Conflict, Gender, and Education in the Democratic Republic of the Congo

Written for ECON 064, Topics in Development Economics, at Dartmouth College, advised by Paul Novosad. The full paper is below and can be downloaded as a PDF.

Abstract

The adverse effects of conflict on human capital are well established. However, the gendered direction of these effects remains contested. This paper examines the effect of conflict exposure in the Democratic Republic of the Congo (DRC) on years of education, with a focus on gender differences. The paper uses data from the Demographic and Health Surveys and ACLED to link respondents to conflict exposure. The results indicate that men exposed to conflict during their school-age years lose more years of education than women, for whom the effect is not detectable. The gender gap in education, therefore, narrows in the DRC, but men's losses primarily drive it.

What the paper does

The question is which gender absorbs more of the educational damage that armed conflict does. Existing work does not settle it, because the answer has run one way in Tajikistan and the other way in Burundi, so it cannot be resolved by theory alone.

Answering it requires knowing how much violence each individual person was exposed to at the age when they should have been in school, and no single dataset records that. The paper builds the measure by combining two. Individual characteristics come from the 2023-24 DRC Demographic and Health Surveys, and violent events come from the Armed Conflict Location and Event Data project. Survey clusters were matched to their GPS coordinates, the country was covered with a grid of ten-kilometre cells in QGIS, and every respondent was assigned to the cell containing them. A person's conflict exposure is then the count of violent events inside their own cell across the years they were aged six to seventeen.

That construction is what makes the comparison possible. Two people of different birth cohorts in the same place, or of the same cohort in different places, lived through different amounts of violence during their school years, and the paper uses that variation to identify the effect.

The estimation follows a difference-in-differences logic with a continuous measure of exposure rather than treated and control groups, using ordinary least squares with grid-cell and birth-cohort fixed effects. The raw sample is 40,264 respondents. Section VI reports descriptive diagnostics and a placebo test, which matters because the design rests on an assumption about parallel trends that cannot simply be asserted.

What it finds

A one-log-unit increase in school-age conflict exposure reduces men's schooling by 0.44 years, significant at the one percent level. For women there is no statistically detectable effect. The result holds across three further specifications, including one restricted to respondents who reported living in the same place throughout their school-age years.

So the education gender gap in the DRC narrows under conflict, and it narrows because men's attainment falls rather than because women's rises. A summary statistic moving in an apparently favourable direction can describe a situation that is getting worse for everyone in it.

The paper