The great divide in scientific productivity: why the average scientist does not exist

Using a panel of individual researchers at the KU Leuven, Belgium, we analyze the impact of a range of productivity drivers on research performance, at the separate quantiles of the productivity distribution. We estimate a correlated random-effects quantile regression model, accounting for unobserved heterogeneity of researchers and applicable to count data. We find that the effect of most regressors, particularly system-factors incentivizing researchers (like promotion record and access to research resources), as well as the gender of the researcher differ significantly at different points in the distribution, yielding strong support for our quantile regression approach. Comparing publications versus citations as dimensions of research performance, we find the incentive factors to work stronger in affecting research quality. Finally, the splitsample regression results emphasize the heterogeneity across scientific disciplines.