Abstract The appropriate allocation of CO 2 emission quotas can build up a solid foundation for future emissions trading. Building CO 2 emissions allocation mechanism has significant practical implications. Based on radial zero sum gains data envelopment (ZSG-DEA) allocation model, this paper uses a non-radial ZSG-DEA model to allocate CO 2 emissions between different Chinese provinces. Unlike previous studies treating CO 2 as the input variable, we treat CO 2 as the undesirable output variable. The modified model can better reflect macro-production process while earlier relevant models treated CO 2 emissions as inputs. This paper contributes to the existing resource allocation method and allocates China's provincial CO 2 emissions from the view of technical efficiency. The results of our paper reveal that (1) between 2006 and 2010 the cumulative optimal amounts of CO 2 emissions were higher than the actual amounts in 12 provinces, and lower in other 18 provinces. (2) Several energy-abundant provinces such as Shanxi and Inter Mongolia need to take more responsibilities in CO 2 emissions reduction. (3) After the ZSG-DEA allocation, all provinces' CO 2 emissions are on ZSG-DEA frontier, which reflects the overall “Pareto Optimality”.