↵Basic regression controls for age dummies, female dummy, race dummy, and pmos*rank*year interactions as well as rain, temperature and temperature squared. Family controls include number of months since last enlistment, total active months in the military, age, college degree, number of dependents, enlisted in the last five years, and a dummy for whether mother was hospitalized for pregnancy-related. Base characteristics include distance to closest city, distance to closest city with 50,000 inhabitants, distance to closest city with 100,000 inhabitants, distance to MTF, dummies for whether closest monitor is within 30 miles, number of sponsors at the base, percent of sponsors that requested base for relocation, Pb, percent officer, percent stepchildren, average number of months in service, average number of dependents, mean ages of sponsors at base, mean age of children at base, percent White non-Hispanic at base, percent of sponsors with some college, percent enlisted within the last five years, percent gone in the next year, percent of children hospitalized for external causes. Standard errors (in parentheses) are clustered at the base level. * significant at 10 percent; ** significant at 5 percent; *** significant at 1 percent