Measuring food waste at the individual household level has been nearly impossible because comprehensive, current data on uneaten food do not exist. By using food acquisition data, this article employs a new approach to estimating household‐level food waste via a stochastic production frontier model in which food waste is identified as input inefficiency. For households in our data, the average household wastes 31.9% of the food it buys, and this figure, using survey weights, translates to annual U.S. consumer‐level food waste valued at $240 billion. In addition, by accommodating heterogeneous wasting behavior, we find that healthier diets and higher income lead to more household food waste, whereas lower household food security, food‐assistance program participation, and larger household sizes are associated with less food waste.