Evaluation with Inadequate Data: The Impact of the French Vending Machine Ban

Abstract

We estimate the effects of the 2005 ban on vending machines in French schools using the 1998 and 2006 INCA nutrition surveys. These surveys contain no information on the presence of vending machines in schools attended by respondents, but the adoption of a Difference-in-Difference design, and a Regression Discontinuity Design enable us to obtain indirect estimates of the policy impact. The application of multiple methods has given promising results. Despite differences imputable to the changes in benchmarks and sample sizes, results are highly consistent, especially when the focus is limited to a single eating occasion, the mid-morning break which we assume to be the most exposed to vending machine consumption. On balance, we found significant, albeit small effects of the policy. These effects tend to disappear when the overall diet is considered, which could hide compensating behaviours or simply the fact that influencing the morning snack is not enough to improve the overall dietary quality.