Hopkins School District in Minnesota implemented an innovative school feeding program, which provides nutritionally sound foods that appeal to students. With access to a unique data set containing students' food service purchases and demographic data from Hopkins High School, we use logit models to analyze the impact of different phases of the program on participation rates, as well as the effect of demographic factors. A relative healthfulness index for the foods purchased is calculated based on information provided by the school dietitian. This index is used to analyze the impact of demographic variables, student lunch expenditures, and program changes on students' lunch consumption. The results of the econometric models indicate that the program innovations have caused positive behavioral changes in most students and that students are choosing healthier foods than under the old program. Moreover, we find that ethnicity and gender play major roles in determining how students react to the new program in Hopkins.