Are food deserts also food monocultures? Part 2: New York City case study.

Following up on my earlier post on comparing the diversity of venues selling food versus the presence of food deserts I would like to thank various readers for their feedback. Through these conversations I was referred to a data set, collected by the Food Census, which had done a recent survey on all of the places selling food in a number of New York neighborhoods. Though, as a number of sociologists and demographers have told me, New York City is the exception to everything in the United States. Referring to the USDA food desert mapturned up no such thing in the entire survey area.

On first glance this isn’t too surprising. New York is a dense city with an extensive transit network. Plus, at least from where I’ve been in Brooklyn and Manhattan, the neighborhoods have a high degree of diversity in shops and restaurants. An analysis of the diversity of food venues found in the Food Census survey does appear to back up this assertion. At least at the level of community districts there is very little commercial repetition.

Of course, without data collected within a known food desert we just have diversity levels for an exceptionally dense and connected part of the country. Though, for the fifteen community districts surveyed, what role does population density and poverty play in the diversity of food venues?

To run this analysis the relative frequency for each food venue in a community district was calculated. If, for example, a district had 70 food venues and three were from the same franchise then that venue’s relative frequency (p) would be 3/70. The overall diversity for that district was then calculated using a Shannon diversity index, where the diversity was the negative sum of all of the p*ln(p) terms. This data was then compared to the population density for each district, as well as the percentage of the population on income assistance (All of this data is posted here). The results, at least for this small data set, are as follows:

1. Assuming a simple linear model approximately ten percent of the variation in the diversity of the available food venues is related to the population density of the district (link). As the population density tends to go up so does the diversity of venues selling food. This isn’t too surprising, although it is not a large effect.
2. Assuming a simple linear model approximately five percent of the variation in the diversity of available food venues is related to the percentage of people in a district on income assistance, a proxy measure for poverty (link). If we assume that the presence of food deserts are correlated with poverty, and that this is in turn is correlated with a lack of food venue diversity, then this result too isn’t surprising. However, none of these neighborhoods fall within a USDA food desert, and the correlation coefficient on this fit is not large. Again, New York is not a typical slice of America.

Where to go from here? The type of data collected by the Food Census people is perfect for this type of analysis, plus it contains even more details on the types of foods sold at each place. For further analysis their collection methodology seems like a way to go if we want to apply this type of study to other parts of the United States.

If you know of already existing data sets such as those collected by Food Census, or are interested in collecting such data in your neighborhood, please contact me at levisimons@gmail.com .

(Originally published herehttps://medium.com/@levisimons/are-food-deserts-also-food-monocultures-part-2-new-york-city-case-study-a859db17c18c on December 16th 2014)