Due to resource scarcity, a paramount concern in ad hoc networks is to utilize the limited resources efficiently. The self-organized nature of ad hoc networks makes the social welfare based approach an efficient way to allocate the limited resources. However, the effect of instability of wireless links has not been adequately addressed in the literature. To efficiently address the routing problem in ad hoc networks, we introduce a new metric, maximum expected social welfare, and integrate the cost and stability of nodes in a unified model to evaluate the optimality of routes. The expected social welfare is defined in terms of expected benefit (of the routing source) minus the expected costs incurred by forwarding nodes. Based on our new metric, we design an optimal and efficient algorithm, and implement the algorithm in both centralized (optimal) and distributed (near-optimal) manners. We also extend our work to incorporate retransmission and study the effect of local and global retransmission restrictions on the selection of routes.