:format(webp)/cdn.vox-cdn.com/uploads/chorus_image/image/70821429/GettyImages_960478834.0.jpg)
Yesterday, the Bureau of Economic Analysis announced that the US economy contracted by 1.4 percent in the first quarter of 2022, the first quarterly drop in GDP since the pandemic hit in 2020.
That’s an alarming-seeming but ultimately misleading number; most of it is accounted for by changes in international trade (imports, which reduce GDP numbers, rose as businesses knew war in Ukraine was coming) and the shrinking of inventories held by businesses as consumers bought them up. “Final sales to domestic purchasers,” which strips out those factors, was up 0.6 percent for the quarter, or 2.6 percent annually.
However you parse the release, though, this seemed as good a time as any to discuss the long-run future of growth in the US and the rest of the world. I’m getting pretty worried about it, to be honest! And I’m worried because of an economics paper by NYU’s Thomas Philippon with the unassuming name “Additive Growth.”
The implication of Philippon’s paper is as simple as it is disturbing: We should expect economic growth to slow down in the long run, and the big leaps forward of the last couple centuries may be an aberration.
This conclusion is far from certain, and it goes against decades of assumptions on how to model economic growth. But Philippon brings a lot of data to bear on his thesis, which makes some intuitive sense, and even the possibility of it being true should alarm us.
The big pile of stuff we know how to do
Philippon’s paper is not concerned with economic growth per se, but with a variable that is central to explaining long-run growth: total factor productivity, or TFP.
Defining or getting a handle on the idea of TFP can be tricky. Technically, it’s just a residual: the annual growth rate of TFP is what you get when you look at annual economic growth, and remove the growth attributable to an increase in labor (more hours worked) or capital accumulation (more factories built, labor-saving machines purchased, etc).
For now, think of TFP as measuring something like “how well humans are able to use labor and tools to do stuff.” If TFP grows, that means we can get more economic output out of the same people and stuff we already have.
That makes TFP the “secret sauce” behind economic growth more generally. Labor inputs can increase, of course — but people can only work so many hours, and don’t really want to work too many hours. Population growth helps there, but less so when it comes to per capita economic growth, which arguably is what matters most.
Capital accumulation — using more labor-saving equipment and tools — helps too, but there’s only so much money to invest. The key is using the resources you have more effectively — and TFP measures, roughly, how effectively we’re using our resources. We can get more effective at using our resources through advancements in science, business management, and other changes.
Economic growth is generally modeled exponentially: our economic output grows by a set percentage every year, and while that percentage varies, it also compounds on itself. TFP is usually modeled the same way. If you have $100 growing 2 percent each year, that’s exponential. If it instead just gains $2 every year, that’s linear.
What Philippon does is attempt to assess whether TFP actually does, in practice, grow exponentially. He first looks at two datasets covering TFP in the US and finds, instead, linear growth since World War II: TFP does not increase by a set percentage each year, but a set amount (0.0245 points, if you’re curious) each year. It doesn’t compound; it just gradually, steadily grows. You’re getting $2 a year, not 2 percent of an ever-increasing pile.
Extending the data back to 1890, he finds linear growth, but with a break: slower growth from 1890 to 1933, and faster after 1933, but steady and non-exponential in each period. He then extends the analysis to 23 relatively wealthy countries, from Japan to Germany to Spain. A linear model fits better here, too.
Linear growth implies, as Philippon writes, that “new ideas add to our stock of knowledge; they do not multiply it.” It also implies a slowing of economic growth over the long run. Or, as Jason Crawford, who writes about the history of science and technology at the Roots of Progress, put it, “GDP per capita can continue to grow without bound, but that growth will slow over time.”
How doomed are we, exactly?
The US and other rich countries have experienced a well-documented decline in productivity growth, especially TFP growth, since 2004 or so. Philippon’s findings could help explain why that is. The slowdown is only there if you assume TFP should be growing exponentially. If you assume mere linear growth, it’s not that things have gotten worse in recent decades. It’s just that they were never that good.
That’s an alarming conclusion, mostly because from the standpoint of human history, the past few centuries have been very good. Before the 17th to 18th century or so, human economies grew extremely slowly. Agriculture showed little productivity growth, meaning there was a fixed population that farming societies could support. Living standards varied mostly based on how many people were around; when the population suddenly shrank (as in the Black Death in Europe) people grew richer on a per capita basis, and when the population swelled the opposite occurred. This is known as the “Malthusian trap.”
“Until about 1800, the vast bulk of people on this planet were poor,” Joel Mokyr, an economic historian at Northwestern, once noted. “And when I say poor, I mean they were on the brink of physical starvation for most of their lives.”
That pattern started to break down in the 17th through 19th centuries, a process sometimes shorthanded as the “Industrial Revolution,” but including a wide variety of cultural, scientific, technological, and economic changes. Long story short: productivity sustainably grew for the first time in human history. And it grew, by historical standards, quite rapidly, such that a far lower share of people alive in 2022 are on the brink of starvation than were in 1800, even though the population needing food has never been greater.
Philippon identifies this break in human history in his paper, looking at TFP data about England. TFP growth is linear throughout English history, he emphasizes, but the rate of linear growth became suddenly higher around 1650 and 1830. These “break points,” he argues, correspond to the so-called first and second industrial revolutions (the first characterized by textile work and the discovery of steam power, the latter by mass production, industrial steel/plastic, electricity, etc.) beginning in England. I’m less persuaded of this than Philippon — the precise timing of the industrial revolution(s) is among the most hotly debated topics in all of history, and 1650 in particular feels a bit early to date the start of it.
But if he’s right, that means that humanity would need another, similarly important break to prevent economic growth from slowing long-term.
Philippon’s paper is bracing, but I’m not 100 percent sold. I share Tyler Cowen’s question about the usefulness of treating TFP as a real, observable attribute of the economy.
What is TFP, anyway? It’s sometimes shorthanded as a measure of “technological progress,” but it doesn’t really measure technological change; TFP can grow or shrink without changes in technology, and technological changes can occur without affecting TFP. The late economist Moses Abramovitz famously dubbed it a ”measure of our ignorance about the causes of economic growth” in a 1956 paper.
We should try to shrink that ignorance — but how meaningful are changes in the unexplainable share of economic growth? They probably mean something, but it’s hard to say what. TFP has been incredibly useful for comparing productivity between, for instance, firms, to identify which are more effective and efficient; here, for example, is a great paper on productivity of smallholder African farms.
But I’m not sure TFP holds the same explanatory weight in explaining the growth of countries over a long period of time.
All the same, Philippon’s paper should at the very least open up an important new direction for research. One observer I respect concluded, after reading the paper, that if true, it means “Human extinction looks much likelier.” I wouldn’t go that far. But it asks an important question, and answering that question correctly matters a lot.
