Measuring accessibility to grocery stores using radiation model and survival analysis

Measuring spatial accessibility to grocery stores and identifying food deserts have been of interest to planners and policy makers. However, measuring and quantifying accessibility in a way that relates to actual travel behaviors and preferences are challenging. This study aims to promote an existing radiation model by incorporating different transportation modes and time use diaries to provide a more realistic estimation of accessibility. We proposed a survival analysis method, such as the Cox proportional hazard model, as a novel approach to develop a cost decay function based on time use diaries. A multinomial logit model was applied to estimate the portion of people walking, traveling by car, and traveling by bus. A case study of grocery stores in Travis County, Texas, is used to illustrate the approach. The results indicate that for walking accessibility to grocery stores, the majority of zip codes fall into the very low, low, and moderate accessibility classes. Moreover, in most western zip codes transit accessibility to grocery stores is found to be very low. However, the spatial accessibility for some of such zip codes has been enhanced because of the presence of cars.