I notice a large number of ‘missing’ or ‘N/A’ values for education (HIGRADE) in the 1950 census. Is this normal?
Yes, this is normal. The 1950 census designated one “sample-line” person within each household who answered additional questions. The Universe Tab of HIGRADE indicates that this question is only asked of the “sample-line” person in the 1950 sample. So, this is the reason for the higher number of “N/A” observations. Additionally, for this same reason, we have a sample-line specific person-level weight variable (SLWT) for using these variables.
I just want to note that Abacus does not use the correct weights for these variables, creating incorrect estimates. You would Abacus would use slwt for these questions in 1950.