Our Methods — How We Conducted This Investigation
How Divest for AR Future obtained, processed, and analyzed nearly 1,100 public records from four Arkansas state agencies.
Transparency about our own methods is as important as the transparency we demand from pension fund managers. Here’s exactly how this investigation was conducted.
Data collection
Public Records Requests
We filed 8 Freedom of Information Act (FOIA) requests across 2 rounds to 4 Arkansas state agencies: the State Treasury, the Arkansas Teacher Retirement System (ATRS), the Arkansas Public Employees’ Retirement System (APERS), and the Auditor of State’s office.
Our requests targeted investment decision records, board meeting materials, correspondence with bond issuers, credit analyses, and internal memoranda related to Israel Bonds purchases.
Processing pipeline
Document Processing
We received 1,098 documents totaling over 4.5 gigabytes across two rounds of FOIA requests. Round 1 (August–October 2025) produced 957 documents from four agencies. Round 2 (February 2026) added 141 documents from three agencies (Treasury: 118; APERS: 16; ATRS: 7). The Auditor of State’s Round 2 response has been received and is under review. Every document was:
- Cataloged with metadata including source agency, date, document type, and FOIA request number
- Processed using optical character recognition (OCR) where necessary to make scanned documents searchable
- Deduplicated to identify and flag documents received from multiple agencies
- Cross-referenced to connect people, organizations, dollar amounts, and events across documents from different agencies
Verification system
Fact Registry
Every factual claim in our evidence materials traces back to a specific source document, page number, and verbatim quote. We maintain a structured database of verified facts extracted from the source documents — not summaries or interpretations, but direct extractions with full citations.
Two-hop citation chain: Source Document → Fact Registry → Published Finding. Every factual claim traces back to a specific source document, page number, and verbatim quote. No AI-generated analysis is used as source material.
Standards
What We Don’t Do
We do not speculate beyond what the documents show. Our findings state what the record contains and what it lacks — not what officials were thinking or what might have happened off the record.
We do not use AI-generated analysis as source material. All factual claims trace to human-readable government documents through a verified citation chain.
We do not cherry-pick. We acknowledge gaps in the documentary record and identify them explicitly. Where agencies have provided exculpatory or contextualizing information, we include it.
Caveats
Limitations
Important: Our evidence is limited to what agencies have provided through FOIA. Some requests are still pending, and agencies may hold responsive documents that were not included in their responses. Where we identify gaps, we note them and file follow-up requests.
Some documents in Round 2 responses — including six sets of APERS board and committee meeting minutes, four Treasury custodial statements (1,561 pages combined), and two ATRS sign-in sheets — were received as scanned images that yielded no extractable text. These documents are prioritized for optical character recognition processing. Until they are recovered, findings from these meetings rely on board packets and other corroborating documents rather than official minutes.
Verify Our Work
We encourage journalists, researchers, and officials to verify any claim we make. Our evidence page cites specific documents for each finding. Our documents archive provides access to key source materials. For questions about specific claims or methodology, contact us at divestforarfuture@proton.me.