I’ve seen two glitches in linking census microsample data to the spatial files. The glitches show up at the second administrative level.
(1) In seeking to link level-2 administrative identifiers in the Benin samples to the corresponding year-specific (not spatially harmonized) shapefiles, I’m finding that for each year “yyyy”, the GEO2_BJyyyy identifier in the census data does not match to IPUMSyyyy in the shapefile (after GEO2_BJyyyy has been left-padded with zeroes). Instead, GEO2_BJyyyy matches directly (without any padding) to the shapefile’s commune identifier COMNyyyy.
(2) For level-2 administrative identifiers in the Bangladesh 2011 sample, there are discrepancies between the census GEO2_BD2011 codes in Rangpur Division (55) and the IPUM2011 code in the 2011 (not spatially harmonized) level-2 shapefile. The level-2 codes in other divisions look just fine: with the usual left-padding, GEO2_BD2011 matches IPUM2011 in the level-2 shapefile perfectly everywhere other than Rangpur.