hgrueso/nigeria-education-returns

Estimating the economic costs of child marriage in Nigeria. Using the 2018–19 Nigeria Living Standards Survey, we estimate Mincerian wage models, Heckman selection corrections, and causal forests to analyse returns to schooling and wage penalties from early marriage, with a focus on Northern Nigeria. # Education, Child Marriage, and Heterogeneous Wage Returns among Women in Nigeria **Authors:** Hernando Grueso, Ritapriya Bandyopadhyay, Chris Desmond, Brendan Maughan-Brown, Lucie Cluver, William Rudgard **Date:** February 2026 ## Overview This repository contains the replication code for the paper *"Education, Child Marriage, and Heterogeneous Wage Returns among Women in Nigeria"*. Using the 2018–19 Nigeria Living Standards Survey (NLSS), we estimate Mincerian wage models with state fixed effects, Heckman selection correction, and causal forests to analyse returns to education for women, and how child marriage moderates those returns. ### Key Findings - Each additional year of schooling raises wages by ~15% in both national and Northern Nigeria samples, with **convex** returns. - Child marriage (before age 18) is associated with a ~15% wage penalty. - Nationally, the interaction is **significant** (p = 4.2.0 - **Quarto** >= 1.4 (for rendering slides and tables PDF) - **LaTeX** distribution (e.g., TinyTeX, TeX Live, or MiKTeX) for PDF output ### R Packages Install all required packages by running: ```r install.packages(c( "tidyverse", "haven", "modelsummary", "kableExtra", "sampleSelection", "marginaleffects", "grf", "survey", "flextable", "officer", "patchwork" )) ``` If you don't have a LaTeX distribution, install TinyTeX from R: ```r install.packages("tinytex") tinytex::install_tinytex() ``` ## Replication Instructions ### 1. Clone the repository ```bash git clone github.com cd nigeria-education-returns ``` ### 2. Add the data Place `NLSS_analysis.dta` in the `data/` folder: ```bash cp /path/to/NLSS_analysis.dta data/ ``` ### 3. Run the full pipeline ```bash Rscript code/03_analysis_NLSS_Final.R quarto render tables_and_figures.qmd quarto render slides_nigeria.qmd ``` The R script takes approximately 5-10 minutes (causal forest grows 4,000 trees per sample). It produces `outputs/analysis_results. …