Counting and documenting conflict-related sexual violence: Data and methods

The paper analyzes how data collection methods shape the understanding of conflict-related sexual violence (CRSV). It compares 'top-down' approaches, like the Sexual Violence in Armed Conflict (SVAC) dataset, which rely on high-threshold fatality counts and a limited number of reputable sources, with 'bottom-up' approaches, such as the Armed Conflict Location & Event Data (ACLED) dataset, which track individual political violence events without fatality thresholds. The author argues that sourcing diversity, language considerations, and the choice between human-led and automated coding significantly impact the prevalence and accuracy of reported CRSV. The document concludes that all data contain biases—including errors of omission, inflation, and misrepresentation—and urges users to rigorously evaluate a dataset's assumptions to avoid inaccurate policy and programmatic conclusions.

• The Sexual Violence in Armed Conflict (SVAC) dataset employs a 'top-down' approach that may overlook peripheral conflicts or emerging actors because it requires a conflict to result in 25 deaths in a calendar year before it is coded.
• The Armed Conflict Location & Event Data (ACLED) dataset uses a 'bottom-up' approach that captures a wider range of sexual violence events because it does not require a fatality threshold for inclusion, though this can make it harder to compare specific conflicts over time.
• Sourcing strategies significantly impact data reliability; while the SVAC dataset relies on a small number of highly reliable sources (US State Department, Amnesty International, and Human Rights Watch), this limits its temporal updates and may omit 'true positives' that lack rigorous evidentiary proof.
• Automated data collection systems using artificial intelligence, such as GDELT or ICEWS, are faster and cheaper than human-led projects but are prone to inaccuracies, including false positives, duplicates, and a lack of strict definitions.
• Data biases can manifest as errors of omission, such as the SVAC dataset excluding sexual violence committed by mercenaries or multilateral entities, or errors of inflation, where automated systems mislocate events to a country's center-point.
• The reliance of Western data collection on English or European languages often leads to the under-reporting of CRSV, as the most reliable local-level reporting is frequently conducted in local languages.

APA

Kishi, R. (2025). Counting and documenting conflict-related sexual violence: Data and methods. Peace Research Institute Oslo. https://cdn.cloud.prio.org/files/001872b9-d950-4ab0-9341-1d403aa24f8e/MPI Paper 4 Kishi FINAL.pdf?inline=true

Chicago

Kishi, Roudabeh. Counting and documenting conflict-related sexual violence: Data and methods. Peace Research Institute Oslo, 2025. https://cdn.cloud.prio.org/files/001872b9-d950-4ab0-9341-1d403aa24f8e/MPI Paper 4 Kishi FINAL.pdf?inline=true.

Wikipedia

{{cite report |last1=Kishi |first1=Roudabeh |title=Counting and documenting conflict-related sexual violence: Data and methods |publisher=Peace Research Institute Oslo |date=December 2025 |url=https://cdn.cloud.prio.org/files/001872b9-d950-4ab0-9341-1d403aa24f8e/MPI Paper 4 Kishi FINAL.pdf?inline=true |access-date=23 October 2026 |via=Climate Insights Directory}}

BibTeX

@techreport{kishi2025counting, author = {Kishi, Roudabeh}, title = {{Counting and documenting conflict-related sexual violence: Data and methods}}, institution = {Peace Research Institute Oslo}, year = {2025}, month = dec, url = {https://cdn.cloud.prio.org/files/001872b9-d950-4ab0-9341-1d403aa24f8e/MPI Paper 4 Kishi FINAL.pdf?inline=true}, urldate = {2026-10-23}, note = {Indexed by Climate Insights Directory} }