Commodity Prices & Foods as Social Networks

You thought “food network” was a television thing, right? No. A new study models foods and their ingredients as implicit social networks in thinking about commodity prices. It’s a fascinating idea, even if limited data made more constrained conclusions.

The Food Network: Explaining Longitudinal Commodity Prices through Ingredient Dependence Graphs

Joshua A. Lospinoso
United States Military Academy, USA

Fluctuations in commodities prices are not currently well understood, and social network analysis techniques can be applied to strengthen our understanding of this interaction between humans and the environment through agricultural endeavors. This study will compile ~10,000 recipes from popular databases on the internet across various countries and cooking styles in order to create a /food network/, where each ingredient is related to another ingredient through the number of recipes in which they are mutually present. This food network is used as a dependence graph in an actor-oriented model (as in Tom Snijder’s actor oriented graphs:, in order to determine whether these network relationships are statistically significant covariates on commodity price fluctuations. In this way, the method hypothesizes that two goods being present in a recipe imply that demand for them will covary to some degree. The method is able to test these hypotheses through the scientific method using the recipe /food network/ as a natural experiment. A better understanding of the fluctuations in commodity prices would have significant ramifications across broad disciplines. The Network Science Center at the United States Military Academy is interested in this area of research for a number of reasons, including creating a comprehensive plan for incentivizing legitimate farming practices in Afghanistan–a policy problem intimately linked with an understanding of commodities markets. This paper hopes to fill in some of the gaps in explaining the human-agriculture interaction from a networked perspective by testing dependence hypotheses through sociocultural (recipes) and financial (commodities prices) data.


As an amusing aside, here are the fifteen most-used ingredients in the recipe bank underlying the study:

1) onion – 35.57%
2) sugar – 35.15%
3) oil – 30.13%
4) egg – 29.25%
5) garlic – 24.51%
6) nut – 21.07%
7) milk – 18.46%
8) tomato – 15.69%
9) lemon – 15.16%
10) corn – 12.22%
11) olive – 11.25%
12) chicken – 10.15%
13) vegetable – 9.08%
14) beef – 8.85%