The Impact of Inflation on Healthy Food Spending: Evidence from Double Machine Learning Analysis

This paper examines how food price inflation affects household healthy food spending in the United States, with attention to heterogeneity across income, racial, and urban-rural groups. Using household-level Consumer Expenditure Survey (CEX) Diary data merged with regional Food-atHome (FAH) Consumer Price Index data from 2015 to 2024, the study exploits temporal and regional price variation to identify causal effects. The empirical strategy combines Fixed Effects OLSwithaDoubleMachineLearning(DML)frameworkemployingLASSO,ElasticNet,Random Forest, and Gradient Boosting algorithms to address high-dimensional controls and nonlinear confounding. While FE OLS yields a positive but insignificant estimate, DML consistently finds a large, significant positive effect of FAH CPI on healthy food spending, with estimates ranging from 1.12 to 1.20 across specifications. Heterogeneity analysis reveals no significant income-based differences, suggesting broad-based spending adjustments. Asian households exhibit a significantly stronger response under LASSO and Random Forest, and urban households increase healthy food spending more than their rural counterparts. These findings have important implications for food assistance program design and nutrition policy, particularly regarding rural food access during periods of macroeconomic volatility.