One of my continuing questions about research in economic growth is why it insists on remaining so focused on manufacturing to the exclusion of the other 70-95% of economic activity in most economies.
He mentions work by Chad Syverson and others. It's true--much of the productivity literature focuses on manufacturing.
Why do we persist in focusing on this particular subset of industries, sectors, and firms? I think one of the main reasons is that our data collection is skewed towards manufacturing, and we end up with a “lamppost” problem.
That is almost certainly the main reason. I guarantee that the people using the microdata would love to be able to carefully study productivity outside of manufacturing. Vollrath describes how industry code schemes are heavily tilted toward detail in manufacturing. Read his post; it's very instructive. (And hey--he's not the only one who blogs industry codes!). I think the switch from SIC to NAICS was a huge improvement on this, though the problem hasn't disappeared.
It's not just about industry codes. The big microdata sources are typically limited to employment information for businesses outside of manufacturing, i.e., there's no capital. And the lack of coverage has persistent consequences. Recent improvements to the data often aren't easy to roll back to earlier years, which means it's hard to study the time series.
The other issue, though, is that even our primary productivity concept--TFP--is really designed for a manufacturing world. The Census Bureau actually has survey microdata for firms in retail and services. But how do we measure capital in those industries? What other inputs are important? This is a much bigger problem than simple lack of coverage. (Obviously, there is a literature).
The further you get from producing widgets with machines, the harder it is to map the TFP concept to the real world. And that's just at the firm (or establishment) level! Facing this challenge in the microdata colors one's views of the TFP concept at the macro level. There is no clean mapping from micro to macro; go down this rabbit hole if you won't take my word for it. Aggregate TFP isn't actually a thing, even if it's still a useful fiction.