Children receiving free or reduced-price school lunches through the National School Lunch Program tend to have worse health outcomes on average than observationally similar children who do not participate, especially in the case of food insecurity. Whether these puzzling correlations reflect causal impacts of the program has become a matter of considerable debate among researchers and policymakers. Much of the empirical literature maintains the untenable exogenous selection assumption, and the systematic classification error problem appears to have been completely ignored. Reviewing the general literature on the causal impacts of food assistance programs, Currie (2003) goes so far as to conclude that ‘‘many studies have . . . simply ‘punted’ on the issue of identification’’. Bhattacharya et al. (2006), in studying the National School Breakfast Program, suggest that ‘‘no study has dealt convincingly with endogenous participation’’ หน้า11