I am combining estimates generated in IPUMS to other 5-year 2019 estimates downloaded from the Census. For MSAs with delineations that have changed (ie, Dayton OH (formerly 19380) and Prescott AZ (formerly 39140)) is there suggested guidance as to whether to rename these areas with their new IDs, or to remove? I understand that this would be based on a judgment on how different the boundaries are, and the researcher’s goals, but in general does IPUMS have guidance? I haven’t tried to map the different boundaries to understand how much they differ, and would rather start here by asking! Interested in general in how people handle this when combining IPUMS generated MSA-level estimates with recent Census estimates (ie, downloaded from tidycensus in R). Thanks.
“Does IPUMS have guidance?” We don’t have any resources designed for this specific issue, and I don’t know of any standard practices. Fortunately, you seem to be on the right track already in thinking through the considerations involved!
The optimal strategy would be one of these:
- Create your own crosswalk from 2010 PUMAs to 2018 MSAs (the MSAs used in 2019 5-year summary data), and then use PUMA info from the microdata to determine 2018 MSA codes.
- Get 2019 5-year summary data for counties, rather than for MSAs, and then group the county-level data according to 2013 MSAs (the MSAs used for the MET2013 variable).
The key files you’d need to take either approach are available through the IPUMS USA 2010 PUMA resource page and the Census Bureau’s CBSA delineation files page. That said, I’m not familiar with how many changes there have been between 2013 and 2018 MSA definitions, so I’m not sure if it’s worth taking these measures!
FWIW, we plan to create a new variable using newer metro area definitions after new PUMA definitions are adopted with the 2022 ACS samples next year, but we currently have no plans to identify metro area vintages from between 2013 & ~2023.