*March 24, 2017 revised

This paper estimates a professor's value added to a postgraduate student's research achievement growth using unique panel data on matched advisor-advisee pairs in a world-leading physics graduate program. To address an identification problem related to the endogenous selection of advisors and advisees, we use professor turnover and estimate a semi-parametric lower bound of the variance in advisor quality affecting advisee research performance. We find that a one-standard-deviation increase in professor quality results in a 0.54 standard deviation increase in a doctoral student's research achievement growth, increasing the number of first-authored papers that are published in top journals by 0.64 at the doctoral level.