Progress Is Measured in Kilowatt-Hours and Error Bars

In 2025, the average American used about 75,000 kilowatt-hours of primary energy. The average person in France used about 37,600. In Ethiopia, in 2024, the figure was under 700, and in the Democratic Republic of the Congo just over 300. These numbers come from Our World in Data, which compiles them from the Energy Institute's Statistical Review of World Energy, and I could build an entire essay on them: a hundredfold gap between the people who can run a refrigerator, a hospital and a steel mill and the people who cannot. That gap is the most important inequality on earth.

But before I build anything on those numbers, I owe you the line printed beneath the chart, in small type: "Traditional biomass is not included." The firewood and charcoal on which hundreds of millions of people still cook do not appear. So the Ethiopian figure does not measure all the energy an Ethiopian burns. It measures the modern energy she can buy: electricity, fuel, gas. That is still the number that matters most for what a life can become. But it is a different number than the one the headline suggests, and anyone who quotes it without the footnote is already, in a small way, lying.

That is the whole of my argument in two paragraphs. Progress is measured in kilowatt-hours, because energy is the currency of everything people do with their hands and machines. And progress is measured in error bars, because a society that cannot be honest about what it knows will spend its kilowatt-hours on the wrong things. I believe in both. I have noticed that the people who love the first often despise the second, and vice versa. I think that divorce is one of the quiet disasters of our time.

The meter on the wall

Start with energy, because it is the part people are now embarrassed to defend. The historian of energy Vaclav Smil has spent a career showing that every civilization can be read as a story of the energy it commands, from muscle to wood to coal to oil to fission. You do not need to accept every word of that story to see its core. A tractor is a hundred farmhands who never tire. A power grid is light, refrigeration, clean water and an intensive-care unit. When I hear that a country "needs less energy," my first question is: less for whom?

Look at the people who still have none. The International Energy Agency counted 730 million people without electricity in 2024, a decline of only 11 million from the year before. The official United Nations tracking report, published in June 2026, puts the same year at 655 million. Two serious institutions, one year, 75 million people apart. The difference is not fraud. The IEA builds its figure from national administrative and utility data and off-grid sales, and it refuses to count the smallest solar lanterns, under 10 watts, as real access; the official report relies on household surveys. Both are measuring something true. They are not measuring the same thing. Keep that in mind; we will need it later.

What both agree on is the direction of the problem. Sub-Saharan Africa accounts for most of the deficit, over 560 million people in the official count. And the IEA adds a detail that should end the romance of "small is beautiful": while electricity access in the region rose from around 30 percent of the population in 2012 to 50 percent in 2024, average household consumption among the connected fell by around a quarter. Tariffs rose faster than incomes, and many new connections are tiny solar systems. More people have a light bulb. Fewer can run a business. A connection that powers a phone charger is better than darkness. It is not development.

Then look at the rich countries, and the story turns strange. In 2000, the average American used about 93,000 kilowatt-hours; by 2025, about 75,000. France went from about 51,600 to 37,600, Germany from about 48,300 to 33,200. Per person, the richest societies on earth now use markedly less energy than they did a generation ago. You can read that two ways. Part of it is genuine efficiency: better engines, better insulation, LEDs instead of incandescent bulbs, the same comfort for fewer kilowatt-hours. That is progress, and I applaud it. Part of it, I suspect, is industry that moved elsewhere, its energy now counted in another country's statistics and its products shipped back to us. The chart does not tell you which share is which. An honest reader says so instead of choosing the reading that flatters his politics.

The footnote under the chart

Which brings me to the second discipline. In 2015, a large team of researchers organized as the Open Science Collaboration tried to replicate 100 studies published in three leading psychology journals, using the original materials wherever possible and much larger samples. Ninety-seven percent of the original studies had reported statistically significant results. Thirty-six percent of the replications did. On average, the effects found in the replications were half the size of the originals. Nobody had to commit fraud to produce that gap. It is what happens when a field rewards novelty, publishes what is surprising, and lets the footnotes shrink.

Set that beside what physics did in June 2025. After measuring the wobble of some 300 billion muons, the Muon g-2 collaboration at Fermilab published its final value of the muon's magnetic anomaly with a precision of 127 parts per billion. The experiment's spokesperson, Peter Winter, compared it to weighing a bison to the nearest 100 milligrams, roughly a sunflower seed. For two decades, that measurement had seemed to disagree with theory, and the disagreement was the most exciting hint of new physics in a generation. Then the theorists did the unglamorous thing. The Muon g-2 Theory Initiative dropped the data-driven calculations that had become internally inconsistent and averaged lattice computations instead. Its prediction, with an uncertainty of 540 parts per billion, now agrees with the experiment within the error bars. The revolution did not come. Physics got something better: a number it can trust. I wrote about that magnet once already, because it refused to flatter anyone, including the people who had hoped it would.

This site keeps returning to that kind of story. A celebrated signal of life on K2-18b appeared in only one of eight ways of slicing the same data. A new model of Thwaites Glacier doubles projected sea-level rise, and the doubling hangs on one parameter. In each case the interesting fact is not the headline number. It is the width of the bar around it.

Where the two disciplines meet

The best place I know to watch kilowatt-hours and error bars work together is a French document most Americans have never heard of and most French people have never read: Futurs énergétiques 2050, the long-range study published by RTE, the operator of France's high-voltage grid, in October 2021 and completed in February 2022. It begins with a correction to a national myth. Nuclear power produces 70 percent of France's electricity, RTE notes, but represents less than 20 percent of the final energy used in the country. Fossil fuels still covered more than 930 terawatt-hours of final consumption a year, against 430 for electricity. The French like to believe they decarbonized. They decarbonized their power plants.

RTE's baseline then runs against the fashionable mood. To leave fossil fuels, France's electricity consumption must rise, by about 35 percent over thirty years, to 645 terawatt-hours in 2050. Total final energy, meanwhile, is supposed to fall by 40 percent, and here RTE is careful about how. Most of that fall is efficiency, not deprivation: an electric motor converts close to 90 percent of its energy into motion, against 25 to 35 percent for a combustion engine; better buildings and appliances do the rest. Then the report addresses "sufficiency," the polite word for consuming less by choice or by rule. It estimates that a sufficiency path could shave another 90 terawatt-hours, about 15 percent, off electricity demand, and it adds a sentence I wish every politician would frame: reaching that goal "would require doing more than sharing slogans about the benefits of consuming less," because it implies radical changes in how society lives and organizes itself.

That is where I part company with the cult of less. Efficiency is engineering: the same life for fewer joules. Sufficiency, when it is imposed, is something else, a social program dressed as physics. A free people may choose to drive less or heat less, and good for them. A state that plans scarcity and calls it virtue is rationing, and the history of rationing is not a history of liberty. I would rather spend public effort making clean energy abundant than making citizens feel guilty for wanting it.

What makes RTE's study a model, though, is not its conclusion. It is its honesty about uncertainty. It compared six generation mixes. Under central assumptions, the annualized full cost of the system in 2060 ranged from roughly €59 billion a year for the most nuclear scenario to roughly €80 billion for the costliest all-renewable one, and RTE concluded "with a good level of confidence" that mixes keeping at least 40 gigawatts of nuclear capacity could be cheaper in the long run than a 100 percent renewable system built on large farms. Then it did what political documents almost never do: it published the stress tests. If new reactors cost as much as the troubled Flamanville EPR, the advantage shrinks but survives. If new nuclear is very expensive and every renewable technology cheap, the gap closes and can even reverse. The report draws the error bars around its own conclusion. That is why I trust its conclusion more, not less.

Optimism with a ruler

People sometimes ask whether I am an optimist. Yes, and a demanding one. Optimistic because the physics allows a world where an Ethiopian child has as much electricity as a French one without cooking the planet; France's own grid proves that abundant low-carbon power is not a fantasy. Demanding because I have watched too many promises sold without their footnotes: a breakthrough that turns out to be a press release, a model that hangs on a single parameter, an AI boom that loses a fight with a 22-meter power pole because nobody budgeted for the wire. The cleanest power on offer is sometimes the least glamorous, as when Google paid to upgrade eleven existing reactors rather than build a new one. Measurement keeps you humble about what you have; energy keeps you ambitious about what you could have.

The two disciplines need each other more than their partisans admit. Energy without honest measurement produces boondoggles: subsidies chasing slogans, projects justified by numbers nobody checked. Measurement without energy produces a clerical civilization that knows exactly how poor it is and has decided to stay that way. The muon needed a superconducting magnet, a particle accelerator and a great deal of electricity; it also needed people willing to say, after twenty years, that the exciting anomaly had gone away.

The same pairing sits underneath the other two pillars of this site. A society that blocks its builders with fees and gates will never deliver abundant energy, which is the argument of The Toll Booth Society. And a nation whose energy runs through someone else's refinery or someone else's switch is not fully its own, which is why I wrote Sovereignty Is a Supply Chain.

So when someone tells me that progress is over, I ask for two numbers. How many kilowatt-hours does the poorest person on earth command, and is that rising? And how wide is the error bar on the claim you just made? A civilization that can answer both honestly, and keeps pushing the first number up while shrinking the second, is making progress, whatever the mood of the decade. One that answers neither is just telling itself stories in the dark.