Why are there so many NIU responses for EMPSAME?

I am trying to use EMPSAME to measure turnover for a particular workforce using a longitudinal match. I used EMPSTAT to ensure respondents were employed in the occupation that I am focused on. I then matched those records to the next month in the sample and tabulated EMPSAME, only to find a large number of responses that were NIU. I tried excluding non-employed respondents in the next month in sample, which had little effect on the NIU group. My approach to matching ensures that MIS 1 and 5 are excluded and that all of the respondents were working in the previous month. Why is this happening?

I was able to replicate what you are observing and it does not appear that you are missing anything from the universe statement. After some digging through documentation on our end, it sounds like a good portion of what is happening may be attributed to individuals who are in the universe (currently employed and employed the previous month) but had a response of don’t know/refused/blank for the name of their employer. In this case, it appears that these individuals may not even be asked the question that drives EMPSAME.

Unfortunately, we don’t have name of employer information so we can’t screen for this. As such, we’ll look into making a modification to the documentation for EMPSAME.

The number of observations month to month that fit what you have described does appear to be fairly consistent, so it may still be a good measure of change over time but is clearly problematic at the individual level.