How to apply weights in subnational-level analysis

I would like to use total family income in IPUMS USA to calculate the percentiles of income distributions in US cities/MSAs. My question is which weight should I apply. (1) There is a household weight and a person weight, but I’m not sure which applies to the families. (2) From reading the documentation, it seems that the weights are used to get a nationally representative sample. I’m not sure if they can be used to estimate city/MSA-level measures.

Thank you in advance of your help!


Since there is no specific weight generated for families, The Census Bureau uses Housing Unit Weights to generate family-level estimates. See this answer for more information.

More descriptive discussions of how weights are constructed can be found in the ACS Accuracy of the PUMS documents. You can also find state-level estimates for verification in the PUMS documentation. For MSAs and Cities, you can estimates against ACS summary tables produced by the Census Bureau (available through American Fact Finder). While it is not likely that estimates generated using the PUMS will perfectly match the Census generated summary tables, as the summary tables are created using a larger sample and specific population controls that can not always be recreated with the PUMS, you should be able to produce similar figures.

I hope this helps.