The long shadow of natural disasters: educational impacts of the 1991 cyclone in Bangladesh

Using the 1991 cyclone in Bangladesh as a natural experiment, this paper examines the long-term educational impacts of early-life disaster exposure. We employ a differences-in-differences approach with data from the 2019 Multiple Indicator Cluster Survey to compare educational outcomes between disaster-affected and unaffected districts across birth cohorts with varying exposure timing. The results reveal substantial and persistent negative effects of early-life cyclone exposure on educational attainment. Children exposed during critical early developmental periods (ages 0–3) experience approximately one year reduction in completed schooling, with secondary completion falling by 12–19 percentage points and higher secondary completion by 10–17 percentage points. Mechanism analysis reveals economic hardship as the primary transmission channel, operating through household budget constraints that force reductions in educational investment. Infrastructure damage creates additional barriers through reduced access, while maternal psychological stress extends impacts to post-disaster birth cohorts. Disaster impacts exacerbate existing inequalities: girls experience roughly double the educational losses of boys, while rural populations face consistently larger impacts. The concentration of effects at secondary and higher secondary education levels suggests that disasters may perpetuate intergenerational poverty by blocking access to the formal labor market, where secondary education is often the minimum requirement. Robustness checks, including threats to identification and placebo tests, confirmed that these results reflected a genuine impact of the cyclone rather than coincidental patterns. These findings are urgent given projected increases in extreme weather frequency under climate change, providing strong justification for integrating disaster resilience into human capital development strategies in vulnerable developing countries.