I’ve just begun working with an extract of ACS 2013-2017 IPUMS. As a test I summed PERWT / 100 (to account for the implied decimal), grouping by state. The values for CA, TX and FL were each in the hundreds of thousands, obviously quite low. I then tried for a national total and got about 6.5 million, again obviously quite low. The import instructions in both txt and xml called for importing at the same position in the file. The record count is about 18.9 million. Any ideas?
I resolved this by re-importing the data using the R readr library, importing the data as a fixed-width format, specifying all column widths and names. I suspect there’s a problem in the IPUMS R file that may have led to my issue. In any case, when I hand-built the import file, the problem vanished.
Hi Ron, glad you were able to find a workaround.
For future reference, you don’t need to divide the weights by 100. The ipumsr program will do this for your automatically. I think this was your problem. I just downloaded an extract using the 2013-17 5-year ACS sample, and PERWT summed to the correct state and national totals.
Thanks. I hadn’t realized that ipumsr could do that. I relied on the old “divide by 100” rule that I’ve used in the past. This is a nice time saver.