I’ve been researching residential mobility using IPUMS-CPS ASEC (and IPUMS-USA, et al.) and I’ve noticed that there are gaps in availability for MIGRATE1 (available 1963-1971, 1976, 1981-1984, 1986-2019), but in one of the years in which it is available, 1995, it seems very wonky.

Here’s a table of year and migrate1 from 1993-1998 using asecwt as the probability weight in Stata:

(The same patterns appear with unweighted counts)

Can anyone provide any insight as to why 1995 has substantially higher counts for NIU and all the mobility types, but substantially lower counts for ‘same house’? It throws the percentages way off, e.g. the mobility within county goes from 10% to 23%.

You’ve pointed out what seems to be a real error in the data. Unfortunately this is present in the original public use files from the Census Bureau and not documented as far as we can tell. There is not much we can do about it at this point, though we’ll make a notice of this in the variable description. For what it’s worth, I noticed that the number staying in the “same house” is exactly the same for “one year ago” (MIGRATE1) and “five years ago” (MIGRATE5) which is awfully suspicious, but not particularly helpful in finding the right number.

Interesting, and a bummer. I did see two other posts on how MIGSTA5 and MIGSTA1 are problematic in 1995, with one about them having the same case counts, and a post about missing values in MIGSTA1 in 1995. Seems like a bad data year for migration. I appreciate the addition of a note on the variable!