To examine how milk consumption varies by milk choice and measure the association of milk choice on the nutritional and energetic content of National School Lunch Program (NSLP) lunches.

An observational plate waste study using the Digital Photography of Foods Method.

Data were collected from two suburban South Carolina schools in one district during February and March 2013.

Totally, 968 NSLP lunches selected by 619 kindergarten to fifth grade students.

Most students chose chocolate milk (75 %). A multinomial logit model indicated milk choice varied significantly by sociodemographic characteristics. An ordinary least square regression indicated that consumption rates of low-fat white milk were 8·5 % lower than fat-free chocolate milk (P = 0·039) and milk consumption rates varied statistically by sociodemographic characteristics. Ordinary least square regressions found that the consumption of energies and nutrients from NSLP lunches varied with sociodemographic characteristics and milk choice; students selecting chocolate milk consumed 58 more energies (P < 0·001) and 10 more grams of total sugar (P < 0·001) than students selecting low-fat white milk from their NSLP lunches. Students consumed statistically similar energies and nutrients from the non-milk components of their meals.

Students selecting chocolate milk consumed more energies and nutrients from their NSLP lunches with the increases in consumption attributed to the milk component of the meal. The findings have implications for recent changes to NSLP guidelines that allow schools to offer both low-fat and fat-free flavoured milk, reversing the previous ban on low-fat flavoured milk under the Healthy, Hunger-Free Kids Act.