Abstract
A method is proposed to calculate the “margin of error” (confidence interval) of a Brazilian Education indicator (Ideb) based on the laws of uncertainty propagation. The method was applied to 29,313 schools and 2,381,722 students in the final years of elementary school in 2015. This year was also compared with the historical series, using the confidence interval. The results depend on the level: at the state level, uncertainty is small and of little relevance; at the school level, uncertainty is high, over 80% of schools did not change significantly from one Ideb edition to another. The results suggest that Ideb has statistical reliability in the Brazilian long-term goals (for 2022) of schools and states, and also for monitoring states even in the short term (2 years). However, Ideb is not reliable enough to monitor schools even in the medium term (6 years). In this sense, it is important that the Ideb per school is accompanied by a confidence interval.
School Quality; Statistical Uncertainty; Educational Indicators; Public Policies
Os círculos representam dados públicos provenientes do Inep, enquanto os retângulos representam valores calculados neste trabalho.Fonte: Elaboração própria (2019)
As cores representam o Ideb da escola (preto) e seu intervalo de confiança (cinza). Intervalo de confiança calculado a partir do erro total (Eptotal).Fonte: Elaboração própria (2019) a partir dos dados do Inep (anos finais do Ensino Fundamental de 2015)
Intervalo de confiança calculado a partir do erro total (Eptotal).Fonte: Elaboração própria (2019) a partir dos dados do Inep (anos finais do Ensino Fundamental de 2015)