Hi IPUMS,
Just a quick clarification – it appears that the union variable in the CPS for available post 1990 years takes on only three values (NIU=0, 1, 3) – doesn’t contain the fourth possible value (2), which denotes those workers who are in a union. Is there an easy way for you to double check I am interpreting this correctly?
Thank you!
I was unable to replicate the problem you encountered with UNION. The value 2 should be present, as this code identifies all respondents who are a member of a labor union. You can see how the frequencies look in each sample on UNION’s codes page.
Since UNION is from the Outgoing Rotation Group (or Earner Study), this question is only asked of civilians age 15 and older in rotation groups 4 or 8 who are not self-employed. These requirements reduce the sample size considerably (in some year-state combinations only 100 people are even asked the UNION question), and depending on your restrictions you may be restricting the variable too much to show any results for UNION=2.
I hope this helps. Please let me know if you are still having trouble with the frequencies for UNION. 
Thanks, will take a look again – maybe something weird happened somewhere in the data cleaning since I didn’t make any odd restrictions on the sample.