CARTELKIT / PUBLIC-SOURCE STUDY
What does the document actually establish? A close reading of institutional reporting on Sinaloa, CJNG, and the response to organized crime.
Who published it, when, and for what purpose?
An assessment, an allegation, and a reported court outcome carry different weight.
Which conclusion would go beyond what this source supports?
01 / ASSESSMENT
What do selected public records establish about Sinaloa and CJNG, and how much do they tell us about the effectiveness of the institutional response?
The reviewed record establishes a dated U.S. threat assessment, formal designations, sanctions, and criminal proceedings. It does not establish the organizations' present strength or the causal effectiveness of the response. Oversight findings and contested modeling show why those questions need separate evidence.
A01 / ANALYSIS
The reviewed record shows sustained U.S. institutional concern and action toward Sinaloa and CJNG across distinct legal instruments. It does not measure their current organizational strength.
An inference about this selected record, not a trend estimate. Designation, sanctions, allegations, and a reported plea have different legal meanings; the DOJ examples are different cases.
A02 / ANALYSIS
The selected sources do not establish whether the response reduced trafficking. GAO's measurement and coordination findings make activity counts an inadequate substitute for an outcome evaluation.
An evidence gap, not a finding that the response failed. The 2025 deficiency is historical; the 2026 field-office sample is nongeneralizable. No causal impact estimate is attempted.
A03 / ANALYSIS
Aggregate academic estimates cannot resolve present group-level capacity in this study. The bounded model comparison and critique leave assumptions and identification questions open; DEA's threat assessment measures a different construct.
Bounded source comparison; no model or agency estimate is independently reproduced.
A04 / ANALYSIS
Growth from 2022 under the intervention is not growth caused by that intervention. In the published 2027 comparison, the doubled-incapacitation scenario is below the unchanged scenario despite remaining above 2022.
Interpretation of the cited v1 scenario and its reference years. It does not settle the critique, independently validate the model, establish an observed policy effect, or imply that the paper never acknowledged containment.
These are hypotheses. This source sample cannot choose among them.
More announced actions could reflect greater enforcement effort or visibility, a continuing threat, or both.
These are competing hypotheses. The sample contains no comparable action series, denominator, or counterfactual that distinguishes them.
Oversight concerns could coexist with useful local results; they could also constrain the response.
The cited findings identify measurement and coordination issues. They do not estimate their effect on trafficking or establish universal institutional failure.
Different estimates could arise from model assumptions, data coverage, or real change in the underlying population.
The bounded full-text comparison does not distinguish these explanations. Consistent definitions and independent replication would be needed.
Revision: generated demonstration records from the earlier CARTELKIT page are excluded. This assessment uses the cited publications below.
02 / READING THE MODEL
The author preprint’s 2027 scenarios compare weekly casualties with 2022. Doubling incapacitation produces an increase of about 8%; unchanged policy produces about 40%. Both rise from the starting year, but the intervention scenario is lower than the same-year counterfactual. Source: §2.2 and Figure 4.
| Scenario | 2027 index | vs. unchanged 2027 |
|---|---|---|
| Unchanged policy | ≈140 | Reference |
| Double incapacitation | ≈108 | ≈22.9% lower |
Arithmetic: (108 − 140) ÷ 140 × 100 ≈ −22.9%. These are normalized published simulation results, not observed casualty counts or a new model run.
Growth despite an intervention does not establish growth caused by it. This comparison clarifies the reported scenario; it does not validate the model or establish a feasible real-world policy.
Rojas challenges the model’s assumptions and identification. His reduced aggregate argument removes saturation; that is not a numerical reproduction of the full calibrated system. The review must examine both the original assumptions and the critique’s simplifications. Read Rojas §3.
The network-derived interaction entries and the four optimized dynamics parameters have different roles. Rojas recognizes both. Full reproduction would require the calibration data, implementation, optimizer settings and sensitivity checks. This revision is a textual comparison.
03 / THE SOURCE RECORD
Each record identifies its source or review date. Citation details keep publication and review dates separate. Reviewing a release does not establish the latest outcome of its underlying court case.
9 records · citations open below each claim
No records match these filters.
DEA's 2025 assessment identifies Sinaloa and CJNG as dominant drug-trafficking threats to the United States.
Limit: An attributed 2025 assessment, not a measured 2026 market share.
Find it: Executive Summary, opening paragraph; publication landing page dated May 13, 2025
Source limits: DEA's dated assessment; it does not independently establish present-day market shares or the guilt of a person.
The State Department's February 2025 notice designated Sinaloa and CJNG as Foreign Terrorist Organizations, effective upon publication.
Limit: A formal U.S. designation; it is not a criminal verdict against everyone associated with a group.
Find it: Pages 1-2: organization list, designation, effective-on-publication language, publication date
Source limits: Evidence of a U.S. administrative designation on that date. Current listing changes and any judicial challenges were not comprehensively reviewed.
Treasury announced sanctions against more than 50 individuals and entities it linked to CJNG on July 23, 2026.
Limit: The affiliations are Treasury's determinations. Sanctions are separate from a criminal conviction.
Find it: Opening paragraph; paragraph identifying Executive Orders 14059 and 13224
Source limits: An administrative sanctions announcement. Agency assertions about affiliations are attributed; sanctions do not substitute for criminal adjudication.
DOJ announced an indictment against four defendants it alleged were affiliated with CJNG in August 2026.
Limit: Charges remain allegations at the cited procedural stage; defendants are presumed innocent. Later case outcomes were not checked.
Find it: Opening announcement, affiliation paragraph, and presumption-of-innocence paragraph
Source limits: The announcement reports allegations. Later dispositions were not verified; no individual profiles or operational allegations are reproduced here.
DOJ reported a CJNG co-founder's guilty plea to a federal narcotics conspiracy in April 2026.
Limit: A reported plea in a different case from C04. The judgment and later sentence were not independently verified.
Find it: Opening paragraph and paragraph beginning 'Valencia-Salazar pleaded guilty to one count'
Source limits: DOJ reports a particular guilty plea. The signed plea, judgment, later sentence, and appeal record were not independently retrieved.
GAO's 2025 review found that DHS lacked the required assessment program and performance goals for evaluating fentanyl-trafficking efforts.
Limit: A finding at publication. The reviewed page records later agency plans; this ledger does not claim the deficiency remains unchanged.
Find it: What GAO Found: assessment program and performance goals; Why GAO Did This Study: FY2021-2024 scope
Source limits: A dated oversight finding about measurement. Seizures and activity counts do not by themselves demonstrate reduced trafficking or causal program impact.
GAO's September 2026 review found uncertainty over DEA and HSI collaboration requirements after a May 2026 DOJ memorandum.
Limit: A nongeneralizable field-office sample supports the oversight review, not a nationwide failure rate.
Find it: What GAO Found: May 2026 memorandum and agreement requirements; Why GAO Did This Study: nine field offices
Source limits: The field-office sample was nongeneralizable. This review does not estimate the nationwide prevalence of collaboration failures or attribute them to a particular cartel.
The author preprint's 2027 scenario reports weekly casualties 8% above 2022 with doubled incapacitation, versus 40% above 2022 under unchanged conditions.
Limit: Rounded published model outputs, not measured effects. Both percentages use 2022 as the reference; the same-year modeled baseline is the unchanged 2027 scenario. No independent reproduction is claimed.
Find it: Author preprint v1, section 2.2 page 5 and Figure 4 page 6; intervention scenarios for 2027 relative to 2022
Source limits: Selected full-text sections reviewed; no independent model reproduction. These are published model outputs, not observed effects or group-level capacity estimates. The v1 preprint was not independently collated against the final Science article; publication date is unasserted because retrieved metadata conflicted.
A separate academic comment challenges that model's identification and central conclusion.
Limit: A methodological critique, not a numerical refutation of the full calibrated model. It recognizes network-derived interaction entries and four optimized dynamics parameters; this study does not adjudicate identification.
Find it: Comment v1, section 2: model and parameter-identification discussion
Source limits: Bounded full-text review. The critique addresses the published Science article; comparison here uses the original arXiv v1. It acknowledges the network construction and four optimized dynamics parameters. This review does not adjudicate full-model identifiability, peer-review status, or later responses.
04 / METHOD & LIMITS
The presentation takes cues from Bellingcat’s published investigation workflow and metadata documentation: make the route from a source to a conclusion visible. This project has no affiliation with Bellingcat or a government agency.
Coverage: selected public sources, not an exhaustive review. Agency statements can share upstream evidence; multiple citations do not automatically provide independent corroboration. No private-person profile or live tracking is included.
05 / REVISION RECORD
The original project tested a synthetic data pipeline. This revision changes the public deliverable: factual records have a publisher, a date, a citation and an explicit evidentiary status. The original history remains available for comparison.
Contribution record: source review, editing, code and presentation were developed with AI assistance. Citations, uncertainties and revision history are preserved in the source ledger.