Can Geographical Factors Determine the Choices of Farmers in the Ethiopian Highlands to Trade in Livestock Markets?

Proximity and affiliation to the local market appear to be two of the most relevant factors to explain farmer’s choices to select a particular trading point. Physical barriers may limit the options, especially in developing countries. A network of villages linked by traders/farmer-traders sharing livestock markets was built with field data collected in 75 villages from 8 kebelles in the Wassona Werna wereda of the Ethiopian Highlands. Two exponential random graph models were fitted with various geographical and demographic attributes of the nodes (dyadic independent model) and three internal network structures (dyadic dependent model). Several diagnostic methods were applied to assess the goodness of fit of the models. The odds of an edge where the distance to the main market Debre Behran and the difference in altitude between two connected villages are both large increases significantly so that villages far away from the main market and at different altitude are more likely to be linked in the network than randomly. The odds of forming an edge between two villages in Abamote or Gudoberet kebelles are approximately 75 % lower than an edge between villages in any other kebelles (p,0.05). The conditional log-odds of two villages forming a tie that is not included in a triangle, a 2-star or a 3-star is extremely low, increasing the odds significantly (p,0.05) each time a node is in a 2-star structure and decreasing it when a node is in a 3-star (p,0.05) or in a triangle formation (p,0.05)), conditional on the rest of the network. Two major constraining factors