vhalulu/kenya-electrification-education

R replication code for IV estimates of household electrification on children's education in Kenya using DHS 2014 & 2022. Girls gain 1.08 years of schooling; no effect on enrollment. # Powering Education: Evaluating Kenya's Last Mile Connectivity Program Using Distance-Based Instrumental Variables --- ## Overview This repository contains the replication code for a study estimating the causal effect of household electrification on children's educational outcomes in Kenya. Using data from the 2014 and 2022 Kenya Demographic and Health Surveys (DHS), I instrument household electricity access with distance to a hypothetical pre-1989 transmission network to address the endogeneity of electrification. **Main findings:** - Electrification has **no significant effect on school enrollment**, consistent with Kenya's already high baseline enrollment rate of 89% - Electrification **increases years of schooling by 0.603 years** (24.3% relative to the sample mean), significant at p 50,000) using minimum spanning tree algorithms. Each DHS cluster's Euclidean distance to the nearest network segment serves as the primary instrument. Least-cost distance — accounting for terrain and river crossings — is used as an alternative instrument in robustness checks. The instrument is valid under the assumption that proximity to this historical, simulated network affects household electrification rates but has no direct effect on children's educational outcomes except through electricity access. --- ## Software | Tool | Version | Purpose | |------|---------|---------| | R | 4.3+ | All analysis | | RStudio | — | IDE | Key packages: `tidyverse`, `haven`, `sf`, `terra`, `fixest`, `flextable`, `officer`, `here`, `patchwork` Install all at once: ```r install.packages(c("tidyverse", "haven", "sf", "terra", "fixest", "flextable", "officer", "here", "patchwork", "conflicted")) ``` --- ## Citation If you use this code, please cite: > Alulu. (2026). *Powering Education: Evaluating Kenya's Last Mile Connectivity Program Using Distance-Based Instrumental Variables*. --- ## Contact Vincent Alulu· [valulu@ucsd.edu]