index.qmd

# Introduction

The ability of people to access quality nutrition has been studied at length in
public health and urban geography for decades [@beaulac2009;@walker2010]. 
This interest is motivated in large part by an observed spatial disparity in
nutrition access in many communities --- though this issue may be particularly
pronounced in the United States [@beaulac2009]. The spatial disparity has been linked at an aggregate level with negative public health outcomes 
[@chen2016; @cooksey-stowers2017], though other complicating factors including prices and habits may be present as well [@ghosh-dastidar2014].


At the same time, access to nutrition and to community resources in general is 
frustratingly hard to define. "Accessibility" is an abstract concept without a
specific quantitative definition [@handy1997]. However, using accessibility as 
a policy measure requires comparative quantification, and transportation and
public health researchers have constructed several quantitative measures, such 
as presence of a store within a travel time buffer, or the distance to the 
nearest store. These measures are relatively easy to calculate using 
readily available GIS software, but elide much useful information [@dong2006; @logan2019].
These types of measures require the researcher to make a series of assumptions
and assertions: why is 30 minutes chosen instead of 40? Is that time by transit
or highway or walking? Should these definitions change for individuals in
different socioeconomic groups? And people do not always go to the closest grocery
store to begin with [@clifton2004; @hillier2011]; how much further are people
willing to travel to go to a store that is cheaper or that has a wider variety
of goods? And perhaps the home location isn't the only spatial point of 
reference [@liu2022]. A measure that potentially combines many of these different
considerations is desirable. 

In this research, we develop and explore an accessibility 
measure based on destination choice models estimated for three distinct 
communities in Utah. This methodology is based on a unique dataset made by 
linking between three extensive data sources:

  1. A detailed survey of the nutrition market in three Utah communities.
  2. Location-based services data derived from mobile phone records revealing
     which grocery stores are frequented by residents of different neighborhoods.
  3. Multi-modal network data providing detailed mobility data by car, walking,
  and public transit.

These data will be combined in order to develop accurate logit models that
demonstrate the variables that are significant to grocery store choice in Utah.
These models could then be used to find accessibility to stores and impact
transportation policy to improve quality of life for all communities in Utah.

This paper is organized in a typical manner. A methodology
for data collection and modeling is described in @sec-methods and a description
of the nutrition environment and choice models estimates follows in
@sec-results. @sec-scenarios presents a series of scenarios to which we apply
the models estimated in @sec-results, illustrating the interrelated elements of
nutrition quality and transportation infrastructure in developing more complete
access to nutrition. @sec-conclude places the findings of this research in 
context, while discussing limitations to the methodology and associated future
research opportunities.