Data offloading is a promising technique for mobile terminals (MTs) for conserving energy in wireless networks. This technique enables MTs to support heavy data traffic by offloading computation-intensive tasks to the cloud. However, due to the long transmission up-link, data offloading may lead to unnecessary energy usage. In this paper, we propose a joint collaborative data offloading framework with channel selection for MTs, in which MTs can first form a coalition and then execute data offloading in a cooperative way. We adopt game theory to formulate the problem as a cooperation game. We then prove that the formulated game is a potential game and can achieve Nash equilibrium. A Markov approximation approach is applied to design a distributed channel selection algorithm for the game so that each MT can self-organize into stability without information exchange across the whole network. Analytical and numerical results show that the algorithm is feasible and efficient in comparison to traditional centralized optimization solutions.