Im looking at Longitudinal for the CPS ASEC data, where each person has 2 surveys, 1 year apart. When looking at individuals who have a “second year” MIGRATE1 value of 4 or 5 (i.e., moved BETWEEN counties, between the 2 surveys) for some reason they have the same COUNTY value in both years (indicating that they still reside in the same county, despite being flagged as movers).
Does anybody happen to know the reason for this discrepancy?
I looked for these cases in the 2017-18 linked ASEC. Some of these are cases where the linkage was not correct (age, sex, and race do not match). Others are cases where the county in both years is “0” (unidentified), so a move between counties would not be properly captured by the COUNTY variable. There are a few hundred remaining anomalies, though. In most cases these have migration status allocated by the Census Bureau (indicated by the variable QMIGRAT1), but this is not always the case. I don’t have a good explanation for these. The IPUMS CPS team will look into these cases in more detail and update the documentation if they find a systematic explanation for the remaining cases, as we do with other minor anomalies in the original CPS microdata affecting a small number of cases.
Thanks for the quick reply Mathew!
I understand the explanation about missing county data for some people, and incorrect linkage for others. However, even when excluding entries with those problems, there is not even a single person in the 2018-18 linked ASEC data that has a different county value for the 2 consecutive years.
In fact, when reviewing the entire linked ASEC data for the decade of 2010s (i.e. 2010/11, 2011/12,… 2019/20), there is not even 1 person that has different (non-zero) county codes between their 2 consecutive years.
This signals to me that there might be a larger problem at play, besides the linkage or missing county data problems, and I would love to know what can be done to overcome this if possible.
This is expected, because the CPS samples housing units (buildings), not households (groups of people). That means that individuals are not followed over time if they move. The only way for an individual to appear in two consecutive years of the ASEC is if they resided in the same house. So the appearance of any values of MIGRATE1 other than “same house” in the second year is the real puzzle here.
Got it. I guess that I assumed that the sample followed households. In this case you are absolutely right, and if I had to guess, instances in which we observe a second year value of MIGRATE1 different from “same house”, it suggests that the original household that was sampled at that housing unit in the first period moved out, and was replaced by a household that migrated into that housing unit, and was thus sampled in the second period.
Thanks for the help Matthew!