Brazilian geographical variables


I am working on the 1980-2010 Brazilian censuses and I have two questions about their geographical variables.

First, I have read that geographical variables have been revised for many countries since the recent data release in October 2015. Does this also concern Brazil (and in particular the variable MUNIBR)?

Second, I want to be sure I have understood properly the difference between the old geographic harmonization implemented by IPUMS (by name) and the new one (using GIS files). Let’s take the example of a municipality. The only way the two harmonizations can differ is when the boundaries of a municipality have changed without involving a merger with some other ones or a division into other ones (ie. without involving the creation or the disappearance of some administrative municipalities). Am I correct?

Thank you for your help,


MUNIBR has not changed since the last data release. A link to the full listing of merged units for the categories in MUNIBR is buried in the comparability description:…

If you are looking for the old MUNIBR2, you may want to consult GEO2_BRX. However, GEO2_BRX will be updated in the coming year.

Finally, your understanding of the two versions of harmonization is correct with one small exception. Yes, when the names of units and their corresponding polygons remain consistent across sample years, the two versions of a variable for those categories (geog units) should be the same. In theory, changes in boundaries (splits, merges, shifts) should be the only circumstance that can lead to different outcomes in the two types of variables. However, changes in names (even without changes in spatial alignment) could also lead to differences across the old and new versions of variables. Newly revised variables have been created after extensive research matching census codes to spatial units. This means that new variables more accurately identify units with identical shapes (even if the names have changed). Such units may not have been reconciled in the old version of the geographic variables. In rare cases, the reverse is also true. A geographic unit name appears the same, but actually refers to a completely different spatial location across sample years. New variables also correct any such mismatches. I hope that makes sense.