Replication materials for When a Digital Audit Trail Becomes Audit Evidence

This record contains the replication materials for When a Digital Audit Trail Becomes Audit Evidence: An Evidence Architecture Map, Regulatory-Disclosure Panel, and Known-Truth Benchmark. The archive supports three distinct components. First, a bounded evidence-architecture map reproduces published descriptives from a 58-article review and maps nine focal articles from AUDITING: A Journal of Practice & Theory to their observed information channels and claim boundaries. Second, a source-linked regulatory-disclosure panel applies a frozen deterministic protocol to 254 Rule 11(g) annual reports recovered from 98 firms within a frozen 100-issuer NIFTY frame and retains 1,166 page-level evidence records with source URLs, excerpts, page references, and SHA-256 hashes. Third, a known-truth computational benchmark evaluates edits, recorded-item deletions, and pre-capture omissions across 45 million generated economic events and 72,000 primary procedure-replication observations. Version 2.1.0 adds 4,500 approximate-linkage replications that remove stable event identifiers and cross exact process matching with relative amount tolerances of ±0.1, ±0.5, and ±1.0 percent and date windows of 0, ±1, and ±3 days. The frozen candidate-presence rule is conservative: a missing event is treated as masked whenever any current-ledger record satisfies all matching tolerances. At the widest frozen rule, combined recall is 85.91 percent across all injected failures: edit recall remains 100.0 percent, while deletion and omission recall fall to 79.0 and 78.7 percent. These outputs are a known-truth implementation sensitivity, not a field match-rate estimate. The licensed-data validation matches all 100 frozen-frame issuers exactly by ISIN to Compustat Global and retains 488 unique 12-month firm-years with no duplicate issuer-year or issuer-date keys. It validates the sample frame; it does not estimate audit quality, fees, restatements, misstatement, or causal effects. The archive contains code, derived public-source data, generated benchmark results, aggregate WRDS diagnostics, retrieval specifications, source lineage, figures, tests, manifests, and SHA-256 checksums. Copyrighted annual-report PDFs, article full texts, and licensed WRDS row-level data are not redistributed. No participants were recruited and no new human-subject data were collected. Published human-subject studies enter only as prior literature.