Abstract: The environmental policy in passive and active phases is different,and some enterprises have started product innovation strategy in the stage of active carbon reduction stage.On the basis of theory of stage division, an exploratory study on innovation strategies in the stage of active carbon reduction is conducted carbon emissions trading and government subsidies constructed are introduced, a Cost-benefit model of enterprises and industry active carbon reduction is, whether there is an optimal strategy is determined by matri positive definiteness, and wether this strategy can realize Pareto optimality and how.The result shows an optimal carbon reduction strategy, enterprises investment in which increases with the increase of corresponding emission reduction effect coefficient, reduction benefit coefficient, and government extra subsidy coefficient on the carbon emissions trading; realization of industrial Pareto optimality needs the subsidy, and the extra subsidy coefficient is an industrial default value, which declines with the increase of enterprises emission reduction effect coefficient and innovation revenue coefficient.Values are assigned for the parameters in the model based on field research and related literature data,and cement industry is taken as an example to validate the model, sensitivity analysis of key parameters is also completed. Example analysis confirms the benefits of innovation and the subsidies are the key factors for enterprises initiative carbon reduction. The results will provide a basis for enterprises selection of optimal initiative carbon reduction strategy and reference for government's fiscal policy. Furthermore, it indicates that government should ensure efficiency of carbon emissions at the same level in the process of Pareto improvement of enterprises carbon-reduction and the enterprises with strong innovation ability should correspond larger emission reduction to achieve Pareto optimality of carbon reduction.

Key words: stage division, active carbon reduction, government policy, Pareto optimality