In this paper I compare the grade transition rates (repetition, promotion and drop out) given by the Education Ministry Statistics Office based in the school census data with rates form PROFLUXO, a mathematical model wich allows us to get theses indicators through the use of IBGE's PNADs (National Household Sample Survery). The picture we get from these indicators shows that MEC's methodology leads to serious errors witch distort the vision of Brazilian educational reality. Opossite to MEC's data it shows that the rates of school drop put are very low in the initial grades of elementary school. This finding leads to the understanting that the notion of child's early drop out from school caused by social on cultural determinants is misleading and that Brazilian families, on the contrary, make an enourmous effort to keep their children a school. On the other hand it shows die low teaching performance which prevents, through high repetition rates, the universalization of Elementary Education in Brazil.
