This paper studies the determinants of scientific productivity from an original database of French academic economists observed in 2009 and 2010. All individuals of this data set were involved in the first and second experiences of the "Prime excellence Scientifique". This academic competition was introduced by the French academic system to select best publishers and PhD supervisors. The context of this competition conforms to the tournament theory in two points. The scientific productivity was observed in a same period from 2005 to 2010 and competitors are selected and ranked by their relative performance in publications using the same criteria (National Scientific Research Center ranking). We construct a scientific production index and estimate productivity regressions using censored data models (Tobit). We control for individual characteristics like age, gender, environmental context, initial publication performance (h index), in line with stylized facts and theoretical foundations like Lotka law, Matthews effect and scientific life cycle productivity. The paper brings two novel dimensions in the literature. The first is done by controlling for the multi-task activities of tenure and associated professor when evaluating their scientific production. We control for the time allowed to alternative occupation like teaching, doctoral supervision, administrative and scientific responsibilities. The second outcome of the paper consists to evaluate the impact of network spillovers externalities induced by local competences accumulated in economics departments (coauthorship, peer externalities...), on each individual performance.