Background Food insecurity is a pressing global challenge with far-reaching consequences for health and well-being. However, little attention has focused specifically on the experiences of children and adolescents over the age of 5 y in food insecure households. Objectives We examine whether the persistence and severity of household food insecurity are negatively associated with children’s educational outcomes. Methods We used data for the younger cohort of the longitudinal Young Lives data from rounds 3 (2009), 4 (2013), and 5 (2016), when children were aged 8 y, 12 y, and 15 y, respectively. Drawing on the Household Food Insecurity and Access Scale, we used descriptive statistics, graphical analysis, and multilevel regressions to document how the persistence and severity of household food insecurity are associated with children’s educational outcomes (years of education, maths, and vocabulary [PPVT] test scores). We controlled for potentially confounding sociodemographic characteristics, including children’s own baseline grade attained and test scores in “value-added” models, to provide robust estimates of household food insecurity in predicting children’s educational outcomes. Results Household food insecurity generally declined between 2009 and 2016. Fewer than 50% of households were food secure across the 3 rounds of data we examined. Our robust, multivariate, value-added models show that the persistence and severity of food insecurity are negatively associated with all 3 children’s educational outcomes we examined. Conclusions We add to a small but growing literature exploring how household food insecurity is associated with children’s educational outcomes in the Global South. Our findings on severity of food insecurity highlight the importance of understanding food insecurity along the severity continuum rather than as a dichotomous state, as previously done in existing literature. Addressing household food insecurity in childhood and adolescence may be a key factor to improve children’s educational outcomes.