Natural disasters, school system resilience, and educational outcomes in Asia

• In Event: Highlighted Session: Climate-induced disadvantage: The effect of extreme weather events and natural disasters on children’s educational outcomes in climate vulnerable countries

Proposal

Background. Natural disasters associated with accelerating climate change may have important implications for children’s educational outcomes. Moreover, children from marginalized groups could be more vulnerable to ill effects when exposed to natural disasters because families in such groups have fewer resources with which to cope with disasters, or because the institutions to which children or families have access are less resilient. However, impacts are just beginning to be conceptualized.

Research questions. We address two questions: First, what are the heterogeneous impacts of natural disasters on grade progression, school enrollment, and cognitive achievements for children along gender, age, and SES gradients? Second, what are the roles of school shutdowns and instructor truancies in mediating heterogeneous natural disaster effects across children and localities?

Data. We focus on countries in Asia—a region that encompasses areas highly vulnerable to climate disasters. We use the 6th round of the Multiple Indicator Cluster Survey (MICS6) for child (age 5 to 17) and school system (closure and teacher truancy) outcome measures for available countries in South, Central, and East Asia: Bangladesh, Kyrgyzstan, Mongolia, Nepal, Pakistan, Thailand, and Turkmenistan. Our natural-disaster variables are drawn from EM-DAT3 (1900-2023) and ERA54 (1959-2023). We match household-specific survey dates as well as the smallest unit of geo-identifier possible from MICS6 with the EM-DAT and ERA5 datasets, which are time- and geo-coded. This allows us to identify natural disasters, especially droughts and floods, and create time-varying intensity variations in disaster exposures.

Methods. In our first empirical model, we estimate the impact of local-level natural disaster shocks on individual-level educational outcome measures. To explore effects heterogeneity as moderated by permanent child- and household-specific factors, we also estimate the model allowing for combinations of interactions between natural disasters and gender, age, and SES variables. In our second empirical model, we estimate the effects of local-level natural disasters on local-level school shutdowns, school-level teacher truancies, and household-level parental-child interaction quality variables. In our third empirical model, we append outcome variables from the second empirical model to the list of regressors from the first empirical model. If our second empirical model acts as a “first-stage” for key channels through which natural disaster disruptions impact child educational outcomes, then we would expect changes in the magnitudes of the estimated direct natural-disaster effects between the first and third empirical models. Given variations across geo-identifiers and survey dates, we aim to identify the effects of disaster exposures on outcomes outlined earlier based on jointly exploiting temporal variations in disaster exposures within the same location as well as intensity variations in disaster exposures across similar locations at the same time. In addition to using linear estimation methods, for discrete choices, we will also deploy a two-stage school enrollment and attendance discrete choice model that accounts for permanent local heterogeneities and time-varying local shocks. Furthermore, for continuous educational achievement outcomes, we will deploy unconditional quantile regression models to decompose the relative roles of disaster intensity and effects heterogeneities on the distributions of achievement measures.

Authors

• Emily Hannum, University of Pennsylvania

• JERE RICHARD BEHRMAN, University of Pennsylvania