G²LM|LIC - Connecting Rural Women to Global Value Chains via Home-Based Work

Eligible women (aged 18–60) living within the workshop catchment areas were identified through door-to-door listing and screening. Enrolled women completed a standardized two-week training at a women-only, child-friendly village workshop (with a training stipend). They were then randomly assigned to an eight-week piece-rate job either from their own homes (Work-from-Home, WfH) or from a women-only workshop inside the village (Work-from-Workshop, WfW).

Independently, women were cross-randomized into an identity-revelation “poster” arm versus no poster, varying the public visibility of women’s participation. In the final implementation batch, a contemporaneous pure-control cohort of eligible, interested women received no job offer during the study window, enabling direct treatment–control comparisons. Data collection includes (i) listing, baseline, endline, and follow-up household surveys; (ii) high-frequency daily monitoring during the production window capturing hours worked, verified outputs, quality checks, and payments for treated participants; (iii) weekly enumerator-assisted time diaries during the intervention window capturing primary and simultaneous activities, location, and co-presence; and (iv) lab-in-the-field productivity exercises conducted under standardized conditions at the end of training and during the final intervention week.

Baseline data collection for Batch 1 began in March 2024; Batch 4 endline was completed in June 2025; and the final follow-up survey round was completed in December 2025. The core experimental sample comprises 297 women randomized to WfH/WfW (with or without poster), and the pure-control cohort includes 131 women enrolled in the final batch. Endline completion for the core experimental sample is approximately 92% and balanced across arms.

The release includes Stata datasets with variable and value labels, Excel codebooks for all instruments, a README describing file structure and recommended merge keys, and an Excel mapping file documenting differences between survey-codebook variable names and final dataset variables as well as variables dropped for de-identification. Direct personal identifiers have been removed; users must not attempt re-identification.