Inconsistencies between yrschl and educbr in the Brazilian Censuses (and illogical answers for educbr).

Hello, yrschl indicates a person’s educational attainment in terms of years of schoolings completed, while educbr indicates a person’s educational attainment in terms of level attained (primary, secondary,…) and years of schooling completed within that level. Yet, I find some inconsistencies between these two variables.

For instance, in 1980 : there are people for whom yrschl indicates « some primary » (91), « some secondary »(93) or « some tertiary »(94) while educbr indicates « None, or never attended school » (0).

Another example, in 1991 : there are people for whom yrschl indicates 4 to 11 years of schooling while educbr indicates «Primary, Grade 3 » (2130) which corresponds to 3 years of schooling, be it under the old or new Brazilian educational systems.

My question then is how to fix these inconsistencies. Right now, I have two possible solutions in mind : a) dropping the inconsistent observations ; b) using only educbr because it is supposed to respect respondents’ answers ; while yrschl was inferred from educbr as well other variables and thus appears to me more prone to errors.

A last question : How do you advise treating “illogical” responses for edubr (eg. persons indicating that they completed ninth grade primary though primary stops at eighth grade )? I don’t know whether to treat them as having attained primary or secondary.

Thank you for your help.

For the 1980 and 1991 samples, there are two sets of questions used to create both EDUCBR and YRSCHL: the current grade/level attended for those still in school and the grade/level completed for those not in school. The inconsistencies you identified between EDUCBR and YRSCHL were due to the way we were handling persons that reported information for both sets of questions. We will fix these discrepancies by consistently using the information provided on current grade/level attended for cases where respondents also provided information for grade/level completed.

Thank you for bringing this issue to our attention! The consistently recoded variables should be available on the IPUMS-I website in the near future.