We study the impact of university-industry research collaborations on academic output, in terms of productivity and direction of research. We report findings from a longitudinal dataset on all the researchers from the engineering departments in the UK in the last 20 years. We control for the endogeneity caused by the dynamic nature of research and the existence of reverse causality. Our results indicate that researchers with industrial links publish significantly more. Productivity, though, is higher for low levels of industry involvement. Moreover, growing ties with the industry skew research towards a more applied approach.