The experimental literature has shown the tendency for experimental trading markets to converge to neoclassical predictions. Yet, the extent to which theory explains the equilibrating forces in markets remains under-researched, especially in the developing world. We set up a laboratory in 94 villages in rural Sierra Leone to mimic a real market. We implement several treatments, varying trading partners and the anonymity of trading. We find that when trading with co-villagers average efficiency is somewhat lower than predicted by theory (and observed in different contexts), and markets do not fully converge to theoretical predictions across rounds of trading. When participants trade with strangers efficiency is reduced more. Anonymizing trade within the village does not affect efficiency. This points to the importance of behavioral norms for trade. Intra-village social relationships or hierarchies, instead, appear less important as determinants of trading outcomes. This is confirmed by analysis of the trader-level data, showing that individual earnings in the experiment do not vary with one's status or position in local networks.