This is much more speculative, but perhaps within-country divergence and convergence explains more of the time-variation in national growth rates than we realise.
One of the very useful reminders of William Easterly’s Tyranny of Experts is that the nation-state is often not the appropriate unit at which to consider big questions of growth and development. For example, if we look at per capita gross world product, there is no slowdown to explain. In fact growth seems to accelerate in the final quarter of the 20th century. The growth-pessimists will respond that this is merely ‘catch-up’, countries far from the technology- and productivity- frontier moving closer to it, rather than a pushing-forward of the frontier itself.
Of course they’re correct, but if that reasoning applies to the late-20th and early-21st century world, why should it not apply within the mid-20th century United States?
In, say, 1930, it’s very likely that some (North-Eastern) states were quite advanced, while some were not; that some factories had adopted Ford’s assembly line techniques, but that many operated little-changed from a century earlier. We know for certain that some early industrial innovations took a long time to reach high levels of consumer adoption, as the following figure shows.
It took around 25 years for electricity to spread from 20% to 80% of US households, and 50 years for telephones to do that. But cellphones achieved the same in around a decade. Adoption of useful technologies appears faster today, and slower in the past.
Here we enter the realm of the unsubstantiated thought experiment. Suppose production technologies did the same thing. Suppose the essential production technologies of the mid-20th century were in place, somewhere, by 1930, but didn’t really achieve widespread adoption until the 1950s (say, as a result of the war economy, or the growth in corporate America). Then what we might see is a pre-war period in which income growth systematically underestimates technological development, followed by a post-war period where income growth systematically overestimates technological development. If you shave the 1928-1972 peak off Gordon’s chart, and relocate the ‘excess’ growth to 1906-1928, you get a much more even chart.
Did this really happen? I have no idea, but it’s a good reminder that aggregate GDP per capita statistics for any country may not meaningfully reflect technological progress.
As William Gibson famously noted, ‘The future is already here — it’s just not very evenly distributed.’