To publish or not to publish? On the aggregation and drivers of research performance