Data-Driven Food Desert Metric to Understand Access to Grocery Stores Using Chi-Square Automatic Interaction Detector Decision Tree Analysis

Abstract

A food desert is a geographic area lacking spatial and socioeconomic access to healthy foods. Disparities in food access and availability hinder public health and individual wellbeing. Access to healthy food has been evaluated by many agencies and researches yet they do not comparably demarcate food deserts or identify vulnerable populations in methodical terms. Existing literature suggests a link between food deserts, income level, and vehicle ownership. This study evaluates the existing methods and proposes a novel data-driven method to identify food deserts in Baltimore, Maryland. This study evaluates responses from 573 respondents for an in-depth analysis of individual grocery store choice and travel decisions. Chi-square automatic interaction detector (CHAID) decision trees are used to develop a user-generated food desert metric. Income level is found to be a key indicator for food desert demarcation, more than vehicle ownership. Network distance was applied to develop the prioritization matrix for a food desert. This research provides a replicable method for determining food-insecure areas in a locality by aggregating individual data to identify such areas. Such a metric can aid policymakers to make investment decisions and to direct resources to areas of need.