I am working with the basic monthly CPS after 2012, and noticed that when I limit my sample to those who are eligible for the ORG earnings questions using ELIGORG, there is still a small portion of observations in each month (~0.1-0.2%) who have a code of 9999.99 for EARNWEEK (not in universe). These observations otherwise look like they should have valid earnings information, i.e., they are from employed wage/salary workers in their fourth/eighth month in sample, etc.
I did some cursory poking around in the basic monthly CPS files off the Census website, and it seems (?) these cases happen in the rare instances when weekly earnings in those files are coded as zero.
If the zero code thing is correct: for my purposes, it does matter whether these observations truly have zero earnings, or if it’s actually a data error or something else.