I’m attempting to use BPL to get birth locations for citizens of large cities using the 2015 ACS figures. Am I correct that aggregating by CITY and BPL, then summing PERWT would get that number? Would there be any concerns other than dealing with CITYERR?
It sounds like your procedure will provide your desired estimates. You probably already know that not all cities are identifiable by CITY in the 2015 ACS sample, but that could be another concern for you to consider. Additionally, depending on the size of the city, your estimates may have fairly large standard errors. This can be ameliorated by using ACS multi-year files. Other than these issues and dealing with CITYERR to impose a more restrictive error limit, there shouldn’t be any additional concerns.