Hello! I am doing research using David Dorn's commuting zone crosswalk files and 1970-2000 5% state samples and a 2011-2015 ACS accessed from IPUMS. The files take PUMAs (or county groups in 1970 and 1980) and apportion them into commuting zones using an m:m merge (joinby in Stata). Then these individuals are re-weighted using the perwt from IPUMS and the "afactor" from Dorn multiplied together. In a sense, individuals are being apportioned using probabilistic matching. However, when I try to us the new weight (perwt x afactor) I get population counts for the 2011-2015 ACS that are much smaller than the actual counts that I should be getting. This process produced accurate totals for all of the decennial years, the only problem is the 5-year pooled sample. Please advise if I am missing something. Thank you!
Dorn's website with crosswalks: https://www.ddorn.net/data.htm
IPUMS weight FAQs: https://usa.ipums.org/usa-action/variables/PERWT#description_section
Census documentation: https://www.census.gov/content/dam/Census/library/publications/2010/acs/Chapter_11_RevisedDec2010.pdf