Mortality specification Clause Samples

Mortality specification. ‌ Given these underlying differences in the demographic characteristics of those affected by tropical cyclones, a simple comparison of birth outcomes will not reveal the causal effect of storm exposure. In order to capture this, my main empirical strategy is a difference-in- differences (DD) approach to estimate the effect of in utero tropical cyclone exposure on birth outcomes. The first source of variation is geographic, based on the mother’s district of residence. The second source of variation is temporal, based on the infant’s date of birth and estimated date of conception. Because exact birth dates are not provided in the births sample, I consider children whose mothers reside in districts whose population centers are within 50 km of a tropical cyclone within the 9 months before their birth to be exposed in utero. I also demonstrate the robustness of these results in Section 7 by executing analysis without births or conceptions in the month of a storm. I use a linear probability model to estimate the effect of in utero exposure on a) mortality within the first 28 days (neonatal mortality) and b) mortality within the first year (infant mortality). The model includes birth cohort, district, and state-by-year fixed effects, and to account for the assignment of exposure to whole districts, standard errors are clustered at the district level. This is modeled by equation (1): Yidsmy = β0 + β1(exposuredmy) + γd + δmy + ηsy + Xi + Zdy + ϵidsmy (1) where the dependent variable Yidsmy denotes the birth outcome for infant i born to a mother residing in district d and state s in month m and year y and exposuredmy is a binary variable taking on a value of 1 if the child’s district d is within 50 km of a cyclone track during their period of gestation, defined as an estimated 9 month period. This model includes three sets of fixed effects: district FEs (γ), month-year birth cohort FEs (δ), and state-by- year FEs (η). Additional control variables are defined as follows: the vector X consists of individual infant and mother characteristics including sex of the child, birth order of the child, whether the child is a ▇▇▇▇▇▇▇▇▇, marital status of the mother, education level of the mother, mother’s ethnicity/ caste membership, total children born to mother, and a household wealth index; the vector Z consists of district-survey year characteristics including female literacy, percentage of residents living in urban areas, and an average measure of wealth index.‌