The Rise and Fall of American Growth: The U.S. Standard of Living Since the Civil War

Robert J. Gordon

Book cover

This is an excellent book, and I fell into the trap of waiting a long time to write my GR review because I wanted to write a really good one, and now the book has receded in my memory a bit (and I didn’t take good notes). However! It was a good enough book that I still have a pretty strong picture of the arguments in it.

Although Gordon doesn’t phrase it as such, the main thing I think he is fighting against in this book is the assumption that the economy is an ergodic system. Roughly speaking, this assumption means that the system has the same probabilistic behavior over time, and that a sample in time contains dynamics that are representative of the whole. This is such a common assumption in economics that it is almost never even mentioned as such. Any time an economist estimates a model on some time series of data and uses that to generalize about the future, she is making this ergodic assumption. Because of the nature of economic data collection, we generally have data available only for the post-Depression or post-war period.

Gordon’s argument is essentially that this is a dangerous and wrong assumption to make. He discourages us from assuming that economic growth is a continuous or time-homogeneous process, where we “can expect” some certain level of average growth per year. He makes this argument by taking a very detailed look at the history of concrete inventions from 1870 to the present. He uses a lot of detailed and creative data sources to do this (the Sears catalog plays an important role!).

A good summary of his arguments is available here:

I won’t go over them in detail because you can read them there, but very briefly: For most of recorded history economic growth was approximately zero. From the late eighteenth to early nineteenth century the “first industrial revolution” occurred, primarily characterized by the development of steam power. This was important for industry and started some major trends in motion, but didn’t have a huge impact on most people’s everyday lives. From 1870 to 1920, the “second industrial revolution” occurred, and this is the one that had the biggest impact on people’s lives: the internal combustion engine, electrification, sewers and public health innovations. The impact of this second IR actually continued through about 1970 as inventions were refined and spread to more or less everyone in developed countries, with jet air travel and air conditioning being about the last echoes of this IR to make a big difference in people’s lives. He recognizes that what some might call a “third IR” of computing and electronic networking occurred in the post-1970 era, but argues that it is not nearly as meaningful as the prior two.

His thesis is essentially that we tend to implicitly assume that the level of growth seen in the 1920-1970 era represents “normal behavior” for an industrial economy and therefore a level that we should generally expect and strive for, but that in fact, it was essentially a “one time dividend” based on the working out of some huge quality of life improvements that could only happen once. This is a bold argument but I think he makes a strong case for it. Although I found his regular use of the phrase “could only happen once” a little annoying, he does a good job of portraying what a complete revolution in ordinary living conditions was effected by these inventions. Moving from only being able to get light and heat by burning stuff (that you had to haul to your house) to getting electricity at the flip of a switch; similarly moving from (women) carrying water back and forth all day to getting it directly from the faucet (and being able to flush the toilet). Moving from horse power to the internal combustion engine. Essentially overcoming infant mortality. He makes a strong case that living in 1870 would seem unimaginably difficult to most of us, but that living in 1940 would seem more or less familiar.

Looking at the economic statistics, he shows that total factor productivity growth was unusually high in the 1920-1970 era, and has subsided since. There was a mild boost beginning in the 1990s that he argues represents the impact of the IT revolution, but this boost has already subsided and Gordon argues that the IT revolution has already run its course in terms of providing economic benefits to average people. He makes a very strong case against the argument that economic statistics do not sufficiently recognize benefits from modern technology–or rather affirms it, but argues that the omissions were far greater regarding technology from the second IR.

One key observation of Gordon’s is that the impacts of technological development are generally forecast-able about 20-25 years out. This is because there tends to be a lag of about this length between discovery or proof of concept and the integration of the invention into everyday life or business practices. Based on the current state of technology, he argues that we should expect very low or stagnant growth over the next couple of decades (and he doesn’t wish to conjecture further out than that). Reading over his discussion of medium-term technology trends, it felt hard to disagree with him–hard to think of anything on the horizon that could make as big of a difference in people’s lives as the inventions of the second IR. (Of course, this is focused on people living in developed countries–there is quite a bit of improvement that could occur for people in developing countries, but that is a matter of distribution rather than of technology, and is in any case irrelevant for US growth.)

He talks specifically about several things that are currently hyped. I don’t have the book in front of me but I think the Powerpoint I linked to mentions most of them. Self-driving cars are probably the most hyped one right now. He argues that they don’t make a huge difference for consumers because you still have to be physically in the car; it’s just a matter of being able to direct your attention elsewhere–which is nice, but not huge. He says that the more significant potential impact is in trucking, but that much of truckers’ work is unloading, stocking, etc., things that are not likely to be replaced in the next couple of decades. He mentions 3-D printing and AI, but sees both as evolutionary or zero-sum developments. In terms of medical developments, it seems obvious that even something huge like curing cancer would not have nearly the impact of essentially curing infant mortality; he also points out that even in developed countries, much of health and mortality is distributional rather than technological, with the wealthy enjoying significantly longer life spans.

I tend to think that Gordon slightly underrates the importance of self-driving cars by not considering knock-on effects, but overall I basically agree with him. I do think they are likely to make it more feasible for most Americans to live without their own car at all. A car is a big expense for most families, although it’s not clear to me how much of the savings from that is likely to be passed on to consumers. I also think it’s likely to reshape our patterns of living, and make it relatively more attractive to live outside of a major metro area, because the commute will be less onerous, and perhaps this will ease the pressure on urban rents. However, one might have said the same thing about IT allowing remote work, and that hasn’t had a significant effect.

It’s an interesting exercise to brainstorm possible future developments that really could have as much impact on the lives of ordinary people in the developed world as the developments of the second IR did–whether or not the technology seems to be on the horizon. I found it quite difficult. Here are the candidates I came up with:

-A nearly costless energy source, such as nuclear fusion. (It’s interesting to note that, while there is a lot of technological development going on in energy right now, most of it is geared toward ameliorating the bad impact of climate change rather than achieving a positive impact.) This would drastically reduce the cost of a lot of things. In some sense it still wouldn’t change our lives that much, since we already have electrical power and just have to pay for it. I do think it would be a plausible candidate for reducing or removing the need for people to work full time, which would be a big change.
-A medical breakthrough that essentially defeats aging. This is the only health-related breakthrough (again, focused on the developed world) that I could think of that would be as significant as defeating infant mortality. Plenty of people are working on this right now, though it’s not at all clear how feasible it will be. It’s also worth noting that it’s not at all clear that achieving this would end up being an unmitigated good; see lots of sci-fi books.

This is honestly about it. Depending on your perspective, this exercise may make you feel good about where we are today, or feel bad about how many of our current problems are political/distributional, and not really solvable with technology. What are the other things people might mention? I think space travel and colonization are cool but wouldn’t really be that beneficial to people’s lives, especially since even an earth ravaged by climate change would be much more habitable than Mars. AI gets hyped a lot, and futurists tend to put a lot of stock into the idea that super-intelligent AIs could come up with great ideas that we can’t even conceive of, but it’s hard for me to deal with that concretely. Improving recommendations or being a digital butler don’t seem like that big of improvements to me.

I’m not sure exactly what the take-aways from the book should be. Certainly inventors should keep working on cool stuff and try to prove him wrong. (And there is plenty of good technological work to be done in combating climate change and extending benefits of already-existing innovations to the global poor.) I suppose the strongest case to be made is toward orientation of policy actors. Trump’s budget has rightly gotten a lot of flak for rosy growth assumptions (and double-counting), but I think in general if budget plans were forced to assume a very limited baseline for near-term growth, policy orientation might change a lot (and inequality might appear to be a much larger issue). I also wonder how much Gordon’s lessons would imply for financial stability. One can obviously argue that the early-2000s tech bubble stemmed in part from assuming that the IT revolution would have economic consequences of a similar magnitude to the second IR. Can something similar be said for the housing bubble and 2008 collapse? Might monetary policy have been different under a drastically reduced assumption on potential output? I would think yes. Policymakers do seem to be slowly coming around to this viewpoint; see for example the consistently lowering estimates of potential output growth (

It’s telling to me that one of the stronger “contra” arguments to the book, by Tyler Cowen (, essentially centered on the possibility that Gordon was underestimating the possibility of totally unforeseen breakthroughs–certainly not impossible, but notably not at all a defense of the importance of recent technological developments. (And note that Cowen sounded a lot of the same notes as Gordon in his extended essay “The Great Stagnation” several years earlier!)

My Goodreads rating: 4 stars