Infrastructure Impact on Women's Empowerment

What this research found

Rather than estimating effects, this project produced a research design: a 20-page framework for measuring how access to water, sanitation and electricity shapes female labor force participation across India's 640 districts. It sets out five research questions, each with its own identification strategy, and works through the data each would need, what it would cost, and how long it would take. Automated retrieval of the district-level Census and health-survey files failed, so no estimates were produced.

• The design joins five government data sources at district level: Census 2011, the fifth National Family Health Survey (NFHS-5), the Periodic Labour Force Survey, the UDISE+ school census covering 14.7 lakh schools, and Reserve Bank of India statistics.
• Each research question is matched to a distinct method — instrumental variables with mediation analysis for the infrastructure-to-participation link, structural equation modeling for education and health pathways, spatial lag and error models with Moran's I for district spillovers, and difference-in-differences for the two policy evaluations.
• Causal identification leans on three sources of variation: 1991 infrastructure levels as an instrument for current access, geographic features such as elevation, terrain and distance to rivers, and the staggered district rollout of the Jal Jeevan Mission and the PMJDY banking scheme.
• Feasibility was costed out rather than assumed. One research question is estimated at 8 to 11 months, all five sequentially at 3 to 4 years, or 1.5 to 2 years with two or three researchers working in parallel, on a 16–32 GB workstation using free R and QGIS tools for $0 to $1,500.
• An automated peer review pass returned accept with minor revisions, flagging that several questions rest on cross-sectional variation that limits causal claims, that 2010–11 Census infrastructure figures may be stale given subsequent Jal Jeevan Mission expansion, and that Reserve Bank economic data is state-level only and would need disaggregating.

How it was done

Six public Indian data sources were catalogued by granularity, key variables and access route, and a joining strategy at the 640-district level was specified alongside two composite measures: an infrastructure index averaging water, sanitation and electricity access, and an empowerment index built as a factor score from NFHS-5 autonomy indicators. Five research questions were then drafted with hypotheses, expected effect magnitudes and matching econometric specifications, running from baseline OLS through two-stage least squares, spatial Durbin models and Callaway-Sant'Anna difference-in-differences estimators. A feasibility review assessed data access timelines, hardware, software and expertise requirements for each. The result is a 20-page report of roughly 12,000 words with 13 verified citations and seven explanatory diagrams.

Data sources

• Census of India 2011 — district-level infrastructure, demographics and employment across 640 districts
• National Family Health Survey (NFHS-5) — district-level health and women's empowerment indicators
• Periodic Labour Force Survey — district-level participation, employment and wage data
• UDISE+ — school infrastructure and enrolment covering 14.7 lakh schools
• Reserve Bank of India — state-level GSDP and financial inclusion statistics
• Jal Jeevan Mission dashboard — district-quarter tap connection coverage

Outputs produced

• 01_data_acquisition.py

  Data acquisition script - attempts to download Census 2011 and NFHS-5 data from public repositories with verification

• 02_data_preprocessing.py

  Data preprocessing pipeline - cleans, standardizes, and merges Census 2011 and NFHS-5 datasets with feature engineering

• 03_eda.py

  Exploratory data analysis - generates descriptive statistics and visualizations (distributions, correlations, scatter plots) for key variables

• 04_statistical_modeling.py

  Statistical modeling - performs OLS regression analysis to quantify infrastructure impact on gender outcomes (FLFPR, Empowerment_Index)

• 05_results_interpretation.py

  Results interpretation - automatically interprets regression results and generates natural language summaries for final report

• MISSING_DATA.md

  Instructions for manual data acquisition - lists failed downloads and provides official sources

• raw

  Directory for raw data files - currently empty, awaiting manual data upload

• processed

  Directory for processed analysis-ready data - will contain merged Census+NFHS dataset after preprocessing