In the magisterial The Rise and Fall of American Growth, Northwestern University economist Robert Gordon describes a “special century” of fast productivity growth from roughly 1870 to 1970. But the apogee of that period was really the 1920–1970 “golden age” period when the economy really felt the impact of the Second Industrial Revolution of the second half of the 19th and early 20th century.

The second IR was a time that produced important and “unrepeatable” inventions flowing from the following five technology clusters: electrification, the internal combustion engine, chemicals, modern communication, and urban sanitation infrastructure. Compared to these “great inventions,” in Gordon’s view, the impact of information technology pales. And since you can only electrify once or widely install indoor plumbing once, another golden age of productivity ain’t happening.

Now plenty of technologists disagree. They think Gordon underestimates the potential impact of new technologies such as artificial intelligence and advanced robotics. Gordon, for instance, doesn’t seem to think autonomous cars will be that big a deal from a productivity standpoint. And VR is no internal combustion engine.

But what if Gordon is overestimating the impact of his great inventions? That is one of the conclusions of the new paper The Sources of Growth in a Technologically Progressive Economy: the United States, 1899-1941 by Gerben Bakker (London School of Economics),  Nicholas Crafts (University of Warwick), and Pieter Woltjer (University of Groningen). (Thanks to Adam Thierer for the pointer.) The researchers found that although “the great-invention sectors were very important,” they only accounted for about 38% of productivity growth from 1899 through 1941. (By comparison, IT-producing sectors were responsible for 54% of growth from 1974 through 2012, though total growth was less .)

This suggests some increased confidence that the age of fast growth isn’t over is warranted. The sources of innovation back were perhaps more widespread than Gordon claims: ” …big unglamorous sectors like Construction, Farming, and Foods sometimes had a substantial impact on growth.” So perhaps a general purpose technology like AI could vastly boost productivity across many areas of the US economy.

(It is also worth noting that the authors “do not agree” with the well-known finding of Alexander Field that the 1930s were the most technologically progressive decade of the century.)

The big question, then, from Bakker, Crafts, and Woltjer: “If the great-inventions were not the sole drivers behind the strong productivity growth in the US during the pre-war era, what then could explain the rapid rate of technological progress?”

And their answer: “[Productivity] growth was not driven indirectly by spillovers from great inventions such as electricity. Instead, the creative destruction-friendly American innovation system was the main productivity driver.” It was the open and competitive nature of the US economy, especially compared to that of Europe, which led to US manufacturing better using new technology.

From the paper:

Finally, it is important to note that competition in product markets was relatively robust in the United States, a large free trade area, compared with its European rivals. A notable contrast is with the United Kingdom, which in the 1930s imposed tariffs and encouraged cartels such that competition was undermined in many sectors and productivity performance was significantly impaired (Crafts, 2012). Although Franklin Roosevelt’s ‘New Deal’, enacted in 1933, which limited competition and vastly increased labor’s bargaining power, threatened to undermine the process of creative destruction, the detrimental effects of the New Deal policies were limited. By 1935 the Supreme Court had ruled that the act’s suspensions of antitrust laws were unconstitutional and by the end of the 1930s Roosevelt switched tack and pursued a pro-competition policy instead (Cole and Ohanian, 2004: 785-8).