After searching around a bit, I can’t find anything that details how (or if) the definition of manufacturing has changed over the years. However, the FR has all of its data available for download. You can see it at here. So I looked at the data available on durable goods, from 1976 to present, which you should be able to download from this link. In particular, with this data I found the year of peak output for each type of durable good:
Wood Product, 2005
Nonmetallic Mineral Product, 2006
Primary Metal, 1978
Fabricated Metal Product, 2008
Machinery, 2000
Computer and Electronic Product, 2008
Electrical Equipment, appliance and component, 2000
Motor vehicles and parts, 2004
Aerospace and Misc. Transport, 2007
Furnature and related product, 2006
Misc., 2007
With the exception of Primary Metal, all categories reached their peak in the last decade. Primary metal reached 84% of its highest output in 2006. Electrical equipment and Machinery also had more recent peaks: Electrical equipment reached 87% of its peak in 2008, while Machinery reached 99% of its earlier peak in 2007.
So it seems clear that the positive numbers for manufacturing output aren’t the result of a shift away from “real manufacturing” in the official definitions.
Finally, it’s not clear to me why we shouldn’t count anything that is“assembled here with parts from China.” Some of the manufacturing done in China is final assembly. We complain when such jobs get moved to China, so why should we claim they don’t count if they get moved back?