Assessment of educational achievement depends on the methodology used and in particular on the statistical indicators of "educational democratization." This term refers to two orders of reality: the historical diffusion of instruction and the equalization of educational opportunities understood as the minimization of the effects of social variables on access to schooling. Examination of the scientific and political stakes involved in the "democratization" of education therefore requires elucidation of the biases behind the choice of the statistical indicators used. The first kind of bias consists in not taking into account the hierarchy of the sections in the school system. The second bias consists in describing educational trajectories exclusively in terms of rates of access to given levels of education, which masks the social conditions of certification. The third bias is connected with the methods used in analyzing this data, which consist in interpreting interdependent variables as if they were isolated and without context. In the end, the present analysis of the use of statistical indicators calls into question the very notion of educational "democratization."