Are food deserts also food monocultures? Part 4: United States case study

Are places with a lower diversity of food venues in a given area more likely to be to be food deserts? This is a question I've been researching (Part 1, Part 2, Part 3) with different commercial data sets as well as the USDA's map of food deserts.

The original supposition in this model is the analogy between the competition of species in an ecosystem and the competition of food venues in a marketplace. Just as the diversity of species in an ecosystem can be described by the Shannon index my second supposition is that the diversity of food venues can be described in an identical way.

One of the questions I’m interested in addressing is how this local measure of diversity relates to the presence of what the US Department of Agriculture (USDA) denotes as food deserts; an idea which I originally posted here. An area, in this case a census tract, can be defined to be a food desert if it falls within the USDA’s definition of low income and one of three different criteria for low access:

Criterion 1. No access to a grocery store within 1 mile of residence, if in an urban area, or 10 miles if in a rural area. This is the original criteria.
Criterion 2. No access to a grocery store within 0.5 miles of residence, if in an urban area, or 10 miles if in a rural area.
Criterion 3. No access to a grocery store within 1 mile of residence, if in an urban area, or 20 miles if in a rural area

To help in addressing my original question on the relationship between food venue diversity I was given access to the United States' business database for 2011. This data set contains information on the location and type of every registered business in the state at that time. Off all of the businesses listed I only selected those with NAICS codes (North American Industry Classification System) which corresponded to food venues (A full list of food venue categories is available on the output data set here.). This includes a number of different food venue categories, such as convenience stores and produce markets. A script was then written to do the following analysis:

1. Count all of the food venues of a specific type for a given area and calculate the Shannon index and the Shannon equitability index. The Shannon index is a raw measure of diversity, but it does not factor in differences in the number of businesses for a given area. The Shannon equitability index adjusts for the number of businesses in a given area by checking how diverse a population is against a theoretical maximum diversity. If this ratio is 0 then the population is a pure monoculture, if it is 1 then every member of the population is unique.

2. This information was then combined with the most recent food desert data from the USDA so that every census tract would have a food desert score, according to each of the three criteria, as well as the number of categorized food venues, their Shannon index, and Shannon equitability index. Maps of this diversity and equitability data can be found here.

3. These data sets were then combined to produce the following table. This table shows the average Shannon equitability index for census tract areas denoted, and not denoted, as food deserts according to all three criteria.

After combing through thousands of entries what do these results mean? Using both the first and third criteria for food deserts shows very little difference in food venue equitability. This implies that census tracts where city dwellers are within a mile of a grocery store, and people living in rural areas are within twenty miles of a grocery store, live in areas with very similar levels of food venue diversity as to people in areas which fail these access thresholds.

Where the difference is notable is when the second criteria for defining a food desert. These results indicate people living in urban areas without access to a grocery story within 0.5 miles of their home, or rural residents within 10 miles, have access to a significantly lower level of food venue diversity. The difference is notable in that food desert census tracts have a Shannon equitability index, on average, only about a quarter of the size for census tracts falling outside of this threshold. This holds true independent of food venue category.

This of course leads to a whole host of other questions, such as what happens between 0.5 and 1 mile in food access which is associated with such a dramatic drop in food venue diversity. This analysis also only includes commercial data. It would be interesting to see what role community gardens and food trucks may also play in this type of geospatial anaysis. I believe much more work should be done on this topic, and I welcome feedback and ideas for collaboration. This analysis shows that the whole United States shows the same pattern as California: food deserts may appear to be food monocultures.

Questions and comments: levisimons@gmail.com .