Replication package for: What willingness-to-pay surveys can tell a community health insurer, and what they cannot

Replication package v1.0.0 for the manuscript "What willingness-to-pay surveys can tell a community health insurer, and what they cannot" (D. M. Dror, sole author), submitted to The Geneva Papers on Risk and Insurance, Issues and Practice (October 2026; manuscript v2.7). Contents: the R scripts (base R, read-only on inputs) that recompute Table 1 from three baseline household surveys of 2008 to 2010 in six Indian districts, 9,199 households, including the pooled elasticity with district fixed effects, the share of declarations at the cap and the censored estimates; the run reports of 6, 7 and 30 September 2026; the coding sheets for the 98 studies read and the 84 in low- and middle-income countries (Tables 2 and 4), with the second coder's sheet and the agreement statistic; the classification and fate trace of the 84 studies with one public source per traced cell (Table 5); the full search record of every arm (the strings and scripts as run, their logs, every record retrieved without abstract text, and the screening sheets), with the PRISMA-S checklist completed item by item; the programme records cited for Kalahandi; the scripts and reports behind the power calculation of Section 5; and Figures 1 and 2 with their scripts. The household-level survey files are not deposited; their provenance and prior use are recorded in the author's reports in 04_source_reports_author, and the derived district and pooled estimates are the deposited results. Every file is listed in MANIFEST.csv with its MD5, and RUN_ALL.R at the package root reproduces every result the package can reproduce from its own files (the hash gate, the Table 2 agreement statistic, the power readings, Figure 1, the consistency of the deposited Table 1 runs, and every count of the search, recounted from the deposited records), reporting SKIPPED with the reason for the steps that need the undeposited household files. Archive MD5: 382d014a9b314fe23f0b6644813a73b4. Licensing: all files other than code are licensed CC BY 4.0; the code (every .R file) is licensed MIT, as the README records. The concept DOI is the stable citation target and always resolves to the latest version; later additions are uploaded as new versions of this record. The DOI is withheld from the anonymised manuscript for the duration of review and given in the author information file.