IPUMS international Tanzania 2012: what weights to use to recover population statistics at the district level?

I am using the IPUMS Tanzania 2012 census to compute statistics at the district level (geo2 level). However, after computing the statistics using PERWT, I noticed that the population estimates don’t line up with those provided by the Tanzania Bureau of Statistics. IPUMS documentation indicates that the Tanzanian 2012 sample uses complex stratification and differential weighting, but I can’t figure out how to combine the individual weight (PERWT), the household weight (HHWT) and the strata (STRATA) to compute the right weight. Is there a formula or any other way to compute the correct weight?

Since population estimates calculated from IPUMS International data are subject to sampling error, whenever possible, we recommend that researchers who need accurate summary population totals should rely on official published tables from the country of origin.

However, we have extensive documentation about sampling error and variance estimation here.

Are you using Stata for your analysis? If so, then this documentation on SVY may be helpful. In particular, with the svyset command, you can specify a weight (i.e. [pweight=perwt]) and the strata (i.e. strata(strata)).