This paper examines the effect of the National School Lunch Program (NLSP) on household food insecurity. We use data from the Early Childhood Longitudinal Survey-Birth Cohort along with instrumental variable techniques to estimate treatment effects of NLSP participation. We utilize state policy variation in kindergarten age cut-offs, individual variation across time and the randomness of individual month of birth as exogenous factors to control for observed and unobserved differences between NSLP participants and non-participants.  Additional falsification tests and controls for child care add to the robustness of our findings.  Our three different analytic strategies all provide consistent support for the contention that the NLSP reduces food insecurity by at least two percentage points.  Sensitivity analysis demonstrates that school entry is not associated with reductions in food insecurity among families whose incomes are above 185 percent of the federal poverty line and ineligible for the NSLP for any of our three strategies.  Additionally, controlling for the reduction in child care hours among low-income households does not diminish the size of the NSLP effect.  The finding that NSLP is associated with reductions in food insecurity contributes to the growing literature suggesting that the federal food program is effective. However, our paper is unique for the focus on the transition to kindergarten, a time period that is especially important for future school success.