I finished reading Robert Gordon’s The Rise and Fall of American Growth a while ago, but finally found the time to write down an extended review. I’m way behind the curve. Tyler Cowen has an excellent review, and Ryan Decker sort of live-tweeted his way through the book recently.

To begin, a comparison. A few months ago I praised Garrett Jones for the sleek nature of Hive Mind. The Rise and Fall of American Growth is, um .. not sleek. It starts out with a 531-page description of the diffusion of major innovations within the U.S. over the last 140 years, then transitions to a 104-page presentation of Gordon’s thesis about what generated differences in productivity growth in that period, and finishes with a 12-page postscript.

Depending on what you’re after, this is either a bug or a feature. As a statement of Gordon’s thesis regarding productivity growth, it’s a bug. You can read Gordon’s thesis in two (free) working papers (here and here). The 531 pages of anecdotes about technology diffusion are ancillary to making this case.

On the other hand, the 531 pages are a feature if you want to understand just how amazing the transition in living standards has been over the last 140 years. For that purpose, the unending onslaught of facts, stories, and comparisons is necessary to make the case. Gordon takes the reader through the transition from the disconnected, unhealthy, underfed, dark lifestyle facing workers in 1880 to a world of lights, abundant food, communication, travel, and health. And that’s just by 1940. The core networks of modern life - electricity, sewers, roads, broadcasting - were in place by the end of the Great Depression. Despite these massive changes in welfare, what Gordon documents is that measured productivity growth up through the 1930’s was actually quite low. Growth in GDP per capita systematically understates growth in welfare; this is not something that has arisen recently.

In the 1950’s, and spilling into the 60’s and early 1970’s, productivity growth jumped appreciably. For Gordon, this is partly a feature of those core networks being extended out into the whole country. Diffusion, more than innovation, is what fundamentally explains the spike in measured productivity growth in the immediate post-war period. Since the 1970’s, productivity growth has retreated back to pre-WWII levels.

One thing I don’t understand about the book is the insistence that the welfare gains of growth after the mid-century are necessarily smaller. Maybe they are, maybe they aren’t. Fundamentally it is a book about economic growth, not welfare growth. Yet all I kept hearing in my head while reading the second half of the 531-page narrative was “you darn kids and your Facebook”. This may just be a problem of tone, but it also pushes Gordon close to a real problem. He veers close to using the assumption that the welfare gains of current innovation are small to conclude that the rate of economic growth in the future must be small as well. That does not follow. It is quite possible that productivity growth could spike to 3.5% per year again in the 2020’s, and yet the welfare gains of this could be trivial.

But will growth spike to 3.5% again? Gordon says no. For him, the innovations in the early 20th century were unique. In networking everyone together, they had a one-time boost to productivity that the computing/internet revolution of the last twenty years cannot match, and more importantly will not match. The productivity growth facts are what they are, productivity has grown much slower 1995-2015 than 1950-1970. But that doesn’t mean productivity growth will stay low. We just don’t know. Tyler Cowen’s review does a great job of laying out the fundamental problem; future technology is unforecastable. And while Gordon makes a passing effort at claiming it is, his anecdotal example of a Ladies Home Journal from 1900 predicting air conditioning does not constitute evidence. We just don’t know what new innovations will bring, or what effect they will have on economic growth.

Meanwhile, there is a very powerful idea running along underneath Gordon’s book, and I don’t know that he draws enough attention to it. Gordon’s narrative shares a feature of Piketty’s Capital; it questions aspects of the Kaldor Facts that have formed the basis for most work on economic growth. Those facts were based on observations made in the 1950’s-70’s. Piketty showed that the idea of a constant labor share is not a universal feature of growth, but an anomaly of immediate post-war period. Gordon shows that rapid, sustained productivity growth was also an anomaly of the immediate post-war period (see his Figure 16-5). Together, these two authors tell us that the stylized facts that generations of growth models have tried to match are probably wrong. So whatever the other merits of Gordon’s book, it is valuable for making this point vividly.

We’ve built generations of growth models on the idea that they have to deliver “balanced growth paths” because those match up to the Kaldor Facts. But if those facts do not hold, then this opens up a much wider set of possible growth models. Growth could run out of steam, growth could accelerate, growth could follow a random walk, etc. etc.. In that sense, Gordon’s book is worth reading, in that it forces you to reconsider - or at least defend - the assumptions underlying our explanations for economic growth. You could probably have gotten there in under 647 pages, but nevertheless it is a good point.