Uppsala Conflict Data Program
The Uppsala Conflict Data Program (UCDP) is a research initiative housed at Uppsala University's Department of Peace and Conflict Research in Sweden, dedicated to systematically collecting, coding, and disseminating empirical data on organized violence globally, encompassing state-based armed conflicts, non-state conflicts between organized groups, and one-sided violence targeting civilians.[1] Operational since the late 1970s with datasets extending back to 1946, UCDP employs an event-based methodology that records individual instances of fatal organized violence—defined as the use of armed force by organized actors resulting in at least one battle-related death per year per dyad—drawing from diverse sources such as news reports, NGO documentation, and academic publications while prioritizing transparency in coding and validation processes.[2] This approach yields disaggregated, georeferenced datasets like the UCDP Georeferenced Event Dataset (GED), enabling granular analysis of conflict dynamics, fatalities, and spatial patterns. UCDP's datasets, including the flagship UCDP/PRIO Armed Conflict Dataset co-developed with the Peace Research Institute Oslo (PRIO), form the backbone for thousands of peer-reviewed studies, policy reports, and visualizations on conflict trends, battle-related deaths, and peace processes, establishing it as the most cited and utilized source for quantitative conflict research worldwide.[3] Key defining characteristics include its commitment to replicable definitions—such as requiring at least 25 battle-related deaths in a calendar year for a state-based conflict—and rigorous source evaluation to mitigate reporting biases, though the program's reliance on media-sourced events may undercount low-visibility violence in remote or censored regions.[2] Notable outputs encompass annual updates revealing escalating global conflict levels, such as the 61 state-involved armed conflicts recorded in 2024—the highest since systematic tracking began—alongside tools like interactive maps, APIs, and battle-death estimates that support causal analyses of violence drivers.[4] While UCDP's empirical focus has advanced first-principles understanding of conflict incidence and intensity over ideological narratives, its academic institutional context within peace research—prone to systemic optimism biases in interpreting data toward conflict resolution agendas—warrants scrutiny of downstream interpretations, though the raw datasets themselves remain a gold standard for causal realism in violence measurement due to their longevity, accessibility, and cross-validation with alternative sources.[5] Achievements include influencing international organizations' conflict monitoring and enabling longitudinal trend analyses that challenge assumptions of perpetual decline in violence, as evidenced by recent upticks in active wars.[3] No major methodological controversies have undermined its core outputs, though debates persist on threshold sensitivities and event underreporting in asymmetric conflicts.[2]History
Founding and Early Development
The Uppsala Conflict Data Program originated in 1978 when Professor Peter Wallensteen, a peace researcher at Uppsala University, began directing efforts to systematically record data on armed conflicts worldwide.[6] This initiative, initially operating under the name Conflict Data Project, aimed to compile annually updated lists of ongoing wars and minor armed conflicts using standardized definitions to enable global comparability and empirical analysis of conflict patterns.[7] By focusing on post-World War II events, the project sought to address gaps in quantitative conflict research, providing verifiable metrics on the incidence and nature of organized violence rather than relying on anecdotal or ideologically driven accounts.[8] The program was formalized as a dedicated project in 1982 and properly established within Uppsala University's Department of Peace and Conflict Research by the mid-1980s, under Wallensteen's continued leadership.[8] Early development emphasized rigorous data collection from diverse sources, including news reports and official records, to track interstate and intrastate conflicts with at least 25 battle-related deaths per year as a threshold for inclusion.[7] This period saw the foundational work of defining armed conflict in terms of incompatibility between organized groups over government or territory, prioritizing causal clarity over expansive categorizations that might inflate conflict counts for advocacy purposes. Wallensteen's approach drew on first-hand access to primary documents and emphasized transparency in methodology to mitigate biases inherent in media-sourced data, which often underreport or sensationalize events based on geopolitical alignments.[7] By the late 1980s and early 1990s, the program's datasets had become a cornerstone for academic studies, influencing analyses of conflict trends amid the Cold War's end, with initial outputs revealing a temporary dip in major wars followed by rising intrastate violence.[7] These early efforts established UCDP's reputation for reliability, as its conservative estimates—grounded in verifiable fatalities rather than unconfirmed claims—contrasted with more inflated figures from sources prone to overgeneralization for humanitarian or policy agendas. The program's expansion in scope during this phase set the stage for later collaborations, such as with the Peace Research Institute Oslo, while maintaining a commitment to data-driven insights over normative interpretations.[7]Key Milestones and Expansions
The Uppsala Conflict Data Program initiated publication of annual armed conflict lists in the SIPRI Yearbook in 1988, marking an early milestone in disseminating its data to the international research community. These lists continued through 2017, providing consistent updates on global conflict trends based on the program's evolving definitions of organized violence. Concurrently, starting in 1993, UCDP began annual reporting in the Journal of Peace Research, which has since become a primary outlet for its trend analyses and dataset updates. A significant expansion in temporal coverage occurred in 2002 through collaboration with the Peace Research Institute Oslo (PRIO), extending the UCDP/PRIO Armed Conflict Dataset backward to 1946 and enabling long-term historical analysis of state-based conflicts.[9] This partnership formalized joint data production, enhancing compatibility and accessibility for researchers studying conflict patterns over decades.[9] In 2003, UCDP broadened its scope beyond state-based armed conflicts to incorporate datasets on one-sided violence—defined as directed attacks by governments or non-state groups against civilians—and non-state conflicts between organized armed groups, addressing previous gaps in coverage of intra-state dynamics excluding state actors. This expansion facilitated more comprehensive tracking of organized violence, with subsequent refinements in data collection methods to include dyadic actor-level details. Further methodological advancement came in 2013 with the release of the UCDP Georeferenced Event Dataset (GED), which introduced event-level granularity with geographic coordinates for incidents of organized violence from 1989 onward, enabling spatial analysis and finer-grained temporal resolution. The GED has since been updated biennially, supporting applications in geographic information systems and predictive modeling.[1]Organizational Structure
Institutional Framework
The Uppsala Conflict Data Program (UCDP) operates as a specialized research unit within the Department of Peace and Conflict Research at Uppsala University in Sweden, functioning under the institution's academic governance and administrative policies. Established in the mid-1980s as the Conflict Data Project, it was restructured and renamed UCDP in autumn 2003 to accommodate an expanded focus on various forms of organized violence beyond civil wars.[7] This embedding ensures alignment with university-wide standards for data transparency, peer-reviewed methodologies, and open-access dissemination, while leveraging departmental resources for staffing and infrastructure.[1] UCDP's internal structure divides responsibilities across leadership, coordination, and analytical roles, with a core team of 10-15 researchers, coordinators, and assistants handling data collection, verification, and regional specialization. This configuration supports an annual budget of 5-10 million Swedish kronor (approximately USD 500,000-1,000,000), directed toward database maintenance, event coding, and tool development such as the Georeferenced Event Dataset launched in 2011.[10][7] The program integrates with broader university initiatives like the DEMSCORE e-infrastructure, which provides computational support for conflict modeling and early-warning systems.[11] Funding sustains UCDP's operations through a mix of institutional and grant-based sources, including baseline support from Uppsala University, the Swedish Research Council (notably grant 2021-00162 for DEMSCORE), and the Riksbankens Jubileumsfond. Historical contributions came from the Swedish International Development Cooperation Agency (Sida), though current reliance on domestic research councils underscores its academic independence from foreign policy influences.[12][7][11] This model prioritizes long-term data continuity over short-term projects, enabling UCDP to maintain its status as a primary empirical resource for global violence trends.[7]Leadership and Key Personnel
The Uppsala Conflict Data Program (UCDP) is currently directed by Magnus Öberg, an associate professor at Uppsala University's Department of Peace and Conflict Research, who assumed the role in 2018.[7][10] Öberg oversees the program's operations, including data collection on organized violence, and conducts research on topics such as the moral psychology of grievances, social conflict dynamics, and the impact of violence on migration patterns.[10] Margareta Sollenberg serves as deputy director, an assistant professor contributing to UCDP's research outputs and publications on armed conflicts and organized violence.[10] The program's core data generation is managed by research coordinators and project leaders, including Stina Högbladh, who handles database maintenance and peace agreement datasets with expertise in the Democratic Republic of Congo; Therese Pettersson, responsible for outreach activities such as lectures and social media engagement, specializing in conflicts in Afghanistan, Ukraine, and Syria; and Lotta Themnér, who oversees planning, personnel, and data processes, focusing on regions like Burkina Faso, Mali, and Ethiopia.[10][13] UCDP was founded in the late 1970s by Peter Wallensteen, who directed the program from 1978 to 2015 and expanded it from initial conflict listings into a comprehensive data collection effort initially known as the Conflict Data Project.[7][7] Subsequent directors include Erik Melander from 2015 to 2017, Kristine Eck from 2017 to 2018, leading to Öberg's tenure.[7] Key affiliated researchers, such as Ralph Sundberg, have contributed to specialized datasets like the Georeferenced Event Dataset (GED), though Sundberg is now primarily at the Swedish Defence University.[11][14] The program's staff typically comprises 10-15 permanent researchers, coordinators, and assistants dedicated to empirical data verification.[7]Methodology
Definitions of Organized Violence
The Uppsala Conflict Data Program (UCDP) defines organized violence as encompassing three mutually exclusive forms: state-based armed conflict, non-state armed conflict, and one-sided violence, each requiring at least 25 fatalities in a given calendar year to meet the threshold for inclusion in datasets.[15] This framework prioritizes empirical tracking of lethal violence involving organized actors, excluding sporadic or unorganized incidents, and relies on verifiable reports of battle-related deaths—defined as fatalities directly resulting from the use of armed force, including manufactured weapons, sticks, stones, fire, or similar means.[16] Organizations qualify as actors if formally structured (e.g., named groups) or informally coordinated through patterns of connected violent incidents, ensuring data captures intentional, collective violence rather than individual crimes.[15] State-based armed conflict occurs when a contested incompatibility over government (e.g., regime type or leadership) or territory (e.g., secession or interstate borders) leads to the use of armed force between at least one state government and an opposition organization or another state, resulting in at least 25 battle-related deaths in a year.[16] A state is operationalized as an internationally recognized sovereign entity or an unrecognized government effectively controlling territory without internal dispute over that control.[17] Opposition organizations must demonstrate sustained armed engagement, distinguishing these events from riots or unorganized unrest.[15] Non-state armed conflict involves the use of armed force between two organized groups, neither affiliated with a state government, yielding at least 25 battle-related deaths annually; unlike state-based conflicts, no specific incompatibility (e.g., over territory or governance) is required.[18] These conflicts often arise from communal, ethnic, or resource disputes, with groups qualifying through evidence of coordination, such as repeated clashes in the same locale.[15] UCDP data from 1989 onward records such events globally, emphasizing fatalities from direct confrontations between the groups.[18] One-sided violence consists of deliberate attacks using armed force by a state government or a formally organized non-state group targeting civilians, causing at least 25 civilian deaths per year per perpetrator.[17] Civilians are unarmed non-combatants unaffiliated with security forces, militias, or opposition groups, excluding deaths from collateral damage in battle or extrajudicial killings in state custody (e.g., prisons).[15] This category captures asymmetric violence like genocidal campaigns or reprisals, with formal organization required for non-state perpetrators to differentiate from mob violence.[17] UCDP's exclusion of incidental civilian deaths in state-based conflicts ensures analytical separation, allowing distinct assessment of intentional civilian targeting.[15]Data Collection and Verification Processes
The Uppsala Conflict Data Program (UCDP) employs a systematic, event-based approach to data collection on organized violence, focusing on incidents of fatal armed conflict involving at least one direct death, with conflicts defined by a minimum threshold of 25 battle-related deaths in a calendar year.[2] Primary data sourcing begins with global news aggregators such as Dow Jones Factiva, utilizing targeted search strings to identify approximately 50,000 reports annually, which are then filtered to 10,000–12,000 potential events for coding.[2] Supplementary sources include reports from non-governmental organizations (NGOs), United Nations agencies, local human rights monitors (e.g., Syrian Observatory for Human Rights or Informal Sector Service Centre in Nepal), academic case studies, and archival materials, accounting for about 20% of events in recent years like 2013–2016.[2] Events are disaggregated spatially and temporally, with unique identifiers assigned, and include details on actors, incompatibility types (e.g., government or territory), and fatality estimates categorized as best, low, and high to reflect source variability.[19] Verification processes emphasize cross-checking multiple independent sources to establish event occurrence, actor involvement, and death tolls, prioritizing primary origins such as eyewitness accounts or on-site journalistic reporting over secondary summaries.[2] Coder teams, comprising staff with regional expertise, apply standardized codebooks and checklists to evaluate source credibility based on transparency, independence from conflict parties, and consistency across reports; discrepancies in fatality figures are resolved by adopting conservative (low-end) estimates when sources conflict equally.[2][19] Each coded event undergoes multi-stage review by project managers, followed by automated scripts in SQL and Python to detect inconsistencies in identifiers, coordinates, or counts, ensuring data integrity before annual releases.[2] This procedure excludes indirect deaths (e.g., from disease or starvation) due to attribution challenges and relies on translated local sources where necessary, with full transparency provided via linked primary references for each event in datasets like the Georeferenced Event Dataset (GED).[19]Strengths and Limitations of Approach
The Uppsala Conflict Data Program (UCDP) employs a systematic coding process involving manual review of thousands of news reports annually from sources like Dow Jones Factiva, supplemented by NGO, UN, and local reports, with events triple-checked for quality control to ensure transparency and consistency in data application across time and cases.[2] This approach facilitates reliable trend analysis in organized violence, defined as the use of armed force by organized actors resulting in at least one direct death, with a 25 battle-related deaths threshold for conflict dyad inclusion, enabling cross-national and longitudinal comparability without conflating violence types.[2] UCDP's emphasis on actor disaggregation and verification of source independence—tracing reports to primary origins like witnesses—provides greater precision in attributing fatalities than datasets with automated or broader event inclusion, such as ACLED, which often lacks detailed actor information.[20][21] Despite these advantages, UCDP's reliance on publicly available reports inherently limits coverage, as unreported events in remote or censored areas result in undercounted fatalities, with estimates serving as conservative baselines rather than comprehensive totals.[2] Media biases and incomplete access to primary data can further skew inputs, particularly in conflicts with restricted information flows, leading to potential gaps in non-state or one-sided violence documentation.[2] The program's strict definitional criteria, while promoting consistency, exclude low-intensity organized violence below thresholds, obscuring patterns of emerging instability that broader datasets capture, as evidenced by divergences in event counts across global comparisons.[22][21] Coding practices have drawn methodological critique, notably for classifying state-inflicted civilian deaths as battle-related or incidental rather than deliberate one-sided targeting, which some analyses attribute to a statist bias favoring government narratives over evidence of intentional violence.[23] Such decisions, rooted in UCDP's focus on organized armed force between actors, may underemphasize asymmetric civilian victimization, as comparisons with detailed victim compilations reveal discrepancies in fatality attribution for specific cases like Kosovo.[24] Annual update cycles introduce time lags, delaying real-time utility compared to more frequent but less verified sources, though this prioritizes accuracy over immediacy.[25] Overall, while UCDP's rigor supports empirical scholarship, its constraints underscore the challenges of deriving causal insights from secondary reporting in opaque conflict environments.Datasets
State-Based Armed Conflicts
The UCDP/PRIO Armed Conflict Dataset records state-based armed conflicts, defined as a contested incompatibility concerning government and/or territory where the use of armed force between two parties—of which at least one is the government of a state—results in at least 25 battle-related deaths in a given calendar year.[15][11] The incompatibility must involve explicitly stated positions, such as control over government structures, replacement of regimes, or territorial claims like secession or autonomy.[15] Parties include the primary government and an organized opposition group capable of effective armed resistance; secondary state parties may intervene with troop commitments, potentially internationalizing internal conflicts.[15] Armed force entails the organized deployment of material weapons by both sides, excluding sporadic or unorganized violence.[15] Conflicts are categorized by type: extrasystemic (e.g., colonial wars, now rare), interstate (between governments), internal (government versus domestic non-state actors), and internationalized internal (internal conflicts with foreign state intervention).[11] Intensity levels distinguish minor armed conflicts (25 to 999 battle-related deaths per year) from wars (1,000 or more deaths per year).[15] The dataset structures data at the conflict-year and dyad-year levels, with variables including conflict identifier, country, year, warring parties (Side A as government, Side B as opposition), conflict type code, intensity level, and onset indicators for new conflicts.[11] Coverage spans 1946 to 2024, with annual updates incorporating verified data from multiple sources such as news reports, academic studies, and official records, ensuring consistency through rigorous coding protocols.[11] Battle-related deaths are estimated with best, low, and high figures derived from cross-verified incident reports, though the dataset focuses primarily on conflict incidence rather than exhaustive fatality counts (the latter handled in complementary UCDP datasets).[11] Version 25.1, released in 2025, maintains compatibility with related UCDP resources like the Battle-Related Deaths Dataset for integrated analysis.[11] This dataset enables tracking of global conflict patterns, such as the persistence of intrastate over interstate violence since the post-World War II era.[11]Non-State Conflicts and One-Sided Violence
The UCDP Non-State Conflict Dataset records instances of organized armed violence between non-governmental groups, defined as the use of armed force by two organized armed groups—neither of which is a state government—resulting in at least 25 battle-related deaths in a given calendar year.[26] Organized groups include formally structured entities with named leadership, as well as informally organized communal or militia groups exhibiting patterned violence.[26] The dataset covers the period from 1989 to 2024 and is structured as a conflict-year format, providing annual aggregates rather than event-level details.[11] Key variables encompass unique conflict and dyad identifiers for cross-dataset compatibility, group names and organizational levels (e.g., formal groups, political supporters, or communal actors), start and end dates, best estimates of fatalities, and indicators for external state troop involvement.[26] Data are derived primarily from the UCDP Georeferenced Event Dataset (GED), aggregating events sourced from global newswires, local news monitoring, and secondary reports from NGOs and academic works, with a two-stage verification process prioritizing reliable fatality estimates.[26] This dataset distinguishes non-state conflicts from state-involved violence by excluding dyads where a government participates, even indirectly, to avoid overlap with intrastate or other state-based categories.[26] Examples include communal clashes over resources in sub-Saharan Africa or inter-militia rivalries in regions like the Middle East, where neither side controls state apparatus.[27] The 25-death threshold ensures focus on sustained, organized violence rather than sporadic incidents, though it may undercount lower-intensity clashes due to definitional rigor.[26] Annual updates, such as version 25.1, incorporate refined estimates and new cases, enabling longitudinal analysis of trends like rising communal conflicts in Africa and Asia.[11] The UCDP One-Sided Violence Dataset captures deliberate attacks on unarmed civilians by state governments or formally organized non-state groups, requiring at least 25 civilian deaths per actor per year for inclusion.[28] Civilians are defined as non-combatants excluding security forces, militias, or government officials, with exclusions for deaths in custody or incidental to military operations.[28] Covering 1989 to 2024, it uses an actor-year structure to track perpetrator-specific violence, including variables such as actor identifiers and names, fatality estimates (best, low, and high), government status flags, locations by country and region, and compatibility codes for merging with other UCDP data.[11] Like the non-state dataset, it aggregates from the GED using a multi-source approach involving newswires, specialized reports, and verification against secondary literature to confirm intent and lethality.[28] One-sided violence encompasses phenomena such as government massacres or rebel group targeting of ethnic minorities, emphasizing direct, intentional civilian harm over collateral damage.[28] The actor-centric design allows disaggregation by perpetrator type, revealing patterns like higher incidences in Africa and Asia, where non-state actors often perpetrate against rival communities.[28] Version updates refine estimates based on emerging evidence, though limitations include reliance on reported events, potentially underrepresenting covert or under-documented atrocities in remote areas.[11] Together, these datasets expand UCDP's scope beyond state-centric conflicts, providing granular data for studying violence dynamics independent of governmental control.[11]Georeferenced Event Dataset (GED)
The Georeferenced Event Dataset (GED) represents the Uppsala Conflict Data Program's (UCDP) most granular data product, disaggregating organized violence into individual fatal events for spatial and temporal precision. It records incidents involving at least one direct death from armed force used intentionally by an organized actor against another organized group or civilians, spanning three categories: state-based conflicts (between governments and non-state groups), non-state conflicts (between organized non-state actors), and one-sided violence (deliberate attacks on non-combatants by organized actors). Events are georeferenced with latitude and longitude coordinates, timestamped (often to the exact day, otherwise month or year), and linked to specific actors, enabling analyses of violence patterns at subnational scales. Global in scope, the GED covers events from January 1, 1989, to December 31 of the prior year, with annual releases incorporating verified data from the latest period.[29][11] Data assembly begins with screening roughly 50,000 news reports annually from databases like Dow Jones Factiva, augmented by non-media sources such as NGO reports (e.g., from INSEC or the Syrian Observatory for Human Rights), UN documents, and local monitoring groups, which collectively identify candidate events. Trained coders, possessing contextual knowledge of the conflicts, extract details including actor identities, fatality counts per side, and locations, with entries rigorously vetted by project coordinators to ensure adherence to UCDP's operational definitions—requiring organized actors with command structures and purposeful use of deadly force. This yields 10,000 to 12,000 coded events per year, with non-media sources accounting for approximately 20% of entries from 2013 onward, though reliance on English-language and accessible reports may introduce gaps in remote or censored areas. Georeferencing employs place names, administrative boundaries, and coordinates, categorized by precision levels (e.g., exact site versus approximate region) to support geographic information system integration.[2] Core variables encompass a unique event identifier, precise date, geographic coordinates, actor dyads (Side A and B for conflicts, perpetrator for one-sided acts), best fatality estimates (disaggregated by actor where possible), and event type, with event IDs ensuring traceability to UCDP's aggregated country-year datasets on conflicts and deaths. The GED's structure allows aggregation into battle-related or civilian fatalities, facilitating compatibility with tools like the ViEWS forecasting system or PRIO collaborations, while its event-level resolution reveals micro-dynamics such as clustering or diffusion of violence overlooked in yearly totals. Version 25.1, issued in October 2025, preserves full compatibility with versions from 17.1 forward, barring minor adjustments, and emphasizes transparency through public codebooks detailing coding rules. Limitations include potential underreporting in low-visibility conflicts due to source dependencies and the exclusion of non-fatal violence or indirect deaths, though cross-verification mitigates biases where multiple reports converge.[30][2]Specialized Datasets
The Uppsala Conflict Data Program (UCDP) maintains several specialized datasets that extend beyond its core conflict typologies, targeting niche forms of organized violence such as attacks on peacekeepers, urban conflict events, and electoral-related fatalities. These datasets leverage the UCDP's foundational event-based coding from the Georeferenced Event Dataset (GED) but apply targeted filters and additional variables to address specific research and policy needs.[11] They emphasize empirical verification through media reports, NGO documentation, and official records, maintaining UCDP's threshold of at least 25 battle-related deaths per year for inclusion where applicable, while adapting definitions for context-specific violence.[31] The UCDP Peacemakers at Risk (PAR) Dataset documents organized violence against international peacekeepers, focusing on sub-Saharan Africa from 1989 to 2009. It records 1,057 events resulting in 1,495 fatalities, primarily from non-state actors like rebel groups, with data disaggregated by perpetrator type, target (e.g., military vs. civilian personnel), and location. Events are coded only if they meet UCDP's organized violence criteria, excluding accidental or criminal incidents, and draw from GED sources plus specialized peacekeeping reports. This dataset highlights vulnerabilities in UN and regional missions, such as the high incidence of ambushes (over 40% of events), aiding analyses of mission effectiveness and actor strategies.[31][32] The Cities and Armed Conflict Events (CACE) Dataset extends the UCDP GED by identifying conflict events occurring within urban areas, covering 1989 to 2017 across global conflicts. It includes over 10,000 urban events from the GED, with variables for city population, event type (state-based, non-state, or one-sided), and fatality counts, revealing that urban violence accounts for approximately 20-30% of total GED events in active conflicts. Coding relies on geospatial matching of GED coordinates to urban polygons from sources like the Global Human Settlement Layer, enabling studies on urbanization's role in intensifying violence dynamics, such as higher civilian casualties in densely populated settings. Limitations include potential underreporting in rapidly urbanizing areas due to GED's media-dependent sourcing.[33][34] The Deadly Electoral Conflict Dataset (DECO) compiles georeferenced events of electoral violence with at least one fatality, spanning 1989 to 2017 in countries holding national elections. It captures 2,456 events linked to 178 elections, primarily one-sided attacks on civilians or non-state clashes, with actors coded per UCDP standards and motives tied to electoral competition (e.g., intimidation or reprisals). Data integrates GED events filtered by temporal proximity to election dates (within 90 days) and verified electoral violence reports, showing spikes in fatalities during pre-election phases in regions like sub-Saharan Africa and South Asia. This dataset supports causal inquiries into how electoral incentives exacerbate organized violence, though it excludes non-lethal incidents and may overlook covert manipulations not meeting fatality thresholds.[35]Collaborations and Partnerships
Partnership with Peace Research Institute Oslo (PRIO)
The Uppsala Conflict Data Program (UCDP) established a formal collaboration with the Peace Research Institute Oslo (PRIO) in the early 2000s, culminating in the joint production of the UCDP/PRIO Armed Conflict Dataset. This partnership originated with the dataset's inaugural release in 2002, developed at PRIO through close coordination with UCDP researchers at Uppsala University's Department of Peace and Conflict Research.[36] The initiative merged UCDP's rigorous data collection methodologies with PRIO's expertise in conflict analysis, focusing on state-based armed conflicts defined by at least 25 battle-related deaths in a calendar year involving a state government as one party.[9] The UCDP/PRIO Armed Conflict Dataset covers conflicts from 1946 to the present, distinguishing between extrasystemic (interstate and colonial), interstate, internal, and internationalized internal types, with annual updates ensuring consistency in coding practices.[11] By 2025, the dataset version 25.1 included over 250 conflict entries, enabling dyadic-level analysis of belligerents and facilitating cross-disciplinary research on conflict patterns, such as state involvement in violence.[36] This collaboration has standardized global conflict tracking, with PRIO hosting mirrored data access and integrating UCDP metrics into its broader research portfolio, including recurrence studies.[37] The partnership extends beyond dataset maintenance to mutual reinforcement of data verification processes, where PRIO researchers contribute to refining UCDP's thresholds for conflict onset and termination, while UCDP provides foundational event data for PRIO's policy-oriented reports.[38] For instance, PRIO's 2025 conflict assessments explicitly rely on UCDP-sourced figures for battle deaths and active conflicts, highlighting a historically elevated global conflict level.[38] This symbiotic arrangement has amplified the datasets' credibility and usability, though it remains anchored in UCDP's primary empirical standards rather than diverging interpretations.[39]Other International and Academic Collaborations
The Uppsala Conflict Data Program (UCDP) has engaged in collaborations with the Human Security Report Project (HSRP) at Simon Fraser University in Vancouver, Canada, focusing on projects to backdate conflict data, analyze non-state conflicts, and document one-sided violence. These efforts have extended the temporal coverage of UCDP datasets and incorporated additional categories of organized violence beyond state-based armed conflicts.[7] UCDP data on armed conflicts has been incorporated into the Stockholm International Peace Research Institute (SIPRI) Yearbook annually from 1988 to 2017, enabling SIPRI's assessments of global conflict patterns to draw directly from UCDP's coding of incompatibilities, actors, and battle-related deaths. SIPRI appendices during this period explicitly reference UCDP definitions and sources for conflict identification, such as contested government or territory leading to at least 25 battle-related deaths per year.[40][41] Beyond these, UCDP maintains cooperative ties with various international research institutes and projects to broaden data comparability and methodological rigor, though specific joint outputs vary by initiative. These partnerships support UCDP's role in providing disaggregated violence data for cross-national studies, emphasizing empirical verification over narrative-driven interpretations.[7]Applications and Impact
Academic and Research Utilization
The Uppsala Conflict Data Program (UCDP) datasets serve as a foundational resource in academic research on armed conflict, providing disaggregated data on events, actors, and fatalities that enable quantitative analyses of conflict dynamics, causes, and outcomes.[5] Scholars frequently employ UCDP's Georeferenced Event Dataset (GED) and battle-related deaths estimates to model spatial patterns of violence, such as urban-rural distributions in intrastate conflicts, by coding events based on news reports and NGO documentation.[42] Similarly, the program's one-sided violence data has been used to examine government targeting of civilians, revealing potential underreporting of deliberate non-combatant deaths classified as incidental.[23] UCDP data's integration spans political science, economics, sociology, and interdisciplinary fields including environmental studies and public health, where it informs regressions on factors like climate variability's role in escalating violence or health impacts from displacement.[5] For instance, extensions of UCDP records have facilitated datasets on external support to belligerents from 1975–2017, allowing tests of hypotheses on foreign intervention's effects on conflict duration and intensity.[43] In conflict termination studies, UCDP's coding of endings—via victory, ceasefire, or unclear cessation—supports evaluations of post-conflict stability, with over 14,000 issues documented in non-state groups' goals aiding causal inferences on grievance-driven escalation.[44] As one of the most cited resources for organized violence, UCDP's rigorous, annually updated metrics underpin meta-analyses comparing global trends, such as rising non-state conflicts since 1989, while its transparency in definitions promotes replicability across peer-reviewed studies.[25][45] This utilization extends to validations against alternative sources, highlighting UCDP's threshold-based approach (e.g., 25 battle-related deaths per year) as a benchmark for reliability in low-intensity violence tracking, though researchers note dependencies on media-sourced events.[24]Policy and Media Influence
The Uppsala Conflict Data Program (UCDP) datasets have informed policy analyses by international financial institutions, including the International Monetary Fund (IMF), which employed UCDP conflict metrics in a 2024 working paper to assess macroeconomic risks and the urgency of conflict prevention strategies.[46] UCDP data also contribute to broader policy frameworks on peacebuilding, as evidenced by its integration into collaborative reports like the World Bank and United Nations' Pathways for Peace, which reference UCDP event data for evaluating societal risks and inclusive approaches to violence reduction.[47] These applications underscore UCDP's role in providing standardized, verifiable indicators of organized violence—such as battle-related deaths exceeding 25 per year—to support evidence-based foreign policy and international relations decision-making, with the program's definitions adopted in outlets like the Stockholm International Peace Research Institute (SIPRI) yearbooks since 1988. In media reporting, UCDP serves as a primary reference for quantifying global conflict trends, with its annual statistics on active armed conflicts—such as the record 61 state-involved conflicts in 2024—frequently cited to contextualize escalations in regions like Ukraine, Gaza, and the Sahel.[4] Major outlets including The New York Times, The Guardian, Reuters, and the Associated Press have invoked UCDP fatality and event data in articles on civilian casualties, with examples encompassing a 2024 New York Times analysis of yearly conflict surges through November and a Guardian examination of civilian death ratios in Gaza drawing on UCDP's global benchmarks since 1989.[48][49] Journalism resources, such as guides from the International Journalists' Network and DataJournalism.com, recommend UCDP for conflict coverage due to its reliance on corroborated sources like news reports and NGO documentation, enabling precise visualizations of violence patterns without inflating figures through unverified claims.[50][51] This usage enhances media accuracy by privileging empirical thresholds over anecdotal reporting, though it may undercount lower-intensity violence below definitional cutoffs.Contributions to Public Understanding of Conflicts
The Uppsala Conflict Data Program (UCDP) contributes to public understanding of armed conflicts by maintaining an open-access repository of datasets, visualizations, and analytical tools that enable non-experts to examine trends in organized violence. Since the 1970s, UCDP has compiled systematic records of state-based, non-state, and one-sided violence, defining armed conflict with consistent thresholds—such as at least 25 battle-related deaths in a calendar year for ongoing conflicts—which provide a standardized benchmark for assessing global patterns.[1] These resources, including the Georeferenced Event Dataset (GED) and dyadic conflict files, are freely downloadable, allowing journalists, educators, and policymakers to reference empirical data rather than anecdotal reports.[11] UCDP further disseminates insights through public-facing outputs like charts, graphs, and maps that illustrate key metrics, such as the number of active conflicts and battle-related deaths over time. For instance, annual updates reveal fluctuations in violence levels; in 2023 data extended to preliminary 2024 figures, UCDP documented over 50 state-based conflicts, highlighting a post-Cold War rise in intrastate warfare.[52] These visualizations counteract sensationalized media narratives by emphasizing verifiable fatalities and incompatibility types (e.g., government vs. territory control), fostering informed discourse on conflict drivers like resource disputes or ethnic tensions.[2] Press releases and bulletins from UCDP amplify these findings for broader audiences, as seen in the June 11, 2025, announcement of a sharp increase in conflicts and wars, marking the highest levels since systematic tracking began.[4] By prioritizing transparency—sourcing from public reports while acknowledging underreporting risks—UCDP equips the public with tools to evaluate claims about "world peace" or escalating threats, such as the role of non-state actors in over 40% of recent conflicts.[2] This approach has positioned UCDP as a primary reference for global violence data, influencing public perceptions through its emphasis on longitudinal trends over isolated events.[1]Criticisms and Methodological Debates
Accuracy and Fatality Reporting Issues
The Uppsala Conflict Data Program (UCDP) has faced scrutiny for potential underreporting of fatalities due to its reliance on open-source media reports, which are often incomplete in remote, high-intensity, or censored conflict zones.[53] This methodological choice prioritizes verifiable events but introduces biases toward larger, more visible incidents, as evidenced by event size discrepancies where smaller-scale killings are systematically underrepresented.[24] In the Kosovo conflict (1998–June 1999), a comparison of UCDP's Georeferenced Event Dataset (GED) with the Kosovo Memory Book's name-by-name compilation of 11,536 victims revealed significant undercounting: GED best estimates captured only about 33% of total fatalities, with high estimates reaching 80.5%, and the most severe gaps occurring during the peak intensity period of March–June 1999, missing thousands of deaths.[24] Only 6–10.4% of Memory Book events matched GED records, highlighting an underreporting bias for smaller events and those in areas with limited media access, though no clear evidence of observer deployment bias was found.[24] Structured expert elicitation involving UCDP coders estimated underreporting rates varying by event scale, with near-100% undercount for single-fatality events dropping to about 30% for those with 100 reported deaths; for instance, in Syria (2020–2022), true fatalities were likely 60% higher than UCDP figures.[54] These probabilistic models, using Gumbel distributions and covariates like violence type, underscore how UCDP's strict inclusion criteria for organized actors and verifiable sources contribute to conservative estimates, potentially understating total tolls without capturing overreporting risks in very large events.[54] Critics highlight exclusionary definitions, such as omitting violence by unnamed or anonymous groups, leading to stark discrepancies; in Mexico (2021), UCDP-GED recorded just 28 civilian fatalities versus over 6,700 in comparable datasets, obscuring patterns of non-state instability.[22] Similarly, UCDP's coding often classifies government-inflicted civilian deaths as incidental or battle-related rather than deliberate one-sided targeting, which may undercount systematic abuses while emphasizing precision over comprehensive recall.[23] Comparisons with datasets like ACLED reveal fatality gaps—e.g., 64,315 more deaths reported in Yemen (2015–2018)—attributed partly to UCDP's stricter auxiliary coding for vague reports (e.g., assigning 2 versus 10 deaths) and avoidance of partisan sources like Houthi media, which ACLED incorporates more liberally.[55] While some attribute this to underreporting from UCDP's higher thresholds, analyses indicate methodological differences in source evaluation and rules explain most variances, with UCDP's conservatism reducing bias but at the cost of lower totals.[55]Allegations of Bias in Coding and Definitions
Scholars have alleged that the Uppsala Conflict Data Program (UCDP) exhibits a statist bias in its coding practices, particularly in classifying civilian deaths inflicted by governments. According to a 2025 analysis by Austin Doctor and Reed M. Wood, UCDP coders consistently categorize civilian fatalities caused by state actors—such as through airstrikes or artillery—as "battle-related" deaths incidental to combat operations, rather than as deliberate "one-sided violence" against civilians, even when evidence suggests intentional targeting.[23] This approach, the authors argue, stems from UCDP's stringent evidentiary requirements for proving intent, which rely heavily on publicly available reports often shaped by government narratives and access restrictions in conflict zones controlled by states.[56] The alleged bias manifests in the application of UCDP's definitions for organized violence categories. UCDP distinguishes "one-sided violence" as the deliberate use of armed force by organized actors (states or rebels) against unarmed civilians, excluding incidental casualties in battle settings.[57] Critics contend that this threshold is applied asymmetrically: non-state actors' attacks on civilians are more readily coded as one-sided due to clearer attribution from media and witness accounts, whereas state actions—especially long-range attacks like drone strikes or shelling—are defaulted to battle-related if any combatant presence is nearby, undercounting deliberate civilian targeting by an estimated margin in datasets like the UCDP Georeferenced Event Dataset (GED) version 23.1.[23] For instance, the study reviewed events where independent reports indicated indiscriminate state bombardment of populated areas, yet UCDP coded them without intent, potentially skewing global estimates of state-perpetrated atrocities.[56] This coding practice has broader implications for research and policy, as it may systematically minimize the scale of government accountability in conflicts, affecting analyses of genocide, mass atrocities, and civilian protection norms. Doctor and Wood propose revising UCDP's one-sided violence definition to encompass "indiscriminate" attacks on civilian areas as presumptively deliberate, arguing that current rules perpetuate a pro-state bias exacerbated by data collection dependencies on official sources.[23] While UCDP maintains that rigorous, consistent definitions prevent subjective over-attribution, detractors highlight that such methodological conservatism, combined with source limitations, favors powerful actors who control information flows.[55] No widespread allegations of partisan political bias (e.g., favoring specific ideologies) have been substantiated in peer-reviewed critiques, though the statist tilt raises questions about neutrality in state-centric academic environments.[23]Comparisons with Alternative Datasets
The Uppsala Conflict Data Program (UCDP) datasets differ from alternatives such as the Correlates of War (COW) project and the Armed Conflict Location & Event Data Project (ACLED) primarily in definitional thresholds, scope of violence included, and methodological rigor for establishing organized armed groups. UCDP defines a state-based armed conflict as involving at least 25 battle-related deaths in a calendar year between a government and an organized non-state group, or between non-state groups over territory or government, emphasizing identifiable organizations with command structures and excluding sporadic or unorganized violence.[58] In contrast, COW employs a higher threshold of 1,000 battle-related deaths per year to classify events as "wars," focusing mainly on interstate wars, extrasystemic (colonial) wars, and civil wars with significant state involvement, which results in fewer but more intense conflicts recorded compared to UCDP's broader "conflict" category that captures lower-intensity organized violence.[59] ACLED, an event-based dataset, adopts looser criteria by coding discrete political violence events without requiring a fixed annual death threshold for inclusion, encompassing a wider array of actors including unidentified armed groups and subnational violence that may not meet UCDP's organization standards, leading to substantially higher event counts—often 5-10 times more in overlapping regions like sub-Saharan Africa.[60] Methodological differences further highlight variances in reliability and coverage. UCDP relies on a multi-source aggregation approach, prioritizing low estimates from credible reports (e.g., NGOs, media, governments) to minimize inflation, with manual coding by trained researchers to verify organization and causality, covering global organized violence from 1989 onward in its core datasets but extending to 1946 for conflicts.[19] COW, established in the 1960s, uses historical archives and emphasizes dyadic interstate relations with less granular event-level data, resulting in retrospective adjustments but potential undercounting of non-state conflicts post-Cold War due to its state-centric focus.[59] ACLED employs near-real-time coding from media and humanitarian reports, enabling rapid updates but introducing higher variability from source biases or unverified actors, as it includes events like riots or clashes without strict fatality verification, which can inflate totals in volatile areas while providing finer spatiotemporal resolution absent in UCDP's annual aggregates.[60] These approaches yield divergent trends; for instance, UCDP reported 56 state-based conflicts in 2022 with conservative death estimates, whereas ACLED logged over 200,000 events globally that year, many below UCDP's thresholds.[11]| Dataset | Death Threshold | Key Focus | Temporal Coverage | Strengths | Limitations |
|---|---|---|---|---|---|
| UCDP | 25+ battle-related deaths/year for conflicts | Organized armed violence (state-based, non-state, one-sided); requires identifiable groups | 1946–present (events from 1989) | High reliability via conservative estimates and organization verification; consistent for trend analysis | May undercount low-intensity or emerging violence due to thresholds |
| COW | 1,000+ for wars | Interstate, civil, and colonial wars with state centrality | 1816–present | Long historical depth; standardized for macro-level war studies | Fewer cases; misses non-war organized violence; less updated for recent non-state dynamics[59] |
| ACLED | No fixed annual threshold; event-based | Political violence events, including unidentified actors | 1997–present (real-time) | Granular, timely data for policy; broad subnational coverage | Potential overcounting from loose actor definitions; reliant on media sources prone to under- or over-reporting[60] |
Recent Developments
Updates and Expansions in the 2020s
In 2020, the UCDP introduced the Candidate Events Dataset, a monthly updated collection of recent organized violence events intended to facilitate timely analysis and serve as input for the Georeferenced Event Dataset (GED).[61] This expansion addressed gaps in near-real-time data availability, covering events from 2019 onward with preliminary coding for actors, fatalities, and locations.[61] The program further expanded in 2022 with the release of the External Support Dataset (ESD), which documents third-party interventions in intrastate conflicts from 1975 to 2017, including troop deployments, logistical aid, and sanctions across 331 conflicts.[43] This dataset enables disaggregated study of external influences on conflict dynamics, distinguishing between support to governments and non-state actors.[43] In 2023, the UCDP Conflict Issues Dataset (CID) was launched, coding primary goals of non-state actors in state-based conflicts since 1946, such as territorial control or regime change, based on manifestos and public statements.[62] These additions complemented ongoing annual updates to core UCDP datasets, including extensions of battle-related deaths and non-state violence series through 2024.[11] Collaborations, such as integration with Our World in Data in early 2024, improved public access to UCDP/PRIO conflict metrics.[63] No major methodological overhauls were reported, but enhanced event-level granularity supported finer-grained empirical research on violence patterns.[45]Trends in Global Conflict Data for 2024
In 2024, the Uppsala Conflict Data Program (UCDP) recorded 61 state-based armed conflicts across 36 countries, marking a historic high since systematic data collection began in 1946 and representing an increase from 59 conflicts in 2023.[64][65] This rise included 11 wars—defined as conflicts with at least 1,000 battle-related deaths—up from 9 the previous year and the highest number since 2016, alongside a doubling of interstate conflicts to 4.[65] The escalation reflected a broader resurgence of large-scale warfare, driven by ongoing high-intensity engagements such as the Russia-Ukraine war and the Israel-Hamas conflict.[64] Fatalities in state-based conflicts totaled approximately 128,400 battle-related deaths, a slight decline from 131,000 in 2023, though 2024 ranked as the fourth most violent year for organized violence since 1989.[65] The Russia-Ukraine conflict remained the deadliest, accounting for around 76,000 battle-related deaths, while the Israel-Hamas conflict contributed over 21,000.[4][65] Overall organized violence fatalities hovered near 160,000, marginally lower than the 160,300 recorded in 2023, with challenges in distinguishing civilian from combatant deaths complicating precise tallies in urban and asymmetric settings.[65] Non-state conflicts decreased modestly to 74 from 80 in 2023, yielding about 17,500 battle-related deaths, amid regional shifts including a sharp rise in Africa—the highest levels on record there—and a decline in the Americas.[64][65] One-sided violence against civilians intensified, involving 49 actors (35 non-state groups and 14 governments) and resulting in 13,900 deaths, an increase from 10,600 the prior year, largely attributable to groups like the Islamic State in Africa and gangs in Haiti.[65] These patterns underscored persistent drivers of violence, including territorial disputes and insurgencies, despite marginal reductions in some categories.[64]| Conflict Type | Number in 2024 | Change from 2023 | Battle-Related Deaths (approx.) |
|---|---|---|---|
| State-based | 61 | +2 (from 59) | 128,400 |
| Non-state | 74 | -6 (from 80) | 17,500 |
| One-sided | 49 actors | Increase | 13,900 (civilians) |