Accuracy of aggregate industry and occupation employment counts in 2011-2015 ACS sample

I’m trying to use the the 2011-2015 ACS 5-year sample to get an estimate of employment and occupation counts by MSA-industry.

To do this I’m using the perwt variable, a crosswalk from pumas to counties to MSAs, and a crosswalk from indnaics to naics codes (which are then further aggregated). I’m counting currently employed individuals; i.e., the ACS variable empstat says employed.

I’ve had no trouble getting people matched to MSAs, and the resulting city population counts look like they match other sources. However, total employment by industry looks very different from what I’ve seen in other sources, such as the BEA.

Should this be expected with the ACS data, or have I made a mistake? The section “Weighting and Estimation” in this document:… made me think that weights are used to target population by age, sex, race, and Hispanic origin status at the county level. So perhaps using perwt to obtain employment by industry counts shouldn’t be expected to give good results.

As a related question, I’ll also need some OCCSOC-based occupation counts by MSA-industry. I’m wondering if using perwt is appropriate for this.