There is a growing body of research on the relationship between education and demographic variables. To the extent that education is an important variable in demographic analysis, it is essential to measure it properly. In Brazil, the literature has neglected the possible effects of education misstatement in demographic, social and economic studies. Although several studies have looked at coverage errors, both vital and census data, little attention has been paid to content errors. In addition, most of the analysis on content errors has been limited to age misreporting, which suggests, based on the high degree of inaccuracy that has been found, that problems of misreporting may affect other variables, including education. Therefore, our goal is to offer evidence of education misstatement by evaluating education reporting among adults in Brazil with data from the 1991 and 2000 censuses. In what follows, we first examine the quality of education data, directly, by looking at three different indicators: the completeness of education data, education misstatement, and educational levels by sex. Next, to provide further indirect evidence of education misstatement both in 1991 and 2000, we calculate the implicit mortality measures by education from intercensal survivorship ratios. Our results gave clear evidences that education census data is not accurate in Brazil.