Global food systems are a major contributor to greenhouse gas emissions, yet strategies to mitigate food-related emissions receive low public acceptance. To better understand this gap, this study pioneers a comprehensive analysis of household-level food carbon footprints using a representative panel survey for France from 2017-2019. Using machine learning techniques to match food purchases and environmental data, I unveil significant emission disparities across sociodemographic groups, with a notable portion of these differences attributable to unobserved heterogeneity. By segmenting households into quartiles based on their current emission levels, I estimate a structural demand model that reveals distinct profiles in consumption behavior and price sensitivity, particularly contrasting low and high-emission households. These findings underscore the need to consider households' heterogeneous reactions to price changes when designing climate-related food policies.