A scarcity chosen

In 1978 Americans used more energy per person than they have in any year since. The country has gone on getting richer on paper while using close to a fifth less energy per head. That is a strange shape for a growth story, and it is the shape of the last half-century in every advanced economy.

Economists have a name for countries that stall on the way up, the middle-income trap. What the rich world fell into is a different trap, and it built this one itself. Of the three terms in the equation, the one that broke was energy, and it broke because people chose to stop expanding it, and then built everything else on the assumption that it would never grow again.

Three pieces of evidence.

Figure 3-1 The climb that stopped at the end of the 1970s

US primary energy consumption per capita, GJ per year

Year	GJ per person
1950	232.1
1951	244.8
1952	238.4
1953	240.6
1954	231.1
1955	249.0
1956	254.3
1957	250.0
1958	244.8
1959	251.1
1960	256.4
1961	256.4
1962	263.8
1963	270.1
1964	277.5
1965	285.9
1966	298.6
1967	303.9
1968	319.7
1969	332.3
1970	339.7
1971	341.8
1972	355.6
1973	367.2
1974	354.5
1975	340.8
1976	357.7
1977	366.1
1978	369.3
1979	369.3
1980	353.4
1981	340.8
1982	322.8
1983	318.6
1984	332.3
1985	329.2
1986	327.1
1987	335.5
1988	349.2
1989	353.4
1990	348.2
1991	342.9
1992	345.0
1993	346.1
1994	349.2
1995	351.3
1996	357.7
1997	355.6
1998	354.5
1999	356.6
2000	361.9
2001	349.2
2002	350.3
2003	348.2
2004	353.4
2005	350.3
2006	343.9
2007	347.1
2008	335.5
2009	315.5
2010	325.0
2011	318.6
2012	308.1
2013	314.4
2014	316.5
2015	311.2
2016	307.0
2017	304.9
2018	314.4
2019	310.2
2020	282.8
2021	296.5
2022	299.6
2023	293.3
2024	293.3

The annual series begins in 1949; the chart starts at 1950.

Source: U.S. EIA, Monthly Energy Review, Table 1.7 (Million Btu per capita, converted at 1.055056 GJ/MMBtu).

First, energy use per person stopped rising. On the EIA's annual series, American primary energy consumption per head climbed from 207 million Btu in 1949 to a peak of 350 in 1978 and 1979, and has since fallen back to 282. Some of that is efficiency, but efficiency gains have historically raised total consumption rather than lowered it, so the fall reads as a halt to expansion, not as its reward.

Second, productivity slowed in the same years. Take productivity in the broad sense, how much output the same labor and capital produce. In Robert Gordon's accounting it grew at roughly 1.9 percent a year in the United States from 1920 to 1970 and at about a third of that pace since. A model that explains growth by the share of energy actually turned into work fits American output until 1973 and 1974 and then starts to fall short.

Third, the country stopped building its energy supply. Every one of the 41 reactors ordered in the United States after 1973 was canceled, and more than two thirds of all reactors ordered after January 1970 were canceled in the end.

Where the difference came from. American reactor costs were already climbing through the late 1960s and went far higher for the units still under construction when Three Mile Island happened. Over the same years France held its cost escalation to a fraction of the American rate and Korea brought its costs down outright, both by building the same standardized design again and again. One technology escalated in one country and fell in another. The difference was regulation and politics, and beneath both, whether a country built the same plant twice or a new one every time.

So the turn of the 1970s was chosen by the people of the advanced economies themselves. There was a reasonable reaction to accidents and to the environment. There was an adjustment to OPEC's price rises. And there was the West's own doubt about whether it should keep growing at all. Whatever the reason, the mistake lay less in stepping down than in replacing nothing. For half a century the rich countries have since rebuilt their economies, their institutions and their expectations on the assumption that energy will never grow again. That assumption is the scarcity I mean.

The exit that was taken. The main new source the advanced countries turned to in that half-century is one whose output does not follow demand. Lion Hirth's study of wind found that its market value, its price relative to the average power price, falls from about 110 percent when there is almost none of it to between 50 and 80 percent once it supplies 30 percent of all electricity. The more wind there is, the more every other turbine is producing in the same hours, so wind lowers the value of the very electricity it makes. German wholesale prices were negative for 573 hours in 2025, a record for the second year running. That is more than three weeks a year in which generators pay to have their output taken away.

I am not against renewables. I am saying they belong on top of power that runs whenever it is needed, from nuclear, gas or hydro, and cannot by themselves be the base a country runs on.

The same choice went further than anyone noticed. Which power plants a country builds sets its electricity price. That price feeds into rent, food and services, and those together set how many children a household can afford. Voters never saw that chain on a ballot, but the price system delivered the bill to every household all the same.

Where the constraint sits now. Half a century on, the bottleneck has moved from generating power to delivering it, from the plant to the people who use it, though most of the debate is still about generation.

Generation itself has grown exponentially. Installed solar capacity has grown about a millionfold since 1975, and China alone now holds close to half of its installed capacity in wind and solar. None of that has made delivered energy grow exponentially, because everything between the plant and the user moves at a different pace: the queue to be connected to the grid, the factories that make transformers and switchgear, the land and permits for transmission lines, the substations. All of it moves in years.

In the United States, the median project that entered operation in 2025 had waited five years from application, and of the projects that applied between 2000 and 2020 only 13 percent, counted by capacity, ever reached operation. Applying for a new connection today buys a chance of being connected rather than a date. A large transformer now takes two to three years to deliver, two to three times what it took in 2010. However many factories are added, each of these lead times stays measured in years, and all of them tightened across the major markets at once from the early 2020s.

Two things follow. Prices in the transition rise before they fall, because demand explodes while supply is held back. And the scarce asset is the site already connected to the grid, and the place in the queue, rather than the power plant itself.

Electricity does not travel far and land does not travel at all. Of the three terms, institutions and labor can be borrowed from a partner abroad; energy has to be where the plant is. So choosing the site is the investment decision.