From the working paper:
> There are several plausible channels for labor measurement difficulties in construction, including a higher than average frequency of employees working irregular hours, contractor labor that may be misclassified by survey respondents, and, especially in more recent decades, labor supplied by undocumented workers.
It then goes on to talk about that in more detail, but I won't copy that in because every time I do, someone goes on with "but what about X?" where X is also in the very next paragraph.
Nothing wrong with coming up with the hypothesis, honestly. But the extraordinary number of responses to your comment that speculate on something that one can immediately find are just so typical of HN comments.
I sometimes see that in software engineering where some engineer will tell me about some library we're using that we have access to source code for: "yeah, I looked at the docs and it's not clear if it does A or B". But...the source code is right there. Why not just use that? At least then if the library is doing it wrong you'll know for sure.