Note that here, Paul Craig Roberts is not discussing the Bush Administration's claims of growth in the past, but rather the assumptions it is making about growth in the future, placing him in the unenviable rhetorical position of deriding an assumption because it is, well, an assumption.
That being said, he does subscribe to the view that GDP and output are overstated (just not here) for the reasons BusinessWeek suggests. He just can't explain himself when someone asks him how much they are overstated . . . which is important, because some economists who have examined his assumptions find no significant impact on the question. In other words, other economists' replies to Roberts' warning about the methodology-problem are basically, "so what? Even if we use your numbers your conclusion is not supported."
This isn't a slam on the BLS or even the GDP and those who formulate the numbers - it's just that there is no real way to measure productivity in the USA anymore since the advent of the flat world economy. They do the best they can - but I am glad there are those who take the time to point out real problems with the official government numbers.
There is a very large problem with our economy at this time - I work for a very large American company who is slowly (doesn't make the papers) ridding themselves of IT jobs (40k - 130k area, probably 1k people let go eventually) but what they have done is to open an 'arm' in India which will house any R&D and IT that can be done there. So in affect, the company will look like it is hiring more employees, which they are, but those employees are Indians, not Americans. They employ more and more H1-B guys too. This is a huge retail supermarket/wholesaler chain that only sells in America and will not be opening stores outside of here. I don't know how stuff like that is measured.