Working Paper: NBER ID: w24892
Authors: Stefan Hut; Emily Oster
Abstract: We use grocery purchase data to analyze dietary changes. We show that households – including those with more income or education - do not improve diet in response to disease diagnosis or changes in household circumstances. We then identify households who show large improvements in diet quality. We use machine learning to predict these households and find (1) concentration of baseline diet in a small number of foods is a predictor of improvement and (2) dietary changes are concentrated in a small number of foods. We argue these patterns may be well fit by a model which incorporates attention costs.
Keywords: diet quality; machine learning; household behavior; public health
JEL Codes: I12
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| attention costs (D91) | dietary improvements (I19) |
| baseline dietary concentration (L72) | dietary improvements (I19) |
| disease diagnoses (I12) | dietary changes (I12) |
| changes in household circumstances (D19) | dietary changes (I12) |
| baseline dietary concentration (L72) | likelihood of being a 'changer' (D91) |