Although haggling over the prices of goods and services is an integral part of life for the world's poor, little work has been done to extend bargaining models to a developing country context. We develop a theoretical model of bargaining that takes into account the initial endowment of the seller. Our model predicts that sellers with lower initial endowment will settle for a lower price. We test our model using data from a bargaining exercise over a real product, a child's shirt, with garment making microenterprises in Ghana. Our data supports the major testable prediction from our model: sellers from poorer households agree to a lower price than rich sellers. This relationship is robust to controlling for a number of potentially confounding firm and owner characteristics, including firm size and product quality. Preliminary findings from a lab-in-the-field experiment, in which we randomly assign initial endowments in a bargaining game that mirrors the shirt exercise, appear to corroborate our results. We conclude that further exploration of this ``need-bargaining'' relationship is a key frontier in future bargaining research aimed at understanding the lives of the poor.