Discrepancy between last year and last week self-employed CPS March Supplement

I’m using the CPS March Supplement via IPUMS to calculate the number of self-employed workers by year. Using the class of worker last year question (classwly), unincorported self-employed workers drop from ~10 million to ~8 million for 1996-1997 while incorporated self-employed jump from ~4 million to ~6 million (see figure below).

I’m defining incorporated self-employed as classwly==(14) and unincorporated as classly==(10,13). From what I’ve read, these definitions don’t change in 1996-97 and neither does the weight variable (wtsupp). Is there a coding mistake where some of the unincorporated are being assigned as incorporated? Thanks so much for your help!

I was able to replicate your issue with both IPUMS-CPS data and with the CPS files released by the Census Bureau. Generally, IPUMS-CPS tries to maintain as close of a relationship with the original data as possible, which means that errors in the source data are also likely to be reflected in the IPUMS-CPS data. Since the values for CLASSWLY come directly from the original CPS files, I recommend that you try contacting the Census Bureau for further information.

Sorry that I could not be of more direct assistance.

I am very late to the question, FFR, Levine 2017 QJE “Smart and Illicit: Who Becomes an Entrepreneur and Do They Earn More” report that the definition changed over time.