When I use CPS data I find too few cases to make an accurate determination of what percent of the workforce is represented by a union for a specific occupation in a specific geographic location (e.g., Philadelphia) in the construction industry. For example, for the entire US, I found only 68 cases of carpenters (Occ2010 #6230) in construction (Ind1990 #60), with 4 represented by a union (68/4=17% union). This is too few cases once it is narrowed to a specific geographic region. Is this methodology correct? Is this too specific of a question for the CPS data?
It looks to me that your methodology is correct. The issue is that this seems like a difficult question to answer with CPS data. Several details make this difficult. First, the CPS has a relatively small sample size which is optimized more for national and state-level analysis. Second, the UNION variable is part of the outgoing rotation group questionnaire, which only asks questions to households who are in their 4th or 8th month in the CPS sample. So, for any analysis of union status in the CPS, the sample is automatically cut down to roughly 1/4th of the total CPS sample size. Finally, when you limit your calculation further by occupation and industry, you are left with even fewer observations. These details limit any sort of analysis like you suggest. I’m sorry that this is the case.
Thank you very much!