Replication Package for "Governance Thresholds in Member-Based Organizations" (v3.0 — corrected)

VERSION: 3.0.0 What's new in v3.0 (11 February 2026) Corrected microinsurance sample size: N=124 → N=113 (79 contemporary + 34 historical case studies). The original count of 45 historical schemes included 8 aggregate entries representing multiple schemes bundled as single rows; the corrected count reflects 34 individually documented case studies with verifiable trust estimates. Added Figure 1 replication script (`08_FIGURE1_REPLICATION.R`): Generates Figure 1 entirely from source data, combining contemporary microinsurance schemes (DOI 10.5281/zenodo.17008749) and historical case studies (DOI 10.5281/zenodo.16749650) with CPR trust data. Added external data directory (`EXTERNAL_DATA/`): Contains microinsurance scheme data and V3 case study files required for Figure 1, with full data provenance documentation. Updated RUN_ALL.R: Pipeline now includes Figure 1 generation as script 08. Updated VERIFIED_RESULTS.json: Added Figure 1 statistics (N=113, success rate below τ*=0.383, above τ*=0.727). All core findings remain unchanged. The trust threshold τ*≈0.68 continues to show convergent evidence across all datasets. V.2 description: Replication package for the manuscript "Governance Thresholds in Member-Based Organizations: Evidence on Trust, Scale, and Formalization from Three Global Datasets" (v.33), submitted to the International Journal of the Commons. Version 2.0 corrects a data processing error present in Version 1.0 (December 2025). The CPR OPLEVEL database export from Arizona State University contains duplicate rows arising from its relational structure, which links operational levels to multiple source documents. Version 1.0 counted all rows (N=98 with valid trust); the correct count after deduplication by unique operational level is N=64. All CPR statistics have been recomputed, and the package has been rebuilt from scratch with a complete end-to-end replication pipeline. Full details are provided in DOCUMENTATION/CORRECTION_NOTE.md inside the package. All materials from Version 1.0 are preserved in the V1_REFERENCE/ directory so that reviewers can compare versions directly. DATASETS The manuscript draws on four independent datasets. This package contains the raw data and replication scripts for two of them (CPR and WCM) and references the other two via their own Zenodo repositories. Dataset 1: Ostrom Common Pool Resource Database (N=64 unique operational levels with valid trust data; 87 subgroups with valid size data). Source: Arizona State University SES Library (https://seslibrary.asu.edu/cpr). The complete CPR database download (all 11 tables plus the coding manual PDF) is preserved in RAW_DATA/GLOBAL_CPR/. The two tables used by the analysis scripts — CPR_OPLEVEL_with_DOC.txt (operational-level data including trust, resource quality, and group size) and CPR_SUBGROUP.txt (subgroup-level data including membership size) — are also placed at RAW_DATA/ for direct access by the pipeline. Data provenance is documented in RAW_DATA/GLOBAL_CPR/DATA_PROVENANCE_CLTTV.txt. Dataset 2: World Cooperative Monitor (100 countries, 5 size categories). Source: International Cooperative Alliance, 2012–2025. File: RAW_DATA/SIZE_DISTRIBUTION_FULL.csv. Contains the cross-national distribution of cooperatives by size band, showing a modal concentration of 29% in the 5,000–20,000 member band. Provenance documented in RAW_DATA/WCM/DATA_PROVENANCE_WCM.txt. Dataset 3: Multi-country microinsurance schemes (N=124, 1995–2017). Archived separately at Zenodo DOI 10.5281/zenodo.17008749. Dataset 4: Consensus organizations (N=274 across three scale levels). Archived separately at Zenodo DOI 10.5281/zenodo.17219731. REPLICATION The package contains eight R scripts (base R only; no additional packages required), forming a single end-to-end pipeline that reproduces all statistics reported in the manuscript. To replicate, run from the package root directory: cd CLTTV_REPLICATION_v2 Rscript SCRIPTS/RUN_ALL.R The pipeline runs in under 30 seconds and writes all outputs to ANALYSIS_OUTPUTS/, including a machine-readable VERIFIED_RESULTS.json compiling every computed statistic. DOCUMENTATION/VERIFICATION_CHECKLIST.md maps each manuscript claim to its corresponding script and output file. The scripts execute in order: (00) deduplicate the raw CPR export, (01) compute trust distribution and success rates, (02) produce Table 2 and Appendix B correlations, (03) run threshold scans with Fisher's exact, logistic regression, and permutation tests, (04) analyse subgroup sizes for the 500-member boundary, (05) verify WCM modal concentration claims, and (06) compile all results into JSON. KEY CORRECTED FINDINGS (v2.0) - CPR sample: 64 unique operational levels (corrected from 98 in v1.0) - Mean trust: 0.23 on a 0–1 scale; only 8 cases (12.5%) exceed the threshold at 0.68 - Table 2: N=29 with joint trust and size data; mid-scale group shows 0% success (0/11) - Threshold effect: +16.2pp at threshold 0.55 (descriptive; not statistically significant due to small N above threshold) - Subgroup sizes: N=87, mean=233.4, median=95; 94.3% below 500 members - WCM: 29% of 100 countries concentrated in 5,000–20,000 member band (unchanged from v1.0) - Cross-dataset convergence: thresholds in the range 0.50–0.68 (mean=0.59, SD=0.09) across three independent datasets (unchanged from v1.0) For questions about replication: [e-mail] ADDITIONAL NOTES Version 2.0 corrects a CPR deduplication error in Version 1.0 and rebuilds the replication package with a complete end-to-end pipeline. All v1.0 materials are preserved in V1_REFERENCE/ for comparison. See DOCUMENTATION/CORRECTION_NOTE.md inside the package for full details.script results