Causal asymmetry suggests productivity underlies scientific collaboration

Abstract

Scientific productivity and collaboration are closely related, yet the causal direction between these two dimensions of academic careers remains unclear. Here we analyze the careers of 26,876 Brazilian researchers holding a nationally prestigious fellowship across multiple disciplines to test whether productivity shapes collaboration or the reverse. Focusing on total publications and distinct coauthors, we assess their directional relationship using a causal discovery framework based on asymmetries in conditional and distributional structure to determine the more plausible generative direction. Collaboration conditioned on productivity exhibits a highly regular mean-variance relation consistent with a geometric model, whereas the reverse conditioning yields a much less coherent pattern. Consistently, publication counts follow a lognormal distribution, whereas the coauthor distribution is reproduced by a geometric-lognormal mixture where productivity determines the expected number of collaborators. Surrogate tests indicate that the data are considerably more compatible with productivity shaping collaboration than with the reverse. The two quantities also accumulate asymmetrically: publication inequality rises steadily with career age, whereas coauthor inequality does not, indicating that advantage compounds more in output than in network breadth. Area-resolved analyses reveal a more nuanced picture, and although the direction where collaboration shapes productivity is never favored, the two directions are most clearly distinguished in areas more organized around laboratory and research-group models, where productivity inequality also grows most clearly. Among this cohort of prominent scientists, larger collaboration networks are thus more likely to reflect cumulative advantage in productivity than to independently generate greater output.