DDR program rearmed during the DDR process, coded as 1 if rearmament occurred and 0 if it did not. “Factions Rearmed during DDR” indicates if some factions within a group rearmed during the DDR process, also coded as 1 for rearmament and 0 otherwise. The “Rearmed post-DDR” variable tracks whether rearmament occurred after the DDR process was completed, with the same binary coding. The “Date of Recurrence” records the year when conflict recurrence was first observed, and “Recurrence Side B” identifies the opposing side in the recurred conflict. “Recurrence Conflict ID” provides the UCDP/PRIO conflict ID for the recurred conflict.
The dataset codes conflict-recurrence outcomes at the program–year level using UCDP/PRIO Armed Conflict (v 24.1) and One-Sided Violence (v 24.1) data (Davies et al. 2024). The variable “Rearmed during DDR” equals 1 in the first year that the armed group remobilizes and surpasses the 25-battle-deaths threshold while the program is still active. That value is then carried forward in every subsequent program-year to capture the fact that the DDR effort is unfolding in the shadow of renewed violence. “Factions Rearmed during DDR” indicates if some factions within a group rearmed during the DDR process, also coded as 1 for rearmament and 0 otherwise. The “Rearmed post-DDR” variable tracks whether rearmament occurred after the DDR process was completed, with the same binary coding. The “Date of Recurrence” records the year when conflict recurrence was first observed, and “Recurrence Side B” identifies the opposing side in the recurred conflict. “Recurrence Conflict ID” provides the UCDP/PRIO conflict ID for the recurred conflict. A post-DDR organization is treated as a remobilization of the original group if three criteria, drawn from UCDP actor-coding rules, are met: (i) continuity of senior leadership or command cadre, (ii) public self-identification as the political–military heir of the earlier force, and (iii) overlap in core constituency or theater of operations. Each successor group was hand-coded: when these criteria are satisfied, the new label is mapped back to the parent program.
Descriptive statistics for outcome variables
Re-armament outcomes by DDR establishment. Bars show the percentage distribution of program-years that experienced no relapse, group-wide rearmament, faction-level rearmament, or renewed violence after DDR completion, across peace-agreement, national, and international initiatives.
Application of the Data
Logistic regression estimates of rearmament during DDR
Note. +p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.
Across all models, the level of disarmament is negatively associated with return to war by the same group or its factions. The coefficient is moderate in the pooled logit (β = −0.489, p < 0.01 in Model 1) and grows in magnitude once the sample is restricted to non-state actors and random effects are introduced (β = −4.789, p < 0.05 in Model 5). Even under the stringent fixed-effect specification, the association remains sizable (β = −2.166, p < 0.01; Model 6). By contrast, demobilization is statistically relevant only in Model 1 and Model 2 (β = −0.373, p < 0.05 and β = −0.492, p < 0.05). The degree of reintegration is not significantly associated with rearmament during DDR in any of the models, while this factor may be decisive when rearmament takes place post-DDR.
The results also suggest that larger programs face bigger challenges: the number of combatants (expressed in thousands) is positively associated with rearmament in every specification (β = 0.005, p < 0.01 in Model 2; β = 0.023, p < 0.01 in the fixed-effects model). Programs that involve multiple armed groups are likewise more prone to relapse (β = 1.427, p < 0.001 in Model 1; β = 5.017, p < 0.001 in Model 6). Introducing a control for group type (state versus non-state) in Model 2 suggests that non-state actors are considerably more likely to rearm (β = 2.579, p < 0.001). A finer disaggregation (Model 3) suggests that integrated (vs. fragmented) non-state groups are less relapse-prone (β = −0.499, p > 0.1); this difference becomes statistically significant once unobserved heterogeneity is modeled (β = −11.396, p < 0.05 in Model 5).
Programs launched as part of a peace agreement are consistently linked to lower odds of rearmament (e.g. β = −1.249, p < 0.01; Model 3). International execution further reduces the risk in the random-effects models (β = −16.762, p < 0.01; Model 4), reinforcing the view that strong international custodianship can contain spoilers and prevent fragile settlements from relapse (Stedman 1997), whereas mixed national–international execution offers no systematic advantage over purely national programs. The log of cantonment duration is positively related to rearmament (β = 0.211, p < 0.05; Model 1) but turns insignificant once additional covariates are introduced. United Nations peacekeeping operations exert no robust effect – positive in Model 5 (β = 12.660, p < 0.05) but null elsewhere – suggesting that their influence is contingent on sample composition and model choice. Moving from pooled to random-effects estimation reduces the Akaike Information Criterion (AIC) from 478 (Model 1) to 139 (Model 4), suggesting better fit for the data.
Random effects logistic regression estimates of peace provisions (implementation levels 1–3)
Note. +p < 0.1; *p < 0.05; **p < 0.01; ***p < 0.001.
Two findings run counter to conventional wisdom. First, a minimal prisoner-release effort actually raises the risk of relapse (β = 27.17, p < 0.10), while full implementation reduces it (β = −21.62, p < 0.01). This suggests that selective releases may embolden spoilers unless followed quickly by a more comprehensive scheme, if the terms of the peace agreement or the DDR program promise prisoner release. Second, full constitutional reform is positively associated with rearmament (β = 11.07, p < 0.05), perhaps because sweeping charter overhauls unsettle pre-existing elite bargains and rights guarantees, creating uncertainty that prompts excluded or distrustful factions to hedge their risks by remobilizing (Sharif et al. 2025). In terms of model fit, the decentralization specification yields the lowest AIC (135.9), with constitutional reform a close second (138.5), indicating that these two provisions carry the greatest explanatory weight once country-level heterogeneity is taken into account. Appendix Section 5 contains the full models.
At this stage, it is important to consider whether disarming, demobilizing, and reintegrating armed groups justifies the significant time, investment, and effort, or if peace processes can succeed without DDR. To answer this question, I employed a random forest model on a subset of variables that capture the implementation levels of peace agreement provisions as well as disarmament, demobilization, and reintegration scores. Conflict scholars have used machine learning in recent works to forecast civil war onset (Blair and Sambanis 2020; Muchlinski et al. 2015), taking into account the inherently interactive nature of numerous explanatory variables. Employing machine learning is especially useful in this preliminary analysis considering the novel nature of the study and lack of prior global theories. The model includes 10,000 trees per forest, 100 observations per tree, and 5 terminal nodes.
Figure 4 ranks variables according to their importance in predicting rearmament during DDR, indicating the number of observations that are incorrectly classified when a particular variable is removed from the ensemble of trees (mean decrease in accuracy). The results suggest that DDR components, especially disarmament and demobilization, play critical roles in reducing the likelihood of rearmament. Consistent with the random-effects logit models, amnesty implementation emerges as the single most influential factor, followed closely by the core DDR components, especially disarmament and demobilization. A second tier of variables mirrors the significant institutional reforms from Table 6: dispute-resolution bodies and executive reform register large accuracy losses when dropped, while decentralization and constitutional change appear with mid-range importance, reflecting their more nuanced logit effects. Variable importance from a random forest model for rearmament during DDR.
The parametric and non-parametric models presented above align with the broader conflict literature, which widely recognizes “demilitarization” through DDR as essential for creating peace (Diehl et al. 2023; Joshi et al. 2017; Walter 2002). Yet this prevailing view assumes that the motivations driving disarmament and participation in peace agreements remain constant, overlooking how they can shift when the peace process falters or when governments fail to implement promised institutional reforms. While individual ex-combatants may disengage due to poor socioeconomic conditions and limited post-conflict opportunities, widespread rearmament often follows the government’s failure to honor provisions on institutional change and citizen protection. The analysis based on the new data suggests that DDR should not be viewed as a standalone, “preliminary” process; rather, it unfolds alongside other peacebuilding initiatives that transform the societal, economic, and political structures that both caused and were shaped by conflict. While demilitarizing politics is essential for later steps such as party formation and elections, DDR’s success depends on sustained peacebuilding after disarmament, as programs often continue well into the implementation of other peace agreement provisions.
The results also underscore the centrality of disarmament and challenge arguments for reintegrating combatants with their arms or commencing the peace process without prior disarmament. In Iraq, for instance, the Coalition Provisional Authority’s 2004 Order 91 formalized the reintegration of militias and other armed forces outside state control (Coalition Provisional Authority 2004), yet delaying clarity on the structure of the new central government until after executive and constitutional reform – changes that would determine which militias disarmed permanently – fueled Sunni disaffection and derailed the reintegration of the Sunni army officer corps (Rathmell et al. 2005). Thus, the postponement of disarmament ultimately undermined state consolidation (Ucko 2009).
Conclusion
This paper introduces DDR-40, a pioneering dataset of the universe of disarmament, demobilization, and reintegration (DDR) programs (1980–2020), covering 83 programs and 407 program-years. The dataset facilitates analysis of DDR programs within and across countries, providing a detailed picture of how DDR advances and why it sometimes results in the rearmament of demobilized groups. It also enables scholars to study DDR programs within the broader context of peacebuilding. The data and analysis provide a better framework for designing future DDR programs, especially when dealing with large and multiple armed groups in less economically developed nations. It offers guidelines for international actors to support or pressure national governments to address issues in the peace process that incentivize a return to conflict. DDR-40 can contribute to scholarly and policy-oriented research on the nexus of demilitarization, peacebuilding, and state-building, particularly in fragile, divided post-conflict environments. The data have been checked for accuracy and consistency in coding. Readers are encouraged to contact the author if they identify any errors.
Beyond the cross-national patterns reported here, DDR-40 opens several avenues for future inquiry. Because every program–year is geo-referenced and keyed to UCDP actor IDs, scholars can merge the dataset with sub-national homicide statistics, illicit-economy indicators, or election returns to examine how the local presence of demobilized fighters affects crime, governance, and state authority after war. The successor–actor links; for example, the mapping of the AUC to the Gaitanist Self-Defense Forces in Colombia, enable researchers to trace when an ostensibly political organization mutates into a criminal enterprise following demobilization, nonetheless retaining the capacity for large-scale violence. Future work can therefore test whether particular reintegration packages, cantonment practices, or peace-accord provisions mitigate (or redirect) this drift into organized crime, and how such effects vary with local resource endowments, wartime peer networks, and state reach, as recent municipal-level studies in Colombia suggest.
Supplemental Material
Supplemental Material - Why do Armed Groups Return to War after Disarmament, Demobilization, and Reintegration? Introducing the DDR-40 Dataset (1980–2020)
Supplemental Material for Why do Armed Groups Return to War after Disarmament, Demobilization, and Reintegration? Introducing the DDR-40 Dataset (1980–2020) by Sally Sharif in Journal of Conflict Resolution.
Supplemental Material
Supplemental Material - Why do Armed Groups Return to War after Disarmament, Demobilization, and Reintegration? Introducing the DDR-40 Dataset (1980–2020)
Supplemental Material for Why do Armed Groups Return to War after Disarmament, Demobilization, and Reintegration? Introducing the DDR-40 Dataset (1980–2020) by Sally Sharif in Journal of Conflict Resolution.
Footnotes
Acknowledgments
For valuable feedback, the author thanks Dave Armstrong, Daniel Arnon, Tatiana Carayannis, Sam Fuller, Madhav Joshi, Oliver Kaplan, Xiaojun Li, Megan MacKenzie, George Mychaliska, Aaron Pangburn, Jason Quinn, Eline de Rooij, Raúl Rosende, Jason Stearns, Mark Ungar, Till Weber, Thomas Weiss, and the officials at the United Nations Disarmament, Demobilization, and Reintegration Section in New York. For excellent research assistance, the author thanks Jody Oetzel, Catherine Bruno, and Dayron Monroy. The author also thanks the audiences at the American Political Science Association annual meetings (2019; 2022), the Latin American Studies Association annual congress (2023), the Comparative Politics Workshop at the Graduate Center of the City University of New York (2019), the Political Science Colloquium at the Universidad de los Andes (2019), the Crime, Conflict, and Violence Workshop at the University of Notre Dame’s Kroc Institute for International Peace Studies (2021), the Colloquium at Simon Fraser University’s School for International Studies (2022), and the IR Colloquium at the University of British Columbia (2024). Finally, the author thanks Paul Huth and the three anonymous reviewers at JCR for their constructive comments.
Ethical Considerations
This study relies on publicly accessible documentary sources and did not involve direct interaction with human subjects. Following best-practice guidance for desk research on political violence (
), all source materials are cited transparently, personally identifying details contained in original documents are not reproduced, and the cleaned dataset is stored and shared in de-identified, aggregate form.
Funding
The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by the Provost’s Digital Innovation Implementation Grant at the Graduate Center, City University of New York. Initial background research was supported by the Social Science Research Council (SSRC) in New York. Additional funding for research assistance was provided by the Kroc Institute for International Peace Studies at the University of Notre Dame.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data Availability Statement
The DDR-40 dataset and the codebook are available on the Harvard Dataverse at https://doi.org/10.7910/DVN/1F4I77. Updated versions of the dataset, codebook, and a detailed changelog documenting all corrections and revisions will be maintained on the author’s website at https://www.sallysharif.com. Suspected errors reported by users will be verified against the original sources, corrected in the underlying scripts, and documented in subsequent releases, .
Supplemental Material
Supplemental material for this article is available online.