4 METHODOLOGY
4.3 Data Collection
4.3.1 Spatial Analysis with GIS: Study Area Qualification
A tool as well as a method of spatial analysis known as geographic information systems (GIS) was a major component of this research. As I am concerned with spatial information of the participants and the location of food stores, the use of GIS is ideal. GIS is a computer-based technology and methodology for collecting, managing, analyzing, modeling, and presenting geographic data for a wide range of applications. The software used in this research is produced by ESRI (Environmental Systems Research Institute or also known as Esri)—ArcMap, version 10.4.1. GIS allows for the representation of tabular data in graphic form, such as maps. This is done with the use of tables, i.e., databases, and those tabular data are then converted into spatial data in the form of digital maps produced by the software system.
For this research, the initial step of locating the food deserts needed to be completed. I obtained a list of grocery stores and Census block data that were then transferred into the GIS software (Figure 4.1). Atlanta actually has multiple areas that qualify as food deserts, based on the USDA’s qualifying criteria of a one mile distance to a food store, (Figure 4.2). Figure 4.2 shows the Atlanta Journal Constitution’s map (AJC 2015) production from a mapping tool provided by the USDA. The USDA mapping program allows the layperson to perform mapping of food deserts across the country. The data are based on the spatial distribution and density of
the food stores overlaid with residents’ demographic data--race and ethnicity and average
household income. In order to perform additional analysis myself, I obtained and mapped Census data and food store locations in order to find those food deserts within the Atlanta study area. These data were then analyzed together to also determine where future recruitment of
participants would be (Figure 4.3).
The use of the Census data was to allow for a more comprehensive analysis of the demographic information alongside the store data. With these data, I chose among the most prominent looking food deserts, i.e. areas that looked to have the sparsest distribution of food stores within the one-mile radius, no major supermarket grocery store, and that also qualified as having a high concentration of fast food establishments. One-mile buffer rings were added as layers on top of available food stores, and spatial analysis was performed in the GIS software to determine those areas in which there was a gap in food store locations, especially major grocery stores (Figure 4.4). Data that were mapped included 1) store locations, available from the North American Classification (NAICS) database of business listings (NAICS code 445110:
commissaries, primarily groceries; delicatessens primarily retailing a range of grocery items and meats; food (i.e., groceries) stores; grocery stores; supermarkets), 2) code 722513: limited service restaurants (i.e. fast food and quick service restaurants), and 3) Census block data showing income levels. One can merely “eyeball” the map to determine the clustering of food stores within the particular area, however, to be more certain, hot spot analysis was also performed as a test to more accurately define those residential areas that qualified based on the income levels (Figure 4.5). This analysis was only performed on the fast food establishments as there were just a handful of major grocery stores in the area.
To add an extra layer of validity, I performed a hot spot analysis in the GIS software. The hot spot analysis tool within GIS helps to determine if there are any statistically significant clustering patterns within the data. The hot spot analysis tool calculates the Getis-Ord Gi* statistic for each feature and dataset and provides p-values and z-scores to indicate which
features with high or low values cluster spatially (Esri 2015). Though the analysis is to highlight hot spots (high values), it could also highlight the “cold spots”—showing areas of low values, making it useful for either purpose. The z-score represents the statistical significance of clustering for a specified distance, in this case, one mile. Figure 4.6 shows these results.
After the hot spot analysis and in analyzing where these stores were located in relation to the population, I was then able to find particular areas within the city that would classify as food deserts. By identifying the food store locations, I was able to measure outward to determine the travel distance away from these locations. Figure 4.5 indicates the proximity measurements (using a one-mile radius) from major supermarkets and analyses already performed. I chose two areas in Atlanta that consisted of lower income and predominantly non- white racial/ethnic makeup. The reasoning behind the particular neighborhood choice is that there are a few major grocery stores and a high concentration of fast food establishments within the area and they also consisted of the qualifying socio-demographic criteria. Also, as one of main themes of this research is to examine residents’ experiences through a critical race perspective, my target group had to consist of those demographics.
For each of the food studies discussed above, it should be noted that there has not been one set parameter in the measurement of the food environment and food access, and this is discussed in more depth in chapter five. Again, food environment is defined as an environment in which people access their food, to include food stores, restaurants, schools, and work sites,
and there may not be a lot of variety within this environment (McKinnon et al. 2009). By its nature, measurements of the food environment take on a geographical element and because there is no systematic compilation of measures of the food environment, these measures can vary depending on the study being performed.
Many researchers have defined those geographic areas that lack access to healthy foods (McKinnon et al. 2009, Wrigley et al., 2002, Cannuscio et al. 2013, Glanz et al. 2005) and have identified food deserts based on their measurement criteria. For this research, the USDA’s (2009) measurement criteria of food desert and access have been utilized: the reasonable distance
allowing for access to supermarkets is roughly one mile. Figure 4 indicates the buffer zones of one mile. As this criterion is still subjective, my hope in this research was to also find out if this distance is in fact reasonable to participants. The idea was that the reasonable distance could
very well be a greater or lesser distance, depending on particular variables that would arise in the research, specifically, what the participants identified as a reasonable distance in their accessing food.