Primary Submission Category: Machine Learning and Causal Inference
Evaluating the impact of popular films on public interest in and consumption of plant-based food
Authors: Anna Thomas, Maya Mathur, Jessica Hope,
Presenting Author: Anna Thomas*
Due to the harms of factory farming to public health, the environment, and animal welfare, many organizations, including the UN and the Intergovernmental Panel on Climate Change, have called for a shift to a plant-based diet. Interventions such as educational campaigns, alternative protein, and policy changes may serve to accelerate this shift. Here we focus on evaluating educational campaigns in the form of documentary films, in order to provide a recommendation to the organizations we work with on whether further promotion of specific existing films (e.g. via screenings or advertisements) or creation of new films are likely to be effective. Several of these films received press on anecdotal reports of behavior change.
We have access to national-level time series on exposures, various outcomes, and time-varying covariates. Using this observational data, we aim to estimate short-term effects of point interventions on film exposure on the outcomes. Results from our pre-registered approach using a doubly robust, propensity score-adjusted regression model, a special case of the structural nested mean model, suggest contemporaneous and lagged effects of some films on Google search volume for plant-based food. However, we do not observe an effect on meat demand. In ongoing work, we are studying the impact on sales of plant-based and animal-based food via the National Consumer Panel, as well as assessing sensitivity of our findings to unmeasured confounding and other assumptions.