HRDAG uses statistics to reveal gaps in human rights records
Published on 18 September 2026 at 02:53 GMT
By Editorial Team SDG16
A record of human rights violations can contain thousands of names and still leave the scale of violence unresolved. The Human Rights Data Analysis Group (HRDAG) examines that gap, using statistical methods to analyse incomplete documentation. Its work in Syria and Colombia shows how recorded cases and estimated totals answer different questions, with consequences for the public record of conflict and the institutions examining it.
The distinction remains relevant after a report is published. HRDAG provides both original and corrected versions of data from its Colombian collaboration, alongside a methodological report updated in March 2025. That publication history illustrates a continuing task in human rights data analysis: establishing what a figure measures, which information produced it and whether subsequent corrections affect its interpretation.
What the records can show
According to HRDAG, documentation is shaped by geography, resources and the willingness of victims or witnesses to speak to a particular organisation. Some communities are accessible to investigators; others are difficult or dangerous to reach. The resulting records are not automatically representative of everyone affected. A larger database can therefore contain extensive evidence while still giving an uneven picture of violence.
Its methods also address repeated reporting. Several organisations may record the same person or incident, making the number of entries different from the number of victims. Matching records helps avoid counting someone more than once. In suitable projects, overlaps between sources also support multiple systems estimation, a method used to estimate violations absent from all the available lists.
These operations serve different purposes. Removing duplicates clarifies the documented count. Estimation addresses the larger population, including cases never recorded. Neither process turns an incomplete archive into a census. Estimates depend on the available data and statistical model, while a documented count describes the cases identified in the sources.
Syria and the difference between counts and estimates
In June 2022, the Office of the United Nations High Commissioner for Human Rights (OHCHR) reported an estimated 306,887 civilian deaths directly caused by the conflict in Syria between 1 March 2011 and 31 March 2021. Its briefing distinguished 143,350 individually documented deaths from a further estimated 163,537 deaths. The total was a statistical estimate, not a list of that many individually identified people.
HRDAG states that its team carried out the statistical analysis in consultation with Syrian partners and UN experts. Its account gives a 95 per cent credible interval of 281,443 to 337,971 for the estimate of direct civilian deaths. The interval expresses uncertainty within the statistical analysis; the central estimate is not an exact, confirmed death toll.
The scope was also limited. OHCHR said the estimate excluded deaths resulting from lost access to healthcare, food, clean water and other necessities. It covered a specified decade, rather than the whole subsequent course of the conflict. Those boundaries matter when comparing it with other figures that cover different periods, populations or causes of death.
The same briefing described displacement, broken information networks, communications outages, surveillance and restrictions on movement as obstacles to casualty recording. Statistical work consequently depended on documentation gathered under conditions that also restricted its coverage. The calculation and the fieldwork were connected parts of the account of civilian harm.
Colombia and the limits of comparison
In Colombia, HRDAG, the Colombian Truth Commission (CEV) and the Special Jurisdiction for Peace (JEP) collaborated on integrating records and estimating violence associated with the armed conflict. Their methodological report describes work on underreporting and patterns across several violations, including homicide, enforced disappearance and kidnapping. It also describes support for the participating institutions' ability to compare sources and implement statistical models.
The report explains why recorded trends can diverge from events outside a database. A change in investigators' activity or access can change the volume of reporting. A rise or fall in entries does not, by itself, establish an equivalent change in violence. That distinction affects comparisons across places and years, even when individual records are credible.
Definitions introduce another boundary. The report states that its homicide analysis could not distinguish civilian deaths from combatant deaths occurring in the armed conflict. Its homicide figures therefore do not describe the same population as the Syria civilian-death estimate. Placing the two side by side without that qualification would conceal a substantive difference in what was measured.
HRDAG's Colombia project page distinguishes data released to replicate the original analysis from a corrected version intended for new analyses. It says the corrections removed some people erroneously included as direct victims and changed estimates of overall underreporting for homicide and enforced disappearance. It reports no change to the results for child recruitment or kidnapping, or to the proportional breakdowns of estimated violations.
The two dataset versions serve different purposes: reproducing the original analysis and conducting new analyses using corrected records. Interpreting a published total also involves identifying its definition, period and dataset version.
Evidence and the public record
OHCHR links its casualty analysis to reporting on SDG indicator 16.1.2, which measures conflict-related deaths per 100,000 people, by sex, age and cause. This places the work within SDG 16, Peace, justice and strong institutions. Distinguishing documented deaths from estimated totals is part of interpreting those measurements.
In both projects, the scope of the records differs from that of the statistical estimates. The Syrian analysis concerns direct civilian deaths over a specified period; the Colombian homicide analysis does not separate civilians from combatants. The published methods and dataset corrections provide context for interpreting the figures and their limitations.
Written by a human author, edited with AI assistance.
Further information:
HRDAG, Core Concepts. Explains incomplete coverage, selection bias, duplicate reporting and multiple systems estimation.
OHCHR, Syria casualty briefing, 28 June 2022. Supports the documented and estimated civilian deaths, period, exclusions, recording obstacles and SDG connection.
HRDAG, Syria. Describes the group's analytical contribution and the credible interval accompanying the 2022 estimate.
JEP, CEV and HRDAG, Methodological report, updated 6 March 2025. Explains the joint project, reporting bias and the scope of its homicide analysis.
HRDAG, Colombia. Documents dataset versions, corrections and their stated effects on the published analysis.
