Abstract By estimating a simultaneous panel data model of environmental innovation and toxic air pollution, this paper identifies bi-directional causal links between the two. We study a panel of 127 manufacturing industries over the period 1989–2004. Pollutant emissions are an implicit measure of policy stringency and environmental patent counts are used to measure environmental innovation. After accounting for the joint endogeneity, we find that environmental innovation is an important driver of reductions in US toxic emissions. Conversely, we find that tightened pollution targets induce environmental innovation. However, our estimates indicate that the “environmental policy multiplier” – the proportionate contribution of induced innovation to long-run emission reduction – is small.