Geographic Distribution of Political Violence Events
The interactive map below shows individual political violence events from the POLVITED dataset. Each circle represents one event; circle size scales with the fatality count. Use the filters to explore events by year, actor, type of violence, attack type, and target type. Click any circle for full event details.
Each circle marks one geo-coded event in POLVITED (1989–2020). Circle radius scales with fatality count — a minimum radius is shown for zero-fatality events. Click a circle to see actor, date, type, attack type, target type, and fatality details. Use scroll to zoom; drag to pan.
Abstract
This study introduces the Political Violence in Türkiye Event Dataset (POLVITED), a resource designed to provide a more comprehensive estimation of political violence incidents and fatalities in Türkiye. Reporting biases, along with variations in the scope and definitions of existing event datasets, significantly affect the quantitative study of conflict and violence. To address these challenges, this study employs the Matching Event Data by Location, Time, and Type (MELTT) methodology to integrate data from the Global Terrorism Database (GTD) and the Uppsala Conflict Data Programme Georeferenced Event Dataset (UCDP-GED).
The findings reveal that UCDP-GED captures approximately 60% of political violence incidents in Türkiye, while GTD records approximately 40%; only 15% of incidents are documented in both datasets. By eliminating duplicate records and mitigating reporting bias, POLVITED offers a more comprehensive representation of political violence in Turkey over the period 1989–2020. Additionally, a comparison with official reports from the Turkish Grand National Assembly (TBMM) demonstrates that while reporting bias cannot be entirely eradicated, data integration significantly enhances dataset coverage and reliability.
Keywords: Political Violence, Terrorism, Civil War, Turkey, MELTT, Dataset Integration, POLVITED
1. Introduction
Research on terrorism and civil war has made significant progress in uncovering the drivers and mechanisms behind political violence, as well as its far-reaching consequences. Scholars have developed sophisticated theoretical frameworks and quantitative methods to analyze these phenomena, deepening our understanding of conflict dynamics. However, an important dimension remains underexplored: the interconnected nature of different forms of political violence.
Current research often isolates terrorism from civil war, leaving unresolved questions about how these forms overlap and influence each other. The lack of universally accepted definitions for various types of political violence has created challenges in distinguishing between them, introducing ambiguity into both conceptual and empirical analyses.
This study leverages advancements in data integration methodologies to address these challenges. Specifically, I apply the Matching Event Data by Location, Time, and Type (MELTT) method (Donnay et al. 2019) to integrate two of the most widely used datasets on political violence: the Global Terrorism Database (GTD) (LaFree and Dugan 2007) and the Uppsala Conflict Data Programme Georeferenced Event Dataset (UCDP-GED) (Sundberg and Melander 2013). The MELTT method facilitates the systematic identification and removal of duplicate records across these datasets, enabling the creation of a novel resource containing uniquely coded events.
I use Türkiye as a case study to estimate the total number of fatalities resulting from political violence and to demonstrate how and when terrorism is used as a tactic by rebel groups within the context of a civil war. Three key factors make Türkiye particularly suitable: (1) the infrastructure for news reporting has largely remained intact, ensuring relatively consistent coverage; (2) the fluctuating levels of political violence help minimize reporting fatigue; and (3) official reports on political violence are accessible, including the Committee on Human Rights Inquiry of the Grand National Assembly of Türkiye (TBMM 2013).
The contributions of this paper are threefold. First, it establishes a theoretical link between terrorism and civil war, emphasizing the overlapping concepts that introduce ambiguities in data collection. Second, it demonstrates how this theoretical framework can inform and guide empirical research. Third, it provides statistical estimates of political violence through POLVITED.
2. Theoretical Framework
2.1 Methodological Concerns of Data Collection on Political Violence
Two primary biases are commonly discussed in the political science literature as critical challenges to the reliability of empirical studies using event datasets. The first is the variation in scope and definitions employed by different datasets, and the second is the reporting bias introduced by news agencies, governmental archives, or non-governmental organizations. Both biases significantly impact the accuracy and reliability of conflict research.
2.2 Differences in Definitions and Scope
Each dataset relies on its own methodology and coding rules. Sambanis (2004) demonstrates how definitional differences among datasets can significantly impact analyses by applying the same probit model to ten civil war datasets and showing these differences lead to over- or underestimation of correlations.
The definition and scope of political violence remain contested. Civil war is defined by UCDP as “a contested incompatibility that concerns government or territory, over which the use of armed force between the military forces of two parties, at least one of which is the government of a state, has resulted in at least 25 battle-related deaths each year” (Gleditsch et al. 2002). The COW project requires “a total of at least 1,000 battle-deaths during each year of the war” (Sarkees and Wayman 2010). The definition of terrorism is even more contested (Weinberg et al. 2004; Ganor 2002), with empirical studies suggesting that guerrilla groups frequently adopt terrorist tactics, blurring the boundary between categories (Stanton 2013; Fortna et al. 2020; Polo and Gleditsch 2016).
Overlap Between Terrorism and Civil War
Both terrorism and civil war share the objective of achieving political goals and often rely on similar means. Several studies highlight that terrorism is frequently employed as a tactic in civil wars by rebel organizations (Fortna et al. 2020; Stanton 2013; Polo and Gleditsch 2016). This suggests that the boundary between terrorism and armed warfare is far from clear-cut — a key motivation for the POLVITED integration approach.
2.3 Reporting and Underreporting Bias
A second major bias arises from the data collection process. Conflict data are compiled from primary or secondary sources and rely on reports aggregated from media, governmental, or NGO sources. Reporting bias — the systematic overreporting or underreporting of certain events — is especially prevalent in media sources (Weidmann 2016) and can be categorized as selection bias (systematic differences in decisions to report specific events) or description bias (variations in how the same event is reported; Davenport and Ball 2002).
Cubukcu and Forst (2018) compare official Turkish terrorism reports with the GTD and uncover significant discrepancies influenced by “incident characteristics and rational factors (especially newsworthiness).” Drakos and Gofas (2006) further show that press freedom in democracies inflates recorded terrorism incidents, while underreporting in autocracies distorts the relationship between regime type and terrorism.
Government reports are also not exempt. The TBMM Committee on Human Rights Inquiry Report (TBMM 2013) — released publicly in 2013 during the Peace Process — provides the only systematic official document on PKK-related fatalities for 1989–2012. It excludes military personnel and police and records only civilian and PKK combatant fatalities, providing a conservative lower bound on PKK-related violence.
2.4 Integration as a Solution
Researchers have increasingly turned to integrating multiple datasets as a methodological solution (Donnay et al. 2019). Polo and Gleditsch (2016) link UCDP actors to the GTD to examine terrorist tactics in civil wars; Fortna et al. (2020) developed the Terrorism in Armed Conflict dataset by integrating UCDP rebel organizations with GTD records. Such integrated datasets reveal patterns that remain obscured when relying on single-source data.
Integration also helps detect reporting bias: different datasets omit variable values in distinct ways. By comparing discrepancies, researchers gain insights into the nature of reporting bias. Furthermore, the GTD records events even when the perpetrator is unknown and when the event results in zero fatalities — providing coverage that UCDP-GED (which requires at least one known fatality and a known organization) cannot.
3. Data and Methodology
3.1 Global Terrorism Database (GTD)
The GTD, maintained by the National Consortium for the Study of Terrorism (START) at the University of Maryland, records terrorist attacks worldwide (LaFree and Dugan 2007). For an incident to be included, it must satisfy three mandatory conditions: (1) the incident must be intentional; (2) it must entail some level of violence or immediate threat of violence; and (3) the perpetrators must be sub-national actors. Furthermore, each incident must meet at least two of three additional criteria related to political goal, coercive intent, and targeting of non-combatants.
| Inclusion Criteria | Number of Events |
|---|---|
| Criteria 1 and 2 | 799 |
| Criteria 2 and 3 | 28 |
| Criteria 1 and 3 | 19 |
| Criteria 1, 2 and 3 | 2,801 |
Note: 28% of GTD events in Türkiye do not meet Criterion 3 (targeting non-combatants), suggesting potential overlap with UCDP-GED armed conflict events.
3.2 UCDP Georeferenced Event Dataset (UCDP-GED)
The UCDP-GED records incidents of organized violence at the sub-national level (Sundberg and Melander 2013). Each incident must result in at least one known fatality and must be perpetrated by organized actors. UCDP-GED categorizes incidents into three types: state-based violence, non-state conflict, and one-sided violence.
| Type of Violence | Number of Events |
|---|---|
| State-based Violence | 4,861 |
| Non-State Conflict | 14 |
| One-sided Violence | 442 |
4. Results
The POLVITED dataset encompasses 8,069 deduplicated events for Türkiye between 1989 and 2020, combining the GTD’s coverage of terrorism with UCDP-GED’s coverage of armed conflict. UCDP-GED captures approximately 60% of total political violence incidents; GTD captures approximately 40%; and 15% are recorded in both datasets. The figures below reproduce and extend the core empirical results of the study.
Figure 4 – Number of Fatalities per Year in POLVITED (by Type of Violence)
Annual fatality counts broken down by type of political violence. Toggle individual categories by clicking legend entries. State-based violence (primarily PKK-related armed conflict) dominates the 1990s period.
Figure 5 – Number of Political Violence Incidents per Year in POLVITED
Annual event counts broken down by type of political violence. The year 2016 recorded the highest number of events (620), combining PKK-related clashes, ISIL attacks, and the coup attempt.
Figure 6 – Number of Fatalities by Actor (2000–2020)
Annual fatalities attributed to the four most prominent actors in the post-1999 period. PKK and affiliated groups account for over 75% of all events and the vast majority of fatalities.
Figure 3 – Matched (Duplicate) Events per Year by Type of Violence
Events identified as duplicates across GTD and UCDP-GED, classified by UCDP violence type and GTD terrorism category. The MELTT algorithm identified – matched event pairs.
Characteristics of Matched Events
The tables below characterise the – events that appear in both the GTD and UCDP-GED, providing insight into the types of incidents most likely to be recorded by both datasets.
Table 6a – Attack Types (GTD side of matched events)
| Attack Type | Count | Share |
|---|
Table 6b – Weapon Types (GTD side of matched events)
| Weapon Type | Count | Share |
|---|
Table 6c – Target Types (GTD side of matched events)
| Target Type | Count | Share |
|---|
Table 5 – Conflict Names (UCDP-GED side of matched events)
| Conflict Name | Count | Share |
|---|
6. Validation: Comparison with TBMM Official Report
As an external validation of the POLVITED dataset, the merged fatality counts are compared against the Terör ve şiddet Olayları Kapsamında Yaşam Hakkı İhlallerini İnceleme Raporu (Report on Violations of the Right to Life Within the Scope of Terrorism and Violence Events) published by the Committee on Human Rights Inquiry of the Grand National Assembly of Türkiye (TBMM) in 2013 (TBMM 2013).
This report provides official data on PKK-related fatalities for the period 1989–2012. It is important to note that the TBMM report excludes military personnel and police, recording only PKK combatant fatalities and civilian victims. It therefore provides a conservative lower bound on PKK-related violence fatalities, while POLVITED captures total fatalities across all violence types from all actors.
Figure 7 – POLVITED Annual Fatalities vs. TBMM Official Report (1989–2012)
The POLVITED merged dataset (green) is compared with the GTD terrorism subset (blue), the UCDP-GED conflict subset (dark green), and the official TBMM parliamentary report (orange). The TBMM figure covers only PKK-related civilian and combatant deaths, excluding security forces. Despite this narrower scope, the POLVITED merged dataset closely tracks the TBMM temporal pattern, providing external validation for the integration approach.
Key Observation
POLVITED fatality counts consistently exceed TBMM figures, as expected given that POLVITED includes security force fatalities not counted in the TBMM report. The temporal pattern — a sharp peak in the mid-1990s followed by a dramatic decline after 1999 (Öcalan’s capture) and another rise post-2015 — is consistent across all datasets and the official TBMM report, lending strong external validity to the POLVITED integration approach.
7. Key Findings and Conclusion
This paper aimed to create a novel dataset on political violence in Türkiye to estimate the number of incidents and fatalities. In the process of constructing POLVITED, the study evaluated matched entries from the GTD and UCDP-GED to theoretically compare the definitions of terrorism and organized violence. The analysis revealed that the distinction between terrorism and organized violence becomes blurred in cases where rebel groups actively engaged in conflict with a government use terrorism as a strategic tool.
📉 Both Datasets Undercount Separately
Neither GTD nor UCDP-GED alone captures the full picture of political violence. GTD misses much of the organized armed conflict recorded in UCDP-GED, while UCDP-GED’s thresholds exclude many smaller terrorist incidents. UCDP-GED captures ~60% of incidents; GTD ~40%; 15% overlap.
📈 1990s: Peak Violence Period
The period 1992–1999 saw the highest levels of political violence, driven by the PKK insurgency. Fatalities peaked at 4,270 in 1994. This period is most completely captured by UCDP-GED’s state-based conflict category.
🔄 2015–2016: Second Escalation
Following the breakdown of the Peace Process (2013–2015), political violence escalated sharply. The year 2016 recorded 620 events, combining PKK clashes, ISIL attacks, and the military coup attempt. The merged data captures both dimensions.
👥 PKK Dominates the Violence Landscape
PKK and affiliated groups (PKK, TAK) account for over 75% of all events and the vast majority of fatalities across the entire study period (1989–2020).
🌎 ISIL Added a New Dimension post-2013
After 2013, ISIL-attributed attacks contributed significantly to terrorism fatalities, including mass-casualty bombings in Suruç (2015), Ankara (2015), and Istanbul (2016). These were well-captured by GTD but largely absent from UCDP-GED.
✅ External Validation via TBMM Report
POLVITED’s annual fatality counts closely track the TBMM parliamentary report for 1989–2012, providing external validity for the integration approach, despite the TBMM’s narrower scope (PKK-related deaths only, excluding security forces).
🏷 Classification Divergence
GTD classifies most PKK incidents as “terrorism,” while UCDP-GED categorizes the same events as “state-based armed conflict.” This definitional divide is why merging — rather than simply combining — the two datasets is methodologically essential.
📍 Geographic Concentration
Political violence is heavily concentrated in southeastern Turkey (Diyarbakır, Şırnak, Hakkari, Mardin), with periodic incidents in major western cities (Istanbul, Ankara) during ISIL and DHKP-C attacks.
📅 2000–2014: Relative Quiet
Following Öcalan’s capture in 1999 and subsequent PKK ceasefires, the period 2000–2012 saw dramatically lower violence. This lull is consistent across all datasets and the TBMM report.
Conclusion
POLVITED provides a robust, comprehensive dataset for analyzing political violence in Türkiye, bridging the gaps left by existing data collection efforts. By including incidents with zero fatalities, unknown perpetrators, and removing arbitrary thresholds for start dates, POLVITED enables tracking of the evolution of terrorist groups and their tactics over time. The dataset opens new avenues for researchers to explore when and why rebel groups employ terrorism as a strategy, or transition from terrorism to civil war.
The comparison with official TBMM figures demonstrates that data integration can significantly enhance dataset coverage and reliability, even if reporting bias cannot be entirely eradicated. The new data will help researchers better understand the nature of violence and the relation between groups.
8. Data Availability and Replication
All data files and analysis code are available if requested.
The dataset was constructed using R with the meltt, ggplot2, dplyr,
tidyverse, ggmap, leaflet, sf,
rnaturalearth, and lubridate packages.
9. Citation and References
How to Cite
References
- Baum, M.A. & Zhukov, Y.M. (2015). Filtering revolution: Reporting bias in international newspaper coverage of the Libyan civil war. Journal of Peace Research, 52(3), 384–400.
- Cubukcu, S. & Forst, B. (2018). Measuring Terrorism. Homicide Studies, 22(1), 94–116.
- Davenport, C. & Ball, P. (2002). Views to a kill: Exploring the implications of source selection in the case of Guatemalan state terror. Journal of Conflict Resolution, 46(3), 427–450.
- Donnay, K., Dunford, E.T., McGrath, E.C., Backer, D. & Cunningham, D.E. (2019). Integrating Conflict Event Data. Journal of Conflict Resolution, 63(5), 1337–1364.
- Drakos, K. & Gofas, A. (2006). The devil you know but are afraid to face: Underreporting bias and its distorting effects on the study of terrorism. Journal of Conflict Resolution, 50(5), 714–735.
- Fortna, V.P., Lotito, N.J. & Rubin, M.A. (2020). Terrorism in armed conflict: new data attributing terrorism to rebel organizations. Conflict Management and Peace Science.
- Ganor, B. (2002). Defining Terrorism: Is One Man’s Terrorist another Man’s Freedom Fighter? Police Practice and Research, 3(4), 287–304.
- Gleditsch, N.P., et al. (2002). Armed conflict 1946–2001: A new dataset. Journal of Peace Research, 39(5), 615–637.
- LaFree, G. & Dugan, L. (2007). Introducing the global terrorism database. Terrorism and Political Violence, 19(2), 181–204.
- Polo, S.M. & Gleditsch, K.S. (2016). Twisting arms and sending messages: Terrorist tactics in civil war. Journal of Peace Research, 53(6), 815–829.
- Sambanis, N. (2004). What is civil war? Conceptual and empirical complexities of an operational definition. Journal of Conflict Resolution, 48(6), 814–858.
- Sarkees, M.R. & Wayman, F. (2010). Resort to War: 1816–2007. CQ Press.
- Stanton, J.A. (2013). Terrorism in the Context of Civil War. The Journal of Politics, 75(4), 1009–1022.
- Sundberg, R. & Melander, E. (2013). Introducing the UCDP Georeferenced Event Dataset. Journal of Peace Research, 50(4), 523–532.
- TBMM. (2013). Terör ve şiddet olayları kapsamında yaşam hakkı ihlallerini inceleme raporu (pp. 1–327). Grand National Assembly of Türkiye, Committee on Human Rights Inquiry.
- Weidmann, N.B. (2016). A Closer Look at Reporting Bias in Conflict Event Data. American Journal of Political Science, 60(1), 206–218.
- Weinberg, L., Pedahzur, A. & Hirsch-Hoefler, S. (2004). The challenges of conceptualizing terrorism. Terrorism and Political Violence, 16(4), 777–794.