Cross-checking values for ACS 2015 IPUMS against FactFinder for California and Arizona I found large discrepancies in the OCC2010 values. I used both downloaded OCC2010 and the SDA tool to make sure it was not operator error. These exceed what I would expect because of the relatively smaller sample size of IPUMS vs. the full ACS.
Example (California)
architecture and engineering – FactFinder: 379,146; IPUMS: 434,540
Legal – FactFinder: 211,828; IPUMS: 250,572
FactFinder lists a total of 18,045,450 in the civilian labor force. If you subtract military and “no occupation” from the IPUMS numbers, you get 22,149,745.
I found a similar problem with Arizona – an ACS civilian labor force of 2,937,510 and an IPUMS-derived civilian labor force of 3,730,522.
What puzzles me is that in checking other IPUMS values, such as nativity and citizenship, they have consistently matched or approached those that I have found in ACS with allowance for the smaller sample size. Is OCC2010 the wrong measure, or should I be using a different guidepost to check my results?
First, OCC2010 is an IPUMS created coding scheme that is harmonized across time to make time series analysis of occupations easier. Basically, it fixed the codes in the 2010 scheme and adjusts the values of OCC accordingly. So, to maximize comparability with FactFinder I recommend using the OCC variable which uses the same occupational coding structure used to create the FactFinder tables for a given year.
Second, the table that I think you are looking at on FactFinder reports the total number of employed civilians who are 16 years or older. This is slightly different than the universe satatements for OCC2010 and OCC, which also include people who had worked within the previous five years. For comparable labor force statistics the EMPSTAT variable can be compared with other Employment Status tables on FactFinder.
I hope this helps.
Thanks very much. I had just discovered that when I filtered my results using EMPSTAT == 1 (reducing the field to those who reported working last week), it brought the results very close to those listed in FactFinder. While the OCC2010 numbering scheme is very convenient, I’m sure I can find a way to summarize the values of OCC. Again, thank you.