This study evaluates the effectiveness of supplementary food program that was aimed to help the children to maintain their health when facing the 1997/98 economic crisis in Indonesia. To do so we apply difference-in-difference method for two different kinds of sample: unmatched and matched one. The results from unmatched (using pooled OLS and fixed-effect) and matched (average treatment effect based on propensity scores) tend to be consistent: during 1997/98 crisis, children who were exposed to the program have better nutritional status relative to those who were not. Yet with matching sample we manage to produce higher estimated program effect on nutritional status of the treated children. This result may suggest that the use of matched sample may even further eliminate the contamination of the program effect from unobserved heterogeneity. In addition, the results also raise the importance of properly pick the control groups for the treatment when it comes to the evaluation of public program using survey data.