We model investment decisions regarding innovation and emissions abatement in a dynamic theoretical framework. Considering knowledge stock as an impure public good, we study the reaction function between one representative agent’s investments in innovation and the other agents’ investments in the public characteristic of the impure public good. We demonstrate that the reaction function has a positive slope under general conditions and that its sensitiveness is affected by assumptions on the elasticity of substitution in the benefit function. The positivity of the reaction function is then empirically tested in an econometric estimation. We exploit an original sector-based database by gathering innovation efforts as well as polluting emissions and economic dimensions over the time span 1996-2005 for 15 European countries and 23 manufacturing sectors. Empirical results show that sector-based innovation investment is positively driven by the public characteristics provided by other sectors. Different reactivity strength for different polluting emissions also allows us to disclose the role of complementarity in agents’ decisions.