Healthy/unhealthy breakfast teenagers' choice: an experiment combining a nutritional training program and the promotion of unhealthy products
The experiment was conducted at 104 schools in the city of Barcelona, with 126 classrooms and 3,291 children participating in the study (average age 13 years, 52% girls, 15% obese or overweighed). Randomized selection of schools was stratified by district and school ownership. A voucher for a free breakfast was provided to each student in both groups, experimental and control. A few weeks later we returned to offer the same voucher. Meanwhile, children in the experimental group had received nutritional training. Vouchers were designed so that one third of students had the opportunity to choose a 2 for 1 promotion of unhealthy food or drink.
We applied Differences in Differences linear models that account for individual, family and school/neighborhood characteristics, and for the influence of peers. We clustered standard errors at class level, and performed robustness analysis to consider the endogeneity of peer effects. We also estimated the model for specific groups of pupils: girls vs. boys, low socioeconomic districts, public schools, receiving pocket money regularly, immigrant condition, ageing within their cohort, overweight condition. In order to improve the estimation efficiency, we estimated the equations for food and beverage choices jointly allowing correlation between their errors due to common unobserved determinants.
Our results show that 1) choices of food and beverages are quite heterogeneous. At baseline, by half of participants chose unhealthy food while approximately two thirds chose unhealthy beverage, most of them coke; 2) the training program was effective to promote changes towards healthy breakfast, and the effect was more intense for food than for beverages. 31% of students in the experimental group moved to a healthier food option compared to 20.2% in the control group whereas these percentages were significantly lower for beverages (19% and 11.6%, respectively). The econometric model corroborates that descriptive result: the estimated reduction is 14.8% in the choice of unhealthy food and 5.9% in unhealthy beverage; 3) students are sensitive to prices (promotion of unhealthy products), particularly in the second round, as some students might had considered the first voucher just “hypothetical”, that is no one would come with the real breakfast the following day. Contrary to the training program, the effect of price promotion is higher for beverages; 4) pupil’s peers have a significant and large influence.
In summary, this pioneering work reveals the heterogeneity of the effects of a nutritional intervention at school. Food and beverages are different. Students are more resistant to good advices against sweetened sodas than to advices on unhealthy sugared food, and they are sensitive to 2 for 1 promotion of unhealthy products and to peers.