In social processes, long-term trends can be influenced or disrupted by various factors, including public policy. When public policies depend on a misrepresentation of trends in the areas they are aimed at, they become random and disruptive, which can be interpreted as a source of disorder. Here we consider policies on the spatial organization of the French Higher Education and Research system, which reflects the authorities’ hypothesis that scientific excellence is the prerogative of a few large urban agglomerations. By geographically identifying all the French publications listed in the Web of Science databases between 1999 and 2017, we highlight a spatial deconcentration trend, which has slowed down in recent years due to a freezed growth of the teaching force. This deconcentration continues, however, to sustain the growth of scientific production in small and medium-sized towns. An examination of the large conurbations shows the relative decline of sites that nevertheless have been highlighted as examples to be followed by the Excellence policies (Strasbourg among others). The number of students and faculty has grown less there, and it is a plaussible explanation for the relative decline in scientific production. We show that the publication output of a given site depends directly and strongly on the number of researchers hosted there. Based on precise data at the French level, our results confirm what is already known at world scale. In conclusion, we question the amount of disorder resulting from policies aligned with poorly assessed trends.