There is a new paper (pdf) on this question by Bakker, Krafts, and Woltjer, here is the abstract:
We present new estimates of TFP growth at the sectoral level and an amount of sectoral contributions to overall productivity growth. We improve on Kendrick (1961) in several ways including expanding the coverage of sectors, extending estimates to 1941, and better accounting for labor quality. The results have important implications including that the pattern of productivity growth was generally ‘yeasty’ rather than ‘mushroomy’, that the 1930s did not experience the fastest TFP growth of the 20th century, and that the role of electricity as a general purpose technology does not explain the ‘yeastiness’ of manufacturing in the 1920s.
They instead suggest that TFP growth is rising throughout the 1920s through the 1960s, a view which I cannot quite agree with. I view the 1960s as a time when previous ideas spread widely, rather than the key years of invention.
My strictly intuitive, historical guess is that TFP growth peaked in the 1890-1930 period, give or take.
More generally, the TFP concept is most useful, and most exact, when TFP growth is low rather than high. The bigger TFP growth might be, the more you have to worry about unmeasured changes in labor quality, and also worry about what is “technical progress embodied in capital goods” as opposed to “sheer accumulation.” When there is less progress, these measurement issues are smaller too.
I am most skeptical of TFP estimates for China, even if you believe the underlying statistics. Compared to “the global frontier,” TFP growth for China has been pretty close to zero, for centuries. Compared to “the frontier within China”…er…Chinese TFP growth and “embodied accumulation of foreign ideas through savings and investment” become pretty much the same thing. The distinction theory was trying to create then has been abolished.
So I’m not convinced by the results of this paper, but they are a useful corrective to excess certainty about Alexander Field and the previous view that the peak of American TFP was the 1930s.
In any case, thumbs up to any paper which uses the word “yeasty.”