The mutual reinforcement of scientific and technological knowledge—a technology-level analysis

While the contribution of science to technological progress has been empirically confirmed, the literature looking at the positive feedback from technology to science remains limited, because of the constrained citation linkages. We apply a novel machine-learning-based method to attribute 2.5 million scientific articles to the categories of WIPO patent taxonomy. We then employ Granger causality analysis to study the coevolution of technological and scientific knowledge over more than 25 years. We demonstrate that the evolution of scientific output is a good predictor of the evolution of technology production, while the opposite effect is weaker. Looking at individual WIPO technology fields, we find significant effects in about one-third of categories.