Replication package for: Trust in Whom? Channel-Specific Trust and Collective Action in Developing Countries

Replication package for the manuscript "Trust in Whom? Channel-Specific Trust and Collective Action in Developing Countries" (version 23, June 2026), under review.The paper distinguishes channel-specific trust — trust in the entity that delivers a collective good — from generic interpersonal trust, and shows that only the former predicts enrollment in voluntary community-based health insurance schemes. The analysis covers 6,939 households across India (4,931 households in 7 states from the ECCP program) and Nepal (2,008 households in Banke and Dhading districts).This package reproduces every numerical claim in §3 and §4 of the manuscript using the raw survey data. It contains:- 22 section scripts, one per manuscript section/subsection (§3.5 through §4.10), each writing a JSON output;- a master runner (00_REPLICATE_ALL.R) that sources all 22 section scripts, then builds an itemized verdict report;- the verdict builder (Tb_v23_build_verdict_report.R), which compares every reproduced value against the manuscript target and tags it LOCKED, CLOSE, MISMATCH, or REPORTED;- the most recent verdict outputs in 09_REPLICATION_PACKAGE/outputs/ (VERDICT_REPORT.md and VERDICT_REPORT.csv), confirming 54 of 54 v23 manuscript headline claims locked within rounding tolerance;- documentation (top-level README.md and 09_REPLICATION_PACKAGE/README.md);- SHA-256 sums for the three raw survey data files (in the top-level README.md), enabling reviewers to verify data integrity.The three raw data files are not redistributed in this deposit (per the terms of access of the original surveys): ECCP_Primary_Jan2007_v1.1.07.sav (India), Banke-22Apr11_CostEstimation.dta and Dhading-22Apr11_CostEstimation.dta (Nepal). Reviewers may request access from the original survey custodians; the master runner resolves these via recursive search on the local filesystem.Tested on Windows 11 / RStudio / R 4.5.2. Total runtime approximately 12 minutes. Auto-installs missing packages (foreign, haven, jsonlite).本复现包对应于手稿"Trust in Whom? Channel-Specific Trust and Collective Action in Developing Countries"（2026年6月第23版，目前处于审稿阶段）。该研究将特定渠道信任（channel-specific trust）——即对提供集体物品的主体的信任——与一般人际信任区分开来，并证实唯有前者可预测自愿社区健康保险计划的参保参与情况。 分析样本覆盖印度与尼泊尔两国共6939户家庭：其中印度地区包含来自ECCP项目7个州的4931户家庭，尼泊尔地区包含班克（Banke）与达丁（Dhading）两县的2008户家庭。 本复现包可复现手稿第3、4章节中的所有数值结论，所用数据为原始调研数据。包内包含： 1. 22个分章节脚本：对应手稿第3.5节至第4.10节的每个章节/子章节，每个脚本均生成一份JSON输出文件； 2. 主运行脚本（00_REPLICATE_ALL.R）：可调用全部22个分章节脚本，并生成分项审核报告； 3. 审核报告构建脚本（Tb_v23_build_verdict_report.R）：将所有复现得到的数值与手稿中的目标值进行比对，并标记为LOCKED、CLOSE、MISMATCH或REPORTED； 4. 存放于09_REPLICATION_PACKAGE/outputs/路径下的最新审核输出结果（VERDICT_REPORT.md与VERDICT_REPORT.csv），结果显示54条v23版手稿核心结论均在舍入误差范围内完全匹配； 5. 文档说明：顶层目录的README.md与09_REPLICATION_PACKAGE/README.md； 6. 三份原始调研数据文件的SHA-256哈希值（存放于顶层README.md中），便于审稿人验证数据完整性。 本存档未随附三份原始数据文件（依据原始调研项目的使用许可条款）：分别为ECCP_Primary_Jan2007_v1.1.07.sav（印度数据）、Banke-22Apr11_CostEstimation.dta与Dhading-22Apr11_CostEstimation.dta（尼泊尔数据）。审稿人可向原始调研数据的托管方申请获取；主运行脚本可通过本地文件系统的递归搜索自动定位这些数据文件。 本复现包已在Windows 11 / RStudio / R 4.5.2环境下完成测试，总运行时长约12分钟，可自动安装缺失的依赖包（foreign、haven、jsonlite）。