Philippine war-on-drugs apparatus

Presidential permission entered an operational plan

Rodrigo Duterte entered the presidency promising lethal action against people associated with illegal drugs. After he took office, his public statements encouraged killing and promised protection to police. On July 1, 2016, the Philippine National Police (PNP) issued Command Memorandum Circular No. 16-2016, launching Project Double Barrel. The circular repeatedly used “negation” and “neutralization” for suspected “drug personalities.” The UN Human Rights Office reported the government's position that no policy authorized killing and that deaths occurred in legitimate operations; it nevertheless concluded that the circular's vague language, joined to encouragement from senior officials, may have emboldened police to treat it as permission to kill.¹

Project Double Barrel made that permission operational. Oplan Tokhang sent police and local representatives to homes to seek “voluntary surrender”; its other component used buy-bust operations, warrants, raids, and checkpoints against alleged high-value targets. In a 2017 Supreme Court submission, the Solicitor General described five-stage Tokhang procedures, including information collection and multi-level validation, and asserted that police did not coerce surrenderers. That is the state's stated design, not an independent finding that validation or consent worked that way in practice.²

The apparatus therefore did not require a written order to kill a named person. It joined leader signals to a police plan, barangay intelligence, home visits, operations, and performance reports. The organizational question is how that chain altered frontline judgment even where formal documents also invoked law and human rights.

Watchlists converted secret classification into immediate exposure

Barangay officials compiled watchlists and passed them to police. The UN Human Rights Office found that a home visit could proceed solely because a person's name appeared on a list, without a search or arrest warrant, and that people had no legal procedure for challenging inclusion before police arrived. It also found that the circular connected refusal of a visit to case-building and “negation,” while mentioning counsel only after a person agreed to make self-incriminating statements.³

That arrangement separated classification from contest. A name could travel from community reporting through validation into a home visit or operation, while the person named could not inspect the allegation. “Voluntary surrender” was therefore an administrative category, not proof that every decision was free, informed, or a judicial admission.

The state reported activity through houses visited, surrenders, arrests, seizures, barangays declared cleared, operations, and deaths. Its final “Real Numbers” release for the Duterte administration reported 345,216 arrests, 239,218 anti-drug operations, and 6,252 alleged drug suspects killed in those operations through May 31, 2022.⁴ These figures establish what the government counted. They do not establish that each classification was correct, each death lawful, or drug-market and community harms durably declined.

The death totals answer different questions

The official 6,252 is a category of deaths during acknowledged government anti-drug operations, not a judicial count of extrajudicial killings and not a total for all drug-related deaths. In 2020, the UN report separately recorded a then-current official total of 5,601 police-operation deaths and the police's classification of 3,062 out of roughly 29,000 “deaths under inquiry” as drug-related; it warned that agencies had published conflicting figures and that it could not independently verify a definitive total.⁵

Independent datasets have different boundaries again. Atun and colleagues coded 5,021 victims reported by news organizations during 508 days of the early campaign, with 55 percent linked in the reporting to acknowledged police operations and the remainder to other perpetrators or unclear attribution. The authors found concentration in urban and poorer areas, but their media-derived dataset was not an exhaustive national register.⁶ Broader totals from advocates and investigators should not be substituted for the government's operational category or presented as a fully audited case list. The absence of one reconciled total is itself evidence of a reporting and accountability failure, not permission to choose the largest or smallest number.

Nanlaban let the operating unit authenticate its own force

Police reports repeatedly said a suspect nanlaban—fought back—and was killed when officers returned fire. In its 2022 national report, the Philippine Commission on Human Rights (CHR) analyzed a probability sample of 882 completed case dockets covering 872 incidents and 1,139 victims. In the drug-related subset, witness accounts in 133 incidents contradicted or did not match police narratives; 201 of 235 victims with wound records were shot in the head or torso. CHR also lacked police documents in 295 drug-related killing incidents and found that precinct-level inquiries were often conducted by the unit involved.⁷

Human Rights Watch examined a narrower early sample: 24 incidents involving 32 deaths from October 2016 through January 2017. It compared police reports with family and witness accounts and described closely repeated self-defense narratives, allegations of victims being taken into custody before death, and allegations that guns or drugs were planted. The report expressly did not claim that its incidents were a statistical sample of every killing.⁸

Those records support a mechanism, not a presumption about every officer or incident. The operating organization selected a target, controlled the scene and physical evidence, wrote the first report, and often investigated itself. Where independent access was absent, a repeated self-defense account could close the informational loop before a contrary witness felt safe enough to speak.

Rewards and local politics changed how the policy traveled

The command environment included more than presidential language. In 2024, retired police colonel Royina Garma testified to a House committee about a reward system for killings. CHR said that testimony was consistent with its own earlier finding that internal reports recommended that operatives be “awarded, rewarded, or recognised,” while calling for the new allegations to be independently validated. The testimony and CHR statement are consequential evidence of an incentive structure; they are not a final criminal judgment about every payment or recipient.⁹

Implementation was also locally mediated. Ravanilla, Sexton, and Haim used a regression-discontinuity design around close mayoral elections and a difference-in-differences analysis. They estimated that politically outsider mayors received less national public-works funding and implemented the drug war more aggressively, including higher rates of anti-drug incidents and killings classified by the authors as police extrajudicial killings. The study's design supports a causal interpretation within its measured elections and assumptions; it does not show uniform behavior by every mayor or a mechanically identical national command.¹⁰

Families preserved the record and absorbed the loss

PhilRights documented 118 alleged extrajudicial-killing victims from August 2017 through July 2019 through interviews with relatives and witnesses checked against media, police material, and death certificates. The people documented were predominantly men in poor households working in low-wage or irregular jobs. Families described difficulty obtaining police and medico-legal records, fear of retaliation, funeral debt, lost income, and stigma. This is a Philippine rights organization's documented case series, not a representative estimate of all victims.¹¹

A separate peer-reviewed study traced downstream effects among Metro Manila households in the conditional cash-transfer program. From a larger incident database, the researchers purposively interviewed 31 affected households with 62 children. They connected the loss of predominantly male breadwinners to food and housing insecurity and schooling disruption; 19 children in those households were out of school, including nine eligible children who left after a drug-related killing in the family. The sample preserves detailed experience among poor beneficiary households but cannot estimate effects for every family or region.¹²

Families, witnesses, Philippine rights groups, and CHR thus operated as a counter-record: testimony, photographs, medical records, case files, and names. Their work also exposes a source boundary. Public testimony is selective because fear, cost, safety, and access prevent many affected people from entering any record at all.

Accountability records are real but incomplete

Domestic courts have established criminal responsibility in particular cases. In People v. Pereda and People v. Oares, G.R. Nos. 268510 and 274142, the Philippine Supreme Court on August 11, 2025 affirmed the murder convictions of three police officers for the 2017 killing of 17-year-old Kian Loyd Delos Santos. The Court held that the officers were not acting within lawful duty and that the killing was neither necessary nor justified. That judgment is case-specific; it does not adjudicate the national campaign.¹³

The International Criminal Court's Philippines situation is a separate proceeding. Philippine authorities arrested Rodrigo Duterte and surrendered him to the Court in March 2025. On April 23, 2026, ICC Pre-Trial Chamber I confirmed all three charges of crimes against humanity—murder and attempted murder—and committed the case to trial. Confirmation means the Chamber found substantial grounds for trial, not guilt; Duterte remains presumed innocent unless convicted.¹⁴

On May 11, 2026, the ICC unsealed a separate arrest warrant for former PNP chief Ronald Dela Rosa based on reasonable grounds concerning alleged responsibility as a co-perpetrator. A warrant is not a conviction or confirmation of charges.¹⁵ Domestic response remained unsettled: the PNP announced a revived investigation covering the more than 6,000 operation deaths and alleged vigilante links, served Dela Rosa a subpoena, and then recalled it after the Senate issued a protective order. Church and civil-society groups separately launched a civilian-led EJK Truth Commission to collect evidence and testimony, and the PNP publicly promised cooperation. These are investigative and record-preservation developments as of July 14, 2026, not adjudicated findings or completed remedies.¹⁶

Killings associated with anti-drug enforcement did not end with the Duterte presidency. The University of the Philippines Diliman's Dahas project recorded 1,273 reported drug-related killings from the start of the Marcos administration through its late-June 2026 update. Dahas compiles incident-level reports from news coverage and police public material; its category includes state agents, unidentified assailants, and other reported perpetrators. It is therefore an ongoing monitoring count, not a finding that all 1,273 deaths were police killings or unlawful.¹⁷

A counter-design must break the self-authenticating loop

The UN High Commissioner recommended repealing Circular No. 16-2016, ending Tokhang, abolishing drug watchlists, assisting victims' families, publishing consistent disaggregated data, and empowering an independent body to investigate killings.¹⁸ Organizationally, those reforms share one purpose: separate the power to classify, raid, use force, preserve evidence, and judge the operation. Contestable notice, judicial authorization, independent scene and forensic control, protected witnesses, health-led support, and measures of safety rather than body counts make it harder for output reporting to substitute for lawful outcome.

Colombia's military false-positives system offers a structural comparison: in both cases, a death could be converted into a successful operational result through evidence initially controlled by the unit being evaluated. This is an analytical comparison of reporting mechanisms, not a claim of historical influence, identical command, or legal equivalence. Operation Condor offers a separate comparison in cross-agency coercive coordination, secrecy, and later evidence recovery. The reviewed sources establish no direct influence or equivalence among the campaigns, jurisdictions, crimes, and affected communities.¹⁹

The public record establishes the written operational chain, incompatible official and independent accounts, and sampled incident patterns.²⁴⁶⁷ It also documents family harms and the status of named domestic and international proceedings through July 14, 2026.^11141516 It does not establish one definitive death total, the legality of every operation, every official's knowledge or motive, or a complete account in the voices of affected regions and families.⁵

Concept fingerprint: the operation authenticated its own lethal output

Purpose, mission, and institutional legitimacy is defining. The campaign claimed community safety and drug control, but its operational measures centered visits, surrenders, seizures, arrests, cleared barangays, and deaths rather than durable reductions in harmful use, market violence, or unequal access to care. A public mission cannot legitimate uncontestable classification or unlawful force.⁴¹⁸

Authority, legitimacy, and acceptance is defining. Presidential speech, police circulars, barangay lists, officers, prosecutors, courts, CHR, families, and international proceedings held different power over naming, visiting, surrender, raid, evidence, force, and review. Administrative “voluntary surrender” did not itself establish informed consent or judicial admission.¹²³

Delegation, decentralization, and responsibility has supporting weight. Central doctrine traveled through national police plans, regional and local units, barangays, mayors, informants, and operation teams. Local political evidence shows that implementation varied, but variation does not erase senior signals or turn each death into an isolated frontline event.¹¹⁰

Coordination, communication, and common understanding has supporting weight. Lists, validation, home visits, buy-bust operations, raids, reports, reward claims, seizure and death totals, and review connected political direction to local action. The same terms could carry lawful language and lethal permission, making ambiguity itself a coordinating feature.²³⁹

Structure, hierarchy, and scale has supporting weight. Presidential authority, PNP hierarchy, barangay intelligence, local politics, forensic and prosecutorial systems, courts, CHR, civil society, and the ICC formed a multilevel apparatus and counter-apparatus. Conflicting death categories and media-derived datasets prevent one precise national size claim.⁵⁶¹⁶

Decision making, judgment, and bounded rationality has supporting weight. Officials and officers judged list inclusion, target priority, surrender, threat, force, evidence, and investigation under leader signals, metrics, local politics, and weak adversarial review. Later domestic convictions and ICC stages apply specific legal standards and cannot decide the legality of every operation.^7131415

Measurement, accounting, and control is defining. Houses visited, surrenderers, arrests, seizures, operations, cleared barangays, deaths, wounds, missing files, media reports, and case stages made selected performance visible. Categories differ: an operation death is not a judicial extrajudicial-killing finding, and a monitoring total is not a count of police killings.⁴⁵⁷¹⁷

The zero score for cooperation, incentives, and organizational equilibrium is retained because threat, coercion, fear, secrecy, and lethal force dominate the documented relationship with named people and communities. Reward evidence and local political incentives matter, but they do not justify describing forced surrender or silence as cooperative equilibrium.³⁹¹¹¹⁹

Work design, productivity, and automation has supporting weight rather than the imported zero. Police and barangay work was organized into information collection, validation, visits, operations, evidence control, reports, self-investigation, and output metrics. The sources do not provide representative officer testimony, workload, pay, refusal, discipline, or mental-health evidence.²⁷¹⁹

Knowledge, expertise, and professional autonomy has limited weight. Police intelligence and forensic authority shaped operations; CHR, courts, journalists, researchers, families, and rights groups built counter-records. When the operating unit controlled the scene, report, and review, professional expertise lacked independent evidentiary separation.⁸⁷¹¹

Learning, quality, and reliability has limited weight. Repeated self-defense narratives and precinct self-review could make lethal action organizationally repeatable without making it lawful or reliable. Court convictions, revived investigations, monitoring, and proposed independent truth work are corrective channels whose completeness remains open.^7131617

Strategy, competition, and adaptation has limited weight. Double Barrel combined household outreach and high-value operations; local politicians adapted implementation; monitoring and killing continued under a later presidency; domestic and international accountability also adapted. The record does not establish a durable health or public-safety strategy beyond enforcement activity.²¹⁰¹⁷

Innovation, entrepreneurship, and renewal remains at score zero as an independent lens. Watchlists, Tokhang, integrated metrics, and counter-record projects altered institutional practice. Novelty does not imply benefit, and the sources do not evaluate renewal outcomes from the later PNP investigation or civilian commission.³¹⁶¹⁹

Governance, stewardship, and accountability is defining. Classification, warrant, force, scene control, forensics, investigation, prosecution, victim support, domestic judgment, ICC procedure, and public data require separated powers and review. Case-specific conviction, charge confirmation, and arrest warrant must retain their different evidentiary standards.^13141518

Culture, informal organization, trust, and voice has supporting weight. Leader threats, nanlaban narratives, neighborhood naming, rewards, fear, stigma, family testimony, witness risk, and institutional denial shaped what could be said and believed. Bounded incident and family studies do not support assigning one motive or culture to every officer and community.⁸⁹¹¹

Executive attention, information, and organizational sensing is defining. National output totals, leader language, police reports, local funding and violence, legislative testimony, CHR findings, and current investigations routed selected information to senior authority. The system made arrests and deaths salient while family harm, treatment access, and evidentiary contradiction often traveled through weaker channels.⁴¹⁰⁹⁷

Organizational ignorance is defining rather than the imported zero. Secret list inclusion, unchallengeable validation, self-authenticating force narratives, missing police documents, unit-level self-investigation, conflicting totals, and frightened witnesses produced official certainty and public uncertainty together. This is an editorial classification, not proof that every official knew or concealed the same facts.³⁵⁷¹⁹

The declared links to executive attention, culture, and governance are conceptual lenses grounded in those mechanisms. The benefit-for-all-life link is an ethical audit of whose safety, health, work, family, future, and unmeasured ecological interests entered the campaign's definition of success. The links do not assert terminology adopted by the cited institutions.¹⁹

Paths into deeper study

• Follow one name from its first source through watchlist validation, home visit, surrender or operation, incident report, evidence review, and any correction.
• Compare police units using activity and death metrics with units measured on lawful case quality, treatment access, community trust, and reduced harm.
• Reconstruct a family's work to preserve evidence and obtain remedy, including every institution that could delay, expose, or protect them.