This paper seeks to measure empirically the effect of supply and demand variables on performance in the Census Student Assessment (ECE). To this end, it used for the first time a database that includes both the socioeconomic conditions in which students live, as well as information about the school they study. It has been estimated a production model of educational attainment and sequential model to explain the passage between the levels of the ECE. The latter model distinguishes the heterogeneous effect that certain variables on educational performance. It is that some inputs are important to explain the performance in math test, but they are not to the same extent, to explain the performance in reading test. Moreover, some variables may be important to get a basic level in the ECE, but are not significant to explain the probability of reaching the highest level. For example, for the math test the mother's education has a significant and positive effect on the probability of moving to the second level on the ECE. However, it is not significant, and even has a negative sign to account for the highest level.